New Results

Probing the evolutionary robustness of two repurposed drugs targeting iron uptake in Pseudomonas aeruginosa

Chiara Rezzoagli, David Wilson, Michael Weigert, Stefan Wyder, Rolf Kümmerli
doi: https://doi.org/10.1101/195974

Abstract

Treatments that inhibit the expression or functioning of bacterial virulence factors hold great promise to be both effective and exert weaker selection for resistance than conventional antibiotics. However, the evolutionary robustness argument, based on the idea that anti-virulence treatments disarm rather than kill pathogens, is controversial. Here we compared the evolutionary robustness of two repurposed drugs, gallium and flucytosine, targeting the iron-scavenging pyoverdine of the opportunistic human pathogen Pseudomonas aeruginosa. After exposing bacteria to treatments for 20 days in human blood serum, as an ex-vivo infection model, we found that resistance against flucytosine quickly arose and spread in all populations. Genetic analysis revealed that mutations in upp, a gene encoding an enzyme required for flucytosine activation, are responsible for resistance evolution. Conversely, resistance against gallium arose only sporadically. Resistance mechanisms were based on mutations in transcriptional regulators, which resulted in the upregulation of pyocyanin, a redox-active molecule promoting siderophore-independent iron acquisition. Our work highlights that mutants resistant against anti-virulence treatments can easily arise, but their selective spreading varies considerably between treatments. This indicates that anti-virulence treatments are not evolutionarily robust per se. Instead, evolutionary robustness is a relative measure, with specific treatments occupying different positions on a continuous scale.

Introduction

There is currently much interest in therapeutic approaches that inhibit the expression or functioning of bacterial virulence factors (Escaich, 2008; Rasko and Sperandio, 2010; LaSarre and Federle, 2013; Maura et al., 2016; Vale et al., 2016; Johnson and Abramovitch, 2017; Rampioni et al., 2017; Dickey et al., 2017). Virulence factors are structures and molecules that allow bacteria to establish and maintain infections (Rahme et al., 1995; Balasubramanian et al., 2013). Examples of virulence factors include flagella and pili to adhere to the host tissue, secreted enzymes, tissue-damaging toxins and siderophores to scavenge iron from the host (Wu et al., 2008). Approaches that target these traits are called anti-virulence treatments. There is great hope that disarming rather than killing pathogens is an efficient and evolutionarily robust way to manage infections (André and Godelle, 2005; Baron, 2010; Rasko and Sperandio, 2010; Pepper, 2012; Allen et al., 2014). In particular, it is assumed that anti-virulence treatments exert weaker selection for resistance than conventional antibiotics because pathogens are not killed directly. However, empirical evidence for the evolutionary robustness of anti-virulence treatments is controversial with positive and negative reports currently balancing each other out (Mellbye and Schuster, 2011; Maeda et al., 2012; García-Contreras et al., 2013; Ross-Gillespie et al., 2014; Gerdt and Blackwell, 2014; Sully et al., 2014).

The controversy entails both conceptual and practical aspects. On the conceptual level, some define anti-virulence approaches as treatments that specifically target virulence factors without affecting pathogen growth (Clatworthy et al., 2007; Rasko and Sperandio, 2010), while others argue that it is unlikely that virulence factors do not effect pathogen fitness, and thus simply use the mechanistic part of the definition (Allen et al., 2014; Vale et al., 2016). On the practical level, there are debates about what exactly a resistance phenotype is (Allen et al., 2014), as it could include restoration of virulence factor production, growth (if affected), and/or the activation of a bypassing mechanism, restoring the virulence phenotype (Ross-Gillespie et al., 2014). Moreover, there is a shortage of studies examining resistance evolution under realistic conditions in replicated populations, both at the phenotypic and genetic level.

Here, we tackle these issues by examining the mechanistic and evolutionary potential of resistance evolution against two repurposed drugs, gallium and flucytosine, which both target the iron-scavenging pyoverdine of the opportunistic human pathogen Pseudomonas aeruginosa (Kaneko et al., 2007; Imperi et al., 2013; Ross-Gillespie et al., 2014). Pyoverdine is an important virulence factor during acute infections (Meyer et al., 1996; Takase et al., 2000; Harrison et al., 2006; Cornelis and Dingemans, 2013; Ross-Gillespie et al., 2014; Bonchi et al., 2014; Granato et al., 2016; Weigert et al., 2017). It is required to obtain iron from host proteins, such as transferrin and lactoferrin (Valenti et al., 2004). Given its importance, it has been proposed that drugs interfering with iron uptake could be effective therapeutics to control infections (Smith et al., 2013). Gallium and flucytosine both fulfill this role, albeit through different modes of action. Gallium, a repurposed cancer drug, is an iron-mimic and binds irreversibly to secreted pyoverdine, thereby rendering the molecules useless for iron uptake (Kaneko et al., 2007; Ross-Gillespie et al., 2014; Weigert et al., 2017). Flucytosine, a repurposed anti-fungal drug, enters the bacterium, where it is enzymatically activated to a fluorinated ribonucleotide. This active form inhibits, via a yet unknown mechanism, the expression of the pvdS iron starvation sigma factor controlling pyoverdine synthesis (Visca et al., 2007; Imperi et al., 2013).

In a first set of experiments, we examined whether these two drugs affect the growth of P. aeruginosa in human blood serum, a medium that has recently been established as an exvivo infection model (Bonchi et al., 2015). We hypothesize that gallium and flucytosine are likely to reduce pathogen fitness as they induce iron starvation (Kaneko et al., 2007; Banin et al., 2008; DeLeon et al., 2009; Ross-Gillespie et al., 2014). In addition, anti-virulence drugs, like any other drugs, might have off-target effects affecting growth. Gallium at high dosage, for instance, can penetrate into bacterial cells, where it interferes with redox-active enzymes (Chitambar, 2012; Hijazi et al., 2017). Flucytosine, once activated, is known to affect RNA synthesis, which might negatively affect growth (Waldorf and Polak, 1983).

In a second experiment, we examined whether mutants, resistant against these two repurposed drugs, evolve and spread through bacterial populations. To this end, we exposed replicated populations of P. aeruginosa to two different concentrations of gallium and flucytosine in human serum. Together with a drug-free control treatment, we let the treated populations evolve for 20 consecutive days in eight-fold replication, by transferring a fraction of the evolving cultures to fresh human serum on a daily basis. Following experimental evolution, we screened evolved populations and clones for possible resistance phenotypes, including the restoration of growth, restoration of virulence factor production, and the evolution of a bypassing mechanism for iron uptake (Allen et al., 2014; Ross-Gillespie et al., 2014). Finally, we sequenced the whole genome of evolved clones to uncover the genetic basis of potential resistance mechanisms.

Resistance evolution requires two processes: the supply of mutations conferring resistance and appropriate selection regimes favoring the spread of these mutants (Hughes and Andersson, 2017). With regard to mutation supply, we predict gallium to show higher evolutionarily robustness than flucytosine because gallium is an ion that targets a secreted virulence factor outside the cell. Thus, common resistance mechanisms including drug degradation, the prevention of drug influx, and increased drug efflux cannot apply for this drug (Ross-Gillespie et al., 2014). As for the spread of mutants, both drugs could be evolutionarily robust because they target a secreted virulence factor, which can be shared as a public good between pathogen individuals (iron-loaded pyoverdine can be taken up by all bacteria with a matching receptor; Griffin et al., 2004; Inglis et al., 2016). Consequently, if resistance entails the resumption of virulence factor production then resistant mutants should not spread because they bear the cost of resumed virulence factor production, whilst sharing the benefit with everyone else in the population, including the drug-susceptible individuals (André and Godelle, 2005; Mellbye and Schuster, 2011; Pepper, 2012; Gerdt and Blackwell, 2014). Conversely, if these drugs have off-target effects, we predict the evolutionary robustness to decline, and accelerated spread of resistance under drug exposure, as for traditional antibiotics.

Material and Methods

Strains and culturing conditions

We used the genetically well-characterized P. aeruginosa PAO1 wildtype strain for all experiments. For some assays, we further used a set of knockout mutants in the PAO1 background as control strains (see Supplementary Table 1). Overnight cultures were grown in 8 ml Lysogeny broth (LB) in 50-ml Falcon tubes, incubated at 37°C, 200 rpm for 18 hours. For all experiments, we washed overnight cultures with 0.8% NaCl solution and adjusted them to OD600 = 2.5. Bacteria were further diluted to a final starting of OD600 = 2.5 × 10−3. All experiments were carried out in human serum, supplemented with HEPES (50 mM) to buffer the medium at physiological pH, and the iron chelator human apo-transferrin (100 μg/ml) and its co-factor NaHCO3 (20 mM) to impose iron limitation. We used gallium (GaNO3) and flucytosine (5-Fluorocytosine) as anti-bacterials. All chemicals, including human serum, were purchased from Sigma-Aldrich, Switzerland.

Growth and virulence factor inhibition curves

To assess the extent to which gallium and flucytosine inhibit PAO1 growth and pyoverdine production, we subjected bacterial cultures to a seven-step antibacterial concentration gradient: 0 - 512 μM for GaNO3 and 0-140 μg/ml for flucytosine. Overnight cultures of bacteria were grown and diluted as described above and inoculated into 200 μl of human serum on 96-well plates. Plates were incubated at 37 °C in a Tecan Infinite M-200 plate reader (Tecan Group Ltd., Switzerland). We tracked growth by measuring OD at 600 nm and pyoverdine-associated natural fluorescence (excitation: 400 nm, emission: 460 nm) every 15 minutes for 24 hours. Plates were shaken for 15 seconds (3 mm orbital displacement) prior to each reading event.

Experimental evolution

We exposed wildtype cultures of PAO1 to experimental evolution for 20 days under five different selective regimes in eight-fold replication. The five regimes included a no-drug control, and a low and a high concentration treatment for both drugs (gallium: 50 μM and 280 μM; flucytosine: 10 μg/ml and 140 μg/ml). The antibacterial concentrations were inferred from the dose-response curves (Figure 1). To initiate experimental evolution, an overnight culture of PAO1 was grown as described above, and individual wells on a 96-well plate were inoculated with 10 μl of culture (diluted to a final density of 106 cells per well) in 190 μl iron-limited human serum. Incubation occurred in the plate reader at 37°C for 23.5 hours, and OD600 was measured every 15 minutes prior to a brief shaking event. Subsequently, cultures were diluted in 0.8% NaCl and transferred to a new plate containing fresh media. We adjusted the dilution factor proportional to the overall growth per treatment; no-drug control: 2 × 10−3 (day 1-10) and 4*10−3 (day 11-20); antibacterial treatments: 10−3 (day 1-10) and 2 × 10−3 (day 11-20). Following transfers, we added 100 μl of a 50% glycerol-LB solution to cultures for storage at −80°C.

Figure 1.
Figure 1. Gallium and flucytosine affect both growth and pyoverdine production of P. aeruginosa in human serum (HS).

Both anti-virulence drugs reduce growth of bacterial cultures in a dose-dependent manner (A, B), albeit following different patterns: gallium curbs bacterial growth only at relatively high concentrations (A), whereas flucytosine already reduces growth at low concentrations (B). Both drugs further affect pyoverdine production (C, D). When increasing gallium exposure, bacteria first upregulate pyoverdine production at intermediate drug concentrations, but then down-scale investment levels at high drug concentrations (C). In contrast, flucytosine administration leads to an almost complete abolishment of pyoverdine production even at the lowest drug concentration. All data are scaled relative to the drug-free treatment. Error bars denote standard errors of the mean across 6 (for flucytosine) and 18 (for gallium) replicates. Dose-response curves were fitted using either Weibull (for gallium) or Gompertz (for flucytosine) functions (see material and methods for details). Vertical lines indicate the drug concentrations used in the experimental evolution.

Quantification of resistance profiles

To test whether populations evolved under antibacterial exposure restored growth and/or pyoverdine production, we exposed evolved lineages to the drug concentrations they experienced during experimental evolution in 5-fold replication. Following shaken incubation at 37° C (160 rpm) for 24 hours, we compared the OD600 and pyoverdine-associated fluorescence of evolved lineages relative to the ancestor wildtype grown under drug and nodrug treatment.

To assess potential resistance profiles of individual clones, we streaked out aliquots of evolved lineages onto LB plates. After overnight incubation at 37°C, we randomly picked 200 clones (five colonies per lineage), and assessed their growth and pyoverdine production in 3-fold replication, as described above. Moreover, we performed an in-depth analysis for 16 randomly picked single clones (four per drug treatment) by quantifying their drug-inhibition curve, following the protocol described above.

To test whether bacteria upregulated alternative iron-acquisition mechanisms, we quantified pyocyanin and protease production of selected clones. For pyocyanin production, overnight bacterial cultures were inoculated into 1 ml of LB (starting OD600 = 10−6), and incubated at 37°C for 24 hours, shaken at 160 rpm. We measured pyocyanin in the cell-free supernatant through absorbance at 691 nm (Ross-Gillespie et al., 2014). For protease production, overnight bacterial cultures were inoculated in human serum (starting OD600 = 2.5 × 10−3), and incubated at 37°C for 24 hours, shaken at 160 rpm. Subsequently, we centrifuged cultures at 3700 rpm for 15 minutes to obtain protease-containing supernatants. To measure proteolytic activity, we adapted the protocol by (Chessa et al., 2000): 0.1 ml azocasein solution (30 mg/ml) were mixed with 0.3 ml 50 mM phosphate buffer (pH 7.5), and 0.1 ml culture supernatant. During incubation at 37°C (2 hours), proteases hydrolyze azocasein and release the azo-dye. Proteolytic reaction was stopped by adding 0.5 ml 20% trichloroacetic acid (TCA), samples centrifuged at 12000 rpm (10 min), and proteolytic activity measured through absorbance of the azo-dye at 366 nm.

Sequencing analysis

We further isolated the genomic DNA of the selected 16 clones and re-sequenced their genomes. We used the GenElute Bacterial Genomic DNA kit (Sigma Aldrich) for DNA isolation. DNA concentrations were assayed using the Quantifluor dsDNA sample kit (Promega). Samples were sent to the Functional Genomics Center Zurich for library preparation (Nextera XT) and sequencing. Sequencing was performed on the Illumina HiSeq 4000 platform with single-end 125 base pair reads. Adapter sequences were clipped using Trimmomatic v0.33 (Bolger et al., 2014) and reads trimmed using Flexbar v2.5 (Dodt et al., 2012). We aligned the reads to the PAO1 reference genome using BWA v0.7.12 (Li and Durbin, 2009). We applied GATK v3.5 (McKenna et al., 2010) indel realignment, duplicate removal and HaplotypeCaller SNP/INDEL discovery according to the GATK Best Practices recommendations. This generated a variant call format (VCF) file, from which the following variants were discarded: (i) coverage < 20 reads; (ii) Fisher Strand (FS) score > 30.0, ensuring that there is no strand bias in the data; (iii) QD value < 2.0 (confidence value that there is a true variation at a given site); and (iv) clustered variants (≥ 3 variants in 35nt window) as they likely present sequencing or alignment artifacts. This filtering process yielded a list of potential SNPs and small INDELs, which we annotated using snpEff 4.1g (Cingolani et al., 2012) and then screened manually, compared to the sequenced genome of our ancestor wildtype for relevant mutations in gene coding and intergenic regions (Supplementary Table 2).

Statistical analysis

We used RStudio for statistical analysis (version 0.99.896, with R version 3.3.0). We analyzed growth curves and pyoverdine production profiles using the grofit package (Kahm et al., 2010). We fitted non-parametric model (Splines) curves to estimate growth yield and integral (area under the curve). To analyze dose-response curves, we used the drc package (Ritz et al., 2015). Model selection included the fitting of different non-linear models and choosing the best fit using the following criteria: the log likelihood value, Akaike’s information criterion (AIC), the estimated residual standard error and the p-value from a lack-of-fit test. Weibull and Gompertz functions provided the best fit for gallium and flucytosine dose-response curves, respectively. For all analyses, we scaled growth yield and pyoverdine production relative to the untreated ancestral wildtype. We used general linear mixed effect models to compare whether growth parameters or pyoverdine profiles differ in evolved cultures treated with or without antibacterials. To test for differences between evolved lines and the ancestral wildtype, we used Welch‘s two-sample t-test. To compare the dose-response curve of evolved clones, we first fitted spline curves to the inhibition curves, then estimated the integrals of these fits, and compared the scaled fits relative to the ancestor wildtype using ANOVA (Analysis of variance). Protease and pyocyanin production of evolved clones and the ancestor wildtype were corrected for cell number (OD600) and analyzed using ANOVA.

Results

Gallium and flucytosine curb growth and pyoverdine production in human serum

To confirm that human serum is an iron-limited media, in which pyoverdine is important for growth, we compared the growth of our wildtype strain PAO1 to the pyoverdine-negative mutant PAO1 ΔpvdD in either pure human serum or human serum supplemented with transferrin (Supplementary Figure 1). As expected for iron-limited media, we observed significantly reduced growth of the siderophore-deficient mutants compared to the wildtype (ANOVA: t49 = −8.13, p < 0.0001) under both conditions.

We then subjected PAO1 to a range of drug concentrations. The resulting dose-response curves revealed that both drugs significantly affected growth and pyoverdine production, albeit following different patterns (Figure 1). For gallium, growth reduction was moderate at low concentrations, and only became substantial at high concentrations (GaNO3 ≥ 256 μM, Figure 1A). Gallium treatment affected pyoverdine synthesis in a complex way (Figure 1C), yet consistent with previous findings (Ross-Gillespie et al., 2014): at intermediate gallium concentrations, pyoverdine is up-regulated to compensate for the gallium-induced pyoverdine inhibition, and down-regulated at higher concentrations, when pyoverdine-mediated signaling becomes impaired (Kaneko et al., 2007). For flucytosine, already the lowest concentration caused a substantial growth reduction (Figure 1B) and completely stalled pyoverdine production, with the reduction remaining fairly constant across the concentration gradient (Figure 1D).

Do bacteria evolve population-level resistance to antivirulence treatments?

We subjected PAO1 wildtype cultures to experimental evolution both in the absence and presence of gallium and flucytosine (two concentrations each). Eight independent lines per treatment were daily transferred to fresh human serum for a period of 20 days. Subsequently, we assessed whether evolved populations improved growth and/or pyoverdine production levels compared to the treated ancestral wildtype, which could provide first hints of resistance evolution.

For growth (Figure 2A), we found that evolved lines grew significantly better under drug exposure than the ancestral wildtype (Welch’s t-tests, gallium low (50 μM): t11.9 = −4.96, p = 0.0003; gallium high (280 μM): t13.3 = −6.48, p < 0.0001; flucytosine low (10 μg/ml): t12.2 = − 5.09, p = 0.0002; flucytosine high (140 μg/ml): t7.5 = −11.79, p < 0.0001). Because growth increase could simply reflect adaptation to media components other than drugs, we also analyzed changes in growth performance of the lines evolved without drugs. It turned out that some of the untreated evolved lineages also showed improved growth compared to the ancestral wildtype, but the overall increase was not significant (t9.1 = −1.61, p = 0.1424, Figure 2A).

Figure 2.
Figure 2. Population level growth and pyoverdine production after evolution in human serum.

PAO1 cultures were exposed to either no treatment, low (50 μM) or high (280 μM) gallium, low (10 μg/ml) or high (140 μg/ml) flucytosine concentrations during a 20-day experimental evolution experiment in eight-fold replication. Following evolution, we assessed growth and pyoverdine production of evolved populations (displayed on the x-axis) and compared their performance relative to the untreated (gray solid line, set to 1) and treated ancestral wildtype (black dashed line). (A) Compared to the treated ancestral wildtype, growth of evolved populations significantly increased under all treatment regimes. Growth also increased in some but not all of the non-treatment lines. (B) Pyoverdine production of evolved populations significantly increased relative to the untreated ancestral wildtype under all conditions, also in the no treatment lines. This indicates that increased pyoverdine production might be a general response to growth in human serum, which makes it difficult to disentangle resistance evolution from media adaptation. Error bars show the standard error of the mean across 5 independent replicates.

For pyoverdine production, we observed no significant change for the lines evolved under low gallium concentration (comparison relative to the treated ancestor, Welch’s t-test: t8.8 = 0.94, p = 0.3719) (Figure 2B). Conversely, lines evolved under the other three drug regimes all showed significantly increased pyoverdine production (Figure 2B) (gallium high: t13.1 = −3.69, p = 0.0026; flucytosine low: t7.2 = −7.64, p = 0.0001; flucytosine high: t9.6 = −54.65, p < 0.0001). While the increase was moderate for the gallium high treatment, there was full restoration of pyoverdine production in both flucytosine treatments (no significant difference relative to the ancestral untreated wildtype, ANOVA, flucytosine low: t88 = −1.31, p = 0.1944; flucytosine high: t88 = 0.42, p = 0.6766). Although pyoverdine restoration might be taken as evidence for resistance evolution, analysis of the control lines shows that a significant increase in pyoverdine production also occurred in the absence of drugs (Welch’s t-test, t9.1= −4.03, p = 0.0047, Figure 2B).

Screening for resistance profiles in evolved single clones

While the above-analyses show that drug resistance and general media adaptation could both contribute to the evolved population growth and pyoverdine phenotypes, we decided to screen individual clones for in-depth analysis. In a first step, we isolated 200 random clones (i.e. 40 per treatment), and individually analyzed their growth and pyoverdine production. These analyses revealed high between-clone variation in growth and pyoverdine production (Supplementary Figure 2), suggesting that most evolved populations were heterogeneous, consisting of multiple different genotypes.

In a second step, we randomly picked 16 single clones (four per drug treatment) and tested whether these evolved clones differ in their drug dose response curve relative to the ancestral wildtype. We observed that three out of eight clones subjected to gallium (Figure 3A-3D) and all eight clones subjected to flucytosine showed a significantly altered dose response (Figure 3E-3H). Clones GL_2 and GL_3, evolved under low gallium, showed a significant increase in pyoverdine production under intermediate gallium concentrations (between 8 and 128 μM), which goes along with an improved growth performance for GL_2, but not GL_3. In contrast, clone GH_1, evolved under high gallium concentration, did not show an altered pyoverdine production response, but grew significantly better when exposed to gallium (Figure 3A-3D). For the eight clones evolved under the flucytosine regime, changes in the dose-response curves were both striking and uniform: growth and pyoverdine production were no longer affected by the drug (Figure 3E-3H). Since these dose-response curves directly include a control for media adaptation (i.e. the no-drug treatment), our results indicate that all eight clones evolved complete resistance to flucytosine. For gallium, on the other hand, our data suggest that three out of the eight clones exhibited a phenotype that is compatible with at least partial resistance.

Figure 3.
Figure 3. Changes in dose-response curves for evolved single clones indicate resistance evolution.

16 randomly picked clones, four per treatment, were exposed to a range of drug concentrations to test whether their dose-response altered during evolution compared to the ancestral wildtype (black circles and lines). (A and B) Growth dose-response curves under gallium treatment show that two evolved clones (GL2 and GH1) are significantly less inhibited than the ancestral wildtype. (C and D) Pyoverdine dose-response curves under gallium treatment show that two evolved clones (GL2 and GL3) make significantly more pyoverdine than the ancestral wildtype. (E and F) Growth dose-response curves under flucytosine treatment show that all evolved clones grow significantly better than the ancestral wildtype, and are in fact no longer affected by the drug. (G and H) Pyoverdine dose-response curves under flucytosine treatment show that all evolved clones produce significantly more pyoverdine than the ancestral wildtype, and are in fact no longer affected by the drug. Growth and pyoverdine production were measured after 24 h. For each clone, values are scaled relative to its performance in human serum without drugs (absolute values of pyoverdine and growth in the absence of treatment are reported in Supplementary Figure 3). We used spline functions to fit dose-response curves, and used the integral (area under the curve) to quantify the overall dose response of each clone across the concentration gradient. Error bars denote standard errors of the mean across 6 replicates. Asterisks represent significance levels: * = p < 0.05; *** = p < 0.0001, based on linear model with df= 45).

Linking phenotypes to genotypes

Our whole-genome re-sequencing of the 16 focal clones revealed a small number of SNPs and INDELs, which have emerged during experimental evolution (Table 1). All the clones evolved under flucytosine treatment had acquired mutations in the coding sequence of upp. There were four different types of mutations, including two different non-synonymous SNPs, a 15-bp deletion and a 1-bp insertion (Supplementary Table 1). The upp gene encodes for a uracil phosphoribosyl-transferase, an enzyme required for the intra-cellular activation of flucytosine (Beck and O’Donovan, 2008; Edlind and Katiyar, 2010).

View this table:
Table 1.

List of mutations in evolved single clones.

For the clones evolved under gallium treatment, the mutational pattern was more heterogeneous (Table 1). No mutations were detected for three clones (GH_2, GH_3, GH_4). In contrast, the three clones with significantly altered dose responses had mutations potentially explaining their phenotypes: clone GH_1 featured a 3-nt deletion in mvaU, whereas the clones GL_2 and GL_3 were mutated in vfr. Both genes encode transcriptional regulators involved in the regulation of virulence factors, including proteases, pyocyanin and pyoverdine.

In addition, several clones had mutations in dipA (dispersion-induced phosphodiesterase A; GL_1, GL_4, FH_4) and morA (motility regulator; GL_3, FH_2). The repeated yet unspecific appearance of these mutations could suggest that they represent non-drug-specific adaptations to human serum. Altogether, our sequencing analysis identified three potential targets explaining resistance evolution: the gene upp for flucytosine, and the genes encoding the transcriptional regulators vfr and mvaU for gallium.

Evolution of bypassing mechanisms for iron acquisition under gallium treatment

It was proposed that bypassing mechanisms, which guarantee iron uptake in a siderophore-independent manner, could confer resistance to gallium (Ross-Gillespie et al., 2014). One such by-passing mechanism could involve the up-regulation of pyocyanin, a molecule that can reduce ferric to ferrous iron outside the cell, thereby promoting direct iron uptake (Cox, 1986; García-Contreras et al., 2013). This scenario indeed seems to apply to the three clones mutated in mvaU or vfr, two regulators that control directly (mvaU) or indirectly (vfr) the expression of pyocyanin (Li et al., 2009; Diggle et al., 2007). These clones displayed significantly increased pyocyanin production compared to the ancestral wildtype (Figure 4A; ANOVA, GH_1: t79 = 9.64, p < 0.0001; GL_2: t99 =6.13, p < 0.0001; GL_3: t99 = 14.8, p < 0.0001).

Figure 4.
Figure 4. Upregulation of pyocyanin or protease production as potential bypassing mechanisms for iron acquisition under gallium treatment.

The eight sequenced clones evolved under gallium treatments (low: 50 μM, high: 280 μM) were screened for their change in the secretion of pyocyanin (A) and proteases (B) relative to the ancestral wildtype. Standard protocols were used for the phenotypic screens in drug free media (see Material and Methods for details). All values are corrected for cell number, and scaled relative to the ancestor wildtype (black line). We included the strains PAO1 ΔrhiR (deficient for pyocyanin production) and PAO1 ΔlasR (deficient for protease production) as negative controls in the respective assays (dashed lines). Error bars denote standard errors of the mean across 3 (for proteases) and 8 (for pyocyanin) replicates. Asterisks represent significance levels: * = p < 0.05; *** = p < 0.0001, based on ANOVA.

A second by-passing mechanism could operate via increased protease production, which would allow iron acquisition from transferrin or heme through protease-induced hydrolysis (Doring et al., 1988; Bonchi et al., 2014). We found no support for this hypothesis. In fact, six of the evolved clones exhibited reduced and not increased protease activity (Figure 4B). Moreover, the two clones with significantly increased protease activity (ANOVA, GL_1: t10 = 13.22, p < 0.0001; GL_4: t10 = 11.60, p < 0.0001, Figure 4B) did not show an altered drug dose-response curve.

Inactivation of Upp is responsible for resistance to flucytosine

Next, we tested whether the mutations in upp are responsible for flucytosine resistance. The enzyme Upp (uracil phosphoribosyl-transferase) is essential for the activation of flucytosine within the cell. The natural function of Upp is to convert uracil to the nucleotide precursor UMP in the salvage pathway of pyrimidine (Figure 5A). However, P. aeruginosa can also produce UMP through the conversion of L-glutamine and L-aspartate (Isaac and Holloway, 1968) (Figure 5A), suggesting that upp is not essential for pyrimidine metabolism. Mutations in this gene could thus prevent flucytosine activation, and confer drug resistance. To test this hypothesis, we compared the flucytosine dose-response curve of the wildtype strain to an isogenic (transposon) mutant (MPAO1Δupp). Consistent with the patterns of the evolved clones (Figure 3G), we found that MPAO1Δupp was completely insensitive to flucytosine, with neither growth (Figure 5B) nor pyoverdine production (Figure 5C) being affected by the drug. These results indicate that upp inactivation is a simple and efficient mechanism to become flucytosine resistant.

Figure 5.
Figure 5. Upp is a non-essential enzyme, and mutations in its gene result in flucytosine resistance.

(A) Flucytosine interferes with the pyrimidine metabolism in P. aeruginosa. The drug enters the cell through the transporter CodB (not shown), where it is first converted to fluorouracil by the cytosine deaminase CodA, and then to fluoro-UMP by the uracil phosphoribosyl-transferase Upp. Fluoro-UMP is a modified nucleotide precursor, the action of which results in RNA molecules with compromised functionality. Through an as yet unknown mechanism, fluoro-UMP also arrests pyoverdine synthesis in P. aeruginosa. Importantly, the nucleotide-precursor UMP can also be produced through an alternative denovo pathway from the amino acids L-glutamine and L-aspartate, making Upp a nonessential enzyme in this bacterium. Experiments with the transposon mutant MPAO1 Δupp (deficient for Upp production) indeed demonstrate that the lack of Upp no longer affects strain growth (B) and pyoverdine production (C). This demonstrates that the inactivation of upp is a simple and efficient way to evolve resistance to flucytosine. Experiments were carried out in human serum across a range of flucytosine concentrations. Growth and pyoverdine production of MPAO1 Δupp (gray-dashed lines) and its corresponding wildtype MPAO1 (black-solid lines) were measured after 24 hours for each treatment separately in 6fold replication. All values are scaled relative to the drug-free treatment. For statistical analysis, we compared the integrals of the dose-response curves between the mutant and the wildtype strain (Welch’s f-test, growth: f4.3 = −5.56, p = 0.0041; pyoverdine production: f5.0 = −48.80, p < 0.0001).

Discussion

New treatment approaches against the multi-drug resistant ESKAPE pathogens, to which P. aeruginosa belongs, are desperately needed (Pendleton et al., 2013; Brown, 2015; Dickey et al., 2017). In this context, treatments that disarm rather than kill bacteria have attracted particular interest, because such approaches have been proposed to be both effective in managing infections and sustainable in the sense that resistance should not easily evolve (André and Godelle, 2005; Baron, 2010; Rasko and Sperandio, 2010; Pepper, 2012; Allen et al., 2014; Vale et al., 2016). Promising approaches include the quenching of toxins (Lu et al., 2014; Henry et al., 2015), siderophores required for iron-scavenging (Kaneko et al., 2007; DeLeon et al., 2009; Kelson et al., 2013; Ross-Gillespie et al., 2014; Imperi et al., 2013), and quorum sensing molecules regulating virulence factor production (LaSarre and Federle, 2013). In our study, we probed the evolutionary robustness argument by focusing on two repurposed drugs (gallium and flucytosine) targeting siderophore production of P. aeruginosa. Using a combination of replicated experimental evolution and phenotypic and genotypic analysis, we show that the often recited argument of anti-virulence drugs being evolutionary robustness is not supported. Instead, we provide a nuanced view on the molecular mechanisms and selective forces that can lead to resistance. For flucytosine, for instance, we found repeated resistance evolution based on a mechanism that prevents drug activation inside the cell, which mitigates off-target effects caused by this drug. For gallium, meanwhile, two types of partially resistant mutants, based on siderophore bypassing mechanisms, arose. However, these mutants only sporadically emerged, indicating that their potential to selective spread in populations is compromised. Our work highlights that evolutionary robustness is a relative measure with specific treatments lying on different positions on a continuum. Thus, our task is not to argue about whether antivirulence drugs are evolutionarily robust or not, but to assess the relative position of each novel treatment on this continuum.

Our findings indicate that it is difficult to define anti-virulence treatments based on fitness effects (Clatworthy et al., 2007; Rasko and Sperandio, 2010; Johnson and Abramovitch, 2017; Dickey et al., 2017). This is because fitness effects might vary in response to the ecological context of the media or the infection. For instance, Imperi et al. (2013) showed that flucytosine does not affect bacterial growth in trypticase soy broth dialysate (TSBD), whereas we found significant fitness effects in human serum. Endorsing the fitness-based definition would mean that flucytosine could only be considered as antivirulence drug in one but not in the other media, which is confusing in a clinical context. For this reason, we support a more generalized definition of antivirulence treatments, intended as drugs targeting bacterial virulence factors, as suggested by Allen et al. (2014) and Vale et al. (2016).

Our results on flucytosine further highlight that negative off-target effects could often be at the source of resistance evolution against anti-virulence approaches. Flucytosine undergoes several enzymatic modifications within the cell, finally resulting in fluorinated ribonucleotides. While flucytosine was shown to inhibit pyoverdine synthesis (Imperi et al., 2013), it certainly also interferes with nucleotide synthesis, which might compromise RNA functionality more generally (Harbers et al., 1959). This sets the stage for selection to favor mutants with alleviated fitness costs under drug exposure. Our results suggest that cells achieved this through mutations in upp. The scheme depicted in Figure 5A shows that the essential pyrimidine nucleotide precursor UMP can be synthesized either through the salvage pathways reutilizing exogenous free bases and nucleosides, or via a de novo biosynthesis pathway using L-glutamine or L-aspartate. While the salvage pathway is typically preferred because it requires less energy, it generates the harmful fluoro-UMP under flucytosine treatment. Thus, the abolishment of the salvage pathway through mutations in upp and the switching to the de novo biosynthesis pathway provides a selective advantage under flucytosine exposure. The notion that off-target effects might compromise the evolutionary robustness of anti-virulence drugs, is also supported by the work of Maeda et al. (2012). They showed that resistance to the quorum-quenching compound C-30 (brominated furanone) evolves repeatedly via upregulation of a drug efflux pump. The spread of these mutants in their experiment can be explained by the fact that quorum quenching did not only inhibit virulence factor production, but also compromised the ability of cells to grow in adenosine medium, which requires a functional quorum sensing system (Dandekar et al., 2012). Their study together with ours shows that negative off-target effects can easily promote resistance evolution.

Our experiments with gallium highlight that it is important to distinguish between the appearance of resistant mutants and their evolutionary potential to spread through populations. At the mechanistic level, we isolated mutants with increased pyocyanin production, a potential mechanism to by-pass gallium-mediated pyoverdine quenching. Pyocyanin is a redox active molecule that can extracellularly reduce ferric to ferrous iron (Cox, 1986; García-Contreras et al., 2013). The upregulation of pyocyanin was associated with mutations in mvaU, encoding a positive regulator of pyocyanin production, and vfr, encoding a global virulence factor regulator (Balasubramanian et al., 2013). Mutations in Vfr can activate PQS (Pseudomonas Quinolon Signal) synthesis, which is known to promote pyocyanin and pyoverdine synthesis (Diggle et al., 2007; Lin et al., 2017). At the evolutionary level, however, the selective advantage of these mutations seemed to be compromised because they occurred only in some but not all clones (Table 1). One plausible explanation for their sporadic appearance is that pyocyanin could serve as a public good, because it reduces iron outside the cell, thereby generating benefits for other individuals in the vicinity, including the drug-susceptible wildtype cells. This scenario would support the argument that anti-virulence strategies should target collective traits, because this would prevent resistant mutants to fix in populations (André and Godelle, 2005; Pepper, 2012; Mellbye and Schuster, 2011; Gerdt and Blackwell, 2014; Ross-Gillespie et al., 2014). The relative success of these mutants is then determined by the viscosity of the environment, determining the shareability of secreted compounds (Weigert and Kümmerli, 2017), and the potential for negative-frequency dependent selection, where strain frequency settles at an intermediate ratio (Ross-Gillespie et al., 2007; Raymond et al., 2012; Yurtsev et al., 2013).

In conclusion, our work advances research on anti-virulence drugs on multiple fronts. First, it shows that resistant phenotypes are difficult to define, as they can involve the restoration of growth, the resumption of virulence factor production, and/or the activation of a bypassing mechanism. Detailed phenotypic and genotypic analyses, as those proposed in our study, are required to disentangle background adaptation from resistance evolution. Second, we show that anti-virulence approaches are neither completely evolution-proof nor does the notion “all roads lead to resistance” apply (Breidenstein et al., 2011). A detailed evolutionary analysis for each individual drug is required to assess its position on the continuum between the two extremes. Third, we advocate the application of more rigorous evolutionary approaches to quantify resistance evolution. While there are rigorous standards to describe the precise molecular mode of action of a novel antibacterial (Ling et al., 2015; Sully et al., 2014), there is much room for improvement for standards regarding the quantification and characterization of resistance evolution (Perron et al., 2006; Hochberg and Jansen, 2015).

Author contributions

CR, DW, MW and RK designed the research. CR and DW conducted the experiments. CR, SW and RK analyzed the data. All authors wrote the manuscript.

Conflict of Interest

The authors declare no conflict of interest.

Data availability

Whole-genome sequencing data have been deposited in the ArrayExpress database at EMBL-EBI (www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-6110.

Acknowledgments

We thank Adin Ross-Gillespie for advice, the Functional Genomics Center Zurich for technical support with the strain sequencing and the Swiss National Science Foundation for funding (grant no PP00P3_165835 to RK).

References

  1. ↵
    Allen RC, Popat R, Diggle SP, Brown SP. (2014). Targeting virulence: can we make evolution-proof drugs? Nat Rev Microbiol 12: 300–308.
    OpenUrlCrossRefPubMed
  2. ↵
    André JB, Godelle B. (2005). Multicellular organization in bacteria as a target for drug therapy. Ecol Lett 8: 800–810.
    OpenUrlCrossRefWeb of Science
  3. ↵
    Balasubramanian D, Schneper L, Kumari H, Mathee K. (2013). A dynamic and intricate regulatory network determines Pseudomonas aeruginosa virulence. Nucleic Acids Res 41: 1–20.
    OpenUrlCrossRefPubMedWeb of Science
  4. ↵
    Banin E, Lozinski A, Brady KM, Berenshtein E, Butterfield PW, Moshe M, et al. (2008). The potential of desferrioxamine-gallium as an anti-Pseudomonas therapeutic agent. Proc Natl Acad Sci U S A 105: 16761–16766.
    OpenUrlAbstract/FREE Full Text
  5. ↵
    Baron C. (2010). Antivirulence drugs to target bacterial secretion systems. Curr Opin Microbiol 13: 100–105.
    OpenUrlCrossRefPubMed
  6. ↵
    Beck DA, O’Donovan GA. (2008). Pathways of pyrimidine salvage in Pseudomonas and former Pseudomonas: Detection of recycling enzymes using high-performance liquid chromatography. Curr Microbiol 56: 162–167.
    OpenUrlCrossRefPubMed
  7. ↵
    Bolger AM, Lohse M, Usadel B. (2014). Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30: 2114–2120.
    OpenUrlCrossRefPubMedWeb of Science
  8. ↵
    Bonchi C, Frangipani E, Imperi F, Visca P. (2015). Pyoverdine and proteases affect the response of Pseudomonas aeruginosa to gallium in human serum. Antimicrob Agents Chemother 59: 5641–5646.
    OpenUrlAbstract/FREE Full Text
  9. ↵
    Bonchi C, Imperi F, Minandri F, Visca P, Frangipani E. (2014). Repurposing of gallium-based drugs for antibacterial therapy. BioFactors 40: 303–312.
    OpenUrlCrossRefPubMedWeb of Science
  10. ↵
    Breidenstein EBM, de la Fuente-Núñez C, Hancock REW. (2011). Pseudomonas aeruginosa: all roads lead to resistance. Trends Microbiol 19: 419–426.
    OpenUrlCrossRefPubMedWeb of Science
  11. ↵
    Brown D. (2015). Antibiotic resistance breakers: can repurposed drugs fill the antibiotic discovery void? Nat Rev Drug Discov 14: 821–832.
    OpenUrlCrossRef
  12. ↵
    Chessa JP, Petrescu I, Bentahir M, Van Beeumen J, Gerday C. (2000). Purification, physicochemical characterization and sequence of a heat labile alkaline metalloprotease isolated from a psychrophilic Pseudomonas species. Biochim Biophys Acta - Protein Struct Mol Enzymol 1479: 265–274.
    OpenUrl
  13. ↵
    Chitambar CR. (2012). Gallium-containing anticancer compounds. Future Med Chem 4: 1257–1272.
    OpenUrlCrossRefPubMedWeb of Science
  14. ↵
    Cingolani P, Platts A, Wang LL, Coon M, Nguyen T, Wang L, et al. (2012). A program for annotating and predicting the effects of single nucleotide polymorphisms, SnpEff. Fly (Austin) 6: 80–92.
    OpenUrl
  15. ↵
    Clatworthy AE, Pierson E, Hung DT. (2007). Targeting virulence: a new paradigm for antimicrobial therapy. Nat Chem Biol 3: 541–548.
    OpenUrlCrossRefPubMedWeb of Science
  16. ↵
    Cornelis P, Dingemans J. (2013). Pseudomonas aeruginosa adapts its iron uptake strategies in function of the type of infections. Front Cell Infect Microbiol 3: 1–7.
    OpenUrl
  17. ↵
    Cox CD. (1986). Role of pyocyanin in the acquisition of iron from transferrin. Infect Immun 52: 263–270.
    OpenUrlAbstract/FREE Full Text
  18. ↵
    Dandekar AA, Chugani S, Greenberg EP. (2012). Bacterial Quorum Sensing and Metabolic Incentives to Cooperate. Science 338: 264–266.
    OpenUrlAbstract/FREE Full Text
  19. ↵
    DeLeon K, Balldin F, Watters C, Hamood A, Griswold J, Sreedharan S, et al. (2009). Gallium Maltolate Treatment Eradicates Pseudomonas aeruginosa Infection in Thermally Injured Mice. Antimicrob Agents Chemother 53: 1331–1337.
    OpenUrlAbstract/FREE Full Text
  20. ↵
    Dickey SW, Cheung GYC, Otto M. (2017). Different drugs for bad bugs: antivirulence strategies in the age of antibiotic resistance. Nat Rev Drug Discov 16: 1–15.
    OpenUrlCrossRef
  21. ↵
    Diggle SP, Matthijs S, Wright VJ, Fletcher MP, Chhabra SR, Lamont IL, et al. (2007). The Pseudomonas aeruginosa 4-Quinolone Signal Molecules HHQ and PQS Play Multifunctional Roles in Quorum Sensing and Iron Entrapment. Chem Biol 14: 87–96.
    OpenUrlCrossRefPubMedWeb of Science
  22. ↵
    Dodt M, Roehr J, Ahmed R, Dieterich C. (2012). FLEXBAR—Flexible Barcode and Adapter Processing for Next-Generation Sequencing Platforms. Biology (Basel) 1: 895–905.
    OpenUrlPubMed
  23. ↵
    Doring G, Pfestorf M, Botzenhart K, Abdallah MA. (1988). Impact of Proteases on Iron Uptake of Pseudomonas aeruginosa Pyoverdin from Transferrin and Lactoferrin. Infect Immun 56: 291–293.
    OpenUrlAbstract/FREE Full Text
  24. ↵
    Edlind TD, Katiyar SK. (2010). Mutational Analysis of Flucytosine Resistance in Candida glabrata. Antimicrob Agents Chemother 54: 4733–4738.
    OpenUrlAbstract/FREE Full Text
  25. ↵
    Escaich S. (2008). Antivirulence as a new antibacterial approach for chemotherapy. Curr Opin Chem Biol 12: 400–408.
    OpenUrlCrossRefPubMed
  26. ↵
    García-Contreras R, Lira-Silva E, Jasso-Chávez R, Hernández-González IL, Maeda T, Hashimoto T, et al. (2013). Isolation and characterization of gallium resistant Pseudomonas aeruginosa mutants. Int J Med Microbiol 303: 574–582.
    OpenUrlCrossRefPubMed
  27. ↵
    Gerdt JP, Blackwell HE. (2014). Competition Studies Confirm Two Major Barriers That Can Preclude the Spread of Resistance to Quorum-Sensing Inhibitors in Bacteria. ACS Chem Biol 9: 2291–2299.
    OpenUrlCrossRefPubMed
  28. ↵
    Granato ET, Harrison F, Kümmerli R, Ross-Gillespie A. (2016). Do Bacterial ‘Virulence Factors’ Always Increase Virulence? A Meta-Analysis of Pyoverdine Production in Pseudomonas aeruginosa As a Test Case. Front Microbiol 7: 1–13.
    OpenUrlCrossRefPubMed
  29. ↵
    Griffin AS, West SA, Buckling A. (2004). Cooperation and competition in pathogenic bacteria. Nature 430: 1024–1027.
    OpenUrlCrossRefPubMedWeb of Science
  30. ↵
    Harbers E, Chaudhuri NK, Heidelberger C. (1959). Studies on fluorinated pyrimidines. VIII. Further biochemical and metabolic investigations. J Biol Chem 234: 1255–62.
    OpenUrlFREE Full Text
  31. ↵
    Harrison F, Browning LE, Vos M, Buckling A. (2006). Cooperation and virulence in acute Pseudomonas aeruginosa infections. BMC Biol 4: 21.
    OpenUrlCrossRefPubMed
  32. ↵
    Henry BD, Neill DR, Becker KA, Gore S, Bricio-Moreno L, Ziobro R, et al. (2015). Engineered liposomes sequester bacterial exotoxins and protect from severe invasive infections in mice. Nat Biotechnol 33: 81–88.
    OpenUrlCrossRefPubMed
  33. ↵
    Hijazi S, Visca P, Frangipani E. (2017). Gallium-Protoporphyrin IX Inhibits Pseudomonas aeruginosa Growth by Targeting Cytochromes. Front Cell Infect Microbiol 7: 1–15.
    OpenUrl
  34. ↵
    Hochberg ME, Jansen G. (2015). Bacteria: Assessing resistance to new antibiotics. Nature 519: 158.
    OpenUrlCrossRef
  35. ↵
    Hughes D, Andersson DI. (2017). Evolutionary Trajectories to Antibiotic Resistance. Annu Rev Microbiol 71: 579–596.
    OpenUrlCrossRef
  36. ↵
    Imperi F, Massai F, Facchini M, Frangipani E, Visaggio D, Leoni L, et al. (2013). Repurposing the antimycotic drug flucytosine for suppression of Pseudomonas aeruginosa pathogenicity. Proc Natl Acad Sci U S A 110: 7458–63.
    OpenUrlAbstract/FREE Full Text
  37. ↵
    Inglis RF, Scanlan P, Buckling A. (2016). Iron availability shapes the evolution of bacteriocin resistance in Pseudomonas aeruginosa. ISME J 10: 2060–2066.
    OpenUrlCrossRef
  38. ↵
    Isaac JH, Holloway BW. (1968). Control of pyrimidine biosynthesis in Pseudomonas aeruginosa. J Bacteriol 96: 1732–41.
    OpenUrlAbstract/FREE Full Text
  39. ↵
    Johnson BK, Abramovitch RB. (2017). Small Molecules That Sabotage Bacterial Virulence. Trends Pharmacol Sci 38: 339–362.
    OpenUrlCrossRef
  40. ↵
    Kahm M, Hasenbrink G, Ludwig J. (2010). grofit: Fitting Biological Growth Curves with R. J Stat Softw 33: 1–21.
    OpenUrlCrossRefPubMedWeb of Science
  41. ↵
    Kaneko Y, Thoendel M, Olakanmi O, Britigan BE, Singh PK. (2007). The transition metal gallium disrupts Pseudomonas aeruginosa iron metabolism and has antimicrobial and antibiofilm activity. J Clin Invest 117: 877–888.
    OpenUrlCrossRefPubMedWeb of Science
  42. ↵
    Kelson AB, Carnevali M, Truong-Le V. (2013). Gallium-based anti-infectives: targeting microbial iron-uptake mechanisms. Curr Opin Pharmacol 13: 707–716.
    OpenUrlCrossRefPubMed
  43. ↵
    LaSarre B, Federle MJ. (2013). Exploiting Quorum Sensing To Confuse Bacterial Pathogens. Microbiol Mol Biol Rev 77: 73–111.
    OpenUrlAbstract/FREE Full Text
  44. ↵
    Li C, Wally H, Miller SJ, Lu C. (2009). The multifaceted proteins MvaT and MvaU, members of the H-NS family, control arginine metabolism, pyocyanin synthesis, and prophage activation in Pseudomonas aeruginosa PAO1. J Bacteriol 191: 6211–6218.
    OpenUrlAbstract/FREE Full Text
  45. ↵
    Li H, Durbin R. (2009). Fast and accurate short read alignment with Burrows–Wheeler transform. Bioinformatics 25: 1754–1760.
    OpenUrlCrossRefPubMedWeb of Science
  46. ↵
    Lin J, Zhang W, Cheng J, Yang X, Zhu K, Wang Y, et al. (2017). A Pseudomonas T6SS effector recruits PQS-containing outer membrane vesicles for iron acquisition. Nat Commun 8: 14888.
    OpenUrlCrossRefPubMed
  47. ↵
    Ling LL, Schneider T, Peoples AJ, Spoering AL, Engels I, Conlon BP, et al. (2015). A new antibiotic kills pathogens without detectable resistance. Nature 517: 455–459.
    OpenUrlCrossRefPubMedWeb of Science
  48. ↵
    Lu C, Maurer CK, Kirsch B, Steinbach A, Hartmann RW. (2014). Overcoming the Unexpected Functional Inversion of a PqsR Antagonist in Pseudomonas aeruginosa □: An In Vivo Potent Antivirulence Agent Targeting pqs Quorum Sensing. Angew Chemie Int Ed 53: 1109–1112.
    OpenUrl
  49. ↵
    Maeda T, García-Contreras R, Pu M, Sheng L, Garcia LR, Tomás M, et al. (2012). Quorum quenching quandary: resistance to antivirulence compounds. ISME J 6: 493–501.
    OpenUrlCrossRefPubMedWeb of Science
  50. ↵
    Maura D, Ballok AE, Rahme LG. (2016). Considerations and caveats in anti-virulence drug development. Curr Opin Microbiol 33: 41–46.
    OpenUrlCrossRef
  51. ↵
    McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Kernytsky A, et al. (2010). The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res 20: 1297–1303.
    OpenUrlAbstract/FREE Full Text
  52. ↵
    Mellbye B, Schuster M. (2011). The sociomicrobiology of antivirulence drug resistance: a proof of concept. MBio 2: e00131-11.
    OpenUrlCrossRefPubMed
  53. ↵
    Meyer JM, Neely A, Stintzi A, Georges C, Holder IA. (1996). Pyoverdin is essential for virulence of Pseudomonas aeruginosa. Infect Immun 64: 518–523.
    OpenUrlAbstract/FREE Full Text
  54. ↵
    Pendleton JN, Gorman SP, Gilmore BF. (2013). Clinical relevance of the ESKAPE pathogens. Expert Rev Anti Infect Ther 11: 297–308.
    OpenUrlCrossRefPubMed
  55. ↵
    Pepper JW. (2012). Drugs that target pathogen public goods are robust against evolved drug resistance. Evol Appl 5: 757–761.
    OpenUrlCrossRefPubMed
  56. ↵
    Perron GG, Zasloff M, Bell G. (2006). Experimental evolution of resistance to an antimicrobial peptide. Proc R Soc B Biol Sci 273: 251 LP-256.
    OpenUrlCrossRefPubMedWeb of Science
  57. ↵
    Rahme LG, Stevens EJ, Wolfort SF, Shao J, Tompkins RG, Ausubel FM. (1995). Common virulence factors for bacterial pathogenicity in plants and animals. Science (80-) 268: 1899–902.
    OpenUrlAbstract/FREE Full Text
  58. ↵
    Rampioni G, Visca P, Leoni L, Imperi F. (2017). Drug repurposing for antivirulence therapy against opportunistic bacterial pathogens. Emerg Top Life Sci ETLS20160018.
  59. ↵
    Rasko DA, Sperandio V. (2010). Anti-virulence strategies to combat bacteria-mediated disease. Nat Rev Drug Discov 9: 117–128.
    OpenUrlCrossRefPubMedWeb of Science
  60. ↵
    Raymond B, West SA, Griffin AS, Bonsall MB. (2012). The dynamics of cooperative bacterial virulence in the field. Science (80-) 337: 85–88.
    OpenUrlAbstract/FREE Full Text
  61. ↵
    Ritz C, Baty F, Streibig JC, Gerhard D. (2015). Dose-Response Analysis Using R. PLoS One 10: 1–13.
    OpenUrlCrossRefPubMed
  62. ↵
    Ross-Gillespie A, Gardner A, West SA, Griffin AS. (2007). Frequency dependence and cooperation: theory and a test with bacteria. Am Nat 170: 331–342.
    OpenUrlCrossRefPubMedWeb of Science
  63. ↵
    Ross-Gillespie A, Weigert M, Brown SP, Kümmerli R. (2014). Gallium-mediated siderophore quenching as an evolutionarily robust antibacterial treatment. Evol Med Public Heal 2014: 18–29.
    OpenUrl
  64. ↵
    Smith DJ, Lamont IL, Anderson GJ, Reid DW. (2013). Targeting iron uptake to control Pseudomonas aeruginosa infections in cystic fibrosis. Eur Respir J 42: 1723–36.
    OpenUrlAbstract/FREE Full Text
  65. ↵
    Sully EK, Malachowa N, Elmore BO, Alexander SM, Femling JK, Gray BM, et al. (2014). Selective Chemical Inhibition of agr Quorum Sensing in Staphylococcus aureus Promotes Host Defense with Minimal Impact on Resistance. PLoS Pathog 10: e1004174.
    OpenUrlCrossRefPubMed
  66. ↵
    Takase H, Nitanai H, Hoshino K, Otani T. (2000). Impact of siderophore production on Pseudomonas aeruginosa infections in immunosuppressed mice. Infect Immun 68: 1834–1839.
    OpenUrlAbstract/FREE Full Text
  67. ↵
    Vale PF, McNally L, Doeschl-Wilson A, King KC, Popat R, Domingo-Sananes MR, et al. (2016). Beyond killing. Evol Med Public Heal 2016: 148–157.
    OpenUrl
  68. ↵
    Valenti P, Berlutti F, Conte MP, Longhi C, Seganti L. (2004). Lactoferrin Functions: Current Status and Perspectives. J Clin Gastroenterol 38: S127–S129.
    OpenUrlCrossRefPubMed
  69. ↵
    Visca P, Imperi F, Lamont IL. (2007). Pyoverdine siderophores: from biogenesis to biosignificance. Trends Microbiol 15: 22–30.
    OpenUrlCrossRefPubMedWeb of Science
  70. ↵
    Waldorf AR, Polak A. (1983). Mechanisms of action of 5-fluorocytosine. Antimicrob Agents Chemother 23: 79–85.
    OpenUrlAbstract/FREE Full Text
  71. ↵
    Weigert M, Kümmerli R. (2017). The physical boundaries of public goods cooperation between surface-attached bacterial cells. Proc R Soc B Biol Sci 284.
  72. ↵
    Weigert M, Ross-Gillespie A, Leinweber A, Pessi G, Brown SP, Kümmerli R. (2017). Manipulating virulence factor availability can have complex consequences for infections. Evol Appl 10: 91–101.
    OpenUrlCrossRef
  73. ↵
    Wu HJ, Wang AHJ, Jennings MP. (2008). Discovery of virulence factors of pathogenic bacteria. Curr Opin Chem Biol 12: 93–101.
    OpenUrlCrossRefPubMed
  74. ↵
    Yurtsev EA, Chao HX, Datta MS, Artemova T, Gore J. (2013). Bacterial cheating drives the population dynamics of cooperative antibiotic resistance plasmids. Mol Syst Biol 9: 683.
    OpenUrlAbstract/FREE Full Text
Back to top
Posted February 26, 2018.
Download PDF

Supplementary Material

Email
Probing the evolutionary robustness of two repurposed drugs targeting iron uptake in Pseudomonas aeruginosa
Chiara Rezzoagli, David Wilson, Michael Weigert, Stefan Wyder, Rolf Kümmerli
bioRxiv 195974; doi: https://doi.org/10.1101/195974
Twitter logo Facebook logo LinkedIn logo Mendeley logo
Citation Tools

Subject Area