|Home | About | Journals | Submit | Contact Us | Français|
Modulation of bacterial chromosomal supercoiling is a function of DNA gyrase-catalyzed strand breakage and rejoining. This reaction is exploited by both antibiotic and proteic gyrase inhibitors, which trap the gyrase molecule at the DNA cleavage stage. Owing to this interaction, double-stranded DNA breaks are introduced and replication machinery is arrested at blocked replication forks. This immediately results in bacteriostasis and ultimately induces cell death. Here we demonstrate, through a series of phenotypic and gene expression analyses, that superoxide and hydroxyl radical oxidative species are generated following gyrase poisoning and play an important role in cell killing by gyrase inhibitors. We show that superoxide-mediated oxidation of iron–sulfur clusters promotes a breakdown of iron regulatory dynamics; in turn, iron misregulation drives the generation of highly destructive hydroxyl radicals via the Fenton reaction. Importantly, our data reveal that blockage of hydroxyl radical formation increases the survival of gyrase-poisoned cells. Together, this series of biochemical reactions appears to compose a maladaptive response, that serves to amplify the primary effect of gyrase inhibition by oxidatively damaging DNA, proteins and lipids.
Bacterial chromosomal topology is maintained by the activities of topoisomerase I, topoisomerase IV and DNA gyrase (topoisomerase II) (Champoux, 2001). The gyrase enzyme is responsible for the introduction of negative DNA supercoils in an ATP-dependent manner, and participates in the processes of replication, transcription, repair, recombination and decatenation (Gellert et al, 1976; Cozzarelli, 1980; Wang, 1996). Gyrase is composed of two subunits, gyrA and gyrB, and complexes with its DNA substrate as an A2B2 tetramer (Mizuuchi et al, 1978; Sugino et al, 1980). GyrA catalyzes the concomitant double-stranded breakage and rejoining of DNA phosphodiester bonds following binding and hydrolysis of ATP by GyrB (Gellert, 1981; Reece and Maxwell, 1991).
The interaction between gyrase inhibitors and the GyrA subunit converts DNA gyrase into a lesion-inducing agent, which results in widespread generation of double-stranded DNA breaks (Maki et al, 1992; Bernard et al, 1993; Chen et al, 1996). DNA damage formation rapidly induces cell division arrest, initially driving entry into bacteriostasis (Hanawalt, 1966; Jaffe et al, 1985; Drlica and Zhao, 1997). In time, the stable DNA–GyrA–inhibitor complex sterically inhibits replication and transcription by establishing a roadblock to DNA and RNA polymerases, respectively, at stalled replication forks (Kreuzer and Cozzarelli, 1979; Bernard and Couturier, 1992; Willmott et al, 1994; Cox et al, 2000). This deadly barrier impedes lesion repair while preventing new DNA synthesis, eventually killing the bacterium (Critchlow et al, 1997; Drlica and Zhao, 1997; Couturier et al, 1998).
In this work, we performed phenotypic, genetic and microarray analyses on gyrase-inhibited Escherichia coli to identify additional contributors to cell death resulting from gyrase poisoning. Taking a systems biology approach, we enriched our gene expression data using transcription factor and functional pathway classifications. This analysis expectedly revealed that the expression of DNA damage response and repair genes was markedly upregulated following inhibition of DNA gyrase and the generation of DNA lesions. Surprisingly, however, we also observed significantly upregulated expression of genes related to superoxide stress, iron–sulfur (Fe–S) cluster synthesis and iron uptake and utilization.
The unfortunate byproduct of increased aerobic respiratory activity and oxidative phosphorylation is the increased formation of the reactive oxygen species (ROS), superoxide (Imlay and Fridovich, 1991). Although superoxide itself presents no additional direct threat to DNA, cytosolic dehydratase proteins containing Fe–S clusters are highly susceptible to oxidative attack (Liochev and Fridovich, 1999; Imlay, 2003, 2006). Various methods have been used to show that superoxide-mediated decomposition of Fe–S clusters, following shifts between anaerobic and aerobic growth states, leads to cytoplasmic release of ferrous (II) iron (Keyer and Imlay, 1996; Liochev, 1996). Elevated intracellular concentrations of ferrous iron have been shown to mediate DNA damage either directly (Henle et al, 1999; Rai et al, 2001) or via the formation of additional oxidative molecules (Touati, 2000; Imlay, 2003). Hydroxyl radicals, a highly destructive ROS, are the product of the Fenton reaction (Imlay et al, 1988; Imlay and Linn, 1988), in which ferrous iron mediates the reduction of hydrogen peroxide. These biologically threatening oxidative species are capable of inducing a wide array of mutagenic single-base substitutions and DNA adducts (Aruoma et al, 1989; McBride et al, 1991; Balasubramanian et al, 1998; Nunoshiba et al, 1999), as well as damaging peptides and lipids (Farr and Kogoma, 1991).
Interestingly, the generation of hydroxyl radicals following DNA damage has been shown in higher order systems, including mammalian cells, and is considered as one of the hallmark features of apoptosis (Jacobson, 1996; Simon et al, 2000). Here, we provide direct evidence that hydroxyl radicals are generated in bacteria following both antibiotic- and peptide-based gyrase poisoning. Our results show that hydroxyl radical production, upon gyrase inhibition, is directly related to the presence and status of Fe–S cluster-containing proteins. In turn, our data suggest that cycling of Fe–S clusters between oxidized and reduced states contributes to a breakdown in endogenous iron regulatory dynamics and global iron management. We thus propose that ROS contribute significantly to bacterial cell death following gyrase inhibition and that the associated series of biochemical reactions represents a maladaptive response to aerobic growth which promotes terminal cell fate.
We initially sought to characterize the in vivo response of the wild-type E. coli strain BW25113 (Wanner, 1983) to gyrase inhibition and to identify additional contributors to cell death following gyrase poisoning. To accomplish this, we employed the synthetic quinolone antibiotic, norfloxacin, and the natural peptide toxin, CcdB (a component of the F plasmid-encoded CcdB–CcdA toxin–antitoxin system; Miki et al, 1992). Upon treatment with norfloxacin, we observed an immediate 2-log reduction in colony forming units (CFU/ml) (Figure 1A). These values steadily declined for the remainder of our experiments. Induction of CcdB expression similarly resulted in consistent CFU/ml reduction over the first 3 h of our experiments (Figure 1A). Wild-type cells expressing CcdB, however, routinely reached their minimum density at the 3 h time point, and then began a sustained increase in cell density that continued over the remainder of the experiment.
Although these results show that, in our system, CcdB is not as efficient as norfloxacin in its ability to induce cellular death in BW25113 cells, our data suggested that both gyrase inhibitors rapidly induce DNA lesion formation. To confirm this hypothesis, we employed an engineered, DNA damage-inducible reporter gene construct that relies upon LexA repression for tight regulation of gfp transcription; fluorescence output is thus reflective of RecA-stimulated autocleavage of LexA following DNA damage recognition (Little, 1991). As anticipated, gyrase poisoning of wild-type bacteria resulted in significant GFP expression from this construct (Figures 1B and C). Norfloxacin treatment uniformly induced high levels of fluorescence (Figure 1B), demonstrating the efficiency and highlighting the irreversibility with which the quinolone traps gyrase and stimulates the formation of DNA breaks. We also observed a large shift in fluorescence upon expression of CcdB, indicative of widespread DNA damage (Figure 1C). In line with our CcdB+ growth data (and the relative inefficiency of CcdB-mediated cell killing compared with norfloxacin in our system), the broad DNA damage-related fluorescence distribution exhibited by CcdB+ cells suggests that DNA damage is being corrected by endogenous repair systems. Taken together, these phenotypic data confirm that norfloxacin and CcdB are both potent inhibitors of DNA gyrase that promote the rapid generation of DNA lesions.
We next examined the transcriptional response of wild-type E. coli cells treated with norfloxacin or expressing CcdB, with the goal of identifying potential secondary contributors to cell death. Analysis of our time-course microarray data revealed that >800 genes exhibited statistically significant upregulation or downregulation at one or more time points as a function of each respective inhibitor (Supplementary Table 1); significance (z-score) was determined on a gene by gene basis, by comparison of mean expression levels to a large (~500) compendium of E. coli microarray data collected under a wide array of conditions (see Materials and methods and Supplementary information for details).
Taking a systems biology approach, we first applied gene ontology (GO) (Ashburner et al, 2000; Camon et al, 2004) classifications to categorize this gene list by biological function, and then enriched these data using GO::TermFinder (Boyle et al, 2004), which enabled us to determine the significance of these classifications (Supplementary Figures 1 and 2; Supplementary Table 2). We next employed the RegulonDB database (Salgado et al, 2006) of known transcription factor connections, in conjunction with a published map of transcriptional regulation (Faith et al, 2007), to focus on transcription factors whose targets were significantly enriched in our data set (Supplementary Table 3). These measures thus allowed us to group genes based on regulatory and functional pathway classifications and to identify coordinated responses to gyrase inhibition.
As expected, LexA, which coordinates the expression of the SOS DNA damage response program, was found to be among the most overrepresented transcriptional regulators following norfloxacin treatment and CcdB expression; this result served to further validate our growth data and DNA damage reporter results. As highlighted in Figure 2, DNA damage-inducible genes comprised one of the largest and most strongly perturbed functional groups in our study. This again is consistent with activation of both the SOS response pathway and the Din (damage-inducible) family of genes (Witkin, 1976; McEntee, 1977; Kenyon and Walker, 1980; Little and Mount, 1982; Walker, 1984; Courcelle and Hanawalt, 2003; Friedberg et al, 2005). The strongest responding among these repair-related genes were error-prone DNA polymerases IV (DinP) and V (composed of UmuC and UmuD) (Elledge and Walker, 1983; Sutton and Walker, 2001).
Interestingly, we also observed statistically significant enrichment for the IscR (iron–sulfur cluster regulator) transcription factor as a function of each perturbation. IscR has been shown to autoregulate expression of the iscRUSA operon of genes, which are primarily involved in de novo Fe–S cluster assembly. The repair of heavily oxidized Fe–S clusters is believed to require the activity of the Isc protein family, and the derepression of the iscRUSA operon following oxidant stress has been previously explored (Schwartz et al, 2000; Djaman et al, 2004).
Along these lines, additional analysis of our microarray results revealed that the oxidative stress response regulator SoxS was also significantly enriched in our data set. Treatment of E. coli with either norfloxacin or CcdB induced the superoxide response operon, soxRS (Figure 2). Transcription of soxS is positively regulated by SoxR and is stimulated upon SoxR activation, the result of superoxide-induced (yet reversible) oxidation of SoxR's Fe–S cluster (Gaudu et al, 1997; Hidalgo et al, 1998). The SoxS transcription factor is known to activate expression of the manganese-superoxide dismutase, sodA, considered as the first responder in the protection of biomolecules from superoxide-induced damage (Hopkin et al, 1992). SodA, like SoxS, was among our list of significantly upregulated genes following norfloxacin treatment or CcdB expression (Figure 2). That damage to Fe–S clusters was occurring as a function of gyrase poisoning was further supported by the observation that each component of the iscRSUA operon exhibited highly increased expression upon inhibition of gyrase (Figure 2).
SoxS has also been shown to activate expression of fur, a ferrous (II) iron-binding transcription factor that regulates the expression of iron uptake, utilization and homeostasis genes (Escolar et al, 1999; Andrews et al, 2003). As highlighted in Figure 2, a number of Fur-regulated iron uptake-related genes were among those genes found to be significantly upregulated by norfloxacin or CcdB. Specifically, we observed increased expression of siderophore-based iron acquisition systems, including components of the Fe3+-siderophore receptor/permease Fec, Fep and Fhu families, as well as the energy transducing TonB–ExbBD complex, which provides the energy required for active siderophore transport.
In the Fenton reaction, free ferrous iron reacts with hydrogen peroxide to generate highly destructive hydroxyl radicals (Imlay, 2003). Superoxide, via oxidative damage to Fe–S clusters and iron leaching, has been shown to play a critical role in the generation of hydroxyl radicals. Based on our microarray results and previous studies, we hypothesized that production of hydroxyl radicals may be the cytotoxic end product of a surge in superoxide and Fe–S cluster cycling following gyrase inhibition. To detect the generation of hydroxyl radicals upon gyrase poisoning by norfloxacin or CcdB, we employed the fluorescent reporter dye, 3′-(p-hydroxyphenyl) fluorescein (HPF) (Setsukinai et al, 2003), which is oxidized by hydroxyl radicals with high specificity.
We detected a significant increase in hydroxyl radical-induced HPF fluorescence at both 3 and 6 h post-treatment of wild-type E. coli with norfloxacin (Figure 3A). Similarly, expression of CcdB in wild-type E. coli also resulted in a large HPF fluorescence increase at both 3 and 6 h post-induction (Figure 3B). Importantly, untreated cultures did not show any detectable increase in HPF fluorescence for the duration of our experiments. These data show that inhibition of DNA gyrase and the generation of double-stranded DNA breaks lead to environmental changes that promote the formation of deleterious hydroxyl radicals.
To demonstrate that generation of hydroxyl radicals was related to the Fenton reaction and an available pool of reactive iron, we repeated our HPF fluorescence measurements in the presence of the iron chelator, o-phenanthroline. This chelator has been previously used to block the effects of the Fenton reaction in bacteria exposed to hydrogen peroxide (Imlay et al, 1988; Asad and Leitao, 1991) and to uncover the role of free radical damage in mitochondria (Yang et al, 1958; de Mello Filho and Meneghini, 1985). As anticipated, o-phenanthroline inhibited the formation of hydroxyl radicals following gyrase inhibition by norfloxacin (Figure 4A) or CcdB (Figure 4B) for the duration of the respective experiments.
To determine the effect of hydroxyl radical-induced oxidative damage on the survival of gyrase-inhibited E. coli, we performed growth studies in the presence of Fenton reaction-neutralizing o-phenanthroline. Addition of the chelator to cultures treated with norfloxacin significantly inhibited cell killing, essentially rendering the drug bacteriostatic (Figure 4C). While the attenuation of cell death in CcdB-expressing cells was not as dramatic, we did observe a greater than 0.5-log increase in CFU/ml before recovery around 2 h (Figure 4C). Our findings thus provide strong evidence for the involvement of hydroxyl radicals in gyrase inhibitor-induced cell death.
Finally, to show that the observed generation of hydroxyl radicals following norfloxacin application was not related to direct redox cycling of the antibiotic, we examined hydroxyl radical formation and survival in two gyrA mutant E. coli strains; these strains, RFS289 (gyrA111) (Schleif, 1972) and KL317 (gyrA17) (Norkin, 1970), are both known to be resistant to the quinolone, nalidixic acid. Treatment with norfloxacin of both mutant strains did not result in loss of cellular viability, proving that these specific gyrA mutations likewise confer resistance to norfloxacin (Supplementary Figure 3). More importantly, we did not observe an increase in hydroxyl radical formation following norfloxacin treatment of these strains (Supplementary Figure 3). These results clearly show that norfloxacin-mediated hydroxyl radical formation occurs only when cell death is observed and as a consequence of the interaction between gyrase and the gyrase inhibitor.
As our hydroxyl radical-sensitive dye experiments revealed, Fenton-reactive iron plays a critical role in cell killing by gyrase inhibitors. To explore whether this hydroxyl radical-generating iron was derived exogenously (Fur-related uptake), endogenously (Fe–S clusters) or from both sources, we first performed growth studies on a series of iron-related single-gene knockouts. To more rapidly perform genetic screening of knockout strains, we initially employed norfloxacin in our phenotypic analyses; the CcdB results are presented in Supplementary information.
Previous work has shown that a deletion of the iron regulator, Fur, leads to an intracellular spike in the concentration of ‘free iron' and potentiates oxidative damage to E. coli DNA (in a ΔrecA background) (Touati et al, 1995). Somewhat surprisingly, we found that a fur deletion mutant responded much slower to gyrase inhibition and exhibited enhanced survival compared with wild-type cultures challenged with norfloxacin over the first 3 h of our experiments (Figure 5A). At our final time point, Δfur cultures exhibited a 1-log increase in cellular survival when compared with the wild-type results (an overall 2-log decrease from time zero). Interestingly, ectopic overexpression of Fur from an isopropyl-β-D-thiogalactopyranoside (IPTG)-inducible ‘rescue' plasmid, in Δfur cultures treated with norfloxacin, served to restore the rapid reduction in viability observed for gyrase-poisoned wild-type cultures (Supplementary Figure 4).
To further explore the role of Fur in the secondary response to gyrase inhibition, and to determine if DNA and oxidative damage promote derepression of iron uptake and utilization genes, we constructed an iron regulation reporter construct that employs the transcription factor Fur as the mediator of gfp expression. Similar in design to our DNA damage reporter construct, fluorescence output from our iron regulatory construct reflects derepression by Fur of an engineered promoter containing Fur operator sites. Following norfloxacin treatment, we observed a large, uniform increase in fluorescence from our Fur-regulated reporter construct, suggesting that gyrase inhibition promotes a breakdown in iron regulatory dynamics (Figure 5B); specifically, norfloxacin induced a 10-fold change, whereas the iron chelator, o-phenanthroline, induced a 100-fold change in fluorescence when compared with untreated fluorescence measurements (Supplementary Figure 5). Importantly, we also observed increased expression of GFP from our Fur-regulated reporter construct following CcdB expression (Supplementary Figure 6). This result supports our microarray analysis, which identified Fur-regulated genes as being among the most significantly upregulated as a function of gyrase poisoning.
To determine if impairment of exogenous iron uptake affects the ability of E. coli to survive gyrase inhibition, we next examined the phenotypic effect of norfloxacin on a deletion mutant of the import complex gene, tonB. Again, the TonB–ExbBD complex controls active iron transport across the inner membrane, in an energy-dependent process, for siderophore uptake systems (Clarke et al, 2001). The ΔtonB knockout had a negligible effect on the viability of norfloxacin-treated E. coli (Figure 5A) compared with our wild-type results (Figure 1A), suggesting that the major iron import system does not play a significant role in gyrase inhibitor-induced cell death. Relative to previous findings, in which exogenous iron import has been implicated in oxidative damage-mediated cell killing (Touati et al, 1995), our results imply that an internal disruption of iron management is a critical factor in gyrase inhibitor-mediated cell death.
To test whether this endogenous iron misregulation was related to Fe–S cluster status, we phenotypically examined single-gene knockouts of the Fe–S cluster synthesis operon, iscRSUA. Although ΔiscR, ΔiscU and ΔiscA deletion mutants showed slower response to application of norfloxacin, survival levels of these cultures were similar to those of our wild-type cultures within 3 h post-treatment (Supplementary Figure 7). Interestingly, deletion of the cysteine desulfurase, iscS, which is critical for efficient Fe–S cluster turnover and plays a role in tRNA modification and thiamine biosynthesis (Schwartz et al, 2000; Djaman et al, 2004), had the most dramatic effect on survival (Figure 5A). Knocking out this protein, which provides atomic sulfur during Fe–S cluster assembly (Zheng et al, 1998), resulted in a near 2-log increase in viability for the majority of the experiment relative to norfloxacin-treated wild-type cultures. ΔiscS cultures had a remarkably slow phenotypic response to gyrase inhibition, exhibiting less than 1-log of cell death within the first 4 h post-treatment. We also screened the effect of norfloxacin-based gyrase inhibition on a deletion mutant of the iscS paralogue, sufS. SufS is a component of the sufABCDSE operon, which is thought to handle Fe–S cluster assembly not catalyzed by isc operon activity during normal growth and is considered to be required for Fe–S cluster assembly in iron-limiting and oxidatively stressful conditions (Djaman et al, 2004; Outten et al, 2004). We found that the ΔsufS knockout had little effect on norfloxacin-mediated cell death and exhibited wild-type levels of survival (Supplementary Figure 7).
It is worth mentioning that ΔiscS cultures grew at a slower rate when compared with wild-type cultures in our study, consistent with previous work (Schwartz et al, 2000; Djaman et al, 2004). However, it is also important to note that the growth rate of ΔiscS was similar to ΔrecA cultures, which showed a near 5-log decrease in viability following norfloxacin application (Supplementary Figure 7). Together, these data suggest that the decreased cell death exhibited by norfloxacin-treated ΔiscS cultures is not simply related to drug killing efficiency and rate of growth.
In order to more directly monitor the overall status of Fe–S clusters following gyrase inhibition, we constructed an iscRUSA promoter-gfp reporter gene fusion construct using the native (−150 to +26) iscRUSA operon promoter region (Giel et al, 2006). This promoter is negatively regulated by the IscR transcription factor and responds to conditions that are rate limiting for Fe–S cluster biogenesis by upregulating expression of the Isc protein family, including IscR itself (Schwartz et al, 2001). Norfloxacin treatment of wild-type cells led to a near five-fold increase in fluorescence at 3 h and more than a 10-fold increase in fluorescence at 6 h when compared with untreated cultures (Figure 6B); CcdB expression also yielded increased fluorescence from our IscR-regulated reporter construct (Supplementary Figure 8).
In light of our Isc-related results, we next sought to determine whether superoxide-based oxidative attack to Fe–S clusters was indeed occurring by monitoring the activation of the superoxide response. To do so, we constructed a soxS promoter-gfp reporter gene fusion using the native soxS promoter region. This promoter is transcriptionally regulated by SoxR, and its activation is directly related to the redox state of SoxR (superoxide-related oxidation activates transcription; Hidalgo et al, 1998). As expected, gyrase inhibition by norfloxacin resulted in GFP expression from this superoxide response sensor (Figure 6A); more specifically, both norfloxacin and the superoxide-generating compound, paraquat (at a concentration of 10 μM; Supplementary Figure 5), induced a 10-fold increase in fluorescence when compared with untreated cultures. We also observed increased fluorescence from our SoxR-regulated reporter construct following CcdB expression (Supplementary Figure 6). Importantly, our baseline (time zero) and untreated control measurements (at 6 h) were remarkably similar (Figure 6A). These data support the phenotypic- and microarray-based hypothesis that oxidation of Fe–S clusters by superoxide is occurring and that the SoxRS response is activated.
We next studied the phenotypic effect of impaired recognition and mitigation of superoxide formation following gyrase inhibition. Whereas expression of the superoxide-sensitive transcription factor SoxS is significantly upregulated upon gyrase inhibition, treatment of a soxS deletion mutant with norfloxacin resulted in no discernable difference in survival compared with our wild-type growth data (Supplementary Figure 7). Moreover, we observed no change in the level of cell death when ΔsodA (manganese-superoxide dismutase) cultures were challenged with norfloxacin (Supplementary Figure 7).
E. coli encode distinct, yet compensatory, superoxide decomposition systems, and in addition to SodA, possess the Fur-regulated iron-superoxide dismutase, SodB (Fridovich, 1997; Dubrac and Touati, 2002). As SodB has been shown to be more proficient at protecting cytoplasmic Fe–S-containing proteins in E. coli (Hopkin et al, 1992), we next examined the effect of gyrase inhibition on a sodB deletion strain. Whereas ΔsodB and wild-type cells treated with norfloxacin showed similar growth behavior between the 0 and 2 h time points, ΔsodB cells exhibited sharply decreased survival over the final 4 h of the experiment, approaching our limit of detection (Figure 6C). The delay in the phenotypic response to norfloxacin treatment in this strain suggests that the generation of superoxide and superoxide-mediated oxidative damage does not occur immediately following gyrase inhibition. Rather, these data imply that oxidative damage is a secondary, yet quite deadly, effect of gyrase poisoning.
Previous work has established a link between DNA damage, SOS repair activity and bursting of ATP (Barbe et al, 1983; Dahan-Grobgeld et al, 1998). Given the connection between the respiratory electron-transport chain activity and superoxide generation (Imlay and Fridovich, 1991; Messner and Imlay, 2002), we hypothesized that a plausible biochemical explanation for superoxide formation and Fe–S oxidation after gyrase inhibitor-induced DNA damage was the increased production of ATP. To address this, we first measured cellular ATP levels in wild-type cells following gyrase poisoning by norfloxacin. Using a luciferase-based assay, we determined that at 1 h norfloxacin treatment induced a greater than 102-fold change in moles of ATP/cell when compared with untreated cells; this difference increased with the duration of the experiment (Supplementary Figure 9).
To further explore our hypothesis, we next studied the effect of disabling ATP synthase assembly, and thus genetic-level uncoupling of oxidative phosphorylation, on gyrase inhibitor-mediated cell death. We challenged the majority of ATP synthase component genes with norfloxacin and observed increased survival (2-log increase in CFU/ml) in each case (Supplementary Figure 7). The most striking phenotypic change following norfloxacin application was exhibited by ΔatpC cultures, where we observed complete reversal of bactericidal activity. In this strain we observed a 3-log increase in survival relative to norfloxacin-treated wild-type cells (Figure 6C). These results clearly demonstrate that efficient norfloxacin-based killing is directly dependent on ATP, supporting published in vitro and in vivo findings (Kampranis and Maxwell, 1998; Li and Liu, 1998).
Lastly, we examined whether disabling the activity of TCA cycle component succinate dehydrogenase, also an aerobic respiratory electron-transport chain constituent, would affect cell survival following gyrase inhibition. The conversion of succinate to fumarate, catalyzed by succinate dehydrogenase, involves three Fe–S clusters (SdhB subunit) and has been shown to result in the generation of superoxide (Messner and Imlay, 2002); accordingly, we chose to study the effect of norfloxacin on the growth of ΔsdhB cells in order to reduce both the level of endogenous Fe–S cluster-bearing proteins and the production of superoxide. Compared with wild-type cells, norfloxacin treatment of ΔsdhB cells led to a 1-log increase in survival at 1 h and a more modest 0.5-log increase at 6 h. We suspect that this effect may be increased by knockout of additional Fe–S cluster-containing proteins and requires further investigation.
Together, these data suggest that gyrase inhibition promotes endogenous environmental changes that support the generation of superoxide radicals. Moreover, our findings help to provide an explanation for why the loss of superoxide dismutase activity (specifically that of SodB) enhances the killing effect of the gyrase poison, norfloxacin.
To validate the results of our genetic screening, we performed hydroxyl radical measurements, again using the reporter dye HPF, in the ΔatpC, ΔiscS and ΔsodB knockout strains. The largest detectable increase in hydroxyl radical-mediated fluorescence was observed following norfloxacin treatment of ΔsodB cells (Figure 7), consistent with the significantly increased cell death we saw following norfloxacin treatment (Figure 6B). In contrast, gyrase-inhibited ΔiscS cultures exhibited a reduced level of hydroxyl radicals compared with wild-type cultures (Figure 7). This was also consistent with our ΔiscS growth data (Figure 5A) and supports the notion that inefficient repair of Fe–S clusters will decrease the available ferrous iron by limiting redox cycling.
As anticipated, we were unable to detect a hydroxyl radical-mediated fluorescence shift in ΔatpC cultures treated with norfloxacin (Figure 7). This result, which is also consistent with phenotypic data (Figure 6C), strongly suggests a connection between gyrase inhibition, oxidative phosphorylation-driven superoxide production and hydroxyl radical production.
Gyrase inhibitors are known to induce cell death by stimulating DNA damage formation, impeding lesion repair and blocking replication processes (Drlica and Zhao, 1997; Couturier et al, 1998). In this study, we sought to identify additional mechanisms that directly or indirectly contribute to gyrase inhibitor-mediated cell killing. Here, we show that oxidative damage by superoxide and hydroxyl radicals contributes to bacterial cell death following gyrase poisoning. Superoxide, generated as a by-product of aerobic metabolism (Imlay and Fridovich, 1991), will attack proteins containing Fe–S clusters and oxidatively destabilize the cluster, often triggering the release of ferrous (II) iron (Keyer and Imlay, 1996). Based on our phenotypic and microarray analyses, we propose that repetitious Fe–S cluster oxidation and repair arises as a deadly and necessary adjunct that potentiates the killing ability of gyrase inhibitors. This cycling should serve to maintain a large pool of ferrous iron, which readily participates as a substrate in the Fenton reaction to produce hydroxyl radicals. Our results demonstrate that highly deleterious hydroxyl radicals, are in fact, generated through the Fenton reaction, upon gyrase poisoning. Based on our findings, we have constructed a model for a complementary bacterial cell death pathway (Figure 8), stimulated by gyrase inhibition, which involves ROS formation and a breakdown in iron regulatory dynamics.
In this work, we employed biochemical pathway and transcription factor regulatory classifications to identify functional modules within the transcriptional response in E. coli to gyrase poisoning and double-stranded DNA break formation.
This analysis revealed, as expected, that DNA lesion remediation and repair systems constitute a significant proportion of the quickest and highest responding gene set (Figure 2). Our microarray analysis also revealed enrichment for genes involved in oxidative stress responses, including the soxRS operon and sodA. Previous work has shown that a loss of chromosomal supercoiling, which occurs when gyrase is inhibited, results in the upregulation of oxidative phosphorylation and ATP bursting (Dahan-Grobgeld et al, 1998); we validate this finding in our present study, showing that gyrase poisoning rapidly results in a large increase in cellular ATP levels. The consequence of heightened ATP synthase and respiratory chain activity in an oxygen-rich environment is the expedited generation of the ROS, superoxide (Imlay, 2003). Here, we have shown that oxidative phosphorylation contributes to gyrase inhibitor-induced cell death, and that superoxide formation indeed occurs when gyrase is poisoned. With regard to oxidative phosphorylation, we found that blockage of succinate dehydrogenase activity led to a modest increase in cell viability whereas genetic-level uncoupling of ATP synthase activity resulted in a striking reduction of cell death following treatment with norfloxacin (Figure 6C). By employing a promoter-reporter gene fusion construct designed to respond to superoxide-related oxidation of SoxR, we were able to monitor the induction of the superoxide stress regulon as a function of gyrase inhibition (Figure 6A). Together, these findings suggest that our proposed oxidative damage cell death pathway (Figure 8) may exist as a maladaptive response to gyrase inhibitor-induced DNA damage in oxygen-rich environments.
Surprisingly, our systems biology analysis of the microarray data revealed that each component of the Fe–S cluster repair operon, iscRUSA, was significantly upregulated, and that the transcriptional regulator of this operon, IscR, was statistically enriched. This finding implied that prolonged exposure of Fe–S clusters to superoxide was occurring, as the Isc family of proteins has been shown to be critical to the repair of heavily oxidized Fe–S clusters (Djaman et al, 2004). This result also suggested that redox cycling, where heavily oxidized clusters are replaced with newly synthesized clusters, was occurring and would thus provide intracellularly abundant superoxide with a steady supply of new substrates. In light of our Δfur and ΔiscS results, however, it is necessary to amend our initial hypothesis. It has previously been shown that deletion of the master iron regulator, Fur, results in increased levels of intracellular iron, a function of increased iron procurement system expression (Touati et al, 1995). Interestingly, it has also been shown that Fe–S cluster formation is related to Fur and that the overall number of Fe–S clusters is decreased in a Δfur background (McHugh et al, 2003), similar to the effect on Fe–S cluster number in a ΔiscS background (Schwartz et al, 2000). Thus, our findings that cellular survival was increased following gyrase poisoning in Δfur and ΔiscS cells imply that cell killing by gyrase inhibitors is intimately related to the number of intracellular Fe–S clusters and not just their status.
These results raise the question—whether the observed generation of hydroxyl radicals is related to repetitious breakdown and repair of damaged Fe–S clusters (and thus an increased endogenous level of unbound ferrous iron) or is it possible that hydroxyl radical production is actually taking place at the exposed active sites of oxidatively-sensitive dehydratases (thus involving intact, yet perhaps unstable, 4Fe–4S clusters). The latter scenario has previously been proposed by Flint et al (1993), and it helps to explain the resistance to gyrase inhibition observed in Δfur, and much more strikingly, in ΔiscS cells. This hypothesis would also help to explain why ectopic overexpression of Fur in the Δfur background reverts norfloxacin-induced cell death to wild-type levels (Supplementary Figure 4). Moreover, this latter scenario would be still consistent with our iron chelator result, in which o-phenanthroline prevents hydroxyl radical formation by sequestering any Fenton-reactive exposed iron in dehydratase Fe–S clusters.
It is quite likely that these events would induce global iron misregulation and a disruption in iron metabolism. Indeed, our engineered iron regulatory reporter gene construct revealed that a breakdown in iron regulatory dynamics occurs when gyrase is inhibited (Figures 5B). Specifically, increased Fur-related GFP expression implies less Fur–Fe(II) complexes; however, this does not necessarily signify low internal concentrations of Fe(II) and may reflect an internal perception that increased uptake is necessary, or simply a sign that internal iron is ‘stuck' in a redox cycle of Fe–S cluster oxidation and repair and cannot be recognized and/or bound by Fur. It is plausible then that the misregulation of Fur-controlled iron uptake, utilization and maintenance genes may serve to amplify the oxidative potential by sustaining a pool of Fenton-ready ‘free iron'.
As noted above, our proposed oxidative damage cell death pathway (Figure 8) appears to be a maladaptive response by aerobically growing E. coli to DNA gyrase inhibition and DNA damage formation by both a synthetic antibiotic and an endogenous toxin. The biochemical reliance of bacteria on Fe–S catalysis, evolutionarily related to the oxygen-free and iron- and sulfur-rich environment that existed billions of years ago (Johnson et al, 2005; Imlay, 2006), has predisposed these organisms to oxidative damage. This ‘backdoor' appears to be exploited by gyrase-inhibiting bactericidal agents in exacting cell death, which may explain the dependence of quinolone antibiotics (i.e. ofloxacin and ciprofloxacin; Morrissey and Smith, 1994) on oxygen and aerobic growth for bactericidal activity. Our findings may contribute to the development of novel antibacterial therapies (Walsh, 2003; McAdams et al, 2004) that exhibit more efficient bactericidal activity.
It is interesting to note that the isc operon, which we implicate in gyrase inhibitor-mediated iron misregulation and redox cycling, is highly conserved in eukaryotes and is involved in mitochondrial Fe–S cluster assembly (Muhlenhoff and Lill, 2000; Johnson et al, 2005). Moreover, the production of ROS as a kill signal has parallels in apoptosis, and the relationship between ROS, specifically hydroxyl radicals, and mitochondrial breakdown is well established (Zamzami et al, 1995; Jacobson, 1996; Green and Reed, 1998; Madeo et al, 1999; Simon et al, 2000; Brookes et al, 2004). Given these connections, it is thus tempting to consider that the maladaptive response to gyrase poisoning by norfloxacin and CcdB we observe in E. coli may have evolved into targeted and efficient killing programs in higher order organisms, including DNA damage-induced apoptosis in eukaryotes.
Basic molecular biology techniques were implemented as previously described (Sambrook and Russell, 2001). All plasmids were constructed using restriction endonucleases and T4 DNA ligase from New England Biolabs. For cloning purposes, we transformed plasmids into the E. coli strain XL-10 (Stratagene; Tetr Δ(mcrA)183, Δ(mcrCB-hsdSMR-mrr)173, endA1, supE44, thi-1, recA1, gyrA96, relA1, lac Hte (F proAB lacIq ZDM15 Tn10 (Tetr) Amy Camr) using standard heat-shock protocols (Sambrook and Russell, 2001). All cells were grown in selective medium: Luria–Bertani (LB) media (Fisher) supplemented with 30 μg/ml of kanamycin (Fisher). Plasmid isolation was performed using the QIAprep Spin Miniprep kits (Qiagen). Subcloning was confirmed by restriction analysis. Plasmid modifications were verified by sequencing using the PE Biosystem ABI Prism 377 sequencer.
Our riboregulated CcdB expression plasmid was built according to our published design (Isaacs et al, 2004). The ccdB gene was amplified by PCR from F plasmid-containing XL-10 cells using the PTC-200 PCR machine (Bio-Rad) with the Expand Long Template PCR System (Roche). Oligonucleotide primers were purchased from Integrated DNA Technologies.
Transcription of cis-repressed ccdB mRNA was controlled using the PLlacO-1 promoter (Lutz and Bujard, 1997), a modified version of the native bacteriophage λ PL promoter containing two LacI operator sites; transcription was thus induced by addition of IPTG (Fisher). The PBAD promoter regulates the production of the trans-activating RNA necessary for CcdB toxin translation; CcdB translation was induced by the addition of arabinose. This system is described in detail in Supplementary information.
Our DNA damage and iron regulation sensors were based on the design of the PLlacO-1 promoter (Lutz and Bujard, 1997), a synthetic variant of the native bacteriophage λ PL promoter. In this design, LacI operator sites were inserted near both the −10 and −35 sequences required for RNA polymerase binding, such that LacI binding would sterically inhibit transcription. In our study, PCR was used to build each novel promoter, which employed LexA and Fur operator sites to regulate expression, respectively, of the green fluorescent protein gene, gfpmut3b (Cormack et al, 1996).
To construct our superoxide response sensor, we PCR-amplified the native soxS promoter from XL-10 cells and cloned it upstream of the gfpmut3b gene. To construct our Fe–S cluster response (IscR) sensor, we PCR-amplified the native iscRUSA promoter from XL-10 cells and cloned it upstream of the gfpmut3b gene.
In our initial experiments, we compared the growth of untreated, norfloxacin-treated (250 ng/ml), CcdB− (uninduced cultures containing the ccdB riboregulator) and CcdB+ (induced cultures containing the ccdB riboregulator) wild-type BW25113 (lacIq rrnBT14 ΔlacZWJ16 hsdR514 ΔaraBADAH33 ΔrhaBADLD78) cultures. In our specific single-gene knockout experiments, we studied the growth of deletion strains contained in a BW25113 deletion library (Baba et al, 2006).
For all experiments, cells were grown overnight, then diluted 1:1000 in 50 ml LB (+30 μg/ml kanamycin for CcdB-expressing cells) for collection of OD600 and CFU/ml samples. For those experiments performed in the presence of iron chelator, growth media were supplemented with 100 μM o-phenanthroline (Sigma). CcdB+ cultures were induced by the addition of 1 mM IPTG and 0.25% arabinose when they reached an OD600 of 0.3–0.4. Measurements of OD600 were taken using a SPECTRAFluor Plus (Tecan). For CFU/ml measurements, 100 μl of culture was collected, washed twice with filtered 1 × PBS, pH 7.2 (Fisher) and then serially diluted in 1 × PBS. A 10 μl portion of each dilution was plated onto individual wells of a 24-well plate, with each well containing 1 ml of LB agar (Fisher) (+30 μg/ml kanamycin for CcdB-expressing cells), and the plate was incubated overnight at 37°C. Only dilutions that yielded between 20 and 100 colonies were counted and CFU/ml values were calculated using the formula ((# of colonies)(dilution factor))/0.01 ml.
To monitor the occurrence of DNA damage, oxidative damage to Fe–S clusters and changes in iron regulation, we employed engineered sensors that respond to these biochemical events by activating expression of gfpmut3b. All data were collected using a Becton Dickinson FACSCalibur flow cytometer (Becton Dickinson) with a 488-nm argon laser and a 515–545 nm emission filter (FL1) at low flow rate. The following PMT voltage settings were used: E00 (FSC), 300 (SSC) and 700 (FL1). Calibrite Beads (Becton Dickinson) were used for instrument calibration.
Flow data were converted to ASCII format using MFI (E Martz, University of Massachusetts, Amherst) and processed with MATLAB (MathWorks) to construct figures. At least 50 000 cells were collected for each sample.
In all experiments, cells were grown overnight and then diluted 1:1000 in 50 ml LB supplemented with 100 μg/ml ampicillin (Fisher) (+30 μg/ml kanamycin for CcdB-expressing cells). CcdB expression was induced by addition of 1 mM IPTG and 0.25% arabinose at an OD600 of 0.3–0.4; 250 ng/ml norfloxacin was added to drug-treated cultures at an OD600 of 0.3–0.4. For those experiments performed in the presence of iron chelator, growth media were supplemented with 100 μM o-phenanthroline (Sigma). For those experiments performed in the presence of paraquat, growth media were supplemented with 10 μM paraquat (Acros Organics). Samples were taken immediately before induction (time zero), then every hour for 6 h. At each time point, approximately 106 cells were collected, washed once and resuspended in filtered 1 × PBS, pH 7.2 (Fisher) and measured on the flow cytometer.
We compared the microarray-determined mRNA profiles of E. coli BW25113 cultures in response to CcdB and norfloxacin with those of untreated cultures. For all experiments, cells were grown overnight and then diluted 1:1000 in 50 ml LB (+30 μg/ml kanamycin for CcdB-expressing cells) for collection of total RNA. CcdB-expressing cultures were induced with 1 mM IPTG and 0.25% arabinose at an OD600 of 0.3–0.4; 250 ng/ml norfloxacin was added to norfloxacin-treated cultures at an OD600 of 0.3–0.4. Untreated cultures were grown in LB with no exogenous inducers or antibiotics added. Samples for microarray analysis were taken immediately before induction (time zero) and then at 30, 60 and 120 min post-induction.
Total RNA was obtained using the RNeasy Protect Bacteria Mini kit (Qiagen) according to the manufacturer's instructions. RNAprotect (Qiagen) was added to culture samples, which were then pelleted by centrifugation at 3000 g for 15 min and stored overnight at −80°C. Total RNA was then extracted using the RNeasy kit and samples were DNase-treated using DNA-free (Ambion). Sample concentration was estimated using the ND-1000 spectrophotometer (NanoDrop).
cDNA was prepared from 10 μg total RNA by random primed reverse transcription using SuperScript II (Invitrogen). The RNA was digested by adding 1 M NaOH and then incubating at 65°C for 30 min. The mixture was neutralized by the addition of 1 M HCl. The cDNA was purified using a QIAquick PCR purification column (Qiagen) following the manufacturer's protocol. The cDNA was fragmented to a size range of 50–200 bases with DNase I (0.6 U/μg cDNA) at 37°C for 10 min, followed by inactivation of the enzyme at 98°C for 10 min. Subsequently, the fragmented cDNA was biotin labeled using an Enzo BioArray Terminal Labeling kit with Biotin-ddUTP (Enzo Scientific). Fragmented, biotinylated cDNA was hybridized to Affymetrix E. coli Antisense Genome arrays for 16 h at 45°C and 60 r.p.m.
Following hybridization, arrays were washed and stained according to the standard Antibody Amplification for ProkaryoticTargets protocol (Affymetrix). This consisted of a wash with non-stringent buffer, followed by a wash with stringent buffer, a stain with strepavidin, a wash with non-stringent buffer, a stain with biotinylated anti-strepavidin antibody, a stain with strepavidin–phycoerythrin and a final wash with non-stringent buffer. The stained GeneChip arrays were scanned at 532 nm using an Affymetrix GeneChip Scanner 3000. The scanned images were scaled and quantified using GCOS v1.2 software.
The resulting *.CEL files were combined with *.CEL files from arrays that comprise the M3D compendium (T Gardner, Boston University, http://m3d.bu.edu) and RMA-normalized (Irizarry et al, 2003) for a total of 505 RMA-normalized E. coli expression arrays. Subsequently, we subtracted the mean normalized expression for each gene from its respective normalized expression for each individual experiment and then divided it by the respective standard deviation of each gene across all experiments. This converts the expression values for all genes across all experiments into estimated z-scores based on the observed expression distribution for each gene across all experiments in the M3D compendium.
To determine statistically significant changes in gene expression owing to norfloxacin treatment or CcdB expression, we subtracted the expression z-score of each gene in our uninduced control data set from the corresponding z-score in our perturbed (norfloxacin-treated or CcdB-expressing) sample data set. This was carried out for each experimental time point (0, 30, 60, 120 and 180 min), for example, the z-score of recA expression from our uninduced sample at 30 min was subtracted from the recA z-score from our norfloxacin-treated sample at 30 min. This allowed us to determine the difference in expression between an uninduced control set and a gyrase inhibitor-treated data set in terms of units of standard deviation, a robust metric. A gene was considered to have significantly changed expression when its z-score difference was greater than two units of standard deviation, with the sign determining overexpression and underexpression. Finally, as a separate analysis of differential gene expression, and to reduce the dimensionality of these data, we conflated the difference of z-scores across all time points (a weighted Z-test, essentially combining values from multiple tests of the same hypothesis) by utilizing the formula, Zaverage=sum_t (Z_t)/sqrt(t) (Whitlock, 2005).
Following the identification of significantly changed genes at each time point, we performed functional enrichment using GO classification terms (Ashburner et al, 2000; Camon et al, 2004) and the GO::TermFinder program (Boyle et al, 2004). In doing this, we were able to group genes by GO annotated pathways and determine statistical significance. We next utilized the transcription factor regulatory information contained in RegulonDB (version 4) (Salgado et al, 2006) together with a transcriptional regulatory map assembled by the program, CLR (Faith et al, 2007). Using both sets of information, we were able to categorize significantly changed genes into functional units to monitor coordinated responses.
To detect hydroxyl radical formation following norfloxacin treatment or CcdB expression, we used the fluorescent reporter dye, 3′-(p-hydroxyphenyl) fluorescein (HPF; Invitrogen), which is oxidized by hydroxyl radicals with high specificity. All data were collected using the Becton Dickinson FACSCalibur flow cytometer described above (Becton Dickinson). The following PMT voltage settings were used: E00 (FSC), 300 (SSC) and 825 (FL1). Calibrite Beads (Becton Dickinson) were used for instrument calibration. Flow data were collected, converted and analyzed as above.
In all experiments, cells were grown overnight, then diluted 1:1000 in 50 ml LB (+30 μg/ml kanamycin for CcdB-expressing cells) and supplemented with 5 μM HPF. For those experiments involving iron chelator, growth media were supplemented with 100 μM o-phenanthroline. CcdB expression was induced by the addition of 1 mM IPTG and 0.25% arabinose at an OD600 of 0.3–0.4; 250 ng/ml norfloxacin was added to drug-treated cultures at an OD600 of 0.3–0.4. Samples were collected and prepared for flow cytometry as above.
To monitor the increase in ATP formation following gyrase inhibition, we used the BacTiter-Glo Microbial Cell Viability Assay (Promega), a firefly luciferase bioluminescence-based assay, according to the manufacturer's instructions. Luminescence measurements were taken using a SPECTRAFluor Plus (Tecan). ATP disodium salt (MP Biomedicals) was used as a control to generate a standard curve, according to the manufacturer's instructions.
In all experiments, cells were grown overnight and then diluted 1:1000 in 50 ml LB (+30 μg/ml kanamycin for CcdB-expressing cells). CcdB expression was induced by the addition of 1 mM IPTG and 0.25% arabinose at an OD600 of 0.3–0.4; 250 ng/ml norfloxacin was added to drug-treated cultures at an OD600 of 0.3–0.4. At each time point, 100 μl of each culture was taken and incubated with 100 μl BacTiter-Glo reagent for 10 min, at 25°C, with shaking; samples were also taken simultaneously for CFU/ml measurements. ATP values (in moles) were calculated based on the standard curve, and moles ATP per cell values were calculated by dividing moles ATP to CFU/ml measurements.
Supplementary Table 1
Supplementary Table 2
Supplementary Table 3
This work was supported by the National Science Foundation and the United States Department of Energy. We are grateful to Marc Lenburg and Norman Gerry for performing microarray experiments, and to Hirotada Mori and colleagues for their donation of the BW25113 E. coli knockout library. We thank Iris Keren, Guillaume Cottarel, Jamey Wierzbowski and Philina Lee for critical review of the manuscript. We also thank the anonymous reviewers of this work for their important advice and suggestions.