|Home | About | Journals | Submit | Contact Us | Français|
Stress-mediated programmed cell death (PCD) in bacteria has recently attracted attention, largely because it raises novel possibilities for controlling pathogens. How PCD in bacteria is regulated to avoid population extinction from transient, moderate stress remains a central question. We report that the YihE protein kinase is a key regulator that protects Escherichia coli from antimicrobial and environmental stressors by antagonizing the MazEF toxin-antitoxin module. YihE was linked to a reactive oxygen species (ROS) cascade, and a deficiency of yihE stimulated stress-induced PCD even after stress dissipated. YihE was partially regulated by the Cpx envelope stress-response system, which, along with MazF toxin and superoxide, has both protective and destructive roles that help bacteria make a live-or-die decision in response to stress. YihE probably acts early in the stress response to limit self-sustaining ROS production and PCD. Inhibition of YihE may provide a new way to enhance antimicrobial lethality and attenuate virulence.
The increasing prevalence of antimicrobial resistance now threatens many aspects of medical treatment, particularly those involving invasive procedures. New approaches that enhance pathogen killing may help improve the efficacy of antimicrobial therapy, since enhanced killing rapidly reduces pathogen population size and thus restricts new resistance from arising (Stratton, 2003). Toward that end, stress-mediated programmed cell death (PCD) has emerged as an important theme in bacterial physiology -- we might be able to manipulate PCD to enhance pathogen killing. The observation that lethality from harsh forms of stress increases with the activation of the MazF toxin suggests that cellular factors, in addition to stress-induced lesions, are involved in the stress-mediated death event (Hazan et al., 2004; Lewis, 2000; Rice and Bayles, 2003; Sat et al., 2001). Harsh stress, in the form of lethal antimicrobials, was subsequently linked to a cascade of reactive oxygen species (ROS), with hydroxyl radical being the lethal agent (Dwyer et al., 2007; Kohanski et al., 2007). Thus, many lethal stressors act through a common biochemical mechanism that is reminiscent of ROS involvement in eukaryotic apoptosis (Jung et al., 2001; Mates and Sanchez-Jimenez, 2000; Simizu et al., 1998). More recently, the Cpx and Arc stress-response systems were implicated as potential stress signal transducers that act upstream from the ROS cascade (Davies et al., 2009; Kohanski et al., 2008). Moreover, antibiotic-induced bacterial cell death appears to exhibit features characteristic of eukaryotic apoptosis (Dwyer et al., 2012). Thus, a case is building for the existence of bacterial PCD as a way to actively eliminate damaged members. However, before we can consider stress-mediated PCD to be a common rather than a rare, developmental property of bacteria, two key questions must be answered. First, how is the lethal stress response regulated? It must be triggered only by situations in which stress is very harsh to avoid unintended elimination of bacterial populations by moderate, transient stress. Second, is there a point of no return after which a self-sustaining death system takes over, even if the initial stress has dissipated? Control of triggers and checkpoints is an essential criterion for PCD.
As a starting point for identifying control elements of PCD, we assumed that stress responses restrict the consequences of damage rather than prevent the occurrence of additional primary lesions. If true, genes involved specifically in the response to lethal consequences of stress can be sought using antimicrobial probes whose lethal effects are readily distinguished from the primary bacteriostatic damage they cause. For example, the rapid lethal activity of the older quinolones can be separated from the formation of the drug-gyrase-DNA complexes that block bacterial growth (Chen et al., 1996). Thus, mutant libraries can be screened for hypersusceptibility to lethal stress and counter-screened for wild-type susceptibility to bacteriostatic stress (Han et al., 2010). The counter-screen eliminates from consideration mutants having altered drug uptake, drug efflux, and target affinity, each of which affects the formation of primary lesions. The outcome of such a screen is a collection of genes involved in protecting bacteria specifically from lethal damage. One of those genes might encode a general regulator of stress-mediated bacterial PCD.
In the present work we used nalidixic acid to screen for mutants that render E. coli hypersusceptible to the lethal effects of the drug without affecting bacteriostatic activity. One of the genes identified was yihE, which encodes a eukaryotic-like serine-threonine protein kinase (Zheng et al., 2007). Although yihE had long been thought to participate in stress responses because its promoter region contains a binding site for CpxR (Pogliano et al., 1997), a positive regulator of the Cpx envelope stress-response system (Raivio and Silhavy, 2001), no other connection had been obvious. We found that YihE restricts stress-stimulated ROS accumulation and bacterial cell death mediated by the MazEF toxin-antitoxin system. These results, plus data with katG and cpx mutants, are explained by YihE serving as a negative regulator of the MazEF-Cpx-ROS pathway that constitutes a live-or-die response of bacteria to stress. Since exposure to antimicrobials and host defense systems constitutes harsh stress to bacterial pathogens, artificially antagonizing YihE may be a new way to improve antimicrobial action and attenuate virulence.
When we examined a Tn5tac1 transposon insertion mutagenesis library of E. coli, screening identified a deficiency of yihE as increasing lethal activity of nalidixic acid without affecting bacteriostatic activity. Insertion of Tn5tac1 into yihE reduced survival of E. coli by 100-fold following nalidixic acid treatment (Figure S1A), but minimal inhibitory concentration (MIC), a common surrogate for measuring growth inhibition, was unaffected (MIC was ~3 μg/ml for both wild-type (strain 1045) and the yihE mutant (strain 2562); see Table S1 for strain descriptions). The hypersusceptibility to lethal action in the absence of an effect on growth, which we term hyperlethality, was readily transferred to other E. coli strains by bacteriophage P1-mediated transduction. To establish that the hyperlethal phenotype was characteristic of a yihE deficiency rather than arising from an intrinsic effect of transposon insertion (Wu et al., 2008), we prepared an in-frame deletion of yihE (Baba et al., 2006) in a different wild-type background (strain 3084) and used the ΔyihE mutant (strain 3086) in subsequent experiments.
As expected, the ΔyihE mutant had the same MIC for nalidixic acid as the corresponding wild-type strain (Table S2), while treatment with various concentrations of nalidixic acid for 2 hrs (Figure 1A), or with 50 μg/ml of nalidixic acid for various times (Figure 1B), reduced mutant survival by >100 fold relative to that of the wild-type strain. This hyperlethality was eliminated by expression of wild-type yihE from plasmid pACYC184 (~10–15 copies per cell (Sambrook et al., 1989)) using the native yihE promoter (Figure 1A). We conclude that the absence of YihE is responsible for the hyperlethal response to nalidixic acid. This result indicates that wild-type YihE normally protects E. coli from the lethal action of this quinolone without affecting factors (drug uptake, efflux, or target affinity) that confer MIC changes.
YihE also lowered the lethal action of other stressor types. With the ΔyihE mutant, tetracycline (Figure 1C), mitomycin C (Figure 1D), and ampicillin (Figure S1B) each exhibited elevated lethality with no change in MIC (Table S2). Moreover, when wild-type and ΔyihE mutant cells were exposed to two environmental stressors, UV light (Figure 1E) and hydrogen peroxide (Figure 1F), the ΔyihE mutation increased lethal susceptibility. However, no ΔyihE-mediated hyperlethality was observed with exposure to high temperature (Figure S1C) or to rifampicin (Figure S1D). Thus, YihE protects E. coli from many, but not all types of stress.
During the course of the work, an X-ray crystal structure of YihE was reported, and the protein was shown to be a novel eukaryotic-like Ser/Thr protein kinase (Zheng et al., 2007). However, little was revealed about the biological function of YihE. The X-ray structure of YihE suggested that amino acids Asp-217 and Ser-36 are important residues in the YihE active site, and Asp-217 was shown to be essential for kinase activity (Zheng et al., 2007). To assess the importance of YihE kinase activity in protecting from lethal stress, we replaced Asp-217 or Ser-36 with Ala in plasmid-borne copies of yihE, and then we examined the effect of the mutations in trans on complementation of the hyperlethal phenotype associated with the ΔyihE mutant. Full complementation of ΔyihE-mediated hyperlethality for nalidixic acid was achieved with plasmid-borne, wild-type yihE (Figure 2A, empty squares). In contrast, single-base substitutions that abolished (D217A) or reduced (S36A) kinase activity (Figure 2B) exhibited either no complementation (D217A, Figure 2A, empty diamonds) or only partial complementation (S36A, Figure 2A, filled diamonds). Western blot analysis showed that protein abundance was unaffected by these amino acid substitutions (Figure 2C). The observation that kinase activity paralleled the ability of plasmid-borne yihE to complement a chromosomal yihE deficiency showed that the kinase activity of YihE is responsible for reducing the effects of lethal stress.
Previous work suggested that MazEF promotes stress-mediated cell death for many of the stressors we examined (Engelberg-Kulka et al., 2006; Hazan et al., 2004; Sat et al., 2001). That work is controversial (Christensen et al., 2003; Pedersen et al., 2002; Tsilibaris et al., 2007), and indeed we failed to detect an effect of a ΔmazEF mutation on lethality (Figure 3A and 3B, empty triangles), perhaps due to differences in strain background, growth medium, and/or severity of the stress. Nevertheless, we combined a ΔyihE mutation with a deletion of mazEF to determine whether YihE acts through MazEF. If such is the case, the absence of a functional mazEF gene pair should eliminate the hyperlethality normally exhibited by the ΔyihE mutant during stress. With both nalidixic acid (Figure 3A) and UV irradiation (Figure 3B) a mazEF deletion suppressed the effect of ΔyihE. These data suggest that the hyperlethality associated with a yihE deficiency results from MazF-mediated toxicity. The loss of hyperlethality seen with the ΔyihE-ΔmazEF double mutant was reversed by transduction of wild-type mazEF back into the double mutant, effectively regenerating the ΔyihE strain (Figure 3A and 3B, diamonds). Thus, suppression of ΔyihE was due to deletion of mazEF, not to spontaneous suppressors that might have arisen during transduction analysis. We conclude that YihE protects E. coli from stress-mediated cell death by having a negative effect on MazF.
MazF has been implicated in stimulating toxic ROS accumulation (Kolodkin-Gal et al., 2008), and many lethal antimicrobials, including the older quinolones, appear to kill bacteria via hydroxyl radical action (Kohanski et al., 2007; Wang and Zhao, 2009; Wang et al., 2010). Thus, we suspected that YihE-mediated restriction of MazF activity might limit stress-stimulated, MazF-mediated surges of ROS. If so, YihE could be in the same genetic pathway as KatG, an enzyme known to protect from surges in ROS (Hassan and Fridovich, 1979). Indeed, hyperlethal antimicrobial susceptibility observed in a ΔkatG mutant was similar to that associated with deletion of yihE ((Wang and Zhao, 2009) and Figure 4A). Moreover, a ΔyihE-ΔkatG double mutant exhibited a hyperlethal phenotype similar to that observed with the ΔyihE and the ΔkatG single mutants (Figure 4A). These data indicate that yihE and katG are epistatic, i.e., they act in the same genetic pathway.
To further evaluate the involvement of ROS in YihE function, we examined the effects of treatment with thiourea plus 2,2′-bipyridyl, compounds that inhibit hydroxyl radical accumulation (Kohanski et al., 2007). Treatment of the wild-type and ΔyihE mutant strains with subinhibitory concentrations of thiourea plus 2,2′-bipyridyl blocked the ability of nalidixic acid to kill E. coli (Figure 4B). These data are consistent with both nalidixic acid-mediated lethality in wild-type cells and the hyperlethal effect in the ΔyihE mutant being achieved through accumulation of hydroxyl radical. As expected, thiourea plus 2,2′-bipyridyl also eliminated killing by nalidixic acid with the ΔkatG mutant (Figure 4B).
If ΔyihE-mediated hyperlethality is caused by excess accumulation of hydroxyl radical, a stressor that does not rely on hydroxyl radical to kill bacteria should not exhibit hyperlethality with a yihE-deficient strain. The investigational fluoroquinolone, PD161144, is such a stressor (Wang et al., 2010). This compound exhibited no hyperlethality with the ΔyihE mutant (Figure 4C).
As another test of the idea that ΔyihE-mediated hyperlethality is due to elevated accumulation of ROS, we treated cells with H2DCFDA, a cell-permeable dye that is converted intracellularly to an impermeable cognate that becomes fluorescent if oxidized by ROS. When wild-type and ΔyihE mutant cells were treated with nalidixic acid in the presence of H2DCFDA, they became fluorescently labeled. Analysis by fluorescence-activated cell sorting (FACS) revealed a peak shift from low to high fluorescence intensity with both wild-type and ΔyihE mutant cells (Figure 4D). The fluorescence intensity increase was greater with the ΔyihE mutant, indicating greater ROS accumulation in the mutant (Figure 4D). This experiment, plus the three described above, indicate that a deficiency of yihE facilitates bacterial killing through a stress-induced cascade of ROS similar to that observed with katG-deficient cells.
yihE is located in an operon that is positively regulated by CpxR (Pogliano et al., 1997), the response regulator of the Cpx envelope stress- response system. Consequently, we expected that deletion of cpxR would increase quinolone lethality, since loss of CpxR would prevent up-regulation of yihE and thereby allow more MazF activity. Surprisingly, a cpxR deficiency reduced rather than increased the lethal effect of nalidixic acid (Figure 5A). With UV irradiation, which was also expected to be more lethal in the cpxR-deficient mutant, no effect was seen with the cpxR-deficient strain (Figure 5B). For both stressors, a ΔyihE ΔcpxR double mutant exhibited the hyperlethality seen with the ΔyihE single mutant, not the protective effect (or lack of effect) seen with the ΔcpxR mutant (Figure 5A and 5B). Thus, the hyperlethality due to ΔyihE overrides the protective effect ofΔcpxR on lethal stress. These data indicate 1) that YihE is an essential element of the MazEF-Cpx-ROS pathway (see Discussion), 2) that yihE may be controlled by regulators other than CpxR, and 3) that the protective effect of the cpxR mutation with nalidixic acid depends on YihE function.
PCD has previously been referred to as an active death process that requires the presence of particular protein(s) (Engelberg-Kulka et al., 2006). Such a definition may be inadequate for stress-mediated bacterial PCD, because the original stressor that triggers death was not removed during the traditional killing assay. Consequently, the total cell death readout may derive from either the primary lesion, lesion-triggered PCD, or both. A more accurate definition of PCD requires cells to continue along the death pathway even after the initial PCD-triggering stressor is removed. To our knowledge, such self-sustaining post-stress PCD has not been demonstrated.
To establish an assay that can distinguish post-stress PCD from overall stress-mediated killing, we assume that bacterial PCD involves ROS, since that is the case with eukaryotic apoptosis (Jung et al., 2001; Mates and Sanchez-Jimenez, 2000; Simizu et al., 1998). If post-stress PCD exists and it relies on ROS to execute killing, removal of the initial stressor should not block PCD that has already been triggered by the initial stress. In contrast, if we remove the initial stressor and also scavenge ROS, post-stress PCD should be blocked. As a test of these ideas, we treated wild-type and ΔyihE mutant cultures with nalidixic acid and then removed the stressor by spreading a small amount of culture on agar, which was expected to dilute the drug to non-inhibitory concentrations (the quinolones are known for rapid reversal of activity upon removal of drug from medium ((Goss et al., 1965) and Malik M, Dlica K unpublished observation). Half of the agar plates contained thiourea to scavenge hydroxyl radical and block ROS-mediated post-stress PCD. Viable counts from these thiourea-containing plates revealed the drop in survival occurring before plating (i.e., overall stress-mediated killing minus post-stress PCD). The remaining agar plates lacked thiourea to reveal the drop in survival occurring both before and after plating (i.e., overall stress-mediated killing). The ratio of percent survival in the presence and absence of thiourea served as an indicator of ROS-mediated post-stress PCD. With the ΔyihE mutant, the ratio was 97 (Figure 6), which strongly indicates the existence of post-stress PCD in the mutant (the ratio would be close to 1 if no post-stress PCD occurred). With wild-type cells under conditions that matched the extent of killing to that of the ΔyihE mutant in the absence of thiourea, the PCD indicator ratio was about 10. Thus, ROS-mediated PCD also exists in wild-type cells, but it is 10-fold lower than with ΔyihE mutant cells. When similar experiments were performed with ciprofloxacin, a more potent quinolone than nalidixic acid, post-stress PCD was 25-fold lower with wild-type than with ΔyihE mutant cells (Figure 6). We conclude that quinolones initiate PCD through a pathway that culminates in cell death via hydroxyl radical accumulation even after removal of the initial stress stimulus.
The work described above identified the YihE protein kinase of E. coli as a central factor in limiting the self-destructive response of bacteria to lethal stress. Moreover, the work provided the first evidence for post-stress PCD (e.g. once triggered by an initial stress event, cells continue along a death pathway even after the initial stress dissipates). Deletion of yihE increased PCD and the lethal action of stressors such as peroxide, ultraviolet irradiation, and several antimicrobial classes. For the antimicrobials, the absence of YihE had no effect on bacteriostatic activity; consequently, YihE must affect the response to the lethal consequences of stress rather than the initial events that block growth. The crystal structure of YihE indicates that the protein is a eukaryotic-like Ser/Thr protein kinase (Zheng et al., 2007); we found that loss of kinase activity paralleled the increase in nalidixic acid-mediated lethality. Therefore, the protective effect of YihE is due to its kinase activity. Since YihE is the first bacterial protein kinase to mitigate lethal stress, we suggest that YihE be renamed stress response kinase A (SrkA). Below we discuss stress-mediated interactions among YihE, the toxin-antitoxin module MazEF, the Cpx two-component envelope stress-response system, superoxide, and KatG, a catalase/peroxidase that limits the cascade of ROS (see Figure 7).
Removal of the mazEF operon from a ΔyihE strain abolished ΔyihE-mediated hyperlethality to stressors; thus, the hyperlethal response seen in the ΔyihE mutant appears to occur through a MazF-mediated death pathway. YihE may normally act as a governor of the MazF endoribonuclease, which could allow cells time to repair stress-mediated damage. The YihE-MazF interaction appears to be indirect, since no phosphorylation of MazE, MazF, or MazG (discussed below) was observed when these proteins were purified and examined as substrates for the kinase activity of YihE (A. Dorsey-Oresto and X. Zhao, unpublished observation). Thus, passage of regulatory information from YihE to MazF is likely to be complex.
Antagonism of MazF function by YihE adds to several other ways in which MazF is negatively controlled. One is inhibition of MazF activity by the binding of MazE (Aizenman et al., 1996), and a second is the negative effect of MazE and MazF on the activity of their own promoter (Marianovsky et al., 2001; Zhang et al., 2003a). A third involves MazG, a protein that acts during nutritional stress (Gross et al., 2006). When bacterial cells are starved for nutrients, upregulation of relA and/or spoT results in the generation of guanosine 3′,5′ bispyrophosphate (ppGpp). This nucleotide, which accumulates rapidly, inhibits transcription and leads to rapid proteolytic degradation of the MazE antitoxin. MazG cleaves ppGpp, elevates levels of MazE, and indirectly neutralizes MazF. A fourth may derive from MazF being able to generate special “stress-ribosomes” that can selectively translate leader-less mRNA generated by MazF endoribonuclease activity (Vesper et al., 2011). Such translation may produce survival or death proteins that mitigate or exacerbate MazF-mediated lethality (Amitai et al., 2009). The existence of multiple negative controls over the MazF toxin is consistent with this protein having a potentially serious detrimental effect in bacteria.
Several lines of evidence indicate that YihE suppresses the effects of lethal stress occurring through accumulation of ROS, especially hydroxyl radical. First, deletion of katG, which encodes catalase/peroxidase, was epistatic to a deletion of yihE when quinolone-mediated cell death was measured. Second, treatment of E. coli with subinhibitory concentrations of thiourea plus 2,2′-bipyridyl, agents that inhibit hydroxyl radical accumulation (Kohanski et al., 2007; Wang and Zhao, 2009), blocked quinolone lethality, even when yihE or katG were deleted. Third, a fluoroquinolone stressor that does not depend on ROS to kill bacteria (Wang et al., 2010), exhibited no hyperlethality due to a yihE-deficiency. Fourth, treatment of E. coli with nalidixic acid increased intracellular ROS levels more for a ΔyihE mutant than for wild-type cells.
MazF also influences the accumulation of ROS. For example, with Bacilus subtilis, deletion of the MazF-like toxin (NdoA) eliminates a surge of hydrogen peroxide associated with kanamycin and moxifloxacin treatment. Moreover, this deletion partially protects from the lethal action of both antimicrobials (Wu et al., 2011). With E. coli, deletion of MazEF does not by itself reduce the lethal action of antimicrobials, probably because negative regulation by factors such as YihE normally keeps MazF at such low levels that the effects of a mazEF deletion are not obvious. Nevertheless, the absence of mazEF does lower the amount of protein carbonylation, a product of ROS-mediated oxidization of cellular proteins (Cattaruzza and Hecker, 2008; Maisonneuve et al., 2009; Wong et al., 2008), that is associated with antimicrobial treatment (Kolodkin-Gal et al., 2008). Moreover, thiourea plus 2,2′-bipyridyl reduce protein carbonylation following treatment of E. coli with antimicrobials (X. Wang and X. Zhao, unpublished observations). Collectively, these results fit with the idea that YihE inhibits MazF-mediated hydroxyl radical accumulation, possibly by reducing degradation of katG mRNA by MazF (a >2- fold reduction in katG mRNA level was observed when the ΔyihE mutant was treated with nalidixic acid (Figure S2)).
Stress leads to mRNA cleavage by MazF and the accumulation of truncated mRNA (Kohanski et al., 2008; Zhang et al., 2003b). Translation of truncated mRNA then generates abnormal proteins, some of which may lodge in the cell membrane and activate the Cpx system (Kohanski et al., 2008). Cpx activation upregulates YihE (Pogliano et al., 1997), which would in turn down-regulate MazF function and reduce the generation of truncated mRNAs and proteins. Cpx also up-regulates the degradative enzyme DegP and the protein-folding facilitator DsbA to remove/renature abnormal proteins lodged in the membrane (Raivio and Silhavy, 2001). These activities would eventually halt the induction of Cpx and reset the response system when stress dissipates. Thus, the MazEF and Cpx systems allow cells to respond protectively to low-to-moderate levels of lethal stress, as seen with the MazEF ortholog of B. subtilis (Wu et al., 2011).
Since a cpxR-deficient mutant exhibits lower susceptibility to lethal stress rather than the hypersusceptibility expected from the absence of a positive regulator (Kohanski et al., 2008), yihE is probably regulated by factors in addition to CpxR. Such factors have not been identified. The protective effect of a cpxR deficiency indicates that Cpx can also play a detrimental role in the response to stress. When lethal stress is high and persistent (e.g. exceeding a point of no-return (Amitai et al., 2004)), continued production of truncated proteins arising from MazF action is likely to lead to an interaction between the Cpx and the Arc (Aerobic Respiration Control) two-component system that perturbs the respiratory chain (Kohanski et al., 2008). Interruption of oxidative phosphorylation is expected to allow accumulation of superoxide, which would increase peroxide levels through spontaneous dismutation, dismutation via superoxide dismutases, and perturbation of iron-sulfur clusters (Kohanski et al., 2007). When the resulting peroxide increase overwhelms the KatG catalase/peroxidase system, levels of lethal hydroxyl radical would rise.
Several other observations fit a connection among YihE, MazF, and Cpx. One is the absence of ΔyihE-mediated hyperlethality with high temperature. If high temperature causes extensive protein denaturation that overwhelms the Cpx system, the contribution of toxin-mediated membrane protein truncation to envelope stress will be small, and the ΔyihE effect will be masked. Another observation is the absence of ΔyihE-mediated hyperlethality for rifiampicin. Rifampin inhibits transcription initiation, which would deplete RNA substrates for MazF-mediated RNA cleavage, reduce the mRNA and protein truncation that triggers the Cpx-Arc-ROS cascade (Kohanski et al., 2008), and thereby nullify a ΔyihE effect. Finally, hydroxyl radical causes damage to DNA, protein, and lipid. Such macromolecular damage could serve as a secondary stress input that triggers more hydroxyl radical accumulation. Such a destructive cycle is expected to be self-amplifying once a critical level of hydroxyl radical accumulates. That would explain how death can continue to occur even after lethal quinolone-mediataed stress is removed unless hydroxyl radical accumulation is blocked by thiourea. MazF may contribute to self-sustaining hydroxyl radical production by degrading katG mRNA, thereby pushing peroxide and hydroxyl radical levels beyond the critical threshold.
Observations described above indicate that the Cpx stress response system participates in both protective and destructive activities. Superoxide also appears to have dual function. While superoxide is part of the ROS cascade that leads to hydroxyl radical production and cell death, accumulation of superoxide also induces the protective SoxRS and MarRAB regulons (Gonzalez-Flecha and Demple, 2000; Liochev et al., 1999; Miller et al., 1994). Activation of SoxS by superoxide may activate MarRAB-AcrAB efflux pumps and other SoxS-controlled protective genes that mitigate stress (Blanchard et al., 2007; Chen et al., 2006; Miller et al., 1994; Pomposiello et al., 2001). A protective effect for superoxide is supported by the finding that a sodA-sodB double mutant, which spontaneously accumulates higher levels of superoxide than wild-type cells (M. Mosel, K. Drlica, and X. Zhao unpublished observation), exhibits increased rather than decreased survival with 3 classes of lethal antimicrobials (Wang and Zhao, 2009). Moreover, sublethal concentrations of plumbagin, a metabolic generator of superoxide, reduce the lethal effects of the DNA-damaging agent bleomycin (Burger and Drlica, 2009). Plumbagin and paraquat, another generator of superoxide, also protect E. coli from the lethal action of antimicrobial classes represented by ofloxacin, ampicillin, and kanamycin ((Wu et al., 2012) and Mosel and Zhao unpublished observation). Moreover, deletion of nfo, a repair gene controlled by soxRS, increases the lethal activity of a variety of lethal stressors (M. Mosel, K. Drlica, and X. Zhao unpublished observation).
The protective effect of superoxide may also derive from the susceptibility of iron-sulfur clusters to this oxygen radical (Flint et al., 1993; Gardner and Fridovich, 1991). Since many dehydrogenases in the TCA cycle contain iron-sulfur clusters that may be damaged by low-to- moderate concentrations of superoxide, moderate exposure to superoxide may halt the aerobic TCA cycle and force cells to undergo glycolysis to avoid lethal ROS accumulation. On the other hand, high concentrations of superoxide may generate excessive hydrogen peroxide and release free iron from damaged iron-sulfur clusters. The accumulation of peroxide and the release of free iron provide two components for the Fenton Reaction, which generates highly toxic hydroxyl radical. Thus, superoxide concentration may help determine whether the effects of lethal stress are dampened by protective pathways or amplified by hydroxyl radical accumulation.
Dual functionality has also been reported with a toxin deficiency in B. subtilis. In this organism, the MazF ortholog, NdoA, protects from low levels of ultraviolet irradiation, but it has a destructive effect when irradiation levels are high (Wu et al., 2011). Moreover, NdoA protects cells from antimicrobial stress, but it sensitizes cells to effects of high temperature and nutrient starvation. The generality of dual-functioning stress-response elements (e.g. TA modules, Cpx, and superoxide) helps resolve the controversy concerning opposing functions of the MazEF TA module (Christensen et al., 2003; Engelberg-Kulka et al., 2006; Tsilibaris et al., 2007): they both occur depending on conditions.
Genetic factors involved in the response to lethal stress can be identified by screening for defective genes that increase lethal stress without affecting bacteriostatic activity. YihE is the first to be studied in detail (14 others have been identified (Han et al., 2010). YihE fits within the general view that lethal stress factors have two functions. At low-to- moderate levels of stress, the MazEF toxin-antitoxin pair, the Cpx response system, and levels of superoxide are protective; at high levels of stress each system contributes to a cascade of ROS that ends with accumulation of toxic hydroxyl radical and cell death. At least three safety valves protect against the ROS cascade: YihE restricts MazF, KatG converts peroxide to water, and superoxide leads to induction of protective pathways. Why exposure to stress would trigger a genetically programmed death pathway in unicellular bacteria is open to speculation. Self-destruction may provide a selective advantage to bacterial populations by eliminating seriously damaged cells that might otherwise utilize scarce resources. It might also reduce the risk of hypermutation and loss of genetic integrity arising from massive, error-prone damage repair. Regardless of the reason, the existence of PCD provides a new opportunity to enhance antimicrobial activity.
E. coli K-12 strains (Table S1) were grown at 37 °C in Luria–Bertani medium. Susceptibility to stress was measured with exponentially growing cultures unless indicated otherwise. A Tn5 mutagenesis library was constructed by infecting strain AB1157 with defective bacteriophage lambda carrying Tn5tac1; kanamycin-resistant colonies were screened for nalidixic acid susceptibility using both killing and growth inhibition assays. Strain construction was by bacteriophage P1-mediated transduction. Recombinant plasmids containing yihE or other target genes were constructed by inserting PCR-amplified gene fragments into appropriate plasmid vectors (Table S3).
MIC was assayed by overnight incubation of E. coli in a series of tubes containing sequential 2-fold increases in drug concentration. Lethal activity was determined by incubating bacteria with various concentrations of antimicrobial or other stressor followed by plating on drug-free agar. Post-stress death was assessed by plating aliquots of stress-exposed cultures on LB agar containing or lacking thiourea at 125 mM (1/2 MIC).
pACYC184-yihE was used as a template in a mutagenic PCR with primers listed in Table S3. Nucleotide sequence determination confirmed the presence of the targeted mutations.
Plasmid carrying either yihE, mazE, mazEF, or mazG (see Table S1 for plasmid list and details) was introduced into competent E. coli BL21 (DE3) or Rosetta 2 cells by bacterial transformation. Transformants were grown to A600 = 0.5 and then treated with isopropyl-I, thio-B-D-galactopyranoside (IPTG) for an additional 2 hrs to induce expression of 6 x His-tagged proteins. Cell lysates were clarified by centrifugation, and proteins were purified from supernatant fluids using Qiagen Ni-NTA Fast Start kits as described by the manufacturer.
Kinase activity was measured by incubating YihE with myelin basic protein in kinase buffer (25 mM Tris-Cl pH 7.5, 2 mM MgCl2, 1 mM dithiothreitol) with or without 10 μCi [γ-33P] ATP at 37 °C for 30 min. Proteins were separated by SDS-polyacrylamide gel electrophoresis. Gels were stained with Coomassie Brilliant Blue to locate proteins and exposed to a PhosphorImager screen to measure incorporation of radioactivity.
E. coli cells were grown to early exponential phase, treated with 10 μM H2DCFDA for 30 min, treated with nalidixic acid (50 μg/ml) for 90 min, and then subjected to FACS analysis (Liu et al., 2012).
We thank Babak Baseri, Liping Li, Yulin Qu, and Yuzhi Hong for technical assistance and Ming Lu, David Perlin, Richard Pine, Issar Smith, and Sanjay Tyagi for critical comments on the manuscript. The work was supported by grants from National Institutes of Health (1-DP2-OD007423, 1-R01-AI 073491, and 1-R21 AI068014-01A2) and National Natural Science Foundation of China (No. 30860012 and 30973596).
Supplemental Information, including 2 figures, 3 tables, and Extended Experimental Procedures can be found with this article online.
The authors declare no conflict of interest.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.