|Home | About | Journals | Submit | Contact Us | Français|
Pathways of mutagenesis are induced in microbes under adverse conditions controlled by stress responses. Control of mutagenesis by stress responses may accelerate evolution specifically when cells are maladapted to their environments, i.e. are stressed. Stress-induced mutagenesis in the Escherichia coli Lac assay occurs either by ‘point’ mutation or gene amplification. Point mutagenesis is associated with DNA double-strand-break (DSB) repair and requires DinB error-prone DNA polymerase and the SOS DNA-damage- and RpoS general-stress responses. We report that the RpoE envelope-protein-stress response is also required. In a screen for mutagenesis-defective mutants, we isolated a transposon insertion in the rpoE P2 promoter. The insertion prevents rpoE induction during stress, but leaves constitutive expression intact, and allows cell viability. rpoE insertion and suppressed null mutants display reduced point mutagenesis and maintenance of amplified DNA. Furthermore, σE acts independently of stress responses previously implicated: SOS/DinB and RpoS, and of σ32, which was postulated to affect mutagenesis. I-SceI-induced DSBs alleviated much of the rpoE phenotype, implying that σE promoted DSB formation. Thus, a third stress response and stress input regulate DSB-repair-associated stress-induced mutagenesis. This provides the first report of mutagenesis promoted by σE, and implies that extracytoplasmic stressors may affect genome integrity and, potentially, the ability to evolve.
Stress-induced mutagenesis is a collection of mechanisms observed in bacterial, yeast and human cells in which mutagenesis pathways are activated in response to adverse conditions, such as starvation or antibiotic stresses (reviewed, Galhardo et al., 2007). These mechanisms enhance mutagenesis specifically during times of stress, and thus have the potential to increase genetic diversity upon which natural selection acts, potentially accelerating evolution, specifically when cells are maladapted to their environments, i.e. are stressed. These mechanisms are potentially important models for mutagenesis that drives antibiotic resistance (Cirz and Romesberg, 2007; Galhardo et al., 2007; Lopez and Blazquez, 2009; Petrosino et al., 2009; Kohanski et al., 2010) and pathogen–host evolutionary arms races (e.g. Prieto et al., 2006; Boles and Singh, 2008).
Although there appear to be multiple molecular mechanisms of stress-induced mutagenesis observed in different strains, organisms and environmental conditions, a strong common theme among them is the requirement for one or more cellular stress response(s). For example, induction of the SOS DNA-damage response is required for SOS ‘untargeted’ mutation of undamaged DNA (Witkin and Wermundsen, 1979), stress-induced reversion of a mutant lac gene in Escherichia coli cells starving on lactose medium (McKenzie et al., 2000), ciprofloxacin-induced resistance mutagenesis in E. coli (Cirz et al., 2005), bile-induced resistance mutagenesis in Salmonella enterica (Prieto et al., 2004; 2006;), and mutagenesis in aging E. coli colonies (Taddei et al., 1995). Similarly, the RpoS (σS) general- or starvation-stress response is required for most of the pathways listed above, except ciprofloxacin-induced mutagenesis (Bjedov et al., 2003; Layton and Foster, 2003; Lombardo et al., 2004; J. Casadesus, pers. comm.) and additionally for stress-induced gene amplifications during starvation in E. coli (Lombardo et al., 2004), stress-induced excisions of coliphage Mu (Gomez-Gomez et al., 1997; Lamrani et al., 1999), and stress-induced point mutagenesis (Saumaa et al., 2002) and transposon movement (Ilves et al., 2001) in Pseudomonas putida. The stringent response to amino acid starvation in E. coli (Wright et al., 1999) and Bacillus subtilis (Rudner et al., 1999b), and the ComK competence response to starvation in B. subtilis (Sung and Yasbin, 2002), are required for mutation pathways induced by starvation stresses. In human cells, two different stress responses to hypoxia provoke two separate mechanisms of genome instability (Bindra et al., 2007; Huang et al., 2007). Coupling of mutation pathways to cellular stress responses appears to be how mutagenesis is targeted specifically to times of stress. The stress responses are thus the cornerstone of the regulation of these mutation pathways, and their identities provide windows into the biological/environmental stressors that elicit the mutagenesis responses.
In the E. coli Lac assay for starvation-stress-induced mutagenesis (Cairns and Foster, 1991), either reversion of a frameshift mutation by compensatory frameshift mutation (Foster and Trimarchi, 1994; Rosenberg et al., 1994) in an F′-borne lac gene, or amplification of the leaky lac allele to multiple copies (Hastings et al., 2000; Powell and Wartell, 2001), allows growth of cells starving on lactose medium. The frameshift (‘point’ mutation) pathway requires double-strand-break (DSB)-repair proteins (Harris et al., 1994; 1996; Foster et al., 1996), TraI, an F-encoded single-strand endonuclease or an I-SceI-generated DSB (Ponder et al., 2005), the DinB error-prone DNA polymerase (Foster, 2000; McKenzie et al., 2001), and two stress responses: the SOS DNA-damage response (McKenzie et al., 2000) and the general-stress response regulated by RpoS (Layton and Foster, 2003; Lombardo et al., 2004), both of which upregulate dinB transcription. Our lab has provided evidence that supports a mechanism in which the point mutations are formed in acts of error-prone DSB repair. The normally high-fidelity DSB-repair mechanism switches to an error-prone mutagenic pathway that uses DinB during the SOS and RpoS responses (Ponder et al., 2005). [Alternative models discussed by Roth et al. (2006) and Galhardo et al. (2007) and below]. The GroEL chaperone also modulates DinB activity (Layton and Foster, 2005). GroEL is an essential protein complex that is both constitutively expressed and upregulated during the RpoH/σ32-controlled cytoplasmic unfolded-protein-stress or heat-shock response, and also directly regulates the activity of σ32 (Guisbert et al., 2004).
The rpoE gene encodes the σE transcription factor, which positively regulates the envelope-stress response to extracytoplasmic unfolded proteins. Originally identified as a heat-shock factor, σE is essential for growth in E. coli at all temperatures although the exact nature of the σE requirement for growth is not known (Hiratsu et al., 1995; Rouviere et al., 1995; De Las Penas et al., 1997b). In E. coli, σE expression during either exponential or stationary phase up- or downregulates as many as 200 genes involved in all areas of metabolism (Kabir et al., 2004; Rhodius et al., 2006). Although many σE-regulated genes encode proteins that promote either maintenance or synthesis of the cell envelope, a significant number encode cytoplasmic proteins that function in transcription, translation, and DNA synthesis and repair (Rhodius et al., 2006).
The regulation of σE, and thus the stress response that it controls, is complex and entails at least two independent pathways (Ades, 2004). First, RseA, the σE anti-sigma factor, is an integral transmembrane protein that binds σE, sequestering it at the cytoplasmic face of the inner membrane (Missiakas et al., 1997; De Las Penas et al., 1997a; Campbell et al., 2003). An inner membrane protease cascade (Ades, 2008), part of a regulatory pathway termed RIP (regulatory intramembrane proteolysis) conserved from bacteria to mammals (Rudner et al., 1999a; Brown et al., 2000), is responsible for degradation of RseA. Upon stress, unfolded outer membrane proteins trigger RseA proteolysis thereby releasing σE into the cytoplasm where it can associate with RNA polymerase and activate transcription of the σE-regulon genes (Kanehara et al., 2003; Walsh et al., 2003). Turnover of RseA is continuous, but upon stress, degradation accelerates until the stress is removed, allowing increased σE levels in the cytoplasm and induction of transcription of the σE regulon (Ades et al., 1999; Alba et al., 2001; 2002; Kanehara et al., 2003; Grigorova et al., 2004).
In addition to activation by unfolded membrane proteins, σE is also regulated growth-phase-dependently by the stringent response, via a less well-defined pathway. This results in stationary-phase induction of σE in response to the alarmone, ppGpp (Costanzo and Ades, 2006; Costanzo et al., 2008).
In this study, we show that activation of the σE stress response is required for stress-induced mutagenesis in the E. coli Lac system. We describe a transposon insertion that disrupts σE expression during stress. We show that stress-inducible rpoE expression (the σE stress response) is not required for viability but the constitutive rpoE expression is, and we correlate the stress response, rather than the constitutive essential function of σE, with formation of stress-induced point mutants and with maintenance (and thus accumulation of) DNA amplification. We show that σE plays at least two roles in the point-mutation pathway, one of them involving formation of DSBs. This adds a third stress response to those controlling stress-induced point mutagenesis, a second to that controlling amplification, and additionally adds genetic and genomic instability to the repertoire of consequences of the σE extracytoplasmic protein stress response in E. coli.
We developed a genetic screen, used it with transposon mutagenesis to identify mutants defective in stress-induced mutagenesis in the Lac assay (Experimental procedures in Supporting information), and identified a new mutation that affects the rpoE gene, designated rpoE2072::Tn10dCam. rpoE2072::Tn10dCam was moved by P1 transduction into a ‘clean’ (non-mutagenized) assay strain and characterized.
We find that the rpoE2072::Tn10dCam insertion lies between the −10 and −35 regions of σE-dependent (P2) gene-proximal promoter (Fig. 1A). Such an insertion might be expected to block transcription entirely and, given that rpoE is an essential gene (De Las Penas et al., 1997b; Alba and Gross, 2004), allow growth only of isolates with additional suppressor mutations that restore viability. rpoE null mutants acquire unlinked suppressor mutations (De Las Penas et al., 1997b). Using co-transduction experiments, we show that rpoE2072::Tn10dCam insertion mutant does not contain or require unlinked suppressor mutations for its viability (Fig. S1). We used a kanamycin-resistance marker in yfhH, a non-essential gene partially linked with rpoE2072::Tn10dCam, to transduce cells carrying neither the rpoE nor yfhH::Kan mutation with phage grown on the yfhH::Kan rpoE2072::Tn10dCam donor (Fig. S1). Co-transductant frequencies of 51 ± 4% CamR/KanR transductants, and 67 ± 2% KanR/CamR transductants (Fig. S1) show that although there is a slight bias against acquisition of the rpoE2072 allele, the rpoE2072::Tn10dCam insertion does not induce lethality when transduced into a clean (suppressor-mutation-free) strain background (Fig. S1). Thus unlinked suppressor mutations are not required for the mutant's growth. Although rpoE2072::Tn10dCam is not a null mutation (Fig. S1) this allele causes a slight temperature-sensitive phenotype with similar numbers of cfu at 30°C and 37°C, but about twofold fewer at 42°C relative to 30°C (not shown).
Two likely explanations for the viability of rpoE2072 cells are either that the 3′ end of the transposon supplies a poorly conserved −35 sequence (Fig. 1B) or that transcription is being initiated from an outward-reading promoter within the transposon allowing some σE expression. Data below indicate that the insertion affects the expression from the σE-dependent P2 promoter.
We find that rpoE2072::Tn10dCam cells do not induce the σE response to a stress peptide signal. We tested whether the transposon insertion in P2 affected stress induction of rpoE specifically by using an rpoHP3::lacZ fusion that is dependent on σE for activity (Mecsas et al., 1993). β-Galactosidase activity was assayed over time following initiation of the σE stress response by induction of plasmid-borne YYF, a peptide homologous to the unfolded outer membrane protein-stress signal that activates DegS protease to degrade RseA (anti-σE), liberating σE to induce transcription of the stress response genes (Walsh et al., 2003). Our data show that stress induction of σE-regulon genes is impaired in the rpoE2072::Tn10dCam mutant as follows. Following IPTG induction of the YYF stress-signal peptide from plasmid pBA166, β-galactosidase activity expressed from the rpoHP3–lacZ fusion increased approximately fivefold in rpoE+ cultures coincident with an increase in growth (Fig. 2A and B), whereas in the rpoE2072::Tn10dCam strain, β-galactosidase activity increased only slightly and the cultures stopped growing (Fig. 2A and B). The data show an inability of rpoE2072::Tn10dCam cells to induce the σE stress response in response to inducer peptide.
We can understand the cessation of growth in the rpoE2072 mutant upon induction of YYF (Fig. 2B) as follows. YYF is expected to activate all σE-regulated promoters, including the promoter immediately upstream of the rseA anti-sigma factor gene, but not the transposon-disrupted rpoEP2. This would be expected to result in induction of rseA but not rpoE, shifting the σE:RseA ratio to favour RseA and decreasing available σE. We suggest that an inability to elevate transcription of the σE-dependent P2 promoter of rpoE in response to stress may result in an imbalance between sigma factor and anti-sigma factor, leading to depletion of active σE and thus the observed cessation of growth, because σE function is required for growth (Hiratsu et al., 1995; Rouviere et al., 1995; De Las Penas et al., 1997b).
First, quantitative stress-induced mutagenesis assays confirmed that rpoE2072::Tn10dCam decreases stress-induced Lac+ colony formation dramatically at 30°C (Fig. 3A) and 37°C (Fig. 3C). This decrease cannot be explained by net death of the population in that the overall survival of the population is decreased little during the stress-induced mutagenesis experiments (results range from no decrease, Fig. 3B, to at most a fourfold decrease at later time points in some experiments on amplification below). The Lac+ reversion rate from multiple experiments was approximately 11-fold lower in the rpoE2072::Tn10dCam than wild-type cells (Fig. 3C).
Second, an rpoE null mutation (in cells with extragenic suppressor mutations) causes a similar severe stress-induced mutagenesis defect (Fig. S2). Stress-induced mutagenesis assays were performed at 30°C, because even with its suppressor mutation, the rpoE null mutant is temperature-sensitive and cannot grow at 37°C (De Las Penas et al., 1997b; Alba and Gross, 2004). Nearly identically to the rpoE2072::Tn10dCam mutant, accumulation of stress-induced Lac+ mutant colonies was greatly reduced in the rpoE null mutant (Fig. S2). Similar results were obtained in another strain set in which the rpoE+, the rpoE null and the rpoE2072::Tn10dCam strains all carried ΔydcQ, a known rpoE suppressor mutation (Button et al., 2007), to ensure their isogenicity (Fig. S2). ΔydcQ increased the rate of accumulation of Lac+ mutant colonies in all strains but did not affect the relative reduction in the null mutant compared with the rpoE+ (Fig. S2). These results generalize the conclusion that functional σE is required for appearance of stress-induced mutants in the Lac assay.
Third, the apparent decrease in Lac+ colonies in the stress-induced mutagenesis assay could result either from a failure to form Lac+ mutations or from an inability of the Lac+ mutant colonies carrying an rpoE mutation to form colonies under the assay conditions. We performed reconstruction experiments the results of which show that cells carrying rpoE2072::Tn10dCam and a Lac+ mutation form colonies nearly normally under selective conditions (Table S1). Therefore, we conclude that formation of the mutants, not their subsequent growth on the selective medium, is impaired by the rpoE2072::Tn10dCam allele.
The results shown in Tables S1 and S2 also address a specific hypothesis for why σE might be required for stress-induced mutagenesis. Lac+ stress-induced mutagenesis is associated with a cell subpopulation with transiently increased mutation rates (a hypermutable cell subpopulation) as seen by data showing that Lac+ mutants possess higher levels of other mutations throughout their genomes than stressed cells that did not become Lac+ (Torkelson et al., 1997; Rosche and Foster, 1999; Godoy and Fox, 2000), and that the mutagenesis process is not easily uncoupled from generation of these unselected ‘secondary’ mutations (Gonzalez et al., 2008). The σE stress response might have been required to allow normal growth rates of cells carrying additional unselected chromosomal mutations, perhaps by upregulating chaperones and proteases in response to resulting misfolded proteins. However, the data in Table S2 show that placing the rpoE2072::Tn10dCam allele into five different Lac+ revertants with accumulated unselected secondary mutations did not inhibit their colony-forming ability. The slightly longer times to form colonies overall reflect the lower growth temperature of 32°C used in these experiments (Table S2), whereas the slightly increased time for colony formation under lactose-selective conditions at 37°C (Table S1) appears to be the result of the rpoE mutation alone. Therefore, failure to ameliorate mutant proteins appears not to underlie the strong mutagenesis defect of the rpoE2072::Tn10dCam mutant (Fig. 3); rather, mutagenesis itself is implicated.
Stress-induced Lac+ colonies arise via either ‘point mutation’ (compensatory lac frameshift mutation) (Foster and Trimarchi, 1994; Rosenberg et al., 1994) or via amplification of the leaky lac allele, with lac amplification representing a minority of the Lac+ colonies until after day 8 of starvation on lactose plates (Hastings et al., 2000; Powell and Wartell, 2001). We found that accumulation of lac-amplified colonies was decreased 320-fold in the rpoE2072::Tn10dCam mutant relative to the rpoE+ strain (Fig. 4A and C). In these experiments, in which lac-amplified cfu were separated from the point mutants, the point mutation rate was decreased ~15-fold by rpoE2072::Tn10dCam relative to rpoE+ (Fig. 4D). Although there is some loss of cell viability in the rpoE strain on later days in some experiments (e.g. Fig. 4B), the at most 75% (≤ 4-fold) loss of viability does not explain the ~300-fold deficiency in accumulation of lac-amplified colonies.
Reconstruction of rpoE2072::Tn10dCam lac-amplified strains showed that the decrease in viable cfu (Fig. 4B) was not caused by slow growth, but rather an inability of rpoE2072::Tn10dCam strains to maintain amplification: first, on average, the transconjugants of lac-amplified DNA into rpoE2072::Tn10dCam recipients were 32-fold less able to form colonies on lactose medium than the rpoE+ transconjugants (Table 1, columns 2–4). Second, this lower recovery was not due to slower growth (5.1 ± 0.2 days for rpoE+ versus 4.9 ± 0.3 days for rpoE2072::Tn10dCam cells to form colonies, Table 1), or to lower efficiency of conjugation (Table 1, last two columns). Third, when F′ factors carrying a Lac+ point mutation were conjugated into rpoE+ and rpoE2072::Tn10dCam F- cells (Table 1), there was no difference in conjugation efficiency between rpoE+ and rpoE2072::Tn10dCam strains, showing that the poor plating efficiency of rpoE cells plated on lactose is unique to amplification (Table 1). Thus, we conclude that this decrease represents a failure to maintain the amplified arrays. In addition to the ~30-fold defect in maintenance of amplified arrays (Table 1) there is a variable ≤ 75% (≤ 4-fold) decrease in viable cell counts of the rpoE2072::Tn10dCam in late days in some experiments (Fig. 4B). These two factors can account for 30 × 4 = about 120-fold of the decrease observed in accumulation of lac-amplified cfu during stress-induced mutagenesis experiments. Because the decrease in accumulation of stress-induced lac-amplified colonies is ~300-fold (Fig. 4C), it appears likely that, in addition to inability to maintain gene amplifications once formed, there was also a decrease in amplification formation caused by the rpoE2072::Tn10dCam mutation.
Fluctuation tests were performed to determine whether induction of the σE response provokes mutagenesis generally in growing cells, as, for example, induction of the RpoS regulon does (Yang et al., 2004). We measured mutation to rifampicin resistance, which occurs by any of a few specific base-substitution mutations in the rpoB gene (Jin and Gross, 1988), and used the plasmid-borne gene for the YYF peptide to induce expression of σE-regulated promoters in rpoE+ cells. Expression of YYF had no effect on mutation rate per generation indicating that the σE response does not provoke mutagenesis in general but its requirement is specific for stress-induced mutagenesis (Table S3).
σE upregulates transcription of homologous-recombination genes recB, recD, recJ and recR (Rezuchova et al., 2003; Kabir et al., 2005; Rhodius et al., 2006). Because recB mutants (Harris et al., 1994) and recD recJ double mutants (Foster and Rosche, 1999) are defective in stress-induced mutagenesis, possible loss of homologous-recombination proficiency in rpoE2072::Tn10dCam cells might explain their reduced stress-induced mutagenesis. However, first, whereas recB single mutants and recD recJ double mutants are hypersensitive to ultraviolet (UV) irradiation (Lovett et al., 1988), rpoE2072::Tn10dCam cells are not (Fig. S3), implying that levels of RecB, RecD and RecJ are not, in general, drastically reduced in this strain. These experiments cannot rule out insufficiency of RecB, RecD or RecJ under a particular stress condition that elicits a σE response. Second, quantitative Hfr-mediated conjugation experiments (Table S4) and quantitative P1 transduction experiments (Fig. S4) showed the rpoE2072::Tn10dCam mutant to be as recombination-proficient as its isogenic rpoE+ parent. These results cannot exclude a possible recombination defect specific to recombination during cellular stress such as may occur during stress-induced mutagenesis. Third, if stress-induced mutagenesis required wild-type stress-inducible σE solely to produce sufficient levels of RecB, then we would expect stress-induced mutagenesis to be σE-independent if a different homologous-recombination (HR) pathway was used to substitute for RecBC. Poteete et al. showed that the phage λ Red HR system can substitute for recBC in stress-induced mutagenesis in the Lac assay (Poteete et al., 2002). As shown previously (Poteete et al., 2002), mutation rate was stimulated more than eightfold in ΔrecBCD::red+ relative to isogenic recBC+ cells (Fig. S5A). However, functional RpoE was still strongly required for stress-induced mutagenesis in ΔrecBCD::red+ cells (Fig. S5A). Therefore, the primary cause of the mutagenesis defect of rpoE cells is not deficient expression of recB or recD. Fourth, E. coli recJ mutant cells are not impaired in stress-induced mutagenesis (Harris et al., 1994; Foster and Rosche, 1999), and we find that stress-induced mutagenesis was unchanged in the recJ or recR single mutants as well as in the recJ recR double mutant relative to the isogenic rec+ strain (Fig. S5B). Therefore, diminished RecJ and/or RecR production is not the cause of the stress-induced mutagenesis defect of rpoE cells.
We wished to distinguish whether the σE response is an independent stress input into mutation, or feeds into the SOS, RpoS or cytoplasmic heat-shock stress responses. We first examined its effect on error-prone DNA polymerase, Pol IV or DinB, which is required for DSB-associated stress-induced point mutagenesis (Foster, 2000; McKenzie et al., 2001) and is upregulated transcriptionally by both the SOS (Kim et al., 1997; Courcelle et al., 2001) and RpoS (Layton and Foster, 2003) responses. dinB upregulation is the only role of the SOS response in stress-induced mutation (Galhardo et al., 2009). Although dinB upregulation by σE has not been reported (Rhodius et al., 2006), a modest increase might have been overlooked.
First, we found that σE does not generally upregulate dinB transcription (Fig. 5A). We measured β-galactosidase activity in cells carrying a plasmid-borne dinB promoter fused to lacZ in rpoE+ and rpoE2072::Tn10dCam cells. No difference in activity was observed in either SOS-proficient lexA+ or lexA(Def) strains in which the SOS response is constitutively derepressed, suggesting that dinB transcription is not controlled by σE (Fig. 5A). Second, Western immunoblot analyses showed that DinB protein levels were unchanged in the rpoE2072::Tn10dCam mutant relative to rpoE+ cell extracts (Fig. 5B). Third, we asked whether the σE stress response might function by simply activating the SOS response, which is required for stress-induced mutagenesis (McKenzie et al., 2000), as follows. lexA(Def) (null) mutants are constitutively derepressed for the SOS/LexA regulon genes resulting in higher levels of SOS-controlled proteins including DinB (Friedberg et al., 2005). The factor(s) regulated by σE appear not to be DinB or any other protein induced by the SOS response because the lexA(Def) allele did not restore the ability to mutate to rpoE2072::Tn10dCam cells (Fig. 5C). Fourth, we used two operator-constitutive mutations of dinB that produce SOS-induced levels of DinB constitutively (Galhardo et al., 2009), and found that these do not restore normal levels of mutagenesis to rpoE2072::Tn10dCam cells (Fig. 5D), although they do to SOS-induction-defective cells (Galhardo et al., 2009). These results show that SOS induction and dinB expression are not the limiting factor for stress-induced mutagenesis in the rpoE2072::Tn10dCam cells defective for inducing the σE response.
The σE stress response upregulates σ32, the transcriptional activator of the cytoplasmic unfolded-protein or heat-shock response, which in turn increases transcription of the genes encoding the GroELS chaperone (Lund, 2001), which is required for efficient stress-induced Lac reversion (Layton and Foster, 2005). We tested whether the requirement for the σE response induction in mutagenesis was in fact a requirement for inducing the σ32 regulon by examining dnaK756 mutant cells, in which the σ32 regulon is expressed at induced levels constitutively (Straus et al., 1990). The data show that the dnaK mutation does not substitute for a functional σE response as follows (Fig. 6). First, the dnaK mutation decreases stress-induced Lac reversion (Fig. 6A). However, a functional RpoE response is still required for the mutagenesis that remains, as seen by the severe decrease in mutagenesis in the dnaK rpoE2072::Tn10dCam double mutant relative to the dnaK mutant cells (Fig. 6A). These data indicate that the σE stress response provides function(s) for mutagenesis other than or in addition to those upregulated by the σ32-activated cytoplasmic heat-shock response.
The general- or stationary-phase-stress response controlled by σS (RpoS) is required for both stress-induced point mutagenesis (Layton and Foster, 2003; Lombardo et al., 2004) and amplification (Lombardo et al., 2004); however, two lines of evidence imply that a defect in inducing the RpoS stress response is not the primary cause of diminished mutagenesis in rpoE mutant cells. First, because katE is induced RpoS-dependently in stationary phase (Mulvey et al., 1990), we assayed β-galactosidase activity in rpoE+ and rpoE2072::Tn10dCam saturated cultures that contained the katE::lacZ fusion. We find that rpoE2072::Tn10dCam cells induce the fusion gene similarly to isogenic rpoE+ cells and more than the rpoS deletion mutant in which β-galactosidase activity is severely reduced (Fig. 6C). Second, the requirement for rpoS in stress-induced mutagenesis is not mitigated by provision of DNA DSBs using a restriction enzyme in vivo, indicating that RpoS acts downstream of the production of DSBs in the mutagenic mechanism (Ponder et al., 2005). We show in the following section that much of the requirement for σE is alleviated by introduction of enzyme-generated DSBs.
A fundamental requirement for stress-induced mutagenesis is a DSB, which, in the F′, is thought to originate most frequently from a single-strand nick made at oriT by the F-encoded TraI endonuclease (Foster and Trimarchi, 1995; Rosenberg et al., 1995). This single-strand break is thought to become a DSB during replication (Kuzminov, 1995; Rosenberg et al., 1995; Rodriguez et al., 2002). TraI is required for stress-induced mutation of lac in the F′ but this requirement can be bypassed entirely, and mutagenesis stimulated an additional ~70-fold more, by DSBs made near lac by the I-SceI double-strand endonuclease in vivo (Ponder et al., 2005). I-SceI-induced DSBs made near lac increased mutation rate ~6000-fold in the absence of traI and ~70-fold in its presence (Ponder et al., 2005). Unlike TraI, the requirements for DSB-repair proteins, an inducible SOS response, RpoS and DinB could not be substituted for by I-SceI cuts. The substitution of I-SceI cuts specifically for TraI supports the model that the role of TraI in mutagenesis is the ultimate generation of a DSB (Ponder et al., 2005).
We found that I-SceI-induced DSBs near lac could relieve much of the mutagenesis defect caused by rpoE2072::Tn10dCam (Fig. 7A). First, in the left panel in Fig. 7A, we see that rpoE2072::Tn10dCam depresses the mutation rate 15-fold in cells that do not make I-SceI-induced DSBs because they have either only the chromosomal inducible I-SceI gene but no cutsite (ISceI and rpoE::Tn ISceI), or have neither enzyme nor cutsite. Second, when DSBs were induced in the rpoE2072::Tn10dCam mutant (Fig. 7A, right panel), the stress-induced Lac+ reversion rate increased nearly 500-fold relative to the control strain that expressed I-SceI in the absence of the I-SceI cutsite (Fig. 7A, left panel). In the DSB-inducing strain (Fig. 7A, right panel), the rpoE2072::Tn10dCam mutation caused only a threefold depression of mutation rate, compared with its 15-fold defect in the absence of I-SceI-made DSBs (Fig. 7A, left panel). This constitutes an ~80% alleviation of the rpoE2072::Tn10dCam mutagenesis defect (3/15 = 20% of the mutagenesis defect remaining). Thus much of the requirement for the σE response in stress-induced mutagenesis is a requirement for creation of the DSBs that provoke mutagenesis, implying that the σE response promotes DSB formation in F. This numerical comparison assumes that the mutation mechanism is the same with I-SceI cuts as with TraI. In all ways testable, this was shown to be so previously: lac mutation sequence spectrum, requirements for RecA, RecB, Ruv, SOS, RpoS and DinB, and fraction amplified and point mutant (Ponder et al., 2005). Thus, the comparison appears justified. Western analyses show that TraI protein levels are reduced only about 30 ± 0.6% in the rpoE2072::Tn10dCam mutant relative to isogenic rpoE+ cells (Fig. 7B). Thus, although DSB formation appears to be limiting in the rpoE2072::Tn10dCam background, it is probably not due to lowered expression of the F plasmid traI. We cannot rule out an effect on TraI activity; however, unaltered conjugation frequency in the rpoE2072::Tn10dCam mutant (data not shown) argues against this possibility. Other possibilities are considered below.
One interesting aspect of the data presented is the discovery that rpoE2072::Tn10dCam, a mutation that allows constitutive expression but not induction of the σE stress response by the YYF inducing peptide (Fig. 2), retains viability in the absence of unlinked suppressor mutations (Fig. S1). This implies that the essential function of σE relates to one or more of the functions it controls constitutively, not to transient expression of the stress response. The rpoE2072::Tn10dCam allele is likely to provide a useful reagent for future genetic studies of the effects of loss of the stress response, which can now be achieved both specifically, without simultaneous loss of the essential function, and cleanly, in cells that do not also carry suppressor mutations.
The data presented show that the stress–response function of σE is required for stress-induced point mutagenesis (Figs 3 and and4D,4D, Tables S1 and S2) and gene amplification (Fig. 4, Table 1). Moreover σE constitutes an independent stress and stress–response input to mutation, independent of known SOS and σS and postulated σ32 heat-shock response involvement (Figs 5–7). Therefore, first, stress-induced point mutagenesis and amplification are controlled by multiple independent stress responses: the SOS DNA-damage (McKenzie et al., 2000), RpoS general-stress (Layton and Foster, 2003; Lombardo et al., 2004) and σE responses for point mutation, and RpoS (Lombardo et al., 2004) and σE for amplification. Input of multiple stress responses is important in regulating the circumstances under which cells increase genetic diversity and their potential to evolve. Second, the results indicate that the σE response influences genome (in)stability generally.
Although both σE and σS responses are necessary for the genomic instability studied here, many stressors are expected to activate both, including starvation (via ppGpp) (Costanzo and Ades, 2006; Costanzo et al., 2008), but also many antibiotics, which may trigger membrane stress (such as β-lactams) and all of which appear to induce reactive oxygen species (reviewed, Kohanski et al., 2010) and so oxidative stress, an inducer of σS. The coupling of mutagenesis to antibiotic stressors is a potentially serious problem with resistance mechanisms instigated by the antibiotics (Cirz and Romesberg, 2007; Galhardo et al., 2007; Lopez and Blazquez, 2009; Cohen and Walker, 2010; Kohanski et al., 2010).
One major role of the σE response in stress-induced mutagenesis is apparent: ~80% of its function is substituted by an I-Sce-induced DSB (Fig. 7), implying that somehow the σE response promotes TraI-generated double-strand-end (DSE) formation, for which I-SceI also substitutes (Ponder et al., 2005). Previous work indicates that the point mutagenesis occurs in acts of error-prone DSB repair, and that in the F′, most of the DSBs originate from the action of TraI endonuclease (Ponder et al., 2005). The breaks are repaired non-mutagenically in unstressed cells and mutagenically, using DinB error-prone DNA polymerase if RpoS is induced either by stress or artificially (Ponder et al., 2005). The σE response could promote either the nicking by TraI or the replication into that nick. Although TraI levels were affected only modestly in the rpoE2072::Tn10dCam mutant (Fig. 7B), it is possible that replication or specifically F replication might be affected by σE, and the effect of σE might be F-specific. The mechanism of the remaining ~20% of the σE-response effect is unknown. Given that one of the ways that the σE response is induced is via starvation using ppGpp and the stringent response (Costanzo and Ades, 2006; Costanzo et al., 2008), it might, like RpoS, facilitate the switch to mutagenic repair of DSBs under stress.
Others favour a model of mutation in the Lac system that does not involve stress or stress responses but instead invokes growth of cells carrying an amplified lac array, which produces more β-galactosidase activity from the weakly functional mutant lac gene (Roth et al., 2006; Roth, 2010). In this model, when about 105 cells with many lac copies are present in a microcolony, a Lac+ point mutation occurs spontaneously, and the point mutant then overgrows the colony. We do not favour this model for several reasons. First, its prediction that point mutants arise from lac-amplified clones was contradicted by experiments showing that: (i) Lac+ microcolonies, as early as the two-cell stage, are pure point mutants, not mostly amplified with point mutants arising later – point mutants did not arise in amplified young colonies (Hastings et al., 2004), (ii) lac-amplified microcolonies do not produce point mutants efficiently under selective conditions (Hastings et al., 2004), and (iii) mutation of the DNA polymerase I gene obliterates amplification without diminishing point mutation (Hastings et al., 2004; Slack et al., 2006). This would be impossible if amplification were a precursor to point mutation, as the model specifies. As far as we are aware, alternative interpretations of these data have not been offered. Other data also contradict this model (Stumpf et al., 2007). Second, the fact that three stress responses are required for point mutation [SOS (McKenzie et al., 2000), RpoS/σS (Layton and Foster, 2003; Lombardo et al., 2004) and σE] and two for amplification [σS (Lombardo et al., 2004) and σE] is not compatible with the amplification model, which does not involve stress or stress responses. Neither are results, reviewed above, that showed that DSB repair switches to a mutagenic mode using DinB if RpoS is expressed, even in the absence of any external stress or selection (Ponder et al., 2005). We do not know of alternative interpretations for those data, and favour their obvious interpretation: that stress responses upregulate mutagenesis when cells are stressed.
The coupling of mutagenesis to stress responses means that cells turn up mutation rate specifically when they are maladapted to their environment, i.e. are stressed, potentially accelerating evolution specifically then. This can be a powerful device for adaptation to adversity including antibiotics and host defences in bacteria (McKenzie and Rosenberg, 2001; Cirz et al., 2005; Galhardo et al., 2007), and hypoxia (Bindra et al., 2007; Huang et al., 2007) and chemotherapeutic drugs in developing cancers. More than 80% of E. coli natural isolates respond to stress with mutagenesis, and modelling supports the potential benefits in enhanced evolvability of this response (Bjedov et al., 2003).
The requirement for (at least) three stress responses for point mutagenesis (SOS, σS and σE) and two for amplification (σS and σE) implies that cells do not instigate potentially dangerous programmes of genomic instability until they sense multiple independent stressors. Similarly in B. subtilis, both the stringent response (Rudner et al., 1999b) and ComK-controlled competence stress response (Sung and Yasbin, 2002) are required for starvation-induced mutagenesis. In Salmonella, bile-induced resistance mutagenesis requires both SOS and σS responses (Prieto et al., 2006; J. Casadesus, pers. comm.). Coupling genome instability to more than one stress response, any of which might not be turned on in all cells in a population, may further restrict or regulate mutagenesis to a cell subpopulation (Galhardo et al., 2007; Gonzalez et al., 2008), providing one way to achieve a potential bet-hedging mechanism such as is seen in bistable subpopulations critical to many bacterial survival strategies (Veening et al., 2008). These are highly regulated programmes, exquisitely tuned to cellular stresses, which regulate mutagenesis, and thus the ability to evolve, temporally. Understanding and targeting the regulatory components is likely to provide powerful new antibiotic strategies.
Strains used are listed in Table S5. Standard genetic techniques were used in strain constructions (Miller, 1992). All M9 minimal media (Miller, 1992) had carbon sources added at 0.1% and thiamine (vitamin B1) at 10 µg ml−1. LBH medium is per, for example, Torkelson et al. (1997). Antibiotic and other additives were used at the following final concentrations: ampicillin, 100 µg ml−1; chloramphenicol, 25 µg ml−1; kanamycin, 50 µg ml−1; tetracycline, 10 µg ml−1; rifampicin, 100 µg ml−1; 5-bromo-4-chloro-3-indolyl-β-d-galactoside (X-gal), 40 µg ml−1; sodium citrate, 20 mM.
Lac+ strains carrying rpoE2072::Tn10dCam were constructed and their growth on lactose quantified (Table S1). The rpoE allele was moved by P1 transduction into three Lac+ day 5 stress-induced mutants that had been characterized with respect to colony-forming ability on lactose medium (Rosenberg et al., 1998). To measure the days required for colony formation on lactose minimal medium, growth and plating of cells was identical to that described for stress-induced mutagenesis experiments, including the addition of lac-deleted scavenger cells in the same numbers. Additionally, the rpoE2072::Tn10dCam allele was transferred by P1 transduction into five Lac+ stress-induced mutants known to carry secondary mutations (Torkelson et al., 1997) and the ability to form colonies in exact reconstruction of experimental selection conditions measured. These experiments were carried out at 32°C due to the temperature-sensitive phenotype of one of the mutants.
Generation-dependent mutation rates to rifampicin resistance were measured in MG1655 cells harbouring either pBA166 or the vector pTrc99a. Single colonies grown on LBH-amp plates were inoculated into tubes containing 5 ml of LBH-amp broth with either 0.1% glucose or 0.1 mM IPTG to repress or induce expression of YYF from pBA166 respectively. Cultures were incubated overnight at 37°C prior to plating on LBH plates containing rifampicin. For each determination, 19 cultures were used and median mutant frequencies used to estimate mutation rates using a modified method of the median (Lea and Coulson, 1949; Von Borstel, 1978).
Stress-induced mutagenesis assays were performed as described at 30°C or 37°C unless otherwise indicated (Harris et al., 1996). SMR5236 cultures were concentrated 10-fold before plating to obtain sufficient numbers of Lac+ colonies. In some experiments amplification of the lac region was quantified as described (Hastings et al., 2004).
In reconstruction experiments for amplification, rpoE+ and rpoE2072::Tn10dCam F- cells were plated under the conditions of stress-induced mutagenesis experiments after conjugation with the same set of five F′ factors carrying lac amplification. The ratio of donor to recipient cells was 20:1. Transconjugants were selected on M9 glycerol tetracycline medium, which selects for TetR, carried in the chromosome of the recipient cells and for Pro+, bestowed by the F′ factor, and also on M9 lactose tetracycline medium, which selects for the recipient chromosome and the donor F′, as above, and also requires the maintenance of amplification of the lac locus. Thus the fraction of the colonies on glycerol medium that forms colonies on lactose gives a measure of the ability of cells of that genotype to support amplification.
Bacterial strains carrying the inducible DSB system consisting of a chromosomally encoded I-SceI enzyme and an F′ cutsite were maintained on medium containing 0.1% glucose to repress I-SceI synthesis (Ponder et al., 2005) and experiments performed as described (Ponder et al., 2005).
Quantitative conjugation and P1 transduction assays were performed as described (Lombardo et al., 2004). P1 stocks grown on SMR6263 were used in transduction experiments. Transductants were selected on LBH tetracycline citrate medium.
Co-transduction experiments (Fig. S1) were conducted with P1 grown on SMR11044 which has a kanamycin marker (yfhH::FRTKan) linked to rpoE2072::Tn10dCam. Following transduction, cells were plated on either LBH-kanamycin plates or LBH-chloramphenicol plates. Resistant colonies were then screened for the second antibiotic resistance. Co-transductant frequency was calculated as the number of screened antibiotic-resistant colonies per the number of selected antibiotic-resistant colonies. If the insertion in rpoE confers a null phenotype, the co-transductant frequency of chloramphenicol-resistant per kanamycin-resistant colonies would be expected to be close to zero.
UV sensitivity was determined in saturated overnight LBH cultures. Varying dilutions were spread on LBH plates, irradiated, and the plates immediately placed in the dark and incubated overnight at 37°C.
For monitoring expression of a σE-regulated gene, the rpoE2072::Tn10dCam allele was moved into strain CAG45114, containing an rpoHP3::lacZ fusion. For monitoring expression of an RpoS-controlled gene, rpoE2072::Tn10dCam was moved into SL590, containing a katE::lacZ fusion. The presence of the transposon insertion was verified by PCR amplification using primers flanking the insertion, RpoE-F (5′-CACTGGAAGGTGGACGACG) or RpoE-R (5′-GAGAAGTTACTGGCTGGTGG), in conjunction with an outward-reading primer (5′-GGTGGTGCGTAACGGCAAAAG) specific for Tn10dCam.
For β-galactosidase assays saturated LBH or M9-glucose cultures were diluted back 1:100 in fresh medium. For induction of peptide synthesis from pBA166, 1.0 mM IPTG was added to LBH cultures at OD600, of 0.1 and 0.5 ml aliquots were removed at various time intervals for β-galactosidase assays. Assays were performed as described (Miller, 1992). β-Galactosidase activity is expressed as a function of culture volume. For monitoring β-galactosidase activity of the katE::lacZ fusion throughout the growth phase, LBH cultures were inoculated from saturated overnight cultures. OD600 and β-galactosidase measurements were determined for each sample.
Western blots were performed with polyclonal antiserum against purified DinB as described (Galhardo et al., 2009). TraI was examined by immunoblot analysis using a 1:5 000 dilution of polyclonal antiserum against TraI as described (Will and Frost, 2006).
We are indebted to Virgil Rhodius and Carol Gross for sharing unpublished data, strains and plasmids, and to Masami Yamada and Takehiko Nohmi for the gift of DinB antiserum. We thank Anthony Poteet for strains and John Doan for assistance in medium preparation. R.S.G. was supported in part by a postdoctoral fellowship from the Pew Latin American Fellows Program. This work was supported by Canadian Institutes of Health Research Grant MT11249 (L.S.F.), Human Frontier Science Program Young Investigator Grant HFSP2006-0060 (C.H.), and US National Institutes of Health Grants F32-GM19909 (M.-J.L.), R01-GM64022 (P.J.H.) and R01-GM53158 (S.M.R.).
Additional supporting information may be found in the online version of this article.
Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.