Cyclic AMP Regulates Bacterial Persistence through Repression of the Oxidative Stress Response and SOS-Dependent DNA Repair in Uropathogenic Escherichia coli

OBSERVATION

Urinary tract infections are a worldwide health concern. They are caused mainly by uropathogenic Escherichia coli (UPEC), which in most cases leads to chronic infection (1). It has been proposed that a persister subpopulation is responsible for generating relapsing infections (2). A full understanding of the molecular mechanisms and genetic regulation involved in the generation of persister cells is currently lacking.

The cyclic AMP (cAMP) receptor protein (CRP) was originally described as a global regulator of genes involved primarily in carbon catabolite repression (3). The DNA binding affinity of the CRP homodimer is increased upon cAMP binding. cAMP-CRP activates some genes while repressing others and is predicted to regulate 378 distinct promoters in E. coli K-12 (4). These include, as mentioned, genes involved in carbon catabolism but also genes involved in additional processes such as virulence, biofilm formation, and the SOS response (3, 5–7). cAMP-CRP has previously been implicated in the negative regulation of persistence in UPEC, wherein transposon insertions that disrupted cyaA (which encodes the cAMP synthase adenylate cyclase) and the promoter for crp resulted in increased survival upon exposure to the β-lactam antibiotic ampicillin (8). Consistent with this, another study found that reduction of the cAMP level via its hydrolysis by the cAMP-specific phosphodiesterase CpdA resulted in increased persistence (9). However, a more comprehensive understanding of the gene networks regulated by cAMP-CRP that contribute to persistence is lacking.

There are at least two classes of bacterial persisters, i.e., type I persisters, which are characterized by a dormant state generated, for example, during stationary phase and preexist at the time of an antibiotic challenge (10), and type II persisters, which are generated during exponential growth upon an antibiotic challenge and are believed to result from a combination of a dormant metabolic state and antibiotic challenge-specific mechanisms (10–12). In the present study, we explored persister formation in exponentially growing UPEC cultures exposed to different antibiotics, as well as the role of cAMP in this process.

Increased survival was observed in ΔcyaA mutant cultures exposed to the cell wall-acting antibiotics meropenem, cefoxitin, and oxacillin (Fig. 1a). We observed no effect after gentamicin exposure and the opposite effect, i.e., decreased survival, in ciprofloxacin-treated cultures. This suggests that the mechanisms affected by cAMP that are involved in the generation of persister cells are specific for different antibiotics and likely antibiotic target dependent, as shown previously (11, 12).

• Open in new tab
• Download powerpoint

FIG 1 

cAMP is an antibiotic-specific regulator of bacterial persistence, and this phenomenon depends on the oxidative stress response. (a) The ability to generate persister cells in UPEC cultures exposed to different antibiotics was assessed by CFU counting (n = 4). (b) Killing kinetics of E. coli cultures exposed to ampicillin and periodically assayed for viable cells (n = 4). (c) Bacterial survival upon exposure to ampicillin in the presence of exogenous cAMP (n = 4). (d) Functional classification of Tn-Seq data by GO analysis. A transposon library constructed in the ΔcyaA mutant background was exposed to ampicillin (n = 4), and the genes under negative selection were classified by biological process. (e) Assays of competition between the ΔcyaA single mutant and double mutants defective in cyaA and another gene detected by Tn-Seq screening. Overnight cultures of the single and double mutants were mixed in a 10:1 ratio, grown to exponential phase, and treated with ampicillin for 6 h. Aliquots were washed, and viable cells were plated and counted in LB medium with or without kanamycin (n = 4). Error bars denote standard errors. *, P = 0.05; **, P = 0.01; ***, P = 0.001; ****, P = 0.0001; ns, not significant.

To further characterize the role that cAMP has in the generation of persister cells, we first compared the killing dynamics of the ΔcyaA mutant with those of the wild type (WT) in cultures exposed to ampicillin. After exposure to a lethal concentration of the antibiotic, we observed a classic biphasic killing curve, which is consistent with the generation of persister cells (13), and increased survival of the ΔcyaA mutant (Fig. 1b). Addition of exogenous cAMP restored WT killing kinetics to the ΔcyaA mutant (Fig. 1c), showing that the effect of the mutation is due to loss of cAMP production. The growth rate of the ΔcyaA mutant was lower than that of the WT strain (see Table S1 in the supplemental material). However, this by itself cannot explain the large increase in persister formation, since ampicillin kills growing bacteria, while metabolically dormant persister cells are spared. To further investigate the mechanism of cAMP inhibition on persister formation, we deleted crp, which encodes the cAMP receptor protein. Similar to what was observed with the ΔcyaA mutant, exponentially growing cultures of the Δcrp mutant exhibited 10-fold higher survival upon exposure to ampicillin than the WT, as described before (8). However, addition of exogenous cAMP did not restore WT killing kinetics to the Δcrp mutant (Fig. 1c), as would be expected if cAMP were acting through its receptor, CRP. These data are consistent with the hypothesis that cAMP-CRP mediates a major inhibitory effect on the formation of persisters in UPEC.

To investigate how the ΔcyaA mutant generates a higher level of persister cells, we used transposon sequencing (Tn-Seq) of ampicillin-treated ΔcyaA mutant and WT cultures to identify genes involved in the phenotype. We identified 346 mutants in the WT background and 234 mutants in the ΔcyaA mutant background that showed decreased survival upon exposure to ampicillin (Tables S2 and S3, respectively). Strikingly, only four genes were common to both data sets, i.e., c1489, c2003, c2602, and c3075, which encode a hypothetical protein, fumarase C, hypothetical protein YegO, and the flavohemoprotein HmpA, respectively. That the vast majority of hits were unique to each data set is consistent with the hypothesis that cAMP plays a major regulatory role in persister formation.

There is increasing evidence that part of the damage generated by bactericidal antibiotics is mediated by reactive oxygen species (ROS), particularly hydroxyl radicals (OH˙) (14–18). We used gene ontology (GO) analysis to categorize the 234 genes identified in the ΔcyaA mutant background and determined that several pathways known to contribute to the oxidative stress response appear to play a role in persister formation in the ΔcyaA mutant (Fig. 1d). This included pathways for indole and glutamate metabolism (19, 20). Indeed, these pathways have previously been linked with tolerance to antibiotics (9, 21, 22). In addition, we hit genes related to the biosynthesis of biogenic amines, which have been shown to protect against oxidative damage (23, 24). We also identified genes involved in the regulation of translation, protein maturation, and protein oligomerization. This may be in response to the previously reported decrease in the ATP level after exposure to oxidative stress, which in turn causes inactivation of the DnaK chaperone and a subsequent increase in protein aggregation (25).

None of the genes associated with oxidative stress that were identified in the ΔcyaA mutant screening were identified in the WT strain screening (Tables S2 and S3). Instead, pathways related mainly to cell envelope biogenesis were found to be important (Fig. S1). These results indicate that different mechanisms are involved in the generation of persister cells in the WT and the ΔcyaA mutant under the conditions tested. Altogether, these results suggest that the increased generation of persister cells in the ΔcyaA mutant depends on different genes involved in oxidative stress response and therefore that cAMP-CRP plays a negative regulatory role in persister cell formation in the WT strain.

To explore the potential for a direct role of cAMP-CRP in the transcriptional regulation of genes identified in our screening, we used the PRODORIC tool (26) and determined that 84 (36%) of the 234 genes identified in the ΔcyaA mutant background have CRP binding boxes in their promoter regions (Table S4). This is a much higher percentage than expected by chance (7.5%) (4). Therefore, the regulation of persister cell formation exerted by cAMP-CRP appears to be mediated both by direct transcriptional regulation of genes and by indirect effects, consistent with previous reports (27, 28).

To validate the hits from the Tn-Seq screening, we conducted competition assays between the ΔcyaA mutant parent strain and double mutants corresponding to ΔcyaA and individual genes hit in our screening. We chose to validate katE, acnA, barA, gsp, and fumC, each of which is related to the oxidative stress response (29–31). Each double mutant exhibited a competitive defect when challenged with ampicillin (Fig. 1e). In contrast, single mutations in katE, acnA, barA, gsp, and fumC in the WT background did not impact the generation of persister cells (Fig. S2), thus supporting the cAMP dependence of these hits. Indeed, three of the five genes (acnA, fumC, and gsp) have a CRP binding box in their promoters (Table S4). The mild competitive defects exhibited by the double mutants are likely due to partial redundancy of the different mechanisms involved in the generation of persister cells in cultures exposed to ampicillin, consistent with a previous report (13).

To further evaluate the contribution of oxidative damage to the generation of persister cells, we challenged WT and ΔcyaA mutant cultures with ampicillin in the presence or absence of oxygen. There was ~100-fold greater ampicillin survival of both strains in the absence of oxygen than under aerobic conditions; however, the ΔcyaA mutant still exhibited approximately 100-fold greater survival than the WT strain (Fig. 2a). This result demonstrates the critical contribution of oxidative damage to the toxicity of ampicillin and the impact of ROS on the generation of persister cells.

• Open in new tab
• Download powerpoint

FIG 2 

Decreased oxidative stress levels contribute to the increased survival in the ΔcyaA background by avoiding the generation of mutations in an SOS-dependent fashion. (a) The role of oxygen in the generation of persister cells was assessed by treating cultures under aerobic or anaerobic conditions. Cultures were grown to the exponential phase and challenged with ampicillin for 6 h aerobically or anaerobically (n = 4). Aliquots were washed and plated to assay viable counts. (b) Total ROS generation in ampicillin-treated cultures was assayed with the probe H₂DCFDA and normalized by CFU counts after 1 h of treatment with ampicillin or with FeSO₄ plus H₂O₂ (50 µM and 10 mM, respectively) as a positive control for OH˙ by the Fenton reaction (n = 3). (c) Generation of OH˙ in persister cells was assayed with the specific probe HPF and normalized by the number of survivors after 1 h of treatment with ampicillin or with FeSO₄ plus H₂O₂ under aerobic conditions (n = 3) and normalized by CFU counts. (d) The contribution of oxygen to the generation of OH˙ in persister cells generated by the exposure of cultures to ampicillin under microaerophilic conditions was assayed with the specific probe HPF and normalized by the number of survivors after 1 h of treatment with ampicillin or with FeSO₄ plus H₂O₂ in the presence of Oxyrase. (e) Frequency of mutagenesis was assayed in cells previously exposed to ampicillin for 1 h. Cultures were washed, grown overnight, and plated on LB agar plates supplemented with rifampin. (d) The contribution of the SOS response to the survival upon exposure to ampicillin was assessed by CFU counting with the WT and the ΔcyaA mutant transformed with a plasmid that generates an SOS⁻ phenotype because of a mutation in LexA (plexA) and compared to the empty vector (EV) (n = 4). Error bars denote standard errors. *, P = 0.05; **, P = 0.01; ***, P = 0.001; ****, P = 0.0001; ns, not significant.

We used the fluorescent probes 2',7'-dichlorodihydrofluorescein diacetate (H₂DCFDA) and hydroxyphenyl fluorescein (HPF) to evaluate the role of cAMP in the generation of total ROS and OH˙, respectively, in cultures exposed to ampicillin. The total ROS and OH˙ levels were much higher in ampicillin-treated WT and ΔcyaA mutant cultures than in the respective untreated controls (Fig. 2b and c, respectively). However, there was ~10-fold less OH˙ generation in the ΔcyaA mutant than in the WT (Fig. 2c). Interestingly, the ΔcyaA deletion also results in a large decrease in the level of OH˙ generated through the Fenton reaction (32) in the presence of FeSO₄ and H₂O₂ (Fig. 2c). These results suggest that oxidative stress responses are sufficiently induced in the ΔcyaA mutant to detoxify much of the ROS that is generated. This idea is supported by our Tn-Seq data showing that genes required to detoxify H₂O₂ are needed to cope with the oxidative damage generated by ampicillin exposure (Fig. 1d; Table S2). In addition, we observed a 5-fold increase in the survival of WT cultures exposed to ampicillin in the presence of dimethyl sulfoxide (DMSO), which is a known scavenger of OH˙ (33, 34). However, no difference was observed in that of the ΔcyaA mutant, suggesting that less OH˙ is produced in this strain during ampicillin exposure than in the WT (Fig. S3). No effect of DMSO itself on survival and/or the growth rate was observed.

We further evaluated the contribution of oxygen to the generation of OH˙ mediated by ampicillin by using Oxyrase, which scavenges oxygen to generate a microaerobic environment. We observed an ~10-fold higher level of OH˙ in the WT strain exposed to the antibiotic than in the ΔcyaA mutant (Fig. 2d). Similarly, we observed greater survival of Oxyrase-exposed cultures of both strains (Fig. S4) than of those treated with ampicillin under fully aerobic conditions, consistent with a contribution to cellular damage by ROS during ampicillin exposure. However, the ΔcyaA mutant still exhibited greater survival than the WT strain under microaerobic conditions (Fig. S4). These results further support our findings indicating a direct correlation between the lack of cAMP and an increased ability to cope with oxidative damage. Together, these results are in agreement with the induction of the oxidative stress response that is generated by the derepression of rpoS generated in cultures lacking cyaA crp, as previously reported (35).

Since the main target of damage by OH˙ is DNA (32, 34, 36), we evaluated changes in the frequency of generation of rifampin-resistant mutants as an indicator of DNA damage. A higher mutagenesis rate was observed in the WT strain exposed to ampicillin than in the untreated control, as demonstrated previously (37–39). However, ΔcyaA mutant cultures challenged with ampicillin showed no detectable difference from the untreated control in the frequency of mutation (Fig. 2d), which is consistent with the reduced level of OH˙ in this strain (Fig. 2c).

Finally, it has been previously established that DNA damage generated by oxidative stress induces the SOS response (40, 41) and that this pathway is regulated by cAMP (42). We evaluated the contribution of the SOS response to survival upon exposure to ampicillin by using a plasmid containing an uncleavable version of the repressor LexA because of a mutation (K156R) that generates an SOS-deficient (SOS⁻) phenotype (43). We observed 4-fold less survival of a ΔcyaA mutant expressing the SOS⁻ phenotype than of a strain with the same background but a functional SOS response. No changes were observed in the WT strain under the same conditions (Fig. 2f). This result indicates that survival of the ΔcyaA mutant upon exposure to ampicillin requires an active SOS response and also suggests that the lower frequency of mutagenesis observed in cultures exposed to ampicillin (Fig. 2e) is dependent on SOS-dependent DNA repair.

Altogether, these results show that in cultures exposed to ampicillin and most likely to other cell wall-acting antibiotics, cAMP is an important negative regulator of persistence in UPEC. Our results also highlight the multifactorial and redundant nature of molecular mechanisms involved in the generation and survival of persister cells. We show that the phenotype of increased persister formation when cAMP is absent is mediated by the establishment of an active response leading to (i) decreased accumulation of ROS and (ii) coping with oxidative damage to DNA generated by hydroxyl radicals, which is dependent on an active SOS response. Thus, UPEC must deal not just with the damage generated by the antibiotic to its canonic target but also with the oxidative damage that bactericidal antibiotics generate as part of their lethality (14). In addition, our results provide mechanistic support for and new insight into the model in which E. coli can form persister cells by decreasing the concentration of indole in a cAMP-dependent fashion (9). The resulting persister state would allow UPEC to better tolerate beta-lactam antibiotic-mediated damage, to deal with the toxic effect of antibiotic-generated OH˙, and to decrease the occurrence of damaging mutations. Finally, these results suggest that a decreased mutagenesis rate might slow down the evolution of antibiotic-resistant strains associated with the persister subpopulation, as recently shown (44).