The Global Regulatory Cyclic AMP Receptor Protein (CRP) Controls Multifactorial Fluoroquinolone Susceptibility in Salmonella enterica Serovar Typhimurium

ABSTRACT

Fluoroquinolone antibiotics are prescribed for the treatment of Salmonella enterica infections, but resistance to this family of antibiotics is growing. Here we report that loss of the global regulatory protein cyclic AMP (cAMP) receptor protein (CRP) or its allosteric effector, cAMP, reduces susceptibility to fluoroquinolones. A Δcrp mutation was synergistic with the primary fluoroquinolone resistance allele gyrA83, thus able to contribute to clinically relevant resistance. Decreased susceptibility to fluoroquinolones could be partly explained by decreased expression of the outer membrane porin genes ompA and ompF with a concomitant increase in the expression of the ciprofloxacin resistance efflux pump gene acrB in Δcrp cells. Expression of gyrAB, which encode the DNA supercoiling enzyme GyrAB, which is blocked by fluoroquinolones, and expression of topA, which encodes the dominant supercoiling-relaxing enzyme topoisomerase I, were unchanged in Δcrp cells. Yet Δcrp cells maintained a more relaxed state of DNA supercoiling, correlating with an observed increase in topoisomerase IV (parCE) expression. Surprisingly, the Δcrp mutation had the unanticipated effect of enhancing fitness in the presence of fluoroquinolone antibiotics, which can be explained by the observation that exposure of Δcrp cells to ciprofloxacin had the counterintuitive effect of restoring wild-type levels of DNA supercoiling. Consistent with this, Δcrp cells did not become elongated or induce the SOS response when challenged with ciprofloxacin. These findings implicate the combined action of multiple drug resistance mechanisms in Δcrp cells: reduced permeability and elevated efflux of fluoroquinolones coupled with a relaxed DNA supercoiling state that buffers cells against GyrAB inhibition by fluoroquinolones.

INTRODUCTION

Salmonella enterica is emerging as one of the most formidable pathogens of our time. The human-restricted typhoidal serovars cause invasive infections, estimated at 27 million illnesses annually, with a high rate of mortality in the absence of antimicrobial therapy (1). Nontyphoidal serovars cause many tens of millions of infections each year, the majority of which are noninvasive and clear without antimicrobial treatment (2). However, small children, the elderly, and persons with compromised immune systems are at greater risk for developing severe infections and require antimicrobial intervention (3). It has recently emerged that invasive nontyphoidal S. enterica serovars can be particularly virulent, causing an estimated 3.4 million infections and over 680,000 deaths in 2010 alone (4). Both nontyphoidal and typhoidal serovars are increasingly resistant to the first-line antibiotic ciprofloxacin (5, 6); for example, resistant typhoidal serovars increased from 0% to 8.6% in the United States between 2004 and 2014 (7) and from 0% to 17% in Canada between 2003 and 2013 (8). Therefore, understanding the genetic changes in Salmonella that decrease susceptibility to ciprofloxacin and other fluoroquinolones is of paramount importance.

Ciprofloxacin is a small, hydrophilic molecule that enters bacterial cells through outer membrane porins (2). Once inside, ciprofloxacin disrupts chromosome integrity by complexing with the type II topoisomerases, DNA gyrase (GyrAB) and topoisomerase IV (ParCE) (9–12). These complexes prevent progression of DNA polymerase, and the resultant stalled replication forks are prone to double-strand DNA breaks, which, in turn, inhibit cell division. GyrAB plays a crucial role in maintaining DNA supercoiling because it alone can introduce negative supercoils into DNA (13, 14). Thus, when ciprofloxacin blocks GyrAB function and causes DNA strand breaks, DNA supercoiling becomes relaxed. Ciprofloxacin also binds ParCE, a GyrAB homolog whose primary function is the decatenation of replicated DNA molecules and unknotting of supercoiled molecules (15, 16).

The primary mechanism of resistance to quinolones is alteration of the drug target through mutation of gyrA to reduce quinolone binding to the GyrAB gyrase (17). A single gyrA mutation is sufficient for clinical resistance to synthetic quinolones, like nalidixic acid, but secondary mutations are required for clinical resistance to fluoroquinolones (18). Secondary mutations can occur in gyrA and/or in gyrB or parC. As well, mutations in outer membrane porins can reduce ciprofloxacin permeability across the cell envelope to contribute to resistance (19), and active efflux by the TolC-AcrAB pump is a major contributor to high-level ciprofloxacin resistance in clinical isolates of Salmonella and Escherichia coli (20–23). Although gyrA is the primary determinant of resistance (6), a few clinical isolates of Salmonella with wild-type (wt) gyrA achieved clinical resistance by overproducing TolC-AcrAB (24, 25). In laboratory-evolved strains of Salmonella enterica serovar Enteritidis, mutations in the transcription factors SoxS, SoxR, MarA, and RamR caused reduced susceptibility to ciprofloxacin by upregulating drug efflux through TolC-AcrAB and downregulating drug entry through OmpF (26). Moreover, treatment failures can occur when Salmonella evolves reduced susceptibility to fluoroquinolone antibiotics during infection (18), and it is probably the high number of possible resistance mutations that accounts for such rapid evolution. Thus, fluoroquinolone resistance arises through combined activities of multiple pathways, including mutations in target proteins and transcriptional changes to cell envelope transporters, which underscores the need for broad understanding and identification of cellular mechanisms that contribute to fluoroquinolone susceptibility.

Studies of S. Typhimurium LT2 in the 1970s found that deletion of either of the genes crp and cya caused elevated resistance to several antibiotics, including the quinolone antibiotic nalidixic acid. These genes encode the global regulatory protein cyclic AMP (cAMP) receptor protein (CRP) and the enzyme adenylate cyclase (Cya), which produces CRP's allosteric effector cAMP, respectively. CRP sits atop the hierarchy of all transcription factors and so orchestrates diverse aspects of cell physiology in E. coli and S. enterica (27, 28). Because mutations in regulatory proteins are poorly characterized as determinants of antibiotic resistance, we sought to understand if and how deletion of crp might reduce susceptibility to clinically important fluoroquinolone antibiotics, such as ciprofloxacin. Our study uncovered a combination of transcriptional changes and alterations to steady-state DNA supercoiling that preadapt Δcrp mutant cells to tolerate ciprofloxacin and other fluoroquinolone antibiotics.

RESULTS

An S. Typhimurium Δcrp mutant is resistant to fluoroquinolones.

It was previously shown that deletion of crp or cya in S. Typhimurium LT2 confers resistance to multiple antibiotics, including aminoglycosides, fosfomycin, colistin, and polymyxin B (29). To test whether S. Typhimurium crp mutants are also resistant to the clinically relevant fluoroquinolones, we constructed a Δcrp mutant by deleting the entire crp open reading frame in the model laboratory strain S. Typhimurium SL1344 and then tested susceptibility to ciprofloxacin, ofloxacin, and levofloxacin. Deletion of crp doubled the MIC of each of the three fluoroquinolones for strain SL1344 (Fig. 1A to C). Curiously, the average culture density of fluoroquinolone-treated Δcrp cells was even higher than that of untreated cells, suggesting that antibiotic blocking of GyrAB and/or ParCE function may be beneficial in a Δcrp background, which we discuss below in the context of DNA supercoiling.

FIG 1.

Reduced sensitivity of Δcrp mutants to fluoroquinolones. (A to C) Culture density after 9 h of growth of SL1344:wt or SL1344:Δcrp in increasing concentrations of ciprofloxacin (A), ofloxacin (B), and levofloxacin (C). (D) Culture density after 9 h of growth of SL1344:gyrA(S83P) or SL1344:gyrA(S83P):Δcrp in increasing concentrations of ciprofloxacin. (E) Culture density after 9 h of growth of 14028:wt or 14028:Δcrp in increasing concentrations of ciprofloxacin. (F) Culture density of 14028:wt or 14028:Δcya after 9 h of culturing in the presence of 75 ng/ml of ofloxacin with (+) or without (−) 1 mM cAMP. Means and ranges from 2 to 4 biological replicates are plotted.

Salmonella strains can become less susceptible to fluoroquinolones during infection through additive evolution of multiple resistance mutation (16); this prompted us to test whether deletion of crp enhances the primary fluoroquinolone resistance mutation in gyrA (Ser83Phe) (30). Resistance was greatly elevated in the Δcrp gyrA(S83F) double mutant, especially at 600 ng/μl, at which the double mutant grew maximally yet the gyrA(S83F) mutant was almost completely inhibited (Fig. 1D). The MIC of ciprofloxacin for the gyrA(S83F) mutant was <800 ng/ml, whereas that for the Δcrp gyrA(S83F) double mutant was >1.0 μg/ml, surpassing the breakpoint for clinical resistance as defined by the Clinical and Laboratory Standards Institute (CLSI).

To confirm that deletion of crp was responsible for the observed resistance phenotypes, crp and cya deletions in the model strain S. Typhimurium 14028 were acquired from the Biodefense and Emerging Infections Research Resources Repository (BEI Resources), described in reference 31. The 14028:Δcrp mutant demonstrated a resistance profile similar to that of the SL1344:Δcrp mutant (Fig. 1E). The inability of the Δcya mutant to synthesize CRP's allosteric effector, cAMP, can be restored by supplementation of growth medium with cAMP, providing an excellent mechanism to restore CRP function to Δcya mutant cells. Figure 1F demonstrates that fluoroquinolone susceptibility can be restored to the 14028:Δcya mutant by addition of cAMP to culture medium, confirming that the loss of CRP function is responsible for the growth phenotypes described above.

Permeation by small-molecule antibiotics.

Subinhibitory concentrations of antibiotics are important for genetic studies of resistance because both wild-type and mutant cells continue to grow and modify gene expression and physiology in response to antibiotic challenge. We selected the subinhibitory concentrations of 15 and 25 ng/ml of ciprofloxacin because all strains could grow, and the SL1344:Δcrp mutant demonstrated the resistance phenotype (Fig. 2A).

FIG 2.

Sublethal concentrations of antibiotic differentially affect growth of SL1344:wt and SL1344:Δcrp cells. (A) Ciprofloxacin; (B) kanamycin; (C) tetracycline. Cell densities were measured as OD₆₀₀ at 6 h of growth (full growth curves in Fig. S1), with exposure to antibiotic at the indicated concentrations. Error bars represent 95% confidence intervals.

Bacteria can survive antibiotic exposure by preventing antibiotics from penetrating the cell envelope or by expelling antibiotics from inside the cell using efflux pumps. In order to test the hypothesis that the SL1344:Δcrp mutant is less permeable by small molecule antibiotics than SL1344:wt, we treated cells with kanamycin and tetracycline. Kanamycin and tetracycline were selected for these experiments because each belongs to a different class of antibiotic than ciprofloxacin; they both must cross the cell envelope to reach ribosome targets in the cytoplasm, and neither is expected to interfere with DNA supercoiling. The MICs of kanamycin and tetracycline were 32 μg/ml and 1 μg/ml, respectively (Table 1). SL1344:wt and SL1344:Δcrp cells were equally susceptible to subinhibitory concentrations of kanamycin and tetracycline (8 μg/ml and 16 μg/ml and 250 ng/ml and 500 ng/ml, respectively) (Fig. 2B and C). This suggests that the SL1344:Δcrp mutant is not less permeable by these small molecule antibiotics, consistent with the equal tetracycline susceptibility of the S. Typhimurium LT2 wild-type, Δcrp, and Δcya strains observed by Alper and Ames (29).

TABLE 1.

Antibiotic concentrations and MICs

Antibiotic	Concn tested (μg/ml)	MIC for SL1344 (μg/ml)	Sublethal concn used in this study (μg/ml)
Kanamycin	0–64	32	8	16
Tetracycline	0–4	1	0.25	0.05
Novobiocin	0–25	NDa	15	25
Ciprofloxacin	0–5	0.045	0.015	0.025

^aND, no data.

Because ciprofloxacin requires porins to enter the cell (32), decreased expression or removal of porins has been shown to result in resistance to ciprofloxacin (33–35). OmpC and OmpF have been linked directly to ciprofloxacin resistance (32), while OmpA is a large porin that allows slow diffusion of molecules at a rate 50 times lower than that of OmpF (36). We used reverse transcription-quantitative PCR (RT-qPCR) to test whether changes in expression of genes encoding these three major outer membrane porins could explain the ciprofloxacin resistance phenotype of SL1344:Δcrp. Expression was measured during exponential growth (2 h) and in stationary phase (12 h). SL1344:Δcrp cells had significantly decreased ompA expression (P < 0.001) at 12 h but not at 2 h (Fig. 3A). Expression of ompC did not differ between SL1344:wt and SL1344:Δcrp at 2 h or 12 h (Fig. 3B), whereas ompF expression was significantly decreased in SL1344:Δcrp at 2 h (P < 0.01) and 12 h (P < 0.001) (Fig. 3C). Next we tested whether porin gene expression changes in response to 25 ng/ml of ciprofloxacin. After 2 h of treatment in exponential growth, expression of ompA, ompC, and ompF was equally decreased in both SL1344:wt and SL1344:Δcrp compared to expression in untreated cells. After 12 h, gene expression was significantly lower in SL1344:Δcrp for ompA (P < 0.001), ompC (P < 0.01), and ompF (P < 0.001) (Fig. 3E, F, and G). The presence of ciprofloxacin appeared to reduce expression of ompA and ompF in SL1344:wt but had little effect on expression in SL1344:Δcrp at 2 h. These data suggest that CRP activates ompF expression in exponential growth and activates ompF and ompA expression during stationary phase but does not directly regulate ompC under these conditions. Thus, the decreased expression of the ompF and ompA porins may contribute to ciprofloxacin resistance in SL1344:Δcrp.

FIG 3.

Expression of outer membrane porins and an efflux pump in SL1344:wt and SL1344:Δcrp cells. Shown is the expression of ompA, ompC, ompF, and acrB in cells left untreated (A to D, respectively) and in cells treated with 25 ng/ml of ciprofloxacin (E to H, respectively). All points are expressed relative to untreated SL1344:wt at 2 h, represented by the dotted gray line. Significance was determined by the Holm-Sidak method. *, P < 0.01; **, P < 0.001. Error bars represent 95% confidence intervals.

The AcrAB-TolC efflux pump confers resistance to ciprofloxacin, and decreased expression and mutation of acrB are linked to increased ciprofloxacin susceptibility (20–23). RT-qPCR revealed that after 2 h of exponential growth, acrB expression was unchanged between SL1344:wt and SL1344:Δcrp. After 12 h, expression of acrB decreased in both SL1344:wt and SL1344:Δcrp, and its expression in SL1344:wt was significantly lower than in SL1344:Δcrp (P < 0.001) (Fig. 3D). The presence of ciprofloxacin had little effect on expression of acrB, and again expression in SL1344:Δcrp was significantly lower than in SL1344:wt (P < 0.01) (Fig. 3H). These data suggest that CRP activates expression of acrB in stationary phase and that upregulation of the AcrAB efflux pump may contribute to the ciprofloxacin resistance of SL1344:Δcrp.

Ciprofloxacin induces filamentous growth but not the SOS response in SL1344:wt.

Ciprofloxacin has been shown to induce filamentation in Gram-negative cells at concentrations near the MIC (37, 38). Therefore, we grew SL1344:wt and SL1344:Δcrp cells in the presence and absence of ciprofloxacin, examined them at a magnification of ×1,000, and quantified cell lengths. Untreated SL1344:wt and SL1344:Δcrp cells were very similar in size (Fig. 4A). After 12 h of treatment with 25 ng/ml of ciprofloxacin, SL1344:wt cells were elongated and filamentous, while SL1344:Δcrp cells were not (Fig. 4A). The treated SL1344:wt cells were significantly longer than untreated cells and treated SL1344:Δcrp cells (P < 0.001) (Fig. 4B). Because filamentation is an indicator of the SOS response (39), these data suggest that SL1344:wt cells were under high stress from ciprofloxacin, while SL1344:Δcrp cells were not.

FIG 4.

Ciprofloxacin does not affect cell growth or SOS induction in SL1344:Δcrp cells. (A) Images of SL1344:wt and SL1344:Δcrp cells left untreated or treated with 25 ng/ml of ciprofloxacin. (B) Plot of cell length measurements from microscopic images, including those shown in panel A. Cells were grown in LB with 0.5% NaCl at 37°C with shaking in a plate reader for 12 h and then observed under a microscope at a magnification of ×1,000. (C) Expression of sulA in SL1344:wt and SL1344:Δcrp cells. (D) Expression of sulA in SL1344:wt and SL1344:Δcrp cells treated with 25 ng/ml of ciprofloxacin. For panels B to D, significance was determined by the Holm-Sidak method. **, P < 0.001. Error bars represent 95% confidence intervals.

Considering that double-strand breaks in DNA caused by ciprofloxacin can initiate the SOS response (40, 41), we used RT-qPCR to detect whether the SOS response was activated by ciprofloxacin treatment. A LexA-repressed gene, sulA, is not constitutively expressed, allowing detection of SOS response activation. There was no significant difference in expression of sulA between SL1344:wt and SL1344:Δcrp when cells were grown exponentially either untreated or treated with 25 ng/ml of ciprofloxacin (Fig. 4C and D). At 12 h there was a small, albeit statistically insignificant, increase in sulA expression in SL1344:wt cells treated with ciprofloxacin. Conversely, sulA expression was slightly, yet significantly, lower in treated SL1344:Δcrp cells than in both treated and untreated SL1344:wt cells and untreated SL1344:Δcrp cells (P < 0.001) (Fig. 4D). The filamentation data suggest that the SL1344:wt cells were under stress from ciprofloxacin treatment, but the RT-qPCR results suggest that the SOS response was only weakly induced and that the SL1344:Δcrp cells were less likely to induce the SOS in response to ciprofloxacin challenge.

Topoisomerase gene expression can explain relaxed DNA supercoiling in SL1344:Δcrp.

Fluoroquinolone and aminocoumarin antibiotics inhibit GyrAB, which causes DNA supercoiling to relax. The topoisomerases GyrAB and ParCE are the targets for ciprofloxacin (42, 43); this prompted us to test whether the expression of topoisomerase genes differs between SL1344:wt and SL1344:Δcrp cells, as a difference in drug target levels could explain the reduced susceptibility of Δcrp mutants. We used RT-qPCR to test whether SL1344:Δcrp cells have altered expression of the gene encoding the primary fluoroquinolone target, gyrA, and its oligomerization partner, gyrB. As expected, expression of both genes decreased in stationary phase (12 h) (42), yet there was no detectable difference in gyrA or gyrB expression between SL1344:wt and SL1344:Δcrp cells (Fig. 5A and B). Exposure to 25 ng/ml of ciprofloxacin resulted in sustained gyrA and gyrB expression into stationary phase (Fig. 5F and G), which may be explained by activation of the gyrA and gyrB promoters by the relaxation of DNA supercoiling caused by inhibition of GyrAB function (44, 45).

FIG 5.

Expression of topoisomerase genes in SL1344:wt and SL1344:Δcrp. Shown is expression of gyrA, gyrB, topA, parC, and parE in cells left untreated (A to E, respectively) and in cells treated with 25 ng/ml of ciprofloxacin (F to J, respectively). All points are expressed relative to untreated SL1344:wt at 2 h, represented by the dotted gray line. Significance was determined by the Holm-Sidak method. *, P < 0.01; **, P < 0.001. Error bars represent 95% confidence intervals.

Both topoisomerase I (topA) and topoisomerase IV (parC and parE) relax DNA supercoils in opposition to GyrAB activity. RT-qPCR revealed that there was no significant difference in expression of topA between untreated SL1344:wt and SL1344:Δcrp cells, regardless of growth phase (2 h or 12 h) (Fig. 5C). Exposure to 25 ng/ml of ciprofloxacin caused a nearly 8-fold decrease of topA expression in both cell types at 2 h in exponential growth, consistent with repression of the topA promoter by relaxed DNA supercoiling (46). By 12 h, the decrease in expression caused by ciprofloxacin was less severe, but the Δcrp mutant had significantly lower topA expression after 12 h of treatment than did SL1344 (Fig. 5H). Topoisomerase IV (ParCE) is not recognized as a major contributor to DNA supercoiling in Salmonella or E coli, yet ParCE function can compensate for loss of topA (47, 48) and ParCE homology with GyrAB means that it can be bound by fluoroquinolone antibiotics (25, 49). Although expression of both parC and parE fell significantly as wild-type cells entered stationary phase, expression remained elevated in SL1344:Δcrp at 12 h (P < 0.001) (Fig. 5D and E). When cells were treated with 25 ng/ml of ciprofloxacin, SL1344:wt remained largely unaffected, while SL1344:Δcrp had reduced yet significantly higher parC (P < 0.001) and parE (P < 0.01) (Fig. 5I and J) expression. These data indicate that CRP represses expression of parC and parE and suggest that the sustained expression of parCE in SL1344:Δcrp could alter the supercoiling state of DNA in Δcrp mutant cells.

DNA supercoiling is relaxed in SL1344:Δcrp.

We hypothesized that SL1344:Δcrp cells maintain a more relaxed state of DNA supercoiling than that of wild-type cells. This predicts that ciprofloxacin (fluoroquinlone) and novobiocin (aminocoumarin) have reduced antimicrobial effects because Δcrp mutant cells operate with a relaxed state of DNA supercoiling in the absence of antibiotic. DNA supercoiling activity in cells can be inferred by electrophoretic separation of plasmid DNA in an agarose gel containing chloroquine (42, 43). Chloroquine intercalates between DNA strands and causes base pairs to separate, which reduces torsional stress. The resulting relaxation of negative supercoils allows plasmids to be separated electrophoretically based on the degree to which they were supercoiled in cells. Highly supercoiled DNA is more compact and so has faster electrophoretic migration than relaxed DNA. To enhance separation of plasmid topoisomers, two-dimensional chloroquine gels employ different chloroquine concentrations in each electrophoretic dimension. This analysis confirmed the hypothesis that DNA supercoiling is more relaxed in SL1344:Δcrp (Fig. 6A). Although SL1344:Δcrp had a wide and relatively uniform distribution of topoisomers, a greater proportion of the topoisomers were less supercoiled than those in SL1344:wt.

FIG 6.

DNA supercoiling in SL1344:wt and SL1344:Δcrp cells. (A) Cells containing the pUC19 plasmid were grown in LB with 0.5% NaCl at 37°C with shaking in a plate reader for 12 h, and then plasmids were extracted and separated by chloroquine gel electrophoresis. The diagram illustrates how supercoiled plasmids migrate in two-dimensional chloroquine gel electrophoresis. An uncharacterized band consistently observed in SL1344 plasmid extracts is labeled with an asterisk. (B) Diagram of interquartile analysis for one-dimensional chloroquine gels. The box represents the range of topoisomers from the 25th to the 75th percentile; the center line represents the 50th percentile. (C) Interquartile analysis of untreated (white boxes) compared to novobiocin-treated (25 μg/ml) (light gray boxes) and ciprofloxacin-treated (25 ng/ml) (dark gray boxes) SL1344:wt and SL1344:Δcrp cells.

To determine how antibiotic treatment affects DNA supercoiling in SL1344:wt and SL1344:Δcrp, one-dimensional chloroquine gel analysis of biological replicates was averaged by quartile analysis (Fig. 6B). Quartile analysis provides a statistical assessment of DNA supercoiling in the population of plasmids extracted from a cell culture, and normalization to the range of electrophoretic separation allows for the averaging of biological replicates (42). As explained in the figure legend, the box in Fig. 6B represents the range of electrophoretic separation of topoisomers from the 25th to 75th percentile, while the center line indicates the median distance migrated; thus, a long box indicates a wide range of topoisomers. Novobiocin caused a characteristic relaxation of DNA supercoiling in SL1344:wt (42), but surprisingly, ciprofloxacin caused a slight increase in DNA supercoiling (Fig. 6C). In SL1344:Δcrp cells, DNA supercoiling increased dramatically when cells were treated with either novobiocin or ciprofloxacin.

Disconnect between ATP concentration and degree of DNA supercoiling.

Both GyrAB and ParCE use ATP hydrolysis to reseal DNA after causing double-strand breaks (50), and the previously observed correlation between high ATP/ADP ratios and highly supercoiled DNA is attributed to ATP being the primary driver of GyrAB activity (51, 52). Therefore, we tested whether supercoiling relaxation in SL1344:Δcrp cells may have been driven by a decreased [ATP]/[ADP] ratio that reduced GyrAB activity.

Surprisingly, the [ATP]/[ADP] ratio was slightly higher in SL1344:Δcrp cells than in SL1344:wt cells during exponential growth, and ciprofloxacin did not affect these ratios (Fig. 7). When growth ceased in stationary phase (12 h), the [ATP]/[ADP] ratio was greatly decreased, as expected (53–55). In stationary phase (12 h), the [ATP]/[ADP] ratio was significantly higher in SL1344:Δcrp than in SL1344:wt. After 12 h of exposure to ciprofloxacin, there was a significant decrease in the [ATP]/[ADP] ratio in both SL1344:wt and SL1344:Δcrp. These results indicate that SL1344:Δcrp cells have more ATP than SL1344:wt throughout growth, the opposite of what we hypothesized based on the observation of higher DNA supercoiling levels in SL1344:wt.

FIG 7.

[ATP]/[ADP] ratio in SL1344:wt and SL1344:Δcrp. Cells were grown in LB with 0.5% NaCl at 37°C with shaking in a plate reader for the indicated times. The [ATP]/[ADP] ratio was measured using a luminescent assay described in Materials and Methods. Open shapes indicate untreated cells; solid shapes indicate cells treated with 25 ng/ml of ciprofloxacin. SL1344:wt cells are shown as blue circles, and SL1344:Δcrp cells are shown as red squares. Error bars represent 95% confidence intervals. Significant results are indicated from a two-way analysis of variance (ANOVA) with Tukey's multiple-comparison test for 2-h samples or 12-h samples. *, P < 0.01; **, P < 0.001.

DISCUSSION

Finding genetic changes that reduce susceptibility to clinically relevant antibiotics is important for developing clinical diagnostics that can detect infections that may resist treatment. Moreover, finding mutations assists in understanding the rates, the evolution, and the mechanisms of resistance. Because fluoroquinolone resistance evolves through the additive accumulation of resistance mutations, MIC breakpoints have been revised down to improve detection and identification of clinical isolates demonstrating intermediate levels of resistance (56). The current breakpoints for ciprofloxacin resistance are ≥1.0 μg/ml (Clinical and Laboratory Standards Institute [CLSI]) and >0.06 μg/ml (European Committee on Antimicrobial Susceptibility Testing [EUCAST]). Thus, the Δcrp and Δcya mutations alone are insufficient to confer resistance to ciprofloxacin but can work synergistically with mutations in gyrA to create resistant cells.

Gene regulatory changes have been implicated in ciprofloxacin resistance, specifically when mutation of transcription factors results in upregulation of drug efflux and a reduction in drug entry (26). Here we report that deleting a global regulatory protein, CRP, confers resistance to ciprofloxacin and other fluoroquinolones by altering expression of genes for drug efflux and drug entry, as well as causing significant changes to the primary drug target, DNA supercoiling. Figure 8 presents a model that collates the experimental results presented above, combining drug influx (OmpF and OmpA porins) and efflux (AcrB); fluoroquinolones appear to be capable of diffusing across membranes (36), but this is not included in the schematic. Figure 8 also illustrates the ParCE topoisomerase modulating DNA supercoiling. Despite the dramatic difference between DNA supercoiling in SL1344:wt and SL1344:Δcrp, we did not detect altered expression of gyrAB or topA (Fig. 5), suggesting that the observed relaxed DNA phenotype in the SL1344:Δcrp mutant is caused by increased parCE (topoisomerase IV) expression (Fig. 5). We posit that the relaxed supercoiling of DNA in SL1344:Δcrp serves as a physiological preadaptation that protects cells from the normally deleterious relaxation of DNA supercoiling caused by ciprofloxacin. In fact, SL1344:Δcrp can benefit from subinhibitory concentrations of ciprofloxacin, ofloxacin, levofloxacin, and novobiocin, as these antibiotics caused an increase in DNA supercoiling and improved growth (Fig. 1, 2, and 6; see also Fig. S1 in the supplemental material).

FIG 8.

Model for increased antibiotic resistance in SL1344:Δcrp mutants. An SL1344:wt cell is shown in blue, and an SL1344:Δcrp cell is shown in red. Ciprofloxacin molecules are shown as yellow circles. The predicted decrease in OmpF and OmpA and concomitant increase in AcrB and topoisomerase IV (ParCE) in SL1344:Δcrp are illustrated. Chromosomal DNA is shown in black. CRP mutants of Salmonella and E. coli lack flagella (71).

It was previously reported that treating E. coli with 500 ng/ml of ciprofloxacin or 5 μg/ml of oxolinic acid had the counterintuitive effect of causing DNA supercoiling to increase, but this occurred only in cells with lower states of DNA supercoiling due to mutations in gyrA and gyrB (43, 57). It is well documented that the gyrA, gyrB, and topA promoters are under homeostatic regulation to maintain stable DNA supercoiling levels, such that more gyrase is produced when DNA supercoiling relaxes and more topoisomerase I is produced when DNA supercoiling increases. Thus, Franco and Drlica (57) proposed that weak inhibition of GyrAB by low-level antibiotic caused upregulation of the gyrase genes in response to relaxation of supercoiling, which, in turn, resulted in increased production of GyrAB and a measurable increase in DNA supercoiling in the presence of antibiotics. Our present study with Salmonella confirmed that subinhibitory concentrations of ciprofloxacin caused upregulation of gyrA and gyrB expression in stationary phase, a time point at which these genes are normally significantly downregulated (Fig. 5). This is consistent with the observed increase in DNA supercoiling in SL1344:wt and SL1344:Δcrp when both cell types were exposed to ciprofloxacin. Conversely, novobiocin caused DNA supercoiling to relax in SL1344:wt but to increase in SL1344:Δcrp. This suggests that at the concentrations tested, ciprofloxacin and novobiocin differ in degrees of gyrase inhibition such that novobiocin is insufficient to upregulate gyrA and gyrB, whereas ciprofloxacin causes sufficient relaxation to trigger gyrase production. We are testing whether each antibiotic has a tipping point where DNA supercoiling is sufficiently perturbed to activate homeostatic regulation.

The set point of DNA supercoiling is usually attributed to the relative activities of TopA and GyrAB. Yet increased expression of parCE in SL1344:Δcrp is consistent with ParCE shifting the DNA supercoiling to a more relaxed state. GyrAB and ParCE have analogous structures and require energy from ATP hydrolysis to perform their respective functions. Thus, while an elevated [ATP]/[ADP] ratio increases DNA supercoiling by driving GyrAB activity in wild-type cells, it is unknown how the [ATP]/[ADP] ratio affects ParCE activity. The DNA supercoiling set points observed in this study are most readily explained by the relative abundances of GyrAB, TopA, and ParCE and not differences in ATP availability between the mutant and wild type. The elevated [ATP]/[ADP] ratio observed in stationary-phase SL1344:Δcrp is likely due to normal growth in the mutant compared to the filamentous phenotype of SL1344:wt.

Ciprofloxacin requires porins to cross the outer membrane (58). OmpF has been experimentally determined to be the most important porin for ciprofloxacin entry into the cell (23, 34, 59, 60); thus, the significant decrease in ompF expression observed in the SL1344:Δcrp mutant is consistent with increased ciprofloxacin resistance (Fig. 3). The efflux pump AcrB is well established to contribute to ciprofloxacin resistance (61), and the increase in acrB expression in the SL1344:Δcrp mutant may enhance resistance (Fig. 3). However, the SL1344:Δcrp mutant did not demonstrate elevated resistance to tetracycline, which in E. coli is also imported by OmpF and exported by TolC-AcrAB (62, 63). Thus, changes in the expression of ompF and acrB may not be the primary determinants of ciprofloxacin resistance in the SL1344:Δcrp mutant, as observed when cells carry gyrA mutations (17).

As genetic tools such as PCR and DNA sequencing are increasingly used for detection of resistance genes and alleles in clinical diagnosis, it is important to find all genes in an organism that can contribute to antibiotic resistance. The collective findings of reduced antibiotic susceptibility in Salmonella presented here, resistance described in reference 29, and aminoglycoside resistance in E. coli (64) suggest that the crp and cya loci should be tested in genetic screens of fluoroquinolone-resistant Enterobacteriaceae. Currently, these nonclassical resistance loci may be overlooked in the clinical setting. The finding that active CRP increases antimicrobial susceptibility makes the CRP regulon a putative target for drug development. Moreover, DNA supercoiling has long been recognized as a regulator of virulence in Salmonella (65–68); thus, the discovery that CRP impacts DNA supercoiling homeostasis raises important questions about the interplay between CRP and DNA supercoiling in the control of virulence gene expression. A benefit of characterizing resistance loci that operate through gene regulation is the potential to discover multiple complementary cellular pathways to resistance; understanding the physiology of resistant cells can assist in the design of better drug combinations for more effective treatment of infections.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

All cells used in this study were derivatives of Salmonella enterica subsp. enterica serovar Typhimurium strains SL1344, 14028, and LT2. Mutations were generated by lambda red recombination (69), and all mutations were backcrossed into wild-type SL1344 or 14028 by transduction with P22 HT 105 bacteriophage. SL1344 mutants were made by the authors; 14028:Δcrp and 14028:Δcya mutants are described in reference 31, and the gene deletions were transduced to wild-type 14028 upon receipt by the authors. For DNA supercoiling analysis, mutant strains were transformed by electroporation with the pUC19 plasmid.

All experiments started with colonies isolated from −80°C stocks and grown overnight on LB plates (1% tryptone, 0.5% yeast extract, 1% NaCl, 1.5% agar) at 37°C. Cultures were grown in LB (0.5% NaCl) with shaking at 200 rpm at 37°C in either 16- by 150-mm glass tubes, 250-ml glass flasks, or 96-well plates. When required, antibiotics were used at the following concentrations: ampicillin, 100 μg/ml, and chloramphenicol, 35 μg/ml. All other antibiotics were used in range of concentrations listed in Table 1.

To measure growth curves, optical density at 600 nm (OD₆₀₀) was measured in a Synergy HT microplate reader (BioTek). Cells were grown overnight in 3 ml of LB, subcultured into 3 ml of fresh LB, and grown to an OD₆₀₀ of 0.2 to 0.3. Cells were equalized to an OD₆₀₀ of 0.005 in each well of a 96-well plate, and absorbance readings were taken every 10 min for 24 h with continuous shaking. Fifty microliters of light mineral oil was overlaid on each well to prevent desiccation. The MIC was determined using the Synergy HT microplate reader (BioTek). Cultures were inoculated into LB (0.5% NaCl) with antibiotic and measured in the plate reader for 24 h. After 24 h, cultures were serially diluted, plated onto LB agar, and grown overnight at 37°C to enumerate cells and determine if treatments were bactericidal.

For gene expression experiments, cells were pregrown in LB for 3 h at densities lower than an OD₆₀₀ of 0.2 to 0.3 to ensure steady-state exponential growth. Exponentially growing cells were then equalized to an OD₆₀₀ of 0.005 in 25 ml of broth in 250-ml glass flasks, which were sampled at 2 h (exponential growth) and 12 h (stationary phase).

Chloroquine gel analysis.

Plasmids were isolated from 50 ml of exponential-phase culture from flasks or 2 ml of stationary-phase culture in tubes using a plasmid miniprep kit (BioBasic); to confirm consistency between culture conditions, plasmids were also extracted by pooling 250-μl samples of stationary or exponential-phase cultures in 96-well plates. All electrophoresis was conducted in 20-cm-long 1% agarose gels with 2× Tris-borate-EDTA (TBE) as gel and running buffer. Approximately 1 to 2 μg of DNA (10 to 20 μl) was loaded into each well using 4 μl of 6× blue gel loading dye (New England BioLabs) diluted to 1× with 100% glycerol. Gels were run in one and two dimensions. Electrophoresis was run at 3 V/cm for 16 h for the first dimension; then if the gel was run in a second dimension, it was rotated 90° and run for 1.5 V/cm for 8 h. The concentration of chloroquine was 2.5 μg/ml for the first dimension and equilibrated to 25 μg/ml for the second. After electrophoresis, chloroquine was removed by gentle rocking in water for 2 h, with fresh water every 20 to 30 min. Gels were subsequently stained with 1 μg/ml of ethidium bromide for at least 1 h and then briefly washed with water and visualized under UV light.

Topoisomer distribution was determined by calculating the interquartile range of each electrophoretic separation in each gel lane, as described in reference 44. Total monomer plasmid DNA was quantified by densitometry, summed and divided into quartiles, and then plotted according to distance migrated during electrophoresis (Fig. 6B).

RT-qPCR.

Total RNA was isolated from cultures using the EZ-10 Spin Column Total RNA Mini-Preps Super kit (BioBasic). For each sample, 2 μg of total RNA was DNase treated in a 50-ml reaction using the Turbo DNA-free kit (Ambion), and cDNA templates were synthesized by random priming of 200 ng of RNA in a 20-ml reaction using the Verso cDNA synthesis kit (Thermo Scientific). qPCR oligonucleotide primers are listed in Table 2. PCRs were carried out in duplicate with each primer set on the StepOne system (Applied Biosystems) using PerfeCTa SYBR green FastMix, ROX (Quanta Biosciences). Standard curves were included in every qPCR run, and melt curves were performed at least once with each primer set. Standard curves were generated for each primer set using six serial 10-fold dilutions of chromosomal DNA.

TABLE 2.

Oligonucleotide primers used in this study

Name	Sequence (5′–3′)	Reference
Topoisomerases
gyrA.RT.F	TGATTGAAGTGAAACGCGATGCGG	42
gyrA.RT.R	GTGATGCAGCGCCACCATGTTAAT	42
gyrB.RT.F	ATATGAGATCCTGGCGAAACGCCT	42
gyrB.RT.R	GATCTTCTTTGCCATCGCGCTTGT	42
Se.topA.RT.F	ATGAAGTGCTGCCCGGTAAAGAGA	This work
Se.topA.RT.R	TTGCGAGATAGATGTGGTCGGCTT	This work
Se.parC.RT.F	TGTACGTGATCATGGATCGTGCGT	This work
Se.parC.RT.R	TAAATTTAGCGGTGGCGTTCAGCC	This work
Se.parE.RT.F	TATCTGGGATCGCTGCGCTTATGT	This work
Se.parE.RT.R	CATCTTTCACCACGCCGGAAACAA	This work
Outer membrane porins
Se.ompA.RT.F	GTTAACCCGTATGTTGGCTTTG	This work
Se.ompA.RT.R	CTGAACGCCCTGAGCTTTAT	This work
Se.ompC.RT.F	CGTAACTACGGCGTAACCTATG	This work
Se.ompC.RT.R	TACCACGCTGCTGCATAAA	This work
Se.ompF.RT.F	GGTCAGTGGGAATACCGTACTA	This work
Se.ompF.RT.R	GAACCCACTTCCGCGTATTT	This work
Other
Se.acrB.RT.F	ACCTGGAAGTAAACGTCGTTAG	This work
Se.acrB.RT.R	GTCTATCCCGTTCTCCGTAATG	This work
Se.sulA.RT.F	GGTCTTCGTGGTAGACAACTTC	This work
Se.sulA.RT.R	TGCAAATCGTTCTTCGTCATTTC	This work
Se.gyrA.Ser83Phe	CGTAATCGGTAAATACCATCCCCACGGCGATTTCGCAGTGTATGACACCATCGTTCGTATGGCGCAGCCA	This work

[ATP]/[ADP] ratio assay.

Cells were lysed using a FastPrep-24 homogenizer with Lysing Matrix B tubes (MP Biomedicals). The extract was centrifuged (9,000 × g for 10 min) at 4°C. ATP and ADP measurements were determined using the bioluminescent ADP/ATP ratio assay kit (Abcam) according to the manufacturer's instructions in white-walled U-bottom Nunc 96-well polypropylene microwell plates (Thermo Scientific) using the Synergy HT microplate reader (BioTek). Cells were plated on LB before and after lysis to determine CFU per milliliter for normalization of data and percentage of cells lysed.

Quantification of filamentous phenotype.

Cells were sampled after 12 h of growth. Ten microliters of culture was transferred to a plastic slide and covered with a plastic coverslip. Cells were visualized with oil immersion at a total magnification of ×1,000 and captured with Infinity Capture (Lumenera Corporation). Cell length was measured in pixels in ImageJ (70) using the segmented line tool. A hemocytometer was photographed, followed by measurement of a 50-μm square in pixels in ImageJ, and used to convert cell lengths to micrometers.