Guanosine tetra- and pentaphosphate increase antibiotic tolerance by reducing reactive oxygen species production in Vibrio cholerae

The pathogen Vibrio cholerae is the causative agent of cholera. Emergence of antibiotic-resistant V. cholerae strains is increasing, but the underlying mechanisms remain unclear. Herein, we report that the stringent response regulator and stress alarmone guanosine tetra- and pentaphosphate ((p)ppGpp) significantly contributes to antibiotic tolerance in V. cholerae. We found that N16961, a pandemic V. cholerae strain, and its isogenic (p)ppGpp-overexpressing mutant ΔrelAΔspoT are both more antibiotic-resistant than (p)ppGpp0 (ΔrelAΔrelVΔspoT) and ΔdksA mutants, which cannot produce or utilize (p)ppGpp, respectively. We also found that additional disruption of the aconitase B–encoding and tricarboxylic acid (TCA) cycle gene acnB in the (p)ppGpp0 mutant increases its antibiotic tolerance. Moreover, expression of TCA cycle genes, including acnB, was increased in (p)ppGpp0, but not in the antibiotic-resistant ΔrelAΔspoT mutant, suggesting that (p)ppGpp suppresses TCA cycle activity, thereby entailing antibiotic resistance. Importantly, when grown anaerobically or incubated with an iron chelator, the (p)ppGpp0 mutant became antibiotic-tolerant, suggesting that reactive oxygen species (ROS) are involved in antibiotic-mediated bacterial killing. Consistent with that hypothesis, tetracycline treatment markedly increased ROS production in the antibiotic-susceptible mutants. Interestingly, expression of the Fe(III) ABC transporter substrate–binding protein FbpA was increased 10-fold in (p)ppGpp0, and fbpA gene deletion restored viability of tetracycline-exposed (p)ppGpp0 cells. Of note, FbpA expression was repressed in the (p)ppGpp-accumulating mutant, resulting in a reduction of intracellular free iron, required for the ROS-generating Fenton reaction. Our results indicate that (p)ppGpp-mediated suppression of central metabolism and iron uptake reduces antibiotic-induced oxidative stress in V. cholerae.

Cholera, the epidemic acute diarrheal disease caused by Vibrio cholerae, occurs in many developing countries that have poor sanitation (1). The toxigenic strains express various pathogenic factors, including toxin-co-regulated pilus and cholera toxin (CT) 3 to acquire the host environmental niche where it survives in the acidic gastric conditions and enters the small intestine (1)(2)(3)(4)(5). These virulence factors permit substantial fluid transport from epithelial cells to lumen, which leads to severe watery diarrhea. Thus, treatments mainly used for cholera patients are oral or intravenous hydration therapy that are effective for reducing stool output (6). According to the cholera treatment guidelines from the World Health Organization and the Centers for Disease Control and Prevention, antibiotic treatment in conjunction with an oral rehydration solution (ORS) is recommended for patients who have severe symptoms or are seriously dehydrated and continue to pass a large amount of stool (7). By evaluating many antibiotics, those that are effective at (i) reducing stool output, (ii) reducing the duration of diarrhea, and (iii) inducing bacterial shedding have been selected for treating cholera patients (8 -12). Antibiotic treatment in conjunction with ORS allows patients to recover more rapidly compared with ORS treatment alone. Tetracycline is the most effective antibiotic to reduce the cholera morbidity (8). However, doxycycline, a proxy for tetracycline, is currently the first-line drug of choice for cholera treatment due to its easy administration and low dosage requirement compared with tetracycline (13). Alternative drugs for treatment include chloramphenicol, erythromycin, azithromycin, and furazolidone. This provides drug treatment flexibility dependent on infected regions or antibiotic resistance rates. These antibiotics are only administered in combination with rehydration therapy and are not permitted to be used for the prophylaxis of cholera infection to prevent the induction of antibiotic resistance (14,15). Recently, most isolates of V. cholerae O1 serotype from patient stools showed resistance to antibiotics that are commonly used for cholera treatment (16 -20), but with no clear underlying mechanisms.
In our previous studies, we revealed that stringent response (SR) regulates V. cholerae viability and virulence by providing the bacterium with increased fitness in unfavorable environments (21,22). SR is characterized as one of the global regulatory systems in bacteria, which is activated by a variety of growth-inhibiting stresses (23). SR induces rapid adaptation under various stress conditions via (p)ppGpp accumulation (24). (p)ppGpp, known as a stress alarmone, forms a complex with RNA polymerase and induces profound reprogramming of global gene expression, which leads to growth arrest (25)(26)(27). In most Gram-negative bacteria, (p)ppGpp production is regulated by two enzymes, RelA and SpoT (24,28). RelA, a monofunctional synthetase, recognizes the uncharged tRNA at the A site of a ribosome, and it starts to synthesize (p)ppGpp. SpoT is the bifunctional enzyme that has both synthetase and hydrolase domains but shows mostly strong hydrolysis and weak synthetase activities. In the case of V. cholerae, it has an additional novel (p)ppGpp synthetase called RelV, which loses its N-terminal hydrolase domain (29,30). Because this enteric pathogen encounters the host-derived immune system during infection, human intestinal environments have the potential to influence nutrient availability to and the viability of V. cholerae (1,31).
Interestingly, recent studies have reported that SR-mediated transcriptional switching impacts bacterial physiology and increases antibiotic resistance (32)(33)(34). For example, (p)ppGpp accumulation raises penicillin resistance by inhibiting peptidoglycan synthesis in both Gram-positive and -negative bacteria (35)(36)(37)(38). This is not limited to ␤-lactam antibiotics, and the resistance to other antimicrobials is also linked to (p)ppGpp accumulation (39,40). It has been suggested that SR-mediated growth defects reduce antibiotic sensitivity, but this hypothesis is not fully understood. Recently, some studies have suggested that decreased levels of superoxide dismutase and catalase activities in SR mutants are associated with susceptibility to multiclasses of antibiotics (41,42). This report described the famous mechanistic model that different bactericidal antibiotics, regardless of their primary drug target, had a common pathway that generated deleterious ROS (43). According to this concept, ROS formation following antibiotic treatment enhances antibiotic lethality as well as the interaction with their traditional targets. Importantly, this ROS stress is derived from alterations of bacterial physiology, including hyperactivation of central metabolism and cellular respiration and disruption of iron homeostasis, by disrupting specific drug targets (43)(44)(45). It might seem strange that antibiotic treatment induces the overflow of metabolism, but to support this idea, many reports have described the transcriptomic and proteomic response to the bactericidal antibiotics (43,44). Therefore, we hypothesized that the potential capacity for reducing oxidative stress is essential for bacteria to survive during antibiotic treatment.
In this study, we showed that (p)ppGpp accumulation induced antibiotic resistance in V. cholerae by suppressing its central metabolism. We investigated the molecular basis of the specific physiology in (p)ppGpp-null V. cholerae that restored their viability against antibiotics when the acnB gene was totally abolished. Additionally, we showed that the (p)ppGpp-nonpro-ducing V. cholerae mutant carried a higher level of intracellular free iron, the crucial source of ROS production. This report provides a novel insight into the stepwise regulation of SR that contributes to defend against oxidative stress following antibiotic treatment.

(p)ppGpp-accumulated V. cholerae exhibits antibiotic tolerance
SR regulates bacterial responses to overcome unfavorable growth conditions in V. cholerae. To explore whether V. cholerae develops antibiotic tolerance in an SR-dependent manner, we measured survival rates of a range of SR-related mutant strains of V. cholerae. Front-line drugs for cholera treatment, tetracycline (Tc), erythromycin (Em), and chloramphenicol (Cp), were used for the bacterial treatments. To elucidate whether antibiotic tolerance was specifically induced by (p)ppGpp accumulation, we first chemically induced (p)ppGpp overproduction in N16961, a seventh pandemic O1 V. cholerae strain. When N16961 was treated for 2 h with serine hydroxamate (SHX), a serine analogue that activates SR by mimicking amino acid starvation (41), a marked increase in (p)ppGpp production was induced (Fig. 1A). Although ppGpp and pppGpp were not clearly distinguished in our TLC assay, SHX-treated N16961 produced dramatically abundant (p)ppGpp compared with untreated controls (Fig. 1A). To assess the effect of (p)ppGpp accumulation on bacterial survival under antibiotic stress, we treated N16961 grown in Luria-Bertani medium (LB) or LB ϩ SHX with varying concentrations of Tc, Em, or Cp for 4 h. Loss of viability was detected in LB-grown N16961. Complete killing was observed after 4 h of treatment with 10 g/ml Tc (Fig. 1B), 60 g/ml Em (Fig. 1C), and 25 g/ml Cp (Fig. 1D). SHX-treated bacterial cells, however, were tolerant to the same antibiotic treatment and maintained their viability under conditions lethal to the control groups ( Fig. 1, B-D). Although the extent to which these changes occurred was similar in all instances, bacterial viability was least affected in the Tc-treated group, with ϳ10 6 cfu/ml recovered even after treatment with the 20 g/ml concentration (Fig. 1B). These results suggest that a clear mechanistic link probably exists between SR and antibiotic tolerance in V. cholerae.
To further elucidate the effect of intracellular (p)ppGpp accumulation on bacterial susceptibility to antibiotics in V. cholerae, we tested isogenic ⌬relA⌬spoT double and ⌬relA⌬relV⌬spoT (termed (p)ppGpp 0 ) triple mutant strains. The ⌬relA⌬spoT mutant showed highly elevated levels of (p)ppGpp production due to (i) the action of RelV, an additional enzyme involved in (p)ppGpp biosynthesis and (ii) the lack of SpoT that hydrolyzes (p)ppGpp (21). The (p)ppGpp 0 mutant, which lacked all enzymes involved in (p)ppGpp metabolism, was found to produce no (p)ppGpp (21). The level of (p)ppGpp produced in the ⌬relA⌬spoT mutant was similar to that in the SHX-treated N16961 cells, whereas no (p)ppGpp was produced in the (p)ppGpp 0 triple mutant (Fig. 1A). We also included a ⌬dksA mutant, defective in DksA, that binds to RNA polymerase to facilitate the function of (p)ppGpp during SR (46). The capability of the ⌬dksA mutant to produce (p)ppGpp was Stringent response and antibiotic tolerance in V. cholerae not affected, as shown in Fig. 1A. Because (p)ppGpp levels are known to accumulate as bacterial cells enter the stationary phase (28), we also examined how each bacterial strain, harvested at either the exponential or stationary phase, responded to antibiotic stresses.
When exponential-phase cells were treated with antibiotics, (p)ppGpp 0 and ⌬dksA mutant strains as well as the WT strain N16961 invariably lost their viability. The ⌬relA⌬spoT mutant, which accumulates intracellular (p)ppGpp, was found to be the only strain that maintained viability (Fig. 1, E-G). In contrast, bacterial strains harvested at the stationary phase displayed marked differences from those grown in the exponential phase. Whereas the ⌬relA⌬spoT mutant continued to be antibiotictolerant, WT N16961 was also found to be tolerant to the same antibiotic treatment (Fig. 1, H-J). Furthermore, these two strains developed tolerance to significantly higher concentrations of antibiotics. Their viabilities were not affected even in the presence of 50 g/ml Tc, 600 g/ml Em, or 500 g/ml Cp, respectively.
To further verify the role of (p)ppGpp in conferring antibiotic tolerance in V. cholerae, we tested the antibiotic-susceptible (p)ppGpp 0 mutant harboring pRelV BAD , a plasmid that can express the relV gene via an arabinose-inducible promoter.
When the relV gene was expressed, the mutant cells became tolerant to Tc treatment (Fig. S1). Restored tolerance was observed in cells harvested at either growth phase (Fig. S1, A and B). Such a restoration was not detected in the mutant transformed with a control plasmid. Together, these results demonstrated that the ability to produce and utilize (p)ppGpp, a SR regulator, was critically important in affecting V. cholerae susceptibility to antibiotics.

Mutations in the acnB gene resulted in increased tolerance to antibiotic stresses
To provide a mechanistic insight into the increased antibiotic susceptibility of the (p)ppGpp 0 mutant, we sought to identify additional mutations that rendered the mutant tolerant to antibiotic treatments. To achieve this goal, we constructed a transposon (Tn) insertion library of the (p)ppGpp 0 mutant and looked for mutants that survived Tc treatment. Experimental procedures are described in Fig. 2A. After three successive subcultures in Tc-containing media, we recovered seven mutants that exhibited unencumbered growth. Subsequent analysis indicated that these mutants harbored Tn insertions in the acnB gene encoding a TCA cycle enzyme, aconitase B (Fig. 2B). Importantly, Tn insertion was found to occur at different loca- Figure 1. Effects of (p)ppGpp on viability of V. cholerae strains against clinically used antibiotics. A, detection of (p)ppGpp accumulation pattern in wildtype N16961 by TLC. Wildtype N16961 was divided into two groups and grown in LB supplemented with or without SHX in the presence of [ 32 P]orthophosphate for 2 h. Cellular extracts were prepared and analyzed in TLC. ⌬relA⌬spoT, (p)ppGpp 0 , and ⌬dksA mutants were also grown with [ 32 P]orthophosphate and processed for the TLC assay. B-J, bacterial viabilities under various clinically used antibiotics were measured in each growth condition. Wildtype N16961 and various (p)ppGpp-associated mutant strains were inoculated in LB and aerobically grown to each growth stage. Aliquots of each culture were resuspended in several concentrations of antibiotic-containing LB broth, sampled at 4 h post-inoculation, and 10-fold serially diluted to assess the number of cfu.

Stringent response and antibiotic tolerance in V. cholerae
tions of the gene in each of these mutants (Fig. 2B), suggesting that independent mutational events resulted in an identical consequence. We then introduced an in-frame deletion of the acnB gene into N16961 and (p)ppGpp 0 mutants and measured bacterial responses to the antibiotic treatments. Notably, the viability of bacterial cells, when harvested at the exponential phase, was dramatically increased in both ⌬acnB and the quadruple (p)ppGpp 0 ⌬acnB mutants compared with their parental strains (Fig. 2, C-E). cfu of up to ϳ10 6 /ml were observed following Tc (Fig. 2C) and Cp (Fig. 2E) treatments, whereas ϳ10 4 /ml was recovered after treatment with Em (Fig. 2D). Bacterial cells remained viable even in the presence of the highest concentrations of antibiotics. When the (p)ppGpp 0 ⌬acnB quadruple mutant, harvested at the stationary phase, was treated with antibiotics, higher levels of bacterial survival were detected (Fig. 2, F-H).
Importantly, when the plasmid-born acnB gene was expressed by the arabinose-inducible promoter, bacterial strains became susceptible to Tc treatments (Fig. S2). Consistent with previous results, sharper decreases in bacterial viability were detected when using exponential-phase cells (Fig. S2, A and C). Together, these results demonstrated that deletion of the acnB gene con-ferred a survival advantage to V. cholerae in the presence of antibiotic stresses.

Intracellular (p)ppGpp levels inversely regulate TCA cycle activity, which affects bacterial growth and cell morphology
Aconitase B, encoded by acnB, is the enzyme that catalyzes the interconversion of citrate and isocitrate in the TCA cycle. To examine whether the bacterial TCA cycle was regulated depending on the intracellular concentration of (p)ppGpp, we monitored transcript levels of genes encoding TCA cycle enzymes by RNA-sequencing analysis. First, acnB gene expression was ϳ3-fold higher in the (p)ppGpp 0 mutant than in the N16961 or in the (p)ppGpp-accumulating ⌬relA⌬spoT mutant (Fig. 3A). It is of note that N16961 also possesses the acnA gene, which encodes a phylogenetically distinct aconitase. However, its expression was not detected in all three strains, indicating that aconitase B is probably the major enzyme in the TCA cycle (Fig. 3A). Expressions of mdh, gltA, icd, sucA, sucC, and sucD genes were also noticeably increased in the (p)ppGpp 0 mutant, whereas their expressions were decreased (albeit to varying degrees) in the ⌬relA⌬spoT mutant (Fig. 3A). Although clear (p)ppGpp dependences were not observed in transcriptional

Stringent response and antibiotic tolerance in V. cholerae
regulation of the sdhA, sdhC, sdhD, frdB, and ttdA genes (Fig.  3A), our RNA-sequencing results strongly suggest that TCA cycle activity is inversely regulated by intracellular (p)ppGpp concentrations.
We then asked whether the (p)ppGpp-dependent regulation of TCA cycle gene expression was reflected in bacterial growth. When wildtype N16961 was grown aerobically in LB, A 600 values were reached at ϳ1.38 at 4 h post-inoculation and gradually increased up to ϳ3.5 for the rest of the experimental period (Fig. 3B). Consistent with a previous finding (21), bacterial growth was significantly retarded in the ⌬relA⌬spoT mutant, which accumulates (p)ppGpp. Final A 600 values were ϳ2.0. In contrast, the (p)ppGpp 0 mutant exhibited faster growth compared with N16961, and final A 600 was reached at ϳ4.5 (Fig. 3B). Similarly, the ⌬dksA mutant grew more vigorously, compared with N16961, further confirming that the ⌬dksA mutant shared growth-associated phenotypes with the (p)ppGpp 0 mutant (Fig. 3B). To reveal the effect of acnB gene deletion on bacterial growth, we also monitored the growth of ⌬acnB and (p)ppGpp 0 ⌬acnB mutants. Both mutants exhibited a significantly retarded growth during the early growth stage. For the first 3 h, their growth was comparable with that of the

Stringent response and antibiotic tolerance in V. cholerae
⌬relA⌬spoT mutant (Fig. 3B). Importantly, the robust growth phenotype of the (p)ppGpp 0 mutant disappeared when the acnB gene was additionally disrupted, further suggesting that derepressed aerobic growth of the (p)ppGpp 0 mutant was associated with increased activity of the acnB gene product (Fig. 3B).
Elevated antibiotic resistance is observed when bacterial cells exhibit a persister phenotype that is often accompanied by cell shape changes (48 -50). To address this possibility, we analyzed digitized images of various bacterial strains that showed distinct antibiotic susceptibility. N16961 cells, harvested at the stationary phase and thus more resistant to antibiotics, were shorter and thinner than those harvested at the exponential phase ( Fig. 3C and Fig. S3). Antibiotic-tolerant ⌬relA⌬spoT mutant and SHX-treated N16961 cells looked thinner and shorter, regardless of when they were harvested ( Fig. 3C and Fig. S3). In contrast, antibiotic-susceptible (p)ppGpp 0 and ⌬dksA mutants maintained their regular curve-shaped morphotype even at the stationary phase ( Fig. 3C and Fig. S3). These results clearly suggest that bacterial cells with reduced size are probably more resistant to antibiotic treatment. Of particular interest is that disruption of the acnB gene also resulted in shorter and thinner morphotypes, a phenotype observed under (p)ppGpp-accumulating conditions ( Fig. 3C and Fig. S3). Together, these results demonstrated that (i) cell shape changes could be induced upon intracellular (p)ppGpp accumulation or metabolic alterations by acnB gene mutation and (ii) such changes are closely related to bacterial responses to antibiotic treatment.

Other TCA cycle mutants also exhibit increased antibiotic tolerance
To see whether other TCA cycle mutants also exhibit phenotypes similar to that of the acnB mutant, we additionally constructed in-frame deletion mutants of TCA cycle enzymes (⌬icd, ⌬sucDC, and ⌬mdh) in both wildtype N16961 and (p)ppGpp 0 mutant background and monitored their responses to the Tc treatment. As shown in Fig. 4, TCA cycle mutants were also found to be resistant to Tc. The ⌬icd mutant showed the most similar resistance to the ⌬acnB mutant, whereas the remainingtwomutants(⌬sucDCand⌬mdh)showedlowerresistance (Fig. 4, A and C). Importantly, when each of these mutations was introduced in the antibiotic-sensitive (p)ppGpp 0 mutant, antibiotic resistance was restored (Fig. 4, B and D). Compared with the (p)ppGpp 0 mutant, (p)ppGpp 0 ⌬acnB and (p)ppGpp 0 ⌬icd mutants maintained their viabilities at ϳ10 8 cells/ml following 4-h Tc treatment at 50 g/ml. The (p)ppGpp 0 ⌬sucDC and (p)ppGpp 0 ⌬mdh survived at ϳ10 6 cells/ml. Taken together, these results suggest that disruptions of the TCA cycle function due to the mutation of either the acnB gene or other constituting genes affected antibiotic resistance in V. cholerae. Our results also show that ablation of the acnB or icd gene, whose products mediate early steps of the TCA cycle, produced a more profound effect on bacterial responses to antibiotic stresses.

Antibiotic-mediated bacterial killing occurs only in the presence of molecular oxygen and iron
Our results so far showed that (i) expression of TCA cycle genes, such as acnB, icd, and mdh, was highly induced in the antibiotic-susceptible (p)ppGpp 0 mutant, (ii) antibiotic-mediated bacterial killing was reduced in each of these TCA cycle mutants, and (iii) when (p)ppGpp was accumulated, bacterial cells shrank and became antibiotic-tolerant. These findings led us to hypothesize that (p)ppGpp accumulation results in metabolic slowdown, rendering bacterial cells unresponsive to antibiotics. To explore whether active metabolism, possibly fueled by aerobic respiration, is necessary for antibiotic-mediated bac-

Stringent response and antibiotic tolerance in V. cholerae
terial killing, we compared bacterial responses under aerobic versus anaerobic environments. Exponential-phase N16961 cells, which were found to be antibiotic-sensitive, maintained their viability when further grown for 4 h anaerobically with Tc (Fig. 5A). Anerobiosis also provided an additional protective effect (Ͼ10-fold) on stationary-phase N16961 cells against 50 g/ml Tc (Fig. 5D). Likewise, only marginal viability loss was observed upon treatment with Tc in two antibiotic-sensitive mutants, (p)ppGpp 0 (Fig. 5, B and E) and ⌬dksA (Fig. 5, C and F), under strict anaerobic environments. These results demonstrated that antibiotic-mediated bacterial killing occurred only during aerobic growth.
Hydroxyl radical (OH ⅐ ), the most bactericidal ROS, was reported to be produced during antibiotic treatments in large quantities (43). Because its production is catalyzed by the ironmediated Fenton reaction (51, 52), we next examined how bac- Figure 5. Recovery of tetracycline resistance of (p)ppGpp-deficient strains under anaerobic and iron-deplete conditions. A-F, tetracycline resistance of N16961 (circles), (p)ppGpp 0 mutant (triangles), and ⌬dksA mutant (squares) strains under anaerobic conditions. Wildtype N16961 and various (p)ppGppassociated mutant strains were inoculated in LB and aerobically grown to each growth stage. Aliquots of each culture were resuspended in several concentrations of antibiotic-containing LB broth and incubated for 4 h in anaerobic conditions. Each aliquot was sampled and 10-fold serially diluted to assess the number of cfu. G-L, tetracycline resistance of N16961 and (p)ppGpp-deficient mutant strains under iron-depleted conditions. Aliquots of each culture were resuspended in several concentrations of tetracycline-containing LB supplemented with 2,2Ј-dipyridyl, sampled at 4 h post-inoculation, and serially diluted for cfu counting.

Stringent response and antibiotic tolerance in V. cholerae
terial responses were altered under iron-depleted conditions. To this end, bacterial strains were treated with Tc in the presence of 2,2Ј-dipyridyl, an iron chelator (43). As shown in Fig. 5 (G-L), bacterial strains remained viable under iron-deficient conditions after being treated with as high as 50 g/ml Tc for 4 h (Fig. 5, J-L). Again, two antibiotic-vulnerable mutants and exponential-phase N16961 cells were completely killed by the same treatment in iron-sufficient LB. We then examined whether bacterial killing was alleviated by co-treatment with N-acetylcysteine (NAC), an ROS scavenger. Bacterial survival was substantially restored when antibiotic-sensitive (p)ppGpp 0 and ⌬dksA mutants were co-treated with 5 mM NAC (Fig. 6, B and C). As expected, NAC treatment exerted no protective effects on stationary-phase N16961 cells already resistant to Tc treatment (Fig. 6A). Together, these results indicated that (i) iron availability determined bacterial survival during antibiotic stress and (ii) ROS, produced aerobically during antibiotic treatment, were responsible for bacterial killing.
Consistent with these results, post-antibiotic ROS production was indeed increased in two antibiotic-sensitive mutants (Fig. 7). When treated with Tc for 1 h, ROS-specific fluorescent signals increased ϳ100and ϳ1,000-fold in (p)ppGpp 0 and ⌬dksA mutant, respectively. In contrast, fluorescent signals increased only ϳ10-fold in stationary-phase N16961 cells and (p)ppGpp-accumulating ⌬relA⌬spoT mutant (Fig. 7). It is of particular importance that post-antibiotic ROS production was not detected in the ⌬acnB single and (p)ppGpp 0 ⌬acnB quadruple mutant (Fig. 7), thereby further demonstrating that the recovered antibiotic resistance by acnB gene deletion is associated with the lack of ROS production.
Recently, we reported a recombinant N16961 strain that harbored an eKatE gene encoding a robust catalase (53). The eKatE gene is derived from a commensal Escherichia coli strain, and this strain was found to be resistant to 2 mM H 2 O 2 concentration. This strain, when harvested at the exponential phase, developed clear resistance to Tc treatment. Whereas the control N16961 strain perished completely during the same treatment, ϳ10 4 cells remained viable (Fig. 6D). This result further supports our finding that ROS resistance can help with antibiotic tolerance.

Larger amounts of intracellular free iron are present in SRnegative mutants
Next, we sought to further elucidate mechanisms by which (p)ppGpp regulates iron-dependent ROS production during aerobic antibiotic treatment. Table 1 lists the top five genes, expression of which was highly induced in the antibiotic-susceptible (p)ppGpp 0 mutant, as normalized with that of each counterpart in the antibiotic-tolerant ⌬relA⌬spoT mutant. Among these are genes encoding heme transport protein (hutA), enterobactin receptor protein (irgA), and periplasmic Fe(III) ABC transporter substrate-binding protein (fbpA).

Stringent response and antibiotic tolerance in V. cholerae
Because these proteins are involved in iron acquisition, these RNA-sequencing data strongly suggest that the iron uptake system may be inversely regulated by intracellular (p)ppGpp levels. Consistent with this notion, production of FbpA protein was markedly increased in two SR-negative mutants, whereas its production was not detected in the (p)ppGpp-accumulating ⌬relA⌬spoT mutant (Fig. 8A).
To understand the role of FbpA in iron uptake and antibiotic tolerance, we introduced an in-frame deletion of the fbpA gene in the N16961, (p)ppGpp 0 , and ⌬dksA mutants (Fig. S4). Bacterial growth was not affected by fbpA gene deletion (data not shown). When we measured intracellular free iron concentrations by whole-cell EPR spectrometry, iron-specific signals were invariably decreased in the N16961, (p)ppGpp 0 , or ⌬dksA strains when the fbpA gene was inactivated (Fig. 8B). The (p)ppGpp 0 ⌬fbpA quadruple and ⌬dksA⌬fbpA double mutants contained ϳ2-fold and ϳ3.3-fold decreased levels of free iron, respectively, when compared with their parental mutants (Fig.  8B). These results showed that FbpA protein played an important role in iron uptake in V. cholerae. It is of particular interest that substantially increased amounts of free iron were detected in two antibiotic-susceptible mutants, as compared with N16961 (Fig. 8B). Together, these results clearly demonstrated that the iron import system was suppressed by (p)ppGpp accumulation, and this down-regulation could contribute to reducing the level of intracellular free iron, the crucial source of hydroxyl radical (OH ⅐ ) formation.
In line with these findings, bacterial survival under Tc treatment was significantly increased in ⌬fbpA mutants (Fig. 8,  C-H). When exponential-phase cells were used for treatment, bacterial viability of up to ϳ10 4 -fold was recovered in ⌬fbpA, (p)ppGpp 0 ⌬fbpA, and ⌬dksA⌬fbpA mutants compared with each of their background strains (Fig. 8, C, E, and G). Moreover, (p)ppGpp 0 ⌬fbpA and ⌬dksA⌬fbpA mutants, harvested at the stationary stage, also exhibited Tc resistance (Fig. 8, F and H).

Discussion
Cholera is characterized by CT-induced profuse watery diarrhea. Rapid loss of body fluids often leads to fatal dehydration (54 -56). Although up to 80% of cholera patients can be treated with oral rehydration therapy, annual deaths up to 120,000 are reported (81). Mortality includes victims who fail to receive immediate interventions and young patients with immature stomach function (82). These cases apparently need prompt antibiotic treatment to reduce the volume of diarrhea and kill the causative pathogen, V. cholerae. Treating cholera patients with antibiotics, however, has been a challenge due to the increased emergence of antibiotic-tolerant V. cholerae strains (16 -20, 57).
In this study, we proposed that SR, a conserved bacterial stress response mechanism, regulates antibiotic tolerance in V. cholerae via mechanisms that eventually suppress ROS production. SR controls metabolic activity and intracellular iron level, both of which affect bacterial growth and thereby antibiotic-induced ROS generation. Our successful demonstration on this important issue was made possible by the availability of the ⌬relA⌬spoT double mutant that spontaneously accumulates intracellular (p)ppGpp. Whereas (p)ppGpp-null phenotypes have been well-documented in various bacterial species (21,41,58,59), cellular phenotypes induced by natural (p)ppGpp accumulation have not been clearly described. The (p)ppGpp-accumulating ⌬relA⌬spoT mutant (i) metabolized slowly, (ii) exhibited a smaller cell-size phenotype, and (iii) produced significantly decreased levels of FbpA protein involved in iron acquisition. All of these phenotypes were invariably reversed in (p)ppGpp 0 and ⌬dksA mutants, determined to be antibiotic-susceptible. Among these phenotypes, FbpA-mediated iron metabolism provided us with a clue that helped us to precisely elucidate the role of ROS in antibiotic-mediated bacterial killing.
V. cholerae is an enteric pathogen that is transmitted through the fecal-oral route. The toxigenic V. cholerae is able to rapidly spread through bacterial shedding by evoking deadly diarrhea. It passes through the esophagus and comes into contact with the acid environment of the stomach, to which V. cholerae is known to be susceptible. Moreover, V. cholerae colonizes the small intestine, where it must compete with diverse species of commensal microbes for nutrients (60). To overcome such unfavorable conditions, it is highly likely that SR is activated in V. cholerae inside the host intestine. Consistent with this idea, our previous work (21) demonstrated that (i) SR-defective mutants are incapable of colonizing mouse intestine and (ii) the (p)ppGpp-accumulating ⌬relA⌬spoT mutant produces a markedly increased level of CT. We therefore proposed that strains that survive and shed out to the environments may have active SR and thereby be more able to infect subsequent hosts and overcome antibiotic stresses. To further validate this hypothesis, it will be important to compare SR activity status in pandemic versus non-virulent environmental strains.
Our results demonstrated that (p)ppGpp, when accumulated, down-regulated expression of many TCA cycle genes, including acnB. Not surprisingly, bacterial metabolism and growth were reported to be suppressed by SR in diverse bacte-

Stringent response and antibiotic tolerance in V. cholerae
rial species (25,61,62). Studies also suggested that (p)ppGppmediated regulation participated in bacterial persister formations by mechanisms that involved toxin-antitoxin modules (63,64). Bacterial persisters are transient phenotypic variants that are stochastically induced within a subset of cells in a given population (64). Persister cells, in general, are more tolerant of antibiotic treatments (65). Ronayne et al. (48) showed that E. coli cells undergo cell elongation during the acquisition of a persister phenotype. In contrast, V. cholerae cells, incubated for a long time in nutrient-poor media, were shown to be smaller in size when compared with normal state cells (50), suggesting that the potential persister-like phenotype may be achieved in a distinct manner in V. cholerae. Our results showed that the (p)ppGpp-accumulating ⌬relA⌬spoT mutant and SHX-treated N16961 cells were thinner and shorter. Interestingly, reduced cell size was also observed in the ⌬acnB single mutant. These results indicated that cell shape change and a resultant increase in antibiotic tolerance ensued from (p)ppGpp accumulation or acnB gene deletion in V. cholerae and presented a new question as to which of these two cellular processes was the primary cause of the phenotype. Because a similar cell-size reduction was also detected in the (p)ppGpp 0 ⌬acnB mutant cells, we postulate that acnB gene deletion is a necessary and sufficient condition for the phenotype. Likewise, cell size reduction in the

Stringent response and antibiotic tolerance in V. cholerae
⌬relA⌬spoT mutant was probably induced by suppressed aconitase activity, not directly by (p)ppGpp accumulation. Cell biological features need to be further explored in terms of how inactive metabolism can lead to changes in cell shape. Periplasmic Fe(III) ABC transporter substrate-binding protein, encoded by fbpA, was highly produced in antibiotic-susceptible (p)ppGpp 0 and ⌬dksA mutants. More importantly, EPR analysis clearly showed that intracellular free iron was concomitantly increased in these two mutants. Like other bacterial species, V. cholerae requires iron for growth and possesses a variety of iron uptake systems (66). The iron acquisition systems in V. cholerae involve synthesis, secretion, and uptake of a range of siderophore molecules, such as vibriobactin (67), enterobactin (68,69), and ferrichrome (67,70). In addition, V. cholerae possesses feo and fbp gene clusters encoding systems for the acquisition of ferrous and ferric iron, respectively (66). The Fbp system is a periplasmic binding protein-dependent ABC transport system and consists of three genes, fbpA (VC0608), fbpB (VC0609), and fbpC (VC0610). The fbpB and fbpC encode membrane-spanning proteins forming a pore across the cell membrane and ATP binding proteins, respectively. The fbpA encodes a substrate-binding protein that carries ferric iron to the membrane-spanning proteins. Studies indicate that the potential Fur box (ferric uptake regulatorbinding motif) exists in the promoter region of the fbpABC operon (66,71), suggesting that expression of the operon is highly induced under iron-deficient conditions. However, in our experiments, bacterial strains were grown in LB, considered to be a nutrient-rich medium. It was of particular interest to us that the Fur protein is inactivated by the presence of ROS (72). ROS oxidizes Fe 2ϩ , a co-factor of Fur, converting active Fur into an inactive apo-form. Therefore, one possible explanation for the antibiotic-mediated bacterial killing in the antibiotic-susceptible (p)ppGpp 0 and ⌬dksA mutants would be that ROS produced during derepressed aerobic growth stimulated FbpA-mediated iron uptake, which in turn amplified further ROS production. In support of this idea, the ⌬relA⌬spoT mutant that grew very slowly produced undetectable levels of FbpA protein. Furthermore, we also found that FbpA production was reduced in the ⌬acnB mutant that exhibited a slow growth phenotype (data not shown).
In our experiments, the first-line drugs for treating cholera patients, tetracycline, erythromycin, and chloramphenicol, were used. We found that SR-deficient mutants of V. cholerae completely lost their viability as the antibiotic concentration increased. Interestingly, however, the antibiotics that turned out to be very effective at killing V. cholerae strains are commonly known as bacteriostatic. Although it is still debatable how to classify antibiotics as bacteriostatic or bactericidal based on their in vitro test results, recent studies described their differences based on their effects on bacterial metabolism. Lobritz et al. (44) found that bacteriostatic and bactericidal have opposing effects on bacterial respiration in E. coli and Staphylococcus aureus. They demonstrated that bacteriostatic antibiotics, such as tetracycline, chloramphenicol, and erythromycin, decelerate cellular respiration, whereas bactericidal antibiotics accelerate basal respiration and lead to the production of ROS as a byproduct. On the other hand, we found that bacteriostatic anti-biotics killed bacteria and produced deleterious ROS. Interestingly, in another study, the antibiotics that did not trigger SOS responses in E. coli caused SOS responses in V. cholerae (73). SOS stress responses were activated when exogenous and endogenous triggers provoked DNA damage, and some antibiotics act as exogenous triggers that stimulate ROS production, the crucial weapon of DNA disruption. Aminoglycosides, tetracycline, and chloramphenicol induce SOS responses in V. cholerae, unlike E. coli, and this result suggests the possibility that these antibiotics have more deleterious effects on V. cholerae. Taken together, our results suggest that effects of bacteriostatic and bactericidal antibiotics on bacteria vary from species to species.
In conclusion, our results revealed that bacterial SR regulates antibiotic tolerance by modulating ROS production. Central metabolism and iron transport systems are subject to (p)ppGppmediated regulation (summarized in Fig. 9). When (p)ppGpp is accumulated, the TCA cycle is down-regulated to slow down bacterial growth. At the same time, FbpA-mediated iron uptake is also suppressed in (p)ppGpp-accumulating cells. These dual cellular events both contribute to physiological changes resulting in metabolic slowdown and therefore antibiotic-tolerant states. Bacterial SR has been a target to be inhibited. Chemical compounds that can suppress (p)ppGpp production, such as relacin (74) and iMAC (75), can potentially be used as antibiotic adjuvants. We anticipate that experimental data provided in the current study will stimulate future investigations that eventually help us come up with better strategies to combat bacterial infections, including one by the deadly enteric pathogen V. cholerae.

Bacterial strains and growth conditions
All of the bacterial strains and plasmids used in this study are listed in Table 2. Bacterial cultures were grown in LB (1% (w/v) tryptone, 0.5% (w/v) yeast extract, and 1% (w/v) sodium chloride) at 37°C, and antibiotics were used at the following concentrations: streptomycin (Duchefa), 200 g/ml; ampicillin (Sigma), 50 g/ml; kanamycin (Duchefa), 50 g/ml. All bacterial single colonies on LB plates were picked and inoculated in LB broth for precultures and grown overnight. Precultures were diluted 100-fold in fresh LB broth for subculture and incu- Figure 9. Summary of (p)ppGpp-mediated regulation of antibiotic resistance in V. cholerae. Antibiotic treatment stimulates ROS production by hyperactivating bacterial central metabolism. The released ROS leads to damage of intracellular DNA, proteins, and lipids, which results in cell death. However, (p)ppGpp negatively regulates TCA cycle and aerobic respiration. The down-regulation of aerobic respiration can reduce oxidative stress and eventually prevent cell death. The (p)ppGpp can maximize this effect by restricting free iron uptake from the iron-transporting system.

Stringent response and antibiotic tolerance in V. cholerae
bated at 37°C and 220 rpm. The incubation time was dependent on experimental procedures.

Antibiotic tolerance assay
All antibiotic tolerance assays in this study were performed as described previously with a few modifications (41). For bacterial drug experiments, we used tetracycline (Sigma), erythromycin (Sigma), chloramphenicol (Sigma).
Exponential-phase bacteria-Bacteria from precultures were diluted 100-fold in LB broth and grown shaking at 37°C until the growth reached A 600 0.5. 5 mM serine hydroxamate (Sigma) was added to some of the cultures after 1 h of growth and further incubated for 1 h. Aliquots were then resuspended in serial diluted antibiotic-LB to A 600 0.05 and incubated statically for 4 h. The cfu were measured by serial dilution of individual aliquots on LB plates for statistical testing.
Stationary-phase bacteria-Overnight subcultures (16 h) were resuspended in serial diluted antibiotic-LB aliquots to A 600 0.3 and statically incubated for 4 h. The survival rate of individual aliquots was also measured by viable cell counting.

Determination of intracellular (p)ppGpp concentration by TLC
Intracellular ppGpp concentration was measured as described previously with a few modifications (22). To detect intracellular (p)ppGpp, bacterial cells were grown aerobically with 100 Ci/ml [ 32 P]orthophosphate (PerkinElmer Life Sciences) at each growth phase. The bacterial cell cultures were extracted with cold 10 mM Tris-HCl buffer (pH 8.0). After centrifugation to remove the cell supernatant, cell pellets were resuspended with cold 10 mM Tris-HCl buffer and 19 M formic acid and then freeze-thawed for three cycles. After centrifugation to remove cell debris, cell supernatants were spotted on PEI cellulose F TLC plates (Merck). The TLC plates were developed in 1.5 M KH 2 PO 4 buffer (pH 3.4) in a humidified chamber and imaged with autoradiography.

Construction of in-frame deletion mutants
V. cholerae mutants were created by allele replacement, as described previously (21,76). To induce mutation, 500-bp flanking sequences located at both ends of the ORF were amplified by PCR with the primers listed in Table 3. The primers used to amplify each flanking region were carried by restriction enzyme sites that were located in multiple cloning sites of pCVD442 suicide vectors. The purified forward flanking sequences were ligated in pCVD442 vector with T4 ligase, and extracted vectors were transformed to heat-shock competent cells, SM10/pir strains. The transformed cells were selected on LB plates containing 100 g/ml ampicillin, and cloning vectors were purified from single survival colonies, following recombination of another flanking sequence going through the same steps to transformation. The SM10/pir strains with cloned vectors and V. cholerae recipient strains were mixed at a ratio of 3:1 onto LB plates and incubated for 6 h at 37°C. The mixed pool was suspended in fresh LB broth and spread on LB plates containing 200 g/ml streptomycin or 100 g/ml ampicillin for the first step in allelic exchange. After overnight culture, for plasmid excision from the chromosome by second cross-over, single colonies were selected and streaked on LB plates with 8% sucrose and 200 g/ml streptomycin without NaCl. Screening of in-frame deletion sites for each colony was preceded by PCR to identify the desired allele with primers that contained flanking sequences and ORF.

Construction of a random Tn-insertion mutant library
A Tn-inserted mutant library was constructed by using the mariner-based Tn, TnKGL3, which contains a kanamycin resistance marker. Tn mutagenesis of ⌬relA⌬relV⌬spoT (i.e. (p)ppGpp 0 ) mutant was performed with SM10/pir, which carries the TnKGL3. SM10/pir strains and the V. cholerae recipient strain were mixed at a ratio of 2:1 onto LB plates and incubated for 6 h at 37°C. The mixed pool was suspended with fresh LB broth and spread on LB plates containing 200 g/ml strep- Table 2 Bacterial strains and plasmids used in this study tomycin, 100 g/ml kanamycin and incubated overnight. The survival bacterial pool was harvested in LB broth and diluted in LB aliquots containing 50 g/ml tetracycline to A 600 0.3 and statically incubated for 4 h. The whole bacterial cells of each aliquot were recruited for centrifugation and resuspended again in LB aliquots containing 50 g/ml tetracycline and incubated for 4 h. These procedures were repeated three times, and final aliquots were partially spread on LB plates containing streptomycin and kanamycin. The surviving bacterial colonies were collected into a library, and the Tn-insertion site for each mutant was determined by arbitrary PCR. The first round of arbitrary PCR was performed by the transposon TnKGL3-specific primer Mar1 and two random primers, Arb1 and Arb6. The following second round of PCR proceeded with Mar2 and Arb2, by using the first-round PCR products for templates. The PCR products were sequenced using the primer Mar2. The Tninserted locations were identified by comparison with the public database of the V. cholerae genome sequence.

Scanning electron microscopy analysis
Characterization of bacterial cell morphology and size were visualized with scanning electron microscopy, following procedures described previously (77). Briefly, for the sample preparation, bacterial cell cultures were fixed with PBS containing 2% glutaraldehyde and 0.1% paraformaldehyde for 2 h and stained with 1% OsO 4 . Samples were then coated with gold by an ion sputter (IB-3, Eiko, Japan) and examined with a scanning electron microscope (FE SEM S-800, Hitachi, Japan) at an acceleration voltage of 20 kV. Images were processed with ESCAN 4000 software (Bummi Universe Co., Ltd., Seoul, Korea). For measuring the cell length and diameter, more than 100 straightlined cells were randomly chosen from the digitized scanning electron microscopy images, and the distance between the two ends was automatically calculated.

RNA-sequencing analysis
Bacterial cultures grown in LB were harvested at 16 h postinoculation. Aliquots of each culture (n ϭ 3) were pooled together in one tube for RNA analysis. To extract high-quality bacterial RNA, an RNeasy Protect kit (Qiagen) was used with an RNeasy minikit (Qiagen) following the manufacturer's protocol. The quantity and quality of total RNA were evaluated using RNA electropherograms (Agilent 2100 Bioanalyzer, Agilent Technologies, Waldbroon, Germany) and by assessing the RNA integrity number. From each sample with an RNA integrity number value greater than 8.0, 8 g of the total RNA was used as a starting material and treated with the MICROBExpress TM mRNA enrichment kit (Invitrogen). The resulting mRNA samples were processed for the sequencing libraries using Illumina mRNA-Seq sample preparation kit (Illumina, San Diego, CA) following the manufacturer's protocols. One lane per sample was used for sequencing with the Illumina Genome Analyzer IIx (Illumina) to generate nondirectional, single-ended, 36-base pair reads. Quality-filtered reads were mapped to the reference genome sequences (NCBI Bio-Project accession number PRJNA57623, identification number 57623) using CLRNASeq version 0.80 (Chunlab, Seoul, Korea). Relative transcript abundance was computed by counting the RPKM (78).

ROS measurement
Chemical hydrolysis of 2Ј,7Ј-dichlorofluorescin diacetate was performed following procedures described elsewhere (79). Briefly, 0.5 ml of 5 mM 2Ј,7Ј-dichlorofluorescin diacetate, dissolved in 100% ethanol, was reacted with 2 ml of 0.1 N NaOH at room temperature for 30 min. The reaction was stopped by adding 7.5 ml of 100 mM PBS, giving a final DCF concentration of 50 M. Bacterial suspensions prepared from stationary-phase

Stringent response and antibiotic tolerance in V. cholerae
cultures were treated with 50 g/ml tetracycline for 1 h. After treatment, the suspensions were centrifuged at 13,000 rpm for 3 min, and cell-free supernatants were removed. Cell pellets were resuspended with 50 M DCF solution and incubated for 30 min. Then the DCF intensity, which is indicative of the intracellular ROS level, was measured with a Victor X4 plate reader (PerkinElmer Life Sciences).

SDS-PAGE analysis and protein identification
Preparation of periplasmic fractions followed procedures described previously (80). V. cholerae strains were grown anaerobically in LB for 16 h. Cell pellets were resuspended with PBS containing 250 g/ml polymyxin B and incubated for 15 min at 4°C. After incubation, the mixtures were centrifuged at 13,000 rpm for 20 min at 4°C, and the supernatant was used for separation of periplasmic proteins. Protein was quantified by the method of Bradford, and 5 g of proteins were separated by 12% SDS-PAGE. The SDS-polyacrylamide gel fractions were submitted to Yonsei Proteome Research Center for protein identification.

EPR analysis
Intracellular free iron levels were measured as described previously with a few modifications (47). Bacterial cells were grown in LB and harvested at 16 h post-inoculation. A bacterial cell pellet was resuspended in 5 ml of prewarmed fresh LB broth that contained 10 mM diethylentriaminepentaacetic acid, pH 7.0, and 20 mM desferrioxamine (pH 8.0). Diethylentriaminepentaacetic acid blocks further iron import, whereas desferrioxamine diffuses into cells and binds unincorporated iron in an EPR-visible ferric form. The concentrated cells were incubated at 37°C for 15 min in a shaking incubator. The cells were washed with ice-cold 20 mM Tris-Cl (pH 7.4) twice. Cells were then resuspended in 200 l of ice-cold 10% glycerol, 20 mM Tris-Cl (pH 7.4). The cell suspension (200 l) then was transferred into an EPR tube and frozen in liquid nitrogen. Ferric sulfate standards were mixed with desferrioxamine and prepared in the same Tris buffer containing glycerol. The spectrometer settings were as follows: microwave power, 1 milliwatt; microwave frequency, 9.64 GHz; modulation amplitude, 10 Gauss at 100 KHz; temperature, 15 K.

Statistical analysis
The data are expressed as the means Ϯ S.D. Unpaired Student's t tests (two-tailed, unequal variance) were used to analyze the differences between experimental groups. p values Ͻ 0.05 were considered statistically significant. All experiments were repeated for reproducibility.