The Stringent Response Controls Catalases in Pseudomonas aeruginosa and Is Required for Hydrogen Peroxide and Antibiotic Tolerance

Abstract

Pseudomonas aeruginosa, a human opportunistic pathogen, possesses a number of antioxidant defense enzymes under the control of multiple regulatory systems. We recently reported that inactivation of the P. aeruginosa stringent response (SR), a starvation stress response controlled by the alarmone (p)ppGpp, caused impaired antioxidant defenses and antibiotic tolerance. Since catalases are key antioxidant enzymes in P. aeruginosa, we compared the levels of H₂O₂ susceptibility and catalase activity in P. aeruginosa wild-type and ΔrelA ΔspoT (ΔSR) mutant cells. We found that the SR was required for optimal catalase activity and mediated H₂O₂ tolerance during both planktonic and biofilm growth. Upon amino acid starvation, induction of the SR upregulated catalase activity. Full expression of katA and katB also required the SR, and this regulation occurred through both RpoS-independent and RpoS-dependent mechanisms. Furthermore, overexpression of katA was sufficient to restore H₂O₂ tolerance and to partially rescue the antibiotic tolerance of ΔSR cells. All together, these results suggest that the SR regulates catalases and that this is an important mechanism in protecting nutrient-starved and biofilm bacteria from H₂O₂- and antibiotic-mediated killing.

INTRODUCTION

Reactive oxygen species (ROS) are spontaneously produced during aerobic respiration. During infections, bacteria are also challenged by ROS produced by host phagocytic cells. Given that ROS can readily damage membranes, DNA, and proteins (1, 2), bacteria possess multiple antioxidant defenses to survive during aerobic growth and in vivo. While it is expected that oxidative stress induces antioxidant defenses, interestingly, other stresses such as nutrient limitation also elicit antioxidant responses. For example, carbon and nitrogen starvation increases resistance to H₂O₂ in different bacterial species, but the mechanism remains poorly understood (3–6). Stationary-phase and biofilm bacteria are also nutrient limited and exhibit high oxidant tolerance (7–9). Does nutrient limitation induce starvation responses that also confer protection against oxidative stress? To investigate the contribution of starvation responses in inducing antioxidant defenses, we examined the role of the stringent response (SR).

The SR, which is controlled by the alarmone (p)ppGpp, is a conserved regulatory mechanism that coordinates physiological adaptations to nutrient starvation and other stresses. In Gram-negative bacteria, synthesis of (p)ppGpp is catalyzed by the RelA and SpoT proteins (10, 11). This alarmone primarily modulates gene transcription to shut down biosynthesis of macromolecules and cell replication while inducing mechanisms required for stress survival (12, 13). Although there are several reports that nutrient starvation confers oxidative stress tolerance (3–6), only a few studies have explored the link between the SR and control of oxidative-stress pathways (14, 15).

Pseudomonas aeruginosa is an opportunistic human pathogen that causes lethal acute infections and chronic biofilm infections in the airways of cystic fibrosis patients. P. aeruginosa has several enzymes that can detoxify H₂O₂, including two monofunctional catalases, KatA and KatB. Catalases are key components of antioxidant defenses: they are enzymatic H₂O₂ scavengers that can be highly induced (2, 16). Constitutively expressed, KatA is the dominant catalase during exponential- and stationary-phase growth. It is also required for H₂O₂ resistance and in vivo virulence (17, 18). In contrast, katB is induced by exogenous H₂O₂ stress and likely contributes to acquired H₂O₂ resistance (19). The P. aeruginosa katE (katC) and katN genes also encode putative catalases, but their functional roles remain unclear (17, 20).

Antioxidant defense genes in bacteria are regulated by several overlapping systems. These systems have been extensively studied in Escherichia coli and, to a lesser extent, in P. aeruginosa (1, 2). For example, in P. aeruginosa, OxyR is a H₂O₂-responsive transcriptional regulator that activates expression of a subset of genes involved in antioxidant defense (katA, katB, ahpB, and ahpCF) as well as in iron homeostasis (21–25). Additional global regulators also control expression of antioxidant enzymes, including the Las and Rhl quorum-sensing systems (26, 27), RpoS (28–30), and the iron uptake regulator Fur (31). Because katA expression is under quorum-sensing control, its expression increases rapidly upon entry into the stationary phase (26, 32). These overlapping regulatory networks allow P. aeruginosa to adapt its oxidative defense systems in response to different environmental conditions such as the growth phase and iron availability.

We have recently reported that SR inactivation in P. aeruginosa dramatically decreases the antibiotic tolerance of nutrient-starved cells and biofilms for multiple classes of antibiotics (33). Since recent studies suggest that oxidative-stress pathways contribute to bacterial cell death caused by bactericidal antibiotics (34–38), we hypothesized that robust antioxidant defenses are required for antibiotic tolerance. Our initial studies showed that the ΔSR (ΔrelA ΔspoT) mutant had decreased superoxide dismutase and catalase activities and was more susceptible to oxidants than wild-type P. aeruginosa (33). In this study, we investigated how the SR mediates H₂O₂ tolerance and regulates catalases, which are highly efficient H₂O₂ scavengers. We explored the role of the stationary-phase alternative sigma factor RpoS as an intermediary mediator of the SR and looked at the contribution of catalases to antibiotic tolerance.

MATERIALS AND METHODS

Growth conditions.

All bacterial strains were grown in lysogeny broth (LB) medium for all experiments, except when serine starvation was required to induce the SR. In order to induce the SR, planktonic bacteria were grown in M9 minimal medium (1 mM MgSO₄, 47 mM Na₂HPO₄·2H₂O, 22 mM KH₂PO₄, 9 mM NaCl, 18 mM NH₄Cl, 10 mM glucose) at 37°C with shaking at 250 rpm. Subsequently, 500 μM serine hydroxamate (SHX; Sigma-Aldrich) was added to mid-log-phase cells (optical density at 600 nm [OD₆₀₀], 0.5) to cause serine starvation (39). Antibiotics were used at the following concentrations for selection: ampicillin at 100 μg/ml, kanamycin at 50 μg/ml, carbenicillin at 250 μg/ml, gentamicin at 10 μg/ml (E. coli) or 50 μg/ml (P. aeruginosa), and tetracycline at 10 μg/ml (E. coli) or 90 μg/ml (P. aeruginosa). When required, the inducers l-(+)-arabinose (L-ara; Sigma) and isopropyl-β-d-thiogalactopyranoside (IPTG; Sigma) were added to cultures at concentrations of 1% (wt/vol) and 4 mM, respectively. Stationary-phase planktonic bacteria were grown in LB medium at 37°C with shaking at 250 rpm for 16 to 18 h. Bacteria were grown in colony biofilms as previously described (40) with minor modifications. Here, 5 × 10⁵ CFU from an overnight culture grown in LB was inoculated onto polycarbonate membrane filters (GE Water and Process Technologies) and placed on LB agar for 24 h at 37°C unless otherwise specified.

Bacterial strains.

All strains and plasmids used in this study are listed in Table 1. The P. aeruginosa ΔrelA ΔspoT (ΔSR) strain and complemented mutant (+SR) strain were obtained from a previous study (33). The katA, katB, and rpoS transposon mutants were obtained from the PAO1 two-allele transposon mutant library (41). To create the ΔSR rpoS triple mutant, the rpoS-B03 ISlacZ/hah allele was moved from strain PW7151 into the ΔSR mutant by transformation of the electrocompetent ΔSR mutant with PW7151 genomic DNA as previously described (42). The rpoS mutation was selected for using the tetracycline (Tc) resistance marker. All transposon mutations were confirmed by PCR. Integration of miniTn7 vectors into the P. aeruginosa chromosome was carried out by electroporation with the helper plasmid pTNS2 (47). Antibiotic selection and confirmation of miniTn7 insertions by PCR were carried out according to established protocols (47). Transformation of P. aeruginosa with replicating plasmids was done by standard electroporation techniques.

Table 1.

Bacterial strains and plasmids

Strain or plasmid	Relevant characteristic(s)a	Source or reference
Escherichia coli strains
DH5α	fhuA2 Δ(argF-lacZ)U169 phoA glnV44 φ80lacZΔM15 gyrA96 recA1 relA1 endA1 thi-1 hsdR17	New England Biolabs
ccdB Survival 2 T1R	F− mcrA Δ(mrr-hsdRMS-mcrBC) φ80lacZΔM15 ΔlacX74 recA1	Invitrogen
	araD139 Δ(ara-leu)7697 galU galK rpsL endA1 nupG fhuA::IS2, Strr
EC100D pir-116	F− mcrA Δ(mrr-hsdRMS-mcrBC) F80lacZΔM15 lacX74 recA1 araD139 Δ(ara-leu)7697 galU galK endA1 nupG λ-rpsL pir-116, Strr	Epicenter Biotechnologies
Pseudomonas aeruginosa strains
PAO1	Wild type	Laboratory archive (33)
ΔSR	PAO1 ΔrelA ΔspoT	33
+SR	ΔSR attCTX::relA attTn7::spoT	33
PW8191	PAO1 katA-B07 ISphoA/hah, Tcr	41
PW8770	PAO1 katB-F08 ISphoA/hah, Tcr	41
PW7151	PAO1 rpoS-B03 ISlacZ/hah, Tcr	41
DN707	ΔSR rpoS-B03 ISlacZ/hah, Tcr	This study
MK298	PAO1 attTn7::miniTn7-Gm-GW-araC-pBAD::katA, Gmr	This study
MK300	ΔSR attTn7::miniTn7-Gm-GW-araC-pBAD::katA, Gmr	This study
MK318	PAO1 attTn7::miniTn7-Gm, Gmr	This study
MK320	ΔSR attTn7::miniTn7-Gm, Gmr	This study
MK325	PAO1 attTn7::miniTn7-Gm-GW-pkatA::lacZ, Gmr	This study
MK327	ΔSR attTn7::miniTn7-Gm-GW-pkatA-lacZ, Gmr	This study
MK361	PAO1 attTn7::miniTn7-Gm-GW-pkatB-lacZ, Gmr	This study
MK363	ΔSR attTn7::miniTn7-Gm-GW-pkatB-lacZ, Gmr	This study
MK331	PAO1 attTn7::miniTn7-Gm-GW-promoterless-lacZ, Gmr	This study
MK335	ΔSR attTn7::miniTn7-Gm-GW-promoterless-lacZ, Gmr	This study
Plasmids
pDONR221P1P5r	Multisite Gateway donor vector with attP1 and attP5r recombination sites, Cmr Kmr	Invitrogen
pDONR221P5P2	Multisite Gateway donor vector with attP5 and attP2 recombination sites, Cmr Kmr	Invitrogen
pUC18-miniTn7T-Gm	For gene insertion in Gms bacteria (aacC1 on mini-Tn7T), Apr Gmr	46
pUC18-miniTn7T-Gm-GW	pUC18-miniTn7T-Gm with a Gateway destination cloning site Apr Gmr Cmr	46
pTNS2	T7 transposase expression vector, Apr	46
mini-CTX2	Source template for a multiple cloning site (MCS), Tcr	48
mini-CTX-lacZ	Source template for lacZ, Tcr	47
pUT-miniTn5Pro	Source template for araC-pBAD, Apr Gmr	45
pJJH167	pDONR221P1P5r with an attL-flanked, 658-bp fragment encoding the P. aeruginosa PAO1 katA promoter region, Kmr	This study
pJJH168	pDONR221P1P5r with an attL-flanked, 298-bp fragment encoding the P. aeruginosa PAO1 katB promoter region, Kmr	This study
pJJH173	pDONR221P5P2 with an attL flanked,-1,463-bp fragment encoding a 14-bp synthetic ribosomal binding site and the 1,449-bp katA ORF from P. aeruginosa PAO1, Kmr	This study
pJJH187	pDONR221P1P5r with an attL-flanked, 1,192-bp fragment encoding the araC repressor and the pBAD promoter, Kmr	This study
pMK71	pDONR221P1P5r with an attL-flanked, 100-bp fragment encoding the multiple cloning site from mini-CTX2, Kmr	This study
pMK314	pDONR221P5P2 with an attL-flanked, 3,276-bp fragment encoding an RNase III site and a ribosomal binding site flanking a lacZ ORF, Kmr	This study
pMK365	pUC18miniTn7T-Gm-GW containing a pkatA::lacZ promoter-reporter construct, Apr Gmr	This study
pMK361	pUC18miniTn7T-Gm-GW containing a pkatB::lacZ promoter-reporter construct, Apr Gmr	This study
pMK297	pUC18miniTn7T-Gm-GW containing araC-pBAD:: katA, Apr Gmr	This study
pMK339	pUC18miniTn7T-Gm-GW containing a promoterless lacZ, Apr Gmr	This study
pMMB67HE	Broad-host expression vector containing lacIq and an ORF-less ptac promoter, Apr	43
pMMB-rpoS1	pMMB67HE ptac::rpoS, Apr	43
pBBR1-MCS5	Broad-host cloning vector, Gmr	78
pDN53	pBBR1-MCS5 with a 2.7-kb fragment carrying P. aeruginosa PAO1 relA and its native promoter, Gmr	This study

^aAbbreviations used for genetic markers: Str, streptomycin; Tc, tetracycline; Km, kanamycin; Gm, gentamicin; Ap, ampicillin.

Plasmids and vector construction.

Molecular biology procedures were carried out by standard methods, and all primers used in this work are listed in Table S1 in the supplemental material. Newly created plasmid constructs were confirmed by sequencing using M13 or other insert-specific primers (see Table S1 in the supplemental material). An IPTG-inducible rpoS construct and its parent plasmid were obtained from a previous study (43). A relA overexpression construct was created by PCR amplification of a 2.7-kb fragment containing the P. aeruginosa PAO1 relA open reading frame (ORF) and upstream promoter region as previously described (44). This PCR fragment was then inserted by HindIII and SpeI restriction and ligation into pBBR1-MCS5, creating the vector pDN53 (Table 1).

To overexpress katA, we generated a chromosomally integrated miniTn7 construct using Multisite Gateway technology (Invitrogen). The PAO1 katA ORF was amplified by PCR (using primers katA-F and katA-R), and this was recombined into pDONR221P5P2 using BP clonase (Invitrogen), generating the entry vector pJJH173. Next, araC-pBAD was amplified by PCR from pUT-miniTn5Pro (45) (using primers araC-pBAD-F and araC-pBAD-R) and recombined into pDONR221P1P5r, creating pJJH187. Subsequently, the entry vectors pJJH173 and pJJH187 were recombined with the destination vector pUC18-miniTn7T-Gm-GW (46) using LR Clonase II Plus (Invitrogen) to generate an arabinose-inducible katA overexpression construct (pMK297). Lastly, this miniTn7 construct was integrated into the PAO1 wild-type and ΔSR strains, creating MK298 and MK300, respectively. As a vector control, the miniTn7 from pUC18-miniTn7T-Gm was integrated into the PAO1 wild-type and ΔSR strains, creating MK318 and MK320, respectively.

We constructed transcriptional promoter-reporter fusions for katA and katB using MultiSite Gateway technology. The PAO1 katA and katB promoter regions were amplified by PCR (using primer pair pkatA-F and pkatA-R and primer pair pkatB-F and pkatB-R, respectively), and these were recombined with BP Clonase II into pDONR221P1P5r, creating pJJH167 and pJJH168, respectively. Next, a DNA fragment containing lacZ along with the adjacent RNase III and ribosomal binding site was amplified by PCR from mini-CTX-lacZ (47) (using primers lacZ-F and lacZ-R), and this was recombined into pDONR221P5P2 using BP clonase II (Invitrogen), generating the entry vector pMK314. Subsequently, each of the vectors pJJH167 and pJJH168 was recombined with pMK314 and the destination vector pUC18-miniTn7T-Gm-GW using LR Clonase II Plus (Invitrogen), generating the promoter-reporter constructs pMK365 and pMK361, respectively. The miniTn7 from pMK365 was integrated into the PAO1 wild-type and ΔSR strains, creating MK325 and MK327, respectively. Repeating this procedure with pMK361 generated strains MK361 and MK363, respectively.

A promoterless lacZ miniTn7 vector was also created as a control for promoter-reporter measurements. Here, a 100-bp sequence containing a multiple-cloning site was amplified by PCR from mini-CTX2 (48) (using primers PC-F and PC-R), and this was recombined with pDONRP1P5r using BP Clonase II, generating pMK71. Subsequently, pMK71 and pMK314 were recombined with pUC18-miniTn7T-Gm-GW to generate the promoterless lacZ fusion pMK339. Finally, the miniTn7 from pMK339 was integrated into the PAO1 wild-type and ΔSR strains, creating MK331 and MK335, respectively.

H₂O₂ challenge.

For H₂O₂ killing of planktonic bacteria, 2.5 × 10⁶/ml stationary-phase planktonic cells were incubated with 3 mM H₂O₂ for 2 h in LB medium at 37°C with shaking at 250 rpm. To induce katB-lacZ expression in planktonic bacteria, cells were grown to mid-log phase (OD₆₀₀ of 0.5) and stimulated with 0.2 mM H₂O₂ pulses every 10 min for 60 min at 37°C (for a final added concentration of 1.2 mM H₂O₂). This sublethal stimulation effectively induces katB expression without significant bacterial killing (data not shown). For H₂O₂ killing of biofilm bacteria, colony biofilms were grown on 1.5% agar plates with dilute LB medium (25% or 5 g/liter). After 24 h of growth at 37°C, biofilms were statically immersed in 2 ml of 25% LB medium and challenged with 150 mM H₂O₂ pulses every 10 min for 30 min at 37°C (for a final added concentration of 450 mM H₂O₂). To induce katB-lacZ expression in biofilm bacteria, cells were grown on 1.5% agar plates with dilute LB medium (25% or 5 g/liter). After 24 h of growth at 37°C, biofilms were statically immersed in 1.5 ml of 25% LB medium and challenged with 5 mM H₂O₂ pulses every 10 min for 60 min at 37°C (for a final added concentration of 30 mM H₂O₂). Repeated pulses of H₂O₂ were used to approximate a continuous challenge. To neutralize any remaining H₂O₂, 0.2% sodium thiosulfate was added at the end of all H₂O₂ killing assays. Biofilm bacteria were resuspended in phosphate-buffered saline (PBS) using a vortex mixer. Viable cell counting was carried out using standard microdilution techniques for both planktonic and biofilm bacteria. The “CFU reduction” was defined as the difference in viable counts between untreated and treated cultures.

Antibiotic challenge.

Colony biofilms were grown as described above on LB agar plates with 1% l-arabinose for 24 h and then transferred onto LB agar plates with 1% l-arabinose and colistin at 300 μg/ml (24 h of incubation) or ofloxacin at 20 μg/ml (4 h of incubation) for incubation at 37°C. Control LB agar plates contained 1% l-arabinose without antibiotics. To enumerate viable cells after antibiotic challenge, biofilm bacteria were resuspended in PBS using a vortex mixer and then plated for viable cell counting as described above.

Catalase activity assays.

Catalase activity was measured using a catalase activity kit (Sigma) with minor modifications according to the supplier (49). Briefly, cell pellets from biofilms or planktonic cultures were washed three times in cold 50 mM potassium phosphate buffer (pH 7.0). Cells were lysed on ice by sonication (six 30-s pulses; amplitude, 20%). The total protein in the lysates was determined by the Bradford assay (Bio-Rad) using a bovine serum albumin standard curve. Catalase activity was measured spectrophotometrically at 240 nm. Aliquots (100 μl) of cell lysate containing 10 μg of protein were added to the assay solution. One unit of catalase activity was defined as the decomposition of 1 μmol · min⁻¹ of H₂O₂. In order to pool results from biological replicates, assay results were calculated as relative catalase activity normalized to the wild-type strain.

Measurement of intracellular ROS levels.

Intracellular levels of ROS were measured using 2′,7′-dichlorodihydrofluorescein (DCFH₂) or hydroxyphenyl fluorescein (HPF) (Molecular Probes-Invitrogen). Cells were grown in biofilms or to stationary phase in planktonic cultures as described above, washed, and resuspended in 1 ml of PBS by vortex mixing. HPF (5 μM) or DCFH₂ (10 μM) was added, and the cells were incubated for 20 min at 37°C in the dark. Cells were then washed and resuspended in PBS, and fluorescence was measured at 485-nm excitation and 535-nm emission (PerkinElmer Victor 2030 Explorer Multilabel Reader). The fluorescence values were normalized to the OD₆₀₀ of each sample. In order to pool results from biological replicates, assay results were calculated as relative fluorescence normalized to the wild-type strain.

β-Galactosidase assays.

β-Galactosidase (LacZ) activity was assayed as described by Zhang and Bremer (50). The background activity for promoter-reporter fusions was corrected by measuring the β-galactosidase activity in strains with a promoterless control construct. This baseline activity was subtracted from all readings, and the corrected β-galactosidase activities were expressed as Miller units.

Statistical analysis.

Statistical differences between two means (with equal variance values) were determined using Student t tests. A Bonferroni correction was used to determine statistical significance when multiple comparisons were performed.

RESULTS

Inactivation of the stringent response (SR) decreases tolerance to hydrogen peroxide in planktonic and biofilm bacteria.

In planktonic culture, the ΔSR mutant was significantly more susceptible to H₂O₂ killing than wild-type bacteria. H₂O₂ caused a 3 log₁₀ reduction in viable ΔSR mutant bacteria compared to ∼0.5 log₁₀ in the wild type (Fig. 1). We also measured the susceptibility of ΔSR mutant biofilms to H₂O₂ killing. Since biofilm bacteria are characterized by a high tolerance of oxidants, we used a 150-fold-higher H₂O₂ concentration to challenge biofilm bacteria (three 150 mM pulses for biofilm cells versus a single 3 mM treatment for planktonic cells), and this caused a reduction in wild-type biofilm viable counts of only 1 log₁₀. Inactivation of the SR led to a massive increase in oxidant killing, with a 6 log₁₀ reduction in viable counts (Fig. 1), suggesting that the SR is critical to biofilm H₂O₂ tolerance. Complementation of the relA and spoT genes restored oxidant tolerance in both planktonic and biofilm cultures, confirming that the increased oxidant susceptibility was due to deletion of these genes (Fig. 1).

Fig 1.

The ΔSR mutant is more susceptible to H₂O₂ killing than the wild-type strain in planktonic and biofilm cultures. Levels of susceptibility to H₂O₂ killing in the wild-type strain, ΔSR mutant, and +SR mutant (complemented strain) are presented as “CFU reduction,” which represents the difference in viable counts between untreated and treated cultures. Stationary-phase planktonic cells grown in LB medium for 16 h were diluted to 2.5 × 10⁶/ml and treated with 3 mM H₂O₂ for 2 h at 37°C. Biofilm cells grown on 25% LB agar were treated with three pulses of 150 mM H₂O₂ every 10 min for 30 min at 37°C. Error bars represent standard deviations (SD). *, P ≤ 0.0001 (versus wild type) (n ≥ 6).

The SR is required for full catalase activity in planktonic and biofilm growth.

Catalases dismutate H₂O₂ to water and O₂ and are potent scavengers of hydrogen peroxide. Given that the ΔSR mutant was highly susceptible to H₂O₂, we next investigated whether it had impaired catalase activity. As shown in Fig. 2 and in Fig S1 in the supplemental material, both planktonic and biofilm stationary-phase ΔSR cells exhibited only ∼35% catalase activity compared to the wild-type strain. Complementation of the relA and spoT genes also rescued catalase activity. As expected, the catalase activity was undetectable in the katA mutant, as KatA is the primary constitutive catalase expressed in the absence of H₂O₂ induction (19, 26). On the other hand, the katB mutant retained wild-type levels of catalase activity under planktonic and biofilm growth conditions. This was not surprising, since katB expression is known to be induced only under conditions of exogenous H₂O₂ stress. Of note, neither katA nor katB mutants had reduced viability in stationary-phase planktonic cultures or biofilm (see Fig. S1 in the supplemental material).

Fig 2.

Catalase activity is reduced in the ΔSR mutant. The relative catalase activities of wild-type, ΔSR, +SR (complemented strain), katA, and katB mutant strains were measured in planktonic and biofilm cultures grown in LB medium to the stationary phase (16 h for planktonic, 24 h for biofilm). Catalase activity in the mutants was normalized to wild-type levels under the same growth conditions. ND, not detected. Error bars represent SD. *, P ≤ 0.01 (versus wild type) (n ≥ 3).

Starvation induces catalase activity through (p)ppGpp signaling.

Since SR inactivation led to a decrease in catalase activity, we hypothesized that SR activation would have the opposite effect. In order to induce the SR in the wild type, we added serine hydroxamate (SHX). This serine analog causes serine starvation in bacteria grown in minimal medium, resulting in rapid accumulation of (p)ppGpp and growth arrest (39). It is worth noting that SHX treatment causes similar levels of growth arrest in both wild-type and ΔSR mutant strains but (p)ppGpp accumulation occurs only in the wild-type strain (33). Following addition of SHX to log-phase planktonic bacteria, catalase levels increased in the wild-type strain but not in the ΔSR mutant (Fig. 3). Since wild-type cells, but not ΔSR cells, synthesize (p)ppGpp in response to SHX, this suggests that SR induction upregulated catalase activity and that this requires (p)ppGpp signaling. Furthermore, catalase levels were also low in the ΔSR mutant not treated with SHX, indicating that (p)ppGpp is required for basal catalase activity. Surprisingly, catalase activity decreased in untreated wild-type bacteria upon entry into the stationary phase in minimal medium (Fig. 3; see also Fig. S2 in the supplemental material) but not in LB medium (see Fig. S2 in the supplemental material), a finding previously reported in E. coli (51, 52). The reason for the decreased catalase activity in M9 medium remains unclear.

Fig 3.

The SR and (p)ppGpp signaling upregulate catalase activity. Relative levels of catalase activity were measured during exponential- and early stationary-phase planktonic growth in M9 minimal medium in wild-type (■) and ΔSR (▲) bacteria without SHX treatment (solid line) or with 500 μM SHX (dashed line). Time t = 3 h corresponds to an OD₆₀₀ of 0.5 (mid-exponential phase). Catalase activity was normalized to wild-type levels at t = 3 h under the control conditions. Error bars represent SD. *, P ≤ 0.05 (versus wild-type control) (n = 3).

Inactivation of the SR causes increased endogenous ROS.

During aerobic growth, ROS are generated within cells through autoxidation of flavoenzymes and redox-active metabolites, such as phenazines, and from the partial reduction of O₂ during aerobic respiration. ROS scavengers, including catalases, are required to prevent excessive accumulation of these potentially toxic molecules. Given that catalase activity is reduced in the ΔSR mutant, we asked whether this was associated with increased endogenous intracellular ROS. Using DCFH₂, a nonspecific probe that detects H₂O₂ in the presence of redox-active iron (53), and HPF, a probe with relative specificity for hydroxyl radicals (54), we found that SR inactivation significantly increased ROS levels in both planktonic stationary-phase and biofilm cells (Fig. 4).

Fig 4.

SR inactivation increases endogenous ROS formation. Relative intracellular levels of ROS were measured in stationary-phase planktonic (16 h) and biofilm (24 h) bacteria grown in LB medium. Cells were stained with DCFH₂ (A) and HPF (B). The relative levels of fluorescence (excitation, 485 nm; emission, 535 nm) were normalized to wild-type levels under the same conditions. The data of at least three independent experiments were pooled, and means are shown. Error bars represent standard errors of the means (SEM). *, P ≤ 0.001; **, P ≤ 0.01 (versus wild type).

The role of RpoS in hydrogen peroxide tolerance.

Previous studies have reported that the stationary-phase alternative sigma factor RpoS is required for H₂O₂ resistance in P. aeruginosa (28, 30). Since the SR is required for full RpoS expression (44) and can indirectly control gene expression by modulating the affinity of sigma factors (55), we explored the contribution of RpoS as an intermediary regulator of catalase activity and H₂O₂ tolerance. Both the ΔSR and rpoS mutants exhibited only 35% of wild-type catalase activity (Fig. 5A), and this was further reduced to 15% in the ΔSR rpoS triple mutant. Although the levels of H₂O₂ killing were comparable in the ΔSR and rpoS mutants, they were also further increased in the ΔSR rpoS triple mutant (Fig. 5B). Together, these results suggest that the SR exerts its effect partially through RpoS but that RpoS-independent mechanisms are also involved. Interestingly, overexpression of RpoS did not restore catalase activity in the ΔSR mutant (see Fig. S3A in the supplemental material), and neither did overexpression of RelA in the rpoS mutant (see Fig. S3B in the supplemental material). This supports the notion that the SR and RpoS are highly interdependent, as neither factor alone is sufficient to overcome a defect in the other.

Fig 5.

Catalase activity and H₂O₂ killing in rpoS mutants. (A) Relative catalase activities of wild-type, ΔSR, rpoS, and ΔSR rpoS triple-mutant strains grown planktonically in LB medium to the stationary phase (16 h). Error bars represent SD. *, P ≤ 0.0005 (versus wild type); **, P ≤ 0.005 (versus the ΔSR rpoS strain) (n ≥ 3). (B) Susceptibility of wild-type, ΔSR, rpoS, and ΔSR rpoS triple-mutant strains to H₂O₂ killing. Cells were grown as stationary-phase planktonic cells, diluted to 2.5 ×10⁶/ml, and then challenged with 3 mM H₂O₂ for 2 h in LB medium at 37°C. Error bars represent SD. *, P ≤ 0.001 (versus wild type); **, P ≤ 0.01 (versus ΔSR rpoS) (n ≥ 3).

The SR regulates katA and katB gene expression levels.

We fused the katA and katB promoters to lacZ to measure transcription of the katA and katB genes. Consistent with previous studies (18, 26), katA-LacZ activity was detected at 3, 8, and 16 h of planktonic growth, with the highest levels during early stationary phase (8 h). In contrast, katA-LacZ activity in the ΔSR mutant minimally increased during the stationary phase and was lower than wild-type levels at both 8 h and 16 h (Fig. 6A). katA expression was also reduced in biofilms of the ΔSR mutant (Fig. 6B). Since katB expression is OxyR dependent and is induced only by H₂O₂ stress (24), wild-type katB-LacZ activity was undetectable in the absence of exogenous H₂O₂ and increased after H₂O₂ challenge in both planktonic and biofilm bacteria (Fig. 6C and D). In contrast, both the noninduced and H₂O₂-induced katB-LacZ activities of the ΔSR mutant remained undetectable. These results thus revealed that the SR is involved in the transcriptional control of both growth-phase-dependent katA expression and H₂O₂-inducible katB expression.

Fig 6.

The SR regulates katA and katB gene expression levels. (A) β-Galactosidase activity of katA-lacZ reporter in wild-type (■) and ΔSR mutant (▲) cells during planktonic growth. Cultures were started at t = 0 h by diluting an overnight culture to an OD₆₀₀ of 0.01. β-Galactosidase activity was measured at t = 3 h (mid-exponential phase), t = 8 h (early stationary phase), and t = 16 h (late stationary phase) during growth in LB medium at 37°C. Error bars represent SD. ND, not detected. *, P ≤ 0.05; **, P ≤ 0.001 (versus wild type) (n ≥ 3). (B) β-Galactosidase activity of katA-lacZ reporter in wild-type and ΔSR mutant cells during biofilm growth. β-Galactosidase activity was measured after 24 h of growth on LB agar at 37°C. *, P ≤ 0.001 (versus wild type) (n ≥ 6). (C) β-Galactosidase activity of katB-lacZ reporter in wild-type and ΔSR planktonic cells with and without H₂O₂ challenge. Cells grown to mid-log phase (OD₆₀₀ = 0.5) in LB medium were challenged with 0.2 mM H₂O₂ pulses every 10 min for 60 min at 37°C. β-Galactosidase activity was measured after H₂O₂ challenge or at the same time in the untreated samples. Error bars represent SD. *, P ≤ 0.0001 (versus untreated wild type) (n ≥ 3). (D) β-Galactosidase activity of katB-lacZ reporter in wild-type and ΔSR biofilm cells with and without H₂O₂ challenge. Biofilm cells grown on 25% LB medium for 24 h were challenged with 5 mM H₂O₂ pulses every 10 min for 60 min at 37°C (for a final added concentration of 30 mM H₂O₂). β-Galactosidase activity was measured after H₂O₂ challenge or at the same time in the untreated samples. Error bars represent SD. *, P ≤ 0.001 (versus untreated wild type) (n ≥ 3).

KatA overexpression restores full H₂O₂ tolerance and partial antibiotic tolerance in the ΔSR mutant.

In a recent study, we hypothesized that impaired oxidative defenses in the ΔSR mutant led to reduced antibiotic tolerance, since antibiotic killing may be ROS mediated (33). To determine the contribution of catalases to antibiotic tolerance in P. aeruginosa, we first measured the ofloxacin tolerance of catalase mutants. The katA and katB single mutants showed a modest increase in susceptibility to ofloxacin compared to the wild type (Fig. 7), suggesting that each catalase provides only a small contribution to antibiotic tolerance. As previously observed, the ΔSR mutant was highly susceptible to ofloxacin. Next, we asked whether KatA overexpression could protect the ΔSR mutant against antibiotic killing. We generated a pBAD-katA construct with arabinose-inducible katA expression. Overexpression of KatA increased catalase activity to similar levels in the wild-type and ΔSR mutant strains (Fig. 8A) and fully restored H₂O₂ tolerance to the ΔSR mutant (Fig. 8B). Furthermore, KatA overexpression reduced killing of ΔSR mutant biofilms by colistin (3 versus 7 log₁₀ CFU reduction) and ofloxacin (4 versus 6 log₁₀ CFU reduction) (Fig. 8C and D) but no significant difference was observed in the wild-type background. Thus, our results indicate that wild-type catalase activity plays a vital role in protecting bacteria against antibiotic killing and KatA overexpression partially restores antibiotic tolerance in the ΔSR mutant.

Fig 7.

The ΔSR mutant is more susceptible to ofloxacin killing. Bacteria were grown as biofilms on LB agar for 24 h and then challenged with 20 μg/ml ofloxacin for 4 h at 37°C. CFU reduction values represent differences in viable counts between treated and untreated biofilms. Error bars represent SD. *, P ≤ 0.0001 (versus wild type) (n ≥ 6).

Fig 8.

KatA overexpression rescues H₂O₂ and antibiotic tolerance in the ΔSR mutant. (A) Catalase activity in wild-type and ΔSR cells containing a pBAD-katA construct or vector control. Cells were grown planktonically to the stationary phase (16 h) in LB medium with 1% l-arabinose. Error bars represent SD. *, P ≤ 0.005 (versus wild type) (n = 3). (B) Susceptibility to H₂O₂ killing in wild-type and ΔSR cells containing a pBAD-katA construct or vector control. Cells were grown planktonically to the stationary phase in LB medium with 1% l-arabinose for 16 h and then challenged with 3 mM H₂O₂ for 2 h. Error bars represent SD. *, P ≤ 0.0001 (versus ΔSR vector control) (n = 6). (C and D) Antibiotic tolerance of wild-type and ΔSR cells containing a pBAD-katA construct or vector control. Cells were grown as biofilm cells on LB agar with 1% l-arabinose for 24 h and were then challenged with 20 μg/ml ofloxacin for 4 h (C) or 300 μg/ml colistin for 24 h (D) at 37°C. Error bars represent SD. *, P ≤ 0.0001 (versus ΔSR vector control) (n ≥ 6).

DISCUSSION

Bacteria experience nutrient limitation during the stationary phase, in biofilm growth, or within the host during infections. While nutrient limitation can confer oxidative stress tolerance, thus protecting bacteria against host-generated or other exogenous ROS, the underlying mechanisms remain poorly defined (3–6). The SR is a global regulatory response induced by stress and starvation. It can reorganize cellular processes to shut down growth and induce protective mechanisms for stress survival (11, 12). Although the SR is a central stress response, only a few studies have examined its role in oxidative stress. For example, the relA mutant of Enterococcus faecalis is more susceptible to H₂O₂ challenge than the wild type (14, 15), while aerobic growth of the relA mutant of Geobacter sulfurreducens is impaired due to oxidative stress (56). Additionally, we and others have recently shown that SR inactivation in P. aeruginosa led to increased oxidant susceptibility (33, 57). Conversely, (p)ppGpp accumulation has been associated with oxidative stress tolerance in Streptococcus mutans (58), and transcriptomic studies of (p)ppGpp-regulated genes in Rhizobium etli identified several genes involved in oxidative stress resistance (59).

The increased susceptibility of the ΔSR mutant to H₂O₂ challenge led us to suspect that the SR regulates catalases. It is well established that katA expression is continuous and increases upon entry into the stationary phase (18, 26). Our katA-LacZ studies showed that katA expression was reduced during the stationary phase in the ΔSR mutant, paralleled by a reduction in catalase activity. The SR is therefore required for full katA expression and catalase activity under normal physiological conditions (including stationary phase), and this control occurs at the transcriptional level.

Since katA is positively regulated by RpoS and quorum sensing (26, 28), it is possible that the SR controls catalases through these intermediary regulators. Notably, the SR controls rpoS expression in P. aeruginosa: RpoS levels are reduced in a relA mutant and induced upon overexpression of RelA (44, 60). In E. coli, rpoS and RpoS-dependent promoters require (p)ppGpp for full activation (61, 62), and (p)ppGpp prevents RpoS proteolysis via IraP control (63). Here we found that catalase activity and H₂O₂ tolerance were equally reduced in the ΔSR and rpoS mutants. Supporting the idea that RpoS-dependent promoters require (p)ppGpp, overexpression of RpoS was sufficient to restore catalase levels in the rpoS mutant but not the ΔSR mutant strain. Furthermore, overexpression of relA was also not sufficient to restore catalase levels in the rpoS mutant, and thus increasing (p)ppGpp alone cannot overcome the RpoS defect. The ΔSR rpoS triple mutant showed further reductions in both catalase activity and H₂O₂ tolerance, suggesting that the SR likely controlled catalase activity through RpoS-independent as well as RpoS-dependent pathways. For example, the Las and Rhl quorum-sensing systems control katA expression, and the SR has been linked to quorum-sensing activation (44). These quorum-sensing systems may therefore mediate SR control of catalases, but this question is further complicated by the crosstalk between the quorum-sensing systems and RpoS (64). Further studies are required to elucidate this complex regulatory network.

Consistent with previous reports (17, 19, 22), we found that KatA is the dominant catalase expressed in the absence of exogenous H₂O₂ stress and is critical to H₂O₂ tolerance in biofilms (65). Reduced katA expression is therefore likely responsible for the H₂O₂ susceptibility of ΔSR mutant biofilms. In contrast, katB is tightly controlled by OxyR and is expressed only upon exogenous H₂O₂ challenge (Fig. 5B). OxyR is a redox-sensitive LysR-type transcriptional activator that is regulated at the posttranslational level by oxidation of its conserved cysteine residues (66). In its oxidized form, OxyR activates katB expression (and that of other P. aeruginosa genes, including ahpB, ahpC, and katA) by binding to their −35 OxyR promoter binding site. Since katB gene expression is also abrogated in the ΔSR mutant, the SR may be involved in OxyR-mediated control of katB expression either directly by affecting transcription or indirectly by modifying the redox state or expression of OxyR.

Since SR inactivation reduced catalase activity, we hypothesized that SR activation would upregulate this activity. Upon (p)ppGpp accumulation caused by SHX-mediated serine starvation, catalase activity increased in the wild-type strain but not in the ΔSR mutant. Thus, nutrient starvation can induce antioxidant defenses via (p)ppGpp signaling. SHX-mediated induction of the SR requires the use of minimal medium. Curiously, we also observed that the catalase activity of the wild type decreased in M9 minimal medium in the absence of SHX but not in LB rich medium. KatA and KatB are both heme-containing catalases requiring iron for activity (67). As iron limitation can significantly reduce catalase activity (68), this may explain the decreased wild-type catalase levels in M9 minimal medium. Furthermore, it has also been reported that catalase activity in E. coli decreases during growth in M9 medium, possibly due to glucose-mediated catabolite repression (51, 52). Whether catabolite repression of catalases also occurs in P. aeruginosa remains unknown. The mechanism underlying the reduced catalase activity in M9 minimal medium is likely multifactorial, and the exact contribution of iron limitation requires further investigation.

Recent studies suggest that bactericidal antibiotics kill bacteria through oxidative-stress-induced damage, in particular through the formation of highly reactive hydroxyl radicals (34–36, 69). Consistent with this model, the E. coli oxyR mutant is more susceptible to aminoglycoside killing (70). Inactivation of katG, katE, and ahpCF, responsible for H₂O₂ scavenging in E. coli, also significantly increases ciprofloxacin killing (69). As shown in our recent report as well as this current study, inactivation of the SR causes increased endogenous ROS (33). We hypothesized that this excessive oxidative burden was due to impaired antioxidant defenses (including catalases) and increased levels of pro-oxidant 4-hydroxyl-2-alkylquinolone (HAQ) molecules and showed that the increased endogenous oxidative burden enhanced antibiotic killing in a dose-dependent manner (33). As catalases are the primary H₂O₂ scavengers, we expected intracellular levels of H₂O₂ and possibly of hydroxyl radicals (generated through the Fenton reaction) to be elevated in the katA or katB mutants. Although the HPF signal was modestly increased in the ΔkatA mutant during planktonic growth, the DCFH₂ fluorescence was not increased in either mutant but in fact slightly decreased. This may have been due to compensatory mechanisms, such as H₂O₂ scavenging by other catalases or alkyl hydroperoxidase reductases (24). For example, AhpCF has catalase-like activity and is required for optimal protection against H₂O₂ in P. aeruginosa, particularly in biofilms (25). Taken together, these results suggest that there are multiple defects in the ΔSR mutant leading to increased endogenous ROS, as the antioxidant defenses are highly redundant.

We also recently demonstrated that SHX-mediated starvation conferred ofloxacin tolerance to the wild-type but not to the ΔSR mutant (33) and now show that it induced catalase activity. Extrapolating from this model, we thus predicted that increasing antioxidant defenses would confer antibiotic tolerance. KatA overexpression led to increased catalase activity and restored wild-type H₂O₂ tolerance to the ΔSR mutant. Most interesting, KatA overexpression was also sufficient to significantly rescue the ΔSR mutant from killing by ofloxacin and colistin, two bactericidal antibiotics with different targets. While disrupting a single catalase gene caused only a modest impairment in ofloxacin tolerance, catalases are sufficient to protect cells against ROS-mediated killing upon both H₂O₂ and bactericidal challenges.

In conclusion, our work revealed that the SR regulates catalases, likely through a complex interplay of regulators, including RpoS. As shown in Fig. 9, our results fit within a proposed model where H₂O₂ and antibiotic tolerance result from a balance of prooxidant stress and antioxidant defenses. Through the regulation of catalases as well as other mechanisms, the SR modulates this equilibrium. Interestingly, nutrient starvation is associated with increased antioxidant defenses in several eukaryotic species, such as yeast, fish, and plants (71–74). Although cellular responses to nutrient limitation are clearly very different between eukaryotic and prokaryotic organisms, starvation may be a universal stress signal that triggers protective antioxidant defense mechanisms necessary for long-term survival. Antioxidant resistance is required for bacterial virulence and persistence in vivo (17, 75–77) as well as for survival of antibiotic treatments. While the notion of targeting bacterial catalases directly may be intriguing, this may be challenging given the redundancy of ROS scavenging systems. On the other hand, a broader targeting of antioxidant defenses or the SR may be a promising approach to compromising key mechanisms of bacterial pathogenicity (78).

Fig 9.

Proposed model for the role of the stringent response in H₂O₂ and antibiotic tolerance. The stringent response mediates antibiotic tolerance by coordinating a balance of antioxidant defenses and pro-oxidant metabolites. These mechanisms include direct or indirect regulation of catalases as well as 4-hydroxy-2-alkylquinoline molecules (HAQ).