Nonnative Disulfide Bond Formation Activates the σ³²-Dependent Heat Shock Response in Escherichia coli

Abstract

Formation of nonnative disulfide bonds in the cytoplasm, so-called disulfide stress, is an integral component of oxidative stress. Quantification of the extent of disulfide bond formation in the cytoplasm of Escherichia coli revealed that disulfide stress is associated with oxidative stress caused by hydrogen peroxide, paraquat, and cadmium. To separate the impact of disulfide bond formation from unrelated effects of these oxidative stressors in subsequent experiments, we worked with two complementary approaches. We triggered disulfide stress either chemically by diamide treatment of cells or genetically in a mutant strain lacking the major disulfide-reducing systems TrxB and Gor. Studying the proteomic response of E. coli exposed to disulfide stress, we found that intracellular disulfide bond formation is a particularly strong inducer of the heat shock response. Real-time quantitative PCR experiments showed that disulfide stress induces the heat shock response in E. coli σ³² dependently. However, unlike heat shock treatment, which induces these genes transiently, transcripts of σ³²-dependent genes accumulated over time in disulfide stress-treated cells. Analyzing the stability of σ³², we found that this constant induction can be attributed to an increase of the half-life of σ³² upon disulfide stress. This is concomitant with aggregation of E. coli proteins treated with diamide. We conclude that oxidative stress triggers the heat shock response in E. coli σ³² dependently. The component of oxidative stress responsible for the induction of heat shock genes is disulfide stress. Nonnative disulfide bond formation in the cytoplasm causes protein unfolding. This stabilizes σ³² by preventing its DnaK- and FtsH-dependent degradation.

INTRODUCTION

Oxidative stress is defined as an imbalance between the generation of reactive oxygen species (ROS), such as superoxide anion, hydrogen peroxide, and hydroxyl radicals, and their detoxification by cellular antioxidant systems. ROS have the potential to attack virtually any cellular macromolecule, including DNA, lipids, and proteins. To counteract the damaging effects, microorganisms respond to oxidative stress in various ways. The classical oxidative stress response is regulated by transcription factors, such as OxyR from Escherichia coli, OhrR from Bacillus subtilis, and Yap1 from Saccharomyces cerevisiae, which all act as reversible redox switches (1–3). Activated by the oxidation of highly conserved cysteines, these regulators induce the expression of antioxidant enzymes, such as peroxiredoxins, glutaredoxins, and thioredoxins.

A separate class of transcription factors which specifically sense disulfide stress has been reported in a number of organisms. These regulators are activated when the thiol-disulfide balance is perturbed, resulting in the formation of inter- and intramolecular disulfides within proteins or thiolation of proteinogenic amino acids with small-molecule thiols such as glutathione. This condition may be caused by an accumulation of naturally occurring electrophiles, such as quinones, and does not necessarily involve radical formation or the presence of ROS (4). In Streptomyces coelicolor, a distinct regulatory system consisting of RsrA and σ^R responds to disulfide stress, while in B. subtilis, the redox-sensing proteins Spx and YodB control parts of the disulfide stress response (5–8). As many as 275 genes have recently been shown to be under the control of Spx (9). In Pseudomonas aeruginosa, the LysR-type transcriptional regulator MexT confers resistance to disulfide stress by upregulating an efflux system for electrophilic compounds (10).

In many organisms, the response to ROS not only triggers an antioxidant response but also typically involves the upregulation of heat shock genes. In the Gram-positive model organism B. subtilis, it has been shown that the induction of class III heat shock genes is mediated by the disulfide stress-sensing transcription factor Spx (9).

In the Gram-negative model organism E. coli, all major chaperone systems, including DnaK, GroEL, and HtpG, along with other classical heat shock proteins, such as Clp proteases and the small heat shock proteins IbpA and IbpB, are upregulated upon hydrogen peroxide-induced oxidative stress (11). However, no specific disulfide stress-sensing mechanism responsible for the upregulation of these heat shock genes has been identified so far, and it is not well understood how the heat shock response is prompted by oxidative stress in this organism. In E. coli the heat shock response is controlled by the heat shock sigma factor σ³² (for a review, see reference 12). Under nonstress conditions, the concentration of σ³² is very low, and the sigma factor has a very short half-life in vivo (13, 14). Upon exposure of cells to heat shock, the concentration of σ³² increases; this increase involves numerous transcriptional, translational, and posttranslational regulatory mechanisms (15–26).

In the present study, we demonstrate that environmental stressors associated with oxidative stress all cause intracellular disulfide bond formation. Using chemical and genetic tools to separate disulfide stress from other effects of oxidative stress, we discovered that it is this disulfide stress component of oxidative stress that induces the heat shock response in E. coli. Upregulation of heat shock genes in E. coli upon oxidative stress directly depends on the heat shock sigma factor σ³², which is stabilized under disulfide stress conditions. The half-life of σ³² increased from below 1.6 min under control conditions to well over 20 min in disulfide-stressed cells. This stabilization coincides with the disulfide stress-induced aggregation of proteins. This mechanism provides E. coli with a strategy to upregulate the heat shock genes in response to oxidative stress without the need for a specialized disulfide stress-sensing regulator.

MATERIALS AND METHODS

Strains and growth conditions.

All bacterial strains and plasmids used in this study are listed in Table 1.

Table 1.

Bacterial strains and plasmids

Strain or plasmid	Relevant genotype or description	Source or reference
Strains
DHB4	F′ lac-pro lacIq Δ(ara-leu)7697 araD139 ΔlacX74 galE galK rpsL phoR Δ(phoA)PvuII ΔmalF3 thi	Lab strain collection
LL014	DHB4/pAID135	This work
WP778	DHB4 gor-522. . . .mini-Tn10 Tc trxB::Km	43
BB7222	MC4100 ara+	58
BB7224	MC4100 ara+ rpoH::Km zhf::Tn10 suhX401	58
Plasmids
pAID135	PhoAΔ2-22 under the control of the tac promoter	28
pEC5217	pUC18 carrying E. coli rpoH Ap	59
pBO110	pBAD24 carrying E. coli lpxC Ap	37

Alkaline phosphatase assays.

E. coli LL014 (harboring pAID135) was grown aerobically in glucose MOPS (morpholinepropanesulfonic acid) minimal medium (27) containing 40 μg/ml l-leucine and 100 μg/ml thiamine at 37°C until an optical density at 600 nm (OD₆₀₀) of 0.2 was reached. To induce the expression of leaderless alkaline phosphatase from pAID135, IPTG (isopropyl-β-d-thiogalactopyranoside) was added to a final concentration of 5 mM. At an OD₆₀₀ of 0.4, disulfide stress was induced by the addition of the indicated concentrations of various oxidative stress-inducing agents. At specific time points, 900 μl cells was harvested directly onto 100 μl 1 M iodoacetamide solution and the cells were incubated for 20 min on ice to stop further disulfide exchange reactions (28). Cells were pelleted by centrifugation (13,000 × g, 2 min, 4°C) and washed twice in 1 ml ice-cold washing buffer (50 mM NaCl, 10 mM NH₄Cl, 10 mM MgCl₂, 10 mM iodoacetamide, 40 mM MOPS–KOH, pH 7.3). The pellet was resuspended in 100 μl lysis buffer (10 mM EDTA, 2 mg/ml lysozyme, 10 mM iodoacetamide, 20 mM Tris-HCl, pH 8.0) and incubated for 30 min on ice. Three freeze-thaw cycles (dry ice-ethanol, 28°C water bath) were used to completely lyse the cells. The lysate was resuspended in 900 μl resuspension buffer (10 mM MgCl₂, 10 mM ZnCl₂, 1 M Tris-HCl, pH 8.0) and incubated in a water bath at 28°C. One hundred microliters 0.4% para-nitrophenylphosphate (pNPP) solution in 1 M Tris-HCl, pH 8.0, was added, and as soon as the sample turned visibly yellow, the incubation time with pNPP was noted and 100 μl stop solution (1 M K₂HPO₄) was added. Then, the sample was stored for 10 min on ice. After centrifugation (13,000 × g, 2 min, 4°C) the absorption of the supernatant at 420 nm and at 550 nm was determined. One hundred microliters lysis buffer in 900 μl resuspension buffer, 100 μl pNPP solution, and 100 μl stop solution was used as a blank. Alkaline phosphatase activity (in arbitrary units) was calculated according to the following equation (29): 1,000 {[A₄₂₀ − (1.75 × OD₅₅₀)]/(time × OD₆₀₀)} × dilution factor, where time is in seconds.

Disulfide stress treatment.

WP778 was grown in LB medium supplemented with 8 mM dithiothreitol (DTT) at 37°C until an OD₆₀₀ of 0.5 was reached. To induce disulfide stress in WP778, the medium was exchanged by spinning down the cells (5,000 × g, 5 min, 37°C) and resuspension of the cells in the same amount of DTT-free LB medium. Strain DHB4 was grown in LB medium at 37°C. At an OD₆₀₀ of about 0.5, diamide was added to a final concentration of 1 mM. A culture volume corresponding to 2 ml of a culture of an OD₆₀₀ of 1 was harvested by centrifugation (10,000 × g, 10 min, 4°C) immediately before and at several time points after removal of DTT in the case of WP778 or after addition of diamide in the case of DHB4. Cells were washed twice with double-distilled H₂O, resuspended in 550 μl two-dimensional (2D) rehydration buffer (7 M urea, 2 M thiourea, 1% [wt/vol] Serdolit MB-1, 1% [wt/vol] dithiothreitol, 4% [wt/vol] CHAPS {3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulfonate}, and 0.5% [vol/vol] Pharmalyte 3-10), and then lysed by sonication on ice (power setting, 3.75; 10 pulses of 1 s on and 1 s off performed 2 times interrupted by 30 s; 550 Sonic Dismembrator; Fisher Scientific, Pittsburgh, PA).

2D gel electrophoresis.

2D gel electrophoresis was performed using immobilized pH gradient strips, pH 4 to 7 (Amersham Biosciences, Piscataway, NJ), as previously described (30). Gels were stained as described previously (31) and scanned using an Expression 1680 scanner with a transparency unit (Epson America Inc., Long Beach, CA) at a resolution of 200 dots per inch with a 16-bit grayscale. Gel images were analyzed using Delta2D software (Decodon GmbH, Greifswald, Germany).

Determination of diamide concentration in medium.

Strain DHB4 was grown aerobically in glucose MOPS minimal medium (27) containing 40 μg/ml l-leucine and 100 μg/ml thiamine at 37°C. At an OD₆₀₀ of 0.4, diamide was added to a final concentration of 1 mM. To monitor the diamide concentration in the medium, 1.5-ml samples were harvested, cells were spun down (13,000 × g, 2 min, 4°C), and the absorption of the filtered (0.2-μm-pore-size syringe filters) supernatant was determined at a maximum λ (diamide) of 296 nm (ε = 3,000 M⁻¹ cm⁻¹ [32]).

Quantitative real-time PCR (RT-PCR).

Strains BB7222 and BB7224 were grown aerobically at 30°C in LB medium. When an OD₆₀₀ of 0.5 was reached, the cultures were shifted to 43°C to induce heat shock conditions. Disulfide stress was induced by addition of 1 mM diamide to the medium. For RNA extraction, 1.2 ml of culture was harvested on 600 μl ice-cold killing buffer (20 mM NaN₃, 5 mM MgCl₂, 20 mM Tris-HCl, pH 7.5) and the RNA was extracted using an RNeasy minikit (Qiagen, Valencia, CA) according to the protocol provided by the manufacturer.

RNA was subjected to a second DNase I treatment using an Ambion DNA-free kit (Ambion Inc., Austin, TX). The RNA concentration and A₂₆₀/A₂₈₀ as well as A₂₆₀/A₂₃₀ ratios were measured with a NanoDrop spectrophotometer (Thermo Scientific, Wilmington, DE) (see Table S1 in the supplemental material). For cDNA synthesis, a total of 400 ng RNA was incubated for 5 min at 65°C with 1 μl of 10 mM deoxynucleoside triphosphates and 1 μl of 50 μM random hexamers (Invitrogen, Grand Island, NY) in a total volume of 13 μl. A mixture containing 4 μl 5× first-strand buffer, 1 μl 0.1 M DTT, 1 μl RNaseOUT recombinant RNase inhibitor (Invitrogen, Grand Island, NY), and 1 μl SuperScript III reverse transcriptase (Invitrogen, Grand Island, NY) was added, and cDNA synthesis was performed for 60 min at 50°C, followed by enzyme inactivation at 70°C for 15 min. Primer design was performed using Primer3 software (v.0.4.0) (33) with the following parameters: product size range, 180 to 230 bp; primer size, 20 to 27 bp; primer melting temperature, 60 to 64°C.

Real-time PCR was performed in a total volume of 10 μl containing 200 ng of cDNA, 5 μl SYBR GreenER (Invitrogen, Grand Island, NY), and 0.5 μl of each gene-specific primer (100 pmol/μl). The primers used are listed in Table 2. After initial denaturation for 4 min at 95°C, amplification was performed for 40 cycles (30 s at 95°C, 30 s at 62°C, 30 s at 72°C) prior to melting curve analysis. Expression of each gene was analyzed in triplicate using two biological replicates. No-template controls were included for each primer pair. Since amplification efficiencies were between 96% and 102%, data analysis was performed using the 2^−ΔΔCT method, where C_T is the threshold cycle (34). Expression of test genes upon stress treatment was normalized against accD expression (first normalization) and expression before stress treatment (second normalization). accD is a gene that has been used as a suitable control in experiments with E. coli exposed to various stress conditions (35). To test the suitability of accD as a gene for normalization, phoP was used as an alternative normalization gene. An assessment of the expression values of hslR under diamide stress yielded comparable results for both normalization genes (see Fig. S2 in the supplemental material).

Table 2.

Primers used for RT-PCR

Primer	Sequence	Use
hslO_F	TGCCGCAACATGACCAATTACATC	Test gene
hslO_R	TCACCATCAAACTTCAGCGTAGCGGT	Test gene
hslR_F	CTGTTGAGGTTCGACTGGATAAATG	Test gene
hslR_R	GCTGTTCAGTAATCGCCTTTACAAT	Test gene
dnaJ_2_FW	GAAGCTTATGAAGTTCTGACCGACT	Test gene
dnaJ_2_RV	GCTCCATGTTATAGCGTAAATCAGC	Test gene
dnaK_F	ACAGCACCCGTAAGCAGGTTGAAGAA	Test gene
dnaK_R	TGGGCGATTTCCATCAGTTTCTGGGA	Test gene
accD_F	CTGGATTGAACGAATTAAAAGCAAC	Reference gene
accD_F	AGGCTTCCTTCATCTAACAGGCTAT	Reference gene

Crude extract aggregation assay.

Thirty milliliters of LB medium was inoculated with E. coli strain BB7222 or BB7224 and incubated overnight at 37°C with shaking. Overnight cultures were subcultured to 2 liters LB medium and incubated at 37°C with shaking until an OD₆₀₀ of 0.6 to 0.8 was reached. Cells were harvested (10 min, 8,000 rpm, 4°C), washed in 25 ml of ice-cold lysis buffer (50 mM HEPES–KOH, pH 8.0, 175 mM NaCl, 1 mM MgCl₂), and resuspended in 10 ml lysis buffer. Cells were lysed by two passages through a French pressure cell at 1,200 lb/in². Cell debris was removed by centrifugation at 3,000 × g for 5 min. The supernatant was aliquoted and stored at −80°C. Cell-free lysates were thawed on ice, and aggregates were removed by centrifugation for 20 min at 16,100 × g at 4°C. Supernatants were treated with diamide (0 to 100 mM) at 30°C (disulfide stress) or incubated at 43°C for 20 min (heat stress). After centrifugation for 20 min at 16,100 × g at 4°C, 10 μl 5× reducing SDS buffer was added to the supernatant. Pellets were rinsed with 100 μl lysis buffer prior to addition of 50 μl 1× reducing SDS buffer. Proteins were analyzed on Mini-PROTEAN TGX precast gels (Bio-Rad).

In vivo stability of σ³².

The in vivo stability of σ³² and LpxC was analyzed as previously described (36, 37). E. coli MC4100 carrying the rpoH (encoding σ³²) expression plasmid pEC5217 or the lpxC expression plasmid pBO110 was grown in LB medium until the OD₆₀₀ reached 0.5. The culture was split into three subcultures, and expression of σ³² or LpxC was induced by the addition of 1 mM IPTG or 0.025% arabinose, respectively. While one culture was cultivated at 30°C (control), 1 mM diamide (disulfide stress) was added to the second culture after 5 min of IPTG or arabinose induction. The third culture was transferred to 43°C (heat stress) after 20 min of IPTG or arabinose induction. Protein synthesis was stopped in all cultures by adding chloramphenicol to a final concentration of 200 μg/ml after a total of 25 min of IPTG or arabinose addition. Samples of 1 ml were collected 0, 1, 3, 5, 7, 9, 11, 15, and 20 min (σ³²) or 0, 3, 6, 9, 12, 30, 60, and 90 min (LpxC) after chloramphenicol addition. Samples were immediately frozen in liquid nitrogen and thawed on ice. Cells were harvested by centrifugation at 4°C for 5 min at 13,200 rpm. Pellets were resuspended in 40 μl TE buffer (10 mM Tris, 1 mM EDTA, pH 8.0). After addition of 10 μl 5× SDS sample buffer, samples were vigorously vortexed and incubated for 10 min at 95°C. Ten microliters was loaded on denaturing SDS gels. Western blotting was performed using an iBlot system according to the manufacturer's instructions (Life Technologies, Darmstadt, Germany). Immunodetection of RpoH and LpxC was performed using polyclonal rabbit anti-RpoH and anti-LpxC antibodies (Stressgene) and a secondary IRDye 680LT goat anti-rabbit IgG (H+L; Li-cor Biosciences, Bad Homburg, Germany). Visualization of RpoH and LpxC was carried out using an Odyssey infrared imaging system (Li-cor Biosciences, Bad Homburg, Germany). Band intensities were quantified using the Adobe Photoshop CS5 analysis/record measurements function.

RESULTS

Oxidative stress causes disulfide bond formation in the E. coli cytoplasm.

Several stressors, including cadmium, hydrogen peroxide, paraquat, diamide, and heat shock itself, have previously been described to induce the expression of heat shock proteins in E. coli (11, 38, 39). Given the known oxidizing effects of these chemicals, we hypothesized that disulfide stress might be the common denominator that induces the heat shock response. To monitor which of these stressors induces disulfide stress in vivo, we used a system based on leaderless alkaline phosphatase. This system was originally introduced by Derman and coworkers to identify mutations that allow disulfide bond formation in the cytoplasm (28). It uses a variant of alkaline phosphatase termed PhoAΔ2-22 as the reporter enzyme. Wild-type alkaline phosphatase is a periplasmic enzyme that needs two disulfide bonds to gain full catalytic activity (40). The PhoAΔ2-22 variant, however, lacks a signal sequence and is not exported into the periplasm. Therefore, PhoAΔ2-22 can gain activity only if disulfide bonds are formed in the cytoplasm, making it a versatile tool to test the disulfide stress-inducing abilities of oxidative stressors. We observed no significant activation of PhoAΔ2-22 under heat shock conditions (Fig. 1). Low millimolar concentrations of hydrogen peroxide (1 to 4 mM), however, transiently activated PhoAΔ2-22 (up to ∼7-fold induction after 10 min of exposure to 4 mM H₂O₂), indicating that peroxide does induce disulfide bond formation in E. coli. The relatively low level of PhoA activity confirms our previous observation that hydrogen peroxide does not lead to a large excess of disulfide bond formation in cytoplasmic proteins (41). Cadmium (2 ppm Cd²⁺) and paraquat (an inducer of superoxide [42]) stress also caused a low level of activation of PhoAΔ2-22 (∼3- and ∼1.5-fold, respectively). This activation was, however, observed only at later time points (50 to 180 min) (Fig. 1). The most potent activator of PhoAΔ2-22 was diamide. Within minutes of addition of 1 mM diamide, an ∼10-fold gain of alkaline phosphatase activity was observed (Fig. 1 and 2). The alkaline phosphatase activity remained high (up to ∼24-fold induction) for more than 180 min, until diamide was cleared from the medium (Fig. 2). The increase of alkaline phosphatase over time indicates the relatively slow kinetics of alkaline phosphatase oxidation by diamide. Western blots using an antibody against alkaline phosphatase confirmed that the gain of alkaline phosphatase activity was not due to additional induction of the PhoAΔ2-22 protein upon disulfide stress (Fig. 2).

Fig 1.

Disulfide bond formation in the cytoplasm of E. coli determined by leaderless alkaline phosphatase activation. PhoAΔ2-22 activity was measured under different stress conditions. Activity is expressed as a percentage of the activity under control conditions. Note the differences in the scales for the 4 mM H₂O₂- and diamide-treated samples.

Fig 2.

Disulfide stress upon diamide treatment of E. coli cells grown in MOPS minimal medium. (A) The formation of disulfide bonds in the cytoplasm parallels the diamide concentration in the medium. The arrows in the growth curve represent the time points at which samples were taken to measure the diamide concentration in the medium (B) and PhoAΔ2-22 activity (C). (D) The Coomassie-stained protein gel and Western blot with alkaline phosphatase antibody show that the expression of PhoAΔ2-22 (arrow) is not significantly induced upon addition of diamide to the medium. The incubation times with diamide are indicated. The gain of activity can be attributed only to disulfide bond formation. Co, control.

Disulfide stress induces heat shock proteins in E. coli.

Diamide is a chemical agent introduced by Kosower and Kosower (32) which specifically triggers disulfide formation in thiol-containing compounds. With diamide, we now have a suitable tool to induce disulfide stress without primarily causing other oxidation reactions initiated by classical oxidative stressors. To investigate the cellular response of E. coli toward disulfide stress, we analyzed the protein expression pattern of diamide-stressed cell cultures using 2D gels. We observed the induction of at least 15 different proteins within 10 min of diamide addition. Most of these proteins were identified to be σ³²-controlled heat shock proteins like DnaK/J, GroEL/ES, and small heat shock proteins IbpA and IbpB, or oxidative stress proteins, such as SodA, which are expressed in response to OxyR or SoxR activation (Table 3; Fig. 3).

Table 3.

Proteins upregulated more than 1.5-fold in 2 independent experiments 10 min after addition of 1 mM diamide

Protein namea	GenBank accession no.	Functiona	Molecular mass (kDa)	No. of peptides matched	Peptide coverage (%)	Regulator
AhpC	EG11384	Alkyl hydroperoxide reductase, subunit C; reduced by the AhpF subunit	20.8			OxyRc
ClpB	EG10157	Bichaperone with DnaK for protein disaggregation	95.6	30	44	RpoHd
DnaK	EG10241	Hsp70 molecular chaperone	69.1	27	52	RpoHe
Dps	EG11415	Stress response DNA-binding protein with protective role	18.7	7	22	OxyRc
GpmA	EG11699	Phosphoglycerate mutase, 2,3-bisphosphoglycerate dependent	28.6	9	36	Furf
GroEL	EG10599	Chaperonin Cpn60	57.3	3	12	RpoHe
GroES	EG10600	Chaperonin Cpn10	10.4	6	61	RpoHe
GrpE	EG10416	Nucleotide exchange factor for the DnaKJ chaperone	21.8	5	25	RpoHe
HtpG	EG10461	Protein refolding molecular cochaperone Hsp90	71.4	27	45	RpoHe
IbpA	EG11534	Chaperone, HSP20 family	15.8	5	29	RpoHe
IbpB	EG11535	Chaperone, HSP20 family	16.0	5	34	RpoHe
PrlC	EG11441	Oligopeptidase A, zinc metalloprotease	77.2	16	23	RpoHg
SodA	EG10953	Superoxide dismutase, Mn	23.1	9	34	SoxRh
DSI1b	NAi	Spot not identified				NA
DSI2b	NA	Spot not identified				NA

^aAccording to Ecogene (60).

^bNot identified disulfide stress-induced protein spots.

^cFrom reference 11.

^dFrom reference 61.

^eFrom reference 62.

^fFrom reference 63.

^gFrom reference 64.

^hFrom reference 65.

ⁱNA, not available.

Fig 3.

Disulfide stress induces oxidative stress and heat shock genes in E. coli. (A) False-colored overlay of 2D gels of cell extracts of E. coli wild-type (DHB4) cells grown under control conditions (green) and 30 min after addition of 1 mM diamide to the medium (red). Protein spots that are induced upon addition of diamide appear red. Proteins that are upregulated more than 1.5-fold 10 min after addition of diamide are labeled. (B) False-colored overlay of 2D gels of cell extracts of E. coli Δtrx Δgor (WP778) cells grown with DTT in the medium (green) and 30 min after removal of DTT (red). WP778 is a strain that lacks thioredoxin reductase and glutathione oxidoreductase and forms disulfide bonds in the cytoplasm when grown in the absence of DTT. Proteins induced under disulfide stress conditions upon removal of DTT appear red. Proteins upregulated more than 1.5-fold 10 min after addition of diamide are labeled to facilitate comparison. (C) Time course of the induction of selected heat shock and oxidative stress genes upon addition of diamide. Incubation times with diamide and induction factors are indicated.

To investigate whether the observed induction of heat shock genes was specific for diamide or a feature of general disulfide stress, we analyzed the protein expression of E. coli mutant cells which are exposed to intrinsic disulfide stress. These mutants lack both disulfide-reducing systems due to a deletion in thioredoxin reductase TrxB and glutathione oxidoreductase Gor. The ΔtrxB Δgor double mutant is able to grow only in the presence of DTT in the medium. As shown by a leaderless alkaline phosphatase assay by Prinz et al., these cells accumulate disulfide bonds in the cytoplasm and cease to grow upon removal of DTT from the medium (43). The protein expression pattern of this mutant strain upon removal of DTT looked strikingly similar to the one that we observed under diamide-induced disulfide stress in wild-type cells (Fig. 3). All proteins that were identified to be upregulated more than 1.5-fold 10 min after addition of diamide were upregulated in this strain as well.

Disulfide stress induction of heat shock genes is rpoH dependent.

Disulfide stress-mediated induction of heat shock genes has also been observed in the Gram-positive model organism B. subtilis (5) and found to be dependent on the disulfide stress-sensing regulator Spx (9). So far, no specific disulfide stress-sensing regulator is known in E. coli. To investigate whether the heat shock response is directly triggered by oxidative stress in a σ³²-dependent manner or requires a specific, yet to be identified regulator, we compared the expression levels of heat shock genes in diamide-treated wild-type and rpoH deletion strains. To directly quantify the induction of heat shock genes upon disulfide stress on the RNA level, we performed quantitative real-time PCR analyses. We selected the genes dnaK and dnaJ, hslR, and hslO, encoding the DnaK/J chaperones, the ribosome-associated heat shock protein Hsp15, and the redox-regulated chaperone Hsp33, respectively (44–46). These genes are organized in two independent transcriptional units, dnaK/J and hslR/O, and are both known to be strongly induced under heat stress conditions (47, 48). Induction of all four genes was observed upon heat stress as well as upon disulfide stress in wild-type E. coli (Fig. 4). As observed before, expression of the heat shock genes peaked within 5 min of heat treatment (Fig. 4B) (49). In contrast, heat shock gene induction upon diamide stress increased gradually over a time period of at least 20 min (Fig. 4A). Comparative analysis in a mutant strain lacking rpoH revealed no significant upregulation of the four genes in response to either heat shock or diamide treatment. These results indicated that σ³² is both required and sufficient for the induction of the disulfide stress-mediated heat shock response in E. coli (Fig. 4).

Fig 4.

Induction of heat shock genes upon disulfide (A) and heat (B) stress in the wild type (light gray bars) and ΔrpoH mutant (dark gray bars). The expression of the hslR, hslO, dnaJ, and dnaK genes upon disulfide and heat stress depends on σ³², as shown by quantitative real-time PCR. Disulfide stress induces a gradual increase in expression over time, while heat stress causes a transient upregulation within 5 min of stress treatment.

σ³² stabilization upon disulfide stress initiates the heat shock response.

The classical heat shock sigma factor is regulated on three different levels. On the transcriptional level, rpoH, the gene for σ³², is transcribed from at least four different promoters, three of which are dependent on the housekeeping sigma factor σ⁷⁰ and one of which is dependent on σ^E (15, 16). A second regulatory mechanism affects the secondary structure of rpoH mRNA. Upon heat shock, the base-pairing pattern in its mRNA changes due to partial melting. This makes the Shine-Dalgarno sequence accessible, which enhances translation initiation (17–20). Yet another regulation of σ³² happens on the level of σ³² protein stability. Under nonstress conditions, σ³² has a very short half-life due to its inactivation with the DnaK chaperone system (24, 50), which mediates σ³²'s degradation by the protease FtsH (26, 51). Heat shock conditions lead to the unfolding of cellular proteins, which become substrates of DnaK and in turn titrate the DnaK system. This causes the release of σ³² from DnaK and a significant increase in its half-life.

In our disulfide stress experiments, the transcription levels of heat shock genes showed a gradual increase, which was in contrast to the transient induction typical for heat stress. This observation suggested to us a mechanism different from the classical induction of the heat shock response. We hypothesized that σ³² stabilization rather than an RNA thermometer-based posttranscriptional regulation of σ³² might be a more plausible mechanism. To test this hypothesis, we analyzed the steady-state stability of plasmid-encoded σ³² upon disulfide and heat stress. Addition of the protein translation inhibitor chloramphenicol stopped the synthesis of the plasmid-encoded σ³². We then quantified the abundance of σ³² with a specific antibody over time. Under nonstress conditions, σ³² was rapidly degraded with a half-life of 1.5 min (Fig. 5). A shift of E. coli cells from 30°C to 43°C resulted in a considerable increase in the initial amount of σ³². Nevertheless, the half-life of σ³² was slightly shorter than that under control conditions, as observed by others (22). In contrast, disulfide stress caused by diamide and H₂O₂ (Fig. 5; see Fig. S3 in the supplemental material) significantly stabilized σ³², and no degradation of σ³² was observed during the time course of this experiment. A 20-min paraquat treatment, which induced only low levels of nonnative disulfide bond formation, did increase the half-life of σ³² only slightly to 1.9 min (see Fig. S3 in the supplemental material).

Fig 5.

Analysis of σ³² and LpxC stability upon heat and disulfide stress. Cell cultures containing a σ³² or LpxC expression plasmid were grown to an OD₆₀₀ of 0.5, and protein expression was induced by the addition of IPTG or arabinose. Cells were either left untreated (control), treated with heat by shifting the cultures from 30°C to 43°C after 20 min of induction, or treated with 1 mM diamide after 5 min of induction. After a total of 25 min of induction, chloramphenicol (Cm) was added (time zero) and the stability of σ³² (A) and LpxC (B) was analyzed by Western blotting using σ³²- and LpxC-specific antibodies. t_1/2, half-life.

To check if stabilization of σ³² upon disulfide stress involves inhibition of FtsH, we analyzed the stability of the well-characterized FtsH substrate LpxC (Fig. 5B). Under control conditions, LpxC was degraded within 12 min. While heat stress increased the half-life of LpxC approximately 3-fold, diamide stress resulted in complete stabilization. This result indicates that the induction of the heat stress response upon disulfide stress is mediated by a stabilization of σ³² through inhibition of FtsH-dependent degradation.

Disulfide stress leads to protein aggregation in the absence of σ³²-controlled chaperones.

In addition to a direct FtsH inhibition, the observed stabilization of σ³² under disulfide stress conditions could be caused by titration of the DnaK system. Unfolded proteins, which might accumulate due to the formation of nonnative disulfide bonds, may be recognized and bound by the DnaK system. If that was the case, proteins would aggregate in the absence of DnaK under disulfide stress conditions. We therefore analyzed the solubility of proteins in wild-type and rpoH crude extracts after diamide and heat treatment. In these experiments, the solubility of proteins gradually decreased with increasing diamide concentrations in extracts obtained from the ΔrpoH strain (Fig. 6). This aggregation was not observed in wild-type extracts, indicating that chaperones under the control of σ³², such as DnaK, can prevent disulfide stress-induced protein aggregation. Interestingly, many protein bands that were visible in the pellet fraction of extracts from the rpoH strain after diamide treatment were identical to those that were found after heat treatment. This finding supports our model indicating that a similar set of σ³²-controlled chaperones is responsible for the protection of cytoplasmic proteins against heat stress- and disulfide stress-mediated protein aggregation in E. coli.

Fig 6.

Aggregation of proteins from E. coli crude extracts. Extracts from the E. coli wild type (A) and ΔrpoH mutant (B) were treated with diamide (0 to 100 mM) or with heat (H) prior to separation of soluble and insoluble proteins by centrifugation, followed by SDS-PAGE. The solubility of proteins is decreased in the rpoH mutant. Numbers on the left are molecular masses (in kDa). Lanes S, soluble protein; lanes P, insoluble proteins.

DISCUSSION

In this study, we show that disulfide stress, the formation of nonnative disulfide bonds in the cytoplasm, is an integral part of oxidative stress in E. coli. In cells exposed to chemically or genetically induced disulfide stress, we observed a strong upregulation of typical oxidative stress genes under the control of the oxidative stress regulators OxyR and SoxR. These two regulators have cysteines in their redox-sensing centers, which are probably modified under these conditions. OxyR is known to be activated by the formation of an intramolecular disulfide bond (52). Global studies in other organisms have also shown that oxidative stress genes are a major part of the disulfide stress response (5, 53).

In addition, however, we discovered that disulfide stress is a potent inducer of the σ³²-dependent heat shock response in E. coli. In our study, we focused on the mechanistic underpinnings of this particular finding. The heat shock response has been shown to be part of the oxidative stress response in other organisms as well. In B. subtilis, a specific part of the heat shock response is affected by diamide-induced disulfide stress (5). The so-called class III heat shock genes under the control of the global regulator CtsR are upregulated in this organism, while other heat shock genes under the control of the CIRCE element and the alternative sigma factor σ^B are affected to a lesser extent or are not affected at all (5). The induction of class III heat shock genes under these conditions is mediated by the disulfide stress-specific regulator Spx (9).

Unlike in this Gram-positive organism, which uses a specific mechanism that does not rely on an alternative sigma factor, we found that in the Gram-negative bacterium E. coli the heat shock sigma factor σ³² is both required and by itself sufficient for the induction of heat shock genes under disulfide stress. Our experiments show that the σ³²-dependent induction of heat shock genes is the consequence of the stabilization of the sigma factor.

This is presumably caused by inhibition of FtsH-dependent degradation under disulfide stress conditions. If direct inactivation of FtsH by disulfide stress was an evolved mechanism, one would expect highly conserved regulatory cysteines in this protease. However, the two cysteines of E. coli FtsH are not highly conserved. Alternatively, substrate proteins with nonnative disulfide bonds could block the active site of FtsH and thus lead to inactivation of the protease. In addition to direct blockage of FtsH, delivery of substrates to the protease by the DnaK system might be disturbed due to binding of unfolded proteins. DnaK might not be able to resolve nonnative disulfide bonds without the help of reducing systems, leading to a longer-lasting and more effective occupation of DnaK under disulfide stress than under heat shock.

The fundamental difference in the regulation of genes under disulfide stress in these Gram-positive and Gram-negative model organisms is reflected in the regulation of target genes. In B. subtilis hslO, the gene for the redox-regulated chaperone Hsp33 is under direct control of the disulfide stress regulator Spx (9). In E. coli, hslO is under σ³² control. Hsp33 is a chaperone that is activated only upon a posttranslational modification involving the formation of disulfide bonds in the regulatory C-terminal domain (45). Hsp33 needs this disulfide bond formation to gain activity. Hsp33 has been shown to take over the role of the DnaK system under oxidative heat stress conditions (54).

Some alphaproteobacteria possess two or more RpoH homologs which may have overlapping and specialized functions (55, 56). RpoH_II from Rhodobacter sphaeroides has recently been shown to be involved in the defense against singlet oxygen, while it does not seem to play a major role in the heat shock response (57). It is tempting to speculate that in this organism gene duplication allowed the divergent evolution of one RpoH with a broader role into two separate heat shock and oxidative stress regulators.

In E. coli, the upregulation of heat shock genes by oxidative stress is regulated by σ³² alone. This σ³²-dependent regulatory mechanism is distinct from heat shock-mediated gene regulation. Under oxidative stress, the formation of nonnative disulfide bonds in the cytoplasm triggers protein unfolding. The unfolded proteins then bind to the DnaK system, which prevents delivery of σ³² to FtsH. Inactivation of FtsH could additionally lead to the lasting stabilization of σ³².