Global Characterization of Disulfide Stress in Bacillus subtilis

Abstract

We used DNA macroarray and proteome analysis to analyze the regulatory networks in Bacillus subtilis that are affected by disulfide stress. To induce disulfide stress, we used the specific thiol oxidant diamide. After addition of 1 mM diamide to an exponentially growing culture, cell growth stopped until the medium was cleared of diamide. Global analysis of the mRNA expression pattern during growth arrest revealed 350 genes that were induced by disulfide stress by greater than threefold. Strongly induced genes included known oxidative stress genes that are under the control of the global repressor PerR and heat shock genes controlled by the global repressor CtsR. Other genes that were strongly induced encode putative regulators of gene expression and proteins protecting against toxic elements and heavy metals. Many genes were substantially repressed by disulfide stress, among them most of the genes belonging to the negative stringent response. Two-dimensional gels of radioactively labeled protein extracts allowed us to visualize and quantitate the massive changes in the protein expression pattern that occurred in response to disulfide stress. The observed dramatic alteration in the protein pattern reflected the changes found in the transcriptome experiments. The response to disulfide stress seems to be a complex combination of different regulatory networks, indicating that redox-sensing cysteines play a key role in different signaling pathways sensing oxidative stress, heat stress, toxic element stress, and growth inhibition.

Aerobically growing cells need to be able to cope with reactive oxygen species that inevitably develop during incomplete electron transfer in the respiratory chain. They appear to have a great variety of negative effects on biological systems. Reactive oxygen species damage not only lipids and DNA but also proteins. One major effect of reactive oxygen species on proteins is the oxidation of thiols, resulting in disulfide bond formation.

Disulfide bonds play a major role in stabilizing protein structures, and extracellular proteins are particularly dependent on disulfide bonds to stabilize their structure (4). Nonnative disulfide bonds, on the other hand, may lead to protein misfolding. In some proteins, disulfide bridges are also found transiently as part of their catalytic cycle. Examples include ribonucleotide reductase, methionine sulfoxide reductase, alkylhydroperoxide reductase, and arsenate reductase (10). Other proteins possess cysteines as molecular redox switches that control their activity; some examples include the transcriptional factors NF-κB in higher eukaryotes, Yap1p in Saccharomyces cerevisiae, and OxyR and the chaperone Hsp33 in Escherichia coli (13, 14, 17, 22, 28). In Bacillus subtilis, the activity of the global repressor of the peroxide regulon PerR, a metalloprotein, also seems to be modulated by redox-active cysteine residues (8, 12).

Up to now, most studies on the response of bacteria to oxidative stress have concentrated on the reaction to hydrogen peroxide, alkylhydroperoxide, and superoxide-induced stress. Very little is, however, known about disulfide stress, a subcategory of oxidative stress that causes the accumulation of nonnative disulfide bonds in the cytoplasm. The disulfide stress regulon was analyzed in Streptomyces coelicolor with a new sigma factor, sigma R, whose activity is controlled by the redox-regulated anti-sigma factor RsrA (14, 20).

The goal of our study was to get a global insight into changes in mRNA expression and protein synthesis that occur in response to an excessive oxidation of free cellular thiols and thus to enhance our understanding of the oxidative stress response. We have used diamide [diazenedicarboxylic acid bis(N,N-dimethylamide)], a specific oxidant for thiols, to induce disulfide stress in the gram-positive soil bacterium B. subtilis. Diamide reacts with free thiols according the scheme presented in Fig. 1; the end products are a disulfide bond and a hydrazine derivative (see reference 15 for a review). The global response of B. subtilis to disulfide stress was examined by transcriptome analyses on DNA macroarrays and by proteome analyses by using two-dimensional polyacrylamide gel electrophoresis (2D-PAGE).

FIG. 1.

Scheme depicting the diamide reaction with free thiols. The end products are a disulfide bond and a hydrazine derivative.

The global response of B. subtilis to disulfide stress showed a close relationship to hydrogen peroxide-induced oxidative stress and, not surprisingly, an even closer relationship to paraquat-induced oxidative stress; there was also a clear overlap with the heat shock response, the stringent response, and the heavy metal response. Our results here provide a new view on the reaction of bacteria to an alteration of the cellular redox state and show that disulfide stress response is not only a subcategory of oxidative stress but also a complex combination of different regulatory networks.

MATERIALS AND METHODS

Strains and growth conditions.

B. subtilis 168 (trpC2) (1) was cultivated aerobically at 37°C in a synthetic medium described previously (3). Diamide was added to an exponentially growing culture (optical density at 540 nm [OD₅₄₀] of 0.4) of B. subtilis 168 (final concentrations of 0.5, 1, 2, and 10 mM). For survival experiments, dilutions of the culture were plated on Luria-Bertani agar plates. The number of surviving cells was monitored immediately before and 10, 30, and 50 min after the addition of 1 mM diamide. Agar plates were incubated overnight at 37°C.

For measurement of diamide absorption in the medium, 2-ml aliquots of the cultures were harvested on ice and immediately centrifuged for 1 min (13,000 × g at 4°C). Subsequently, the supernatant was filtered. The amount of diamide was measured by absorption at 303 nm.

Preparation of the l-[³⁵S]methionine-labeled cytoplasmic protein fraction.

Exponentially growing B. subtilis cells in synthetic medium were pulse-labeled for 5 min with 50 μCi of l-[³⁵S]methionine per 5 ml of culture when the OD₅₄₀ reached 0.4 immediately before and 10, 30, 50, and 125 min after the addition of 1 mM diamide. Incorporation was stopped by the addition of chloramphenicol to a final concentration of 1 mg/ml, with an excess of unlabeled l-methionine (10 mM final concentration), and by transferring the cultures to ice. Cells were washed twice in 10 mM Tris HCl (pH 7.5)-1 mM EDTA and then sonicated in 800 μl of 10 mM Tris HCl (pH 7.5)-1 mM EDTA-1 mM phenylmethylsulfonyl fluoride. The soluble cell fraction was separated from cell debris and insoluble proteins by centrifugation for 30 min at 12,000 × g at 4°C.

2D-PAGE and data analysis.

2D-PAGE was performed as previously described (5). Aliquots (80 μg for analytical gels and 250 μg for preparative gels) of crude protein extracts were loaded into 8 M urea, 2 M thiourea, 20 mM dithiothreitol, 10 mg of CHAPS {3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulfonate}/ml, and 5.2 μl of Pharmalyte 3-10 (Amersham Biosciences Europe GmbH, Freiburg, Germany)/ml on immobilized pH gradient gel strips covering a pH range of 4 to 7 (Immobiline IPG-Strips; Amersham Biosciences). After fixation in 50% (vol/vol) methanol-7% (vol/vol) acetic acid, preparative gels were stained with Sypro Ruby (Bio-Rad Laboratories GmbH, Munich, Germany) according to the recommendations of the manufacturer. Analytical gels were silver stained as previously described (5), dried on blotting paper, and exposed to phosphor screens (Molecular Dynamics). These autoradiographs were collected with the PhosphorImager SI Scanner (Molecular Dynamics), and data were analyzed by quantitative comparison of the control 2D protein pattern with that of the diamide-treated cells by using Delta2D software (version 2.1; Decodon GmbH, Greifswald, Germany).

Peptide mass fingerprinting.

For identification of proteins from preparative 2D gels, in-gel digestion and protein extraction were performed (D. Becher and K. Büttner, unpublished method). The peptide solutions were applied to a sample template and analyzed by matrix-assisted laser desorption ionization-time of flight-mass spectrometry (MALDI-TOF-MS; Voyager DE-STR; Perseptive Biosystems, Foster City, Calif.). Afterward, peptide mass fingerprints were analyzed by using MS-Fit software [http://prospector.ucsf.edu/ucsfhtml4.0/msfit.htm] (9).

RNA isolation and Northern blotting.

Immediately before and 10, 30, and 50 min after the addition of 1 mM diamide to an exponentially growing culture of B. subtilis, aliquots were removed, rapidly chilled by the addition of an equal volume of ice-cold 20 mM Tris-HCl (pH 7.5)-5 mM MgCl₂-20 mM NaN₃, and centrifuged (3 min, 5,000 × g, 4°C). The supernatant was rapidly removed, and cells were disrupted in liquid nitrogen in a grinding mill (Mikro-Dismembrator S; B. Braun Biotech International GmbH, Melsungen, Germany). The RNA was isolated as previously described (21). Northern blot analysis was performed according to the method of Wetzstein et al. (27). Digoxigenin-labeled probes for clpE, ylnD, gsiB, and trxA were obtained in vitro with T7 RNA polymerase from T7 RNA promoter containing PCR products of the respective genes.

DNA macroarray experiments and data analysis.

DNA macroarray analysis was performed with ³³P-labeled cDNA on Panorama B. subtilis gene arrays (Sigma Genosys, Ltd., Pampisford, United Kingdom) as previously described (11). The hybridized arrays were exposed to phosphor screens. The autoradiograph was analyzed by using the STORM 840 scanner (Molecular Dynamics). Hybridization signals were quantified with ArrayVision software (version 6.0; Imaging Research, St. Catharines, Ontario, Canada). The background was defined as the median of signals surrounding local spot groups. For every gene, an “artifact-reduced” volume of blackness was obtained by replacing pixels in a spot clipping 10% above or below the median blackness of the surrounding pixels. For these artifact-reduced spots, a normalized volume was evaluated by (i) calculating the volume of an average spot by dividing the sum of the volumes of all artifact-reduced spots on the array by the total number of spots and (ii) dividing the volume of the individual artifact-reduced spot by the volume of this average spot. For the control values and each time point, mRNA was prepared from two independent cultivations and then used for independent cDNA synthesis and DNA array hybridization experiments. For each cultivation and every time point, an induction ratio for every individual gene was calculated by dividing the average of the normalized, artifact-reduced volume of each replica spot on the array after treatment with 1 mM diamide by the corresponding volume on the control array. Only genes that showed (i) at least a threefold induction in both experiments and (ii) whose volume was at least two standard deviations above that of the local background level upon diamide treatment were considered to be significantly induced.

The genes were flagged if (i) the volume of at least one replica spot on at least one control array was less than a twofold standard deviation above the background level as “minimal induction value” (as an improper detection of a spot in the control array would possibly lead to a higher volume than the actual volume and thus to a lower induction rate) and (ii) the induction factor differed by more than 100% in both experiments and the value closer to 1 was chosen as a “safe value” for the induction factor. Otherwise, the average of both experiments was stated.

Lists for significantly repressed genes were prepared accordingly (but volumes had to be at least twofold standard deviations above that of the local background on the control array and were flagged as a “minimal repression value” if close to the detection limit upon diamide treatment). This very stringent method should exclude the majority of false-positive results. For the same reason, however, it is likely that not all genes affected by disulfide stress were detected.

RESULTS AND DISCUSSION

Effects of diamide treatment on growth and survival.

We first studied the effects of diamide-induced disulfide stress on the growth of exponentially growing cultures of B. subtilis to find adequate concentrations for the global analyses. Various concentrations of diamide were added to exponentially growing cultures of B. subtilis (OD₅₄₀ of 0.4), and growth was monitored until the stationary phase was reached. Concentrations of up to 2 mM led to growth inhibition for various lengths of time, and a 10 mM concentration of diamide eventually led to cell lysis (Fig. 2).

FIG. 2.

Effect of diamide on growth of B. subtilis 168 in minimal medium. Cells were grown aerobically at 37°C. At an OD₅₄₀ of 0.4, different concentrations of diamide were added to the cultures: •, no diamide (control); ○, 0.5 mM diamide; ▪, 1 mM diamide; □, 2 mM diamide; ▴, 10 mM diamide.

To examine whether the growth inhibition that was observed upon addition of 1 mM diamide causes a reduction in viability, survival experiments were performed. The number of CFU did not alter significantly during the time of growth arrest (Fig. 3).

FIG. 3.

(A) Growth (▪) and survival (bars) of B. subtilis 168 in minimal medium treated with 1 mM diamide. Cells were grown aerobically at 37°C. Diamide was added from a 100 mM aqueous stock solution to the exponential growing culture at an OD₅₄₀ of ca. 0.4. The number of CFU per ml was determined by plating dilutions of the culture on Luria-Bertani agar plates. Agar plates were incubated overnight at 37°C. (B) Growth (▪) of B. subtilis 168 and relative concentration of diamide (bars). Diamide was added from a 100 mM aqueous stock solution to the exponential growing culture at an OD₅₄₀ of ca. 0.4 to a final concentration of 1 mM. Absorption of diamide was measured at λ = 303 nm in the filtered medium.

Diamide reacts with free thiols in a stoichiometric manner, resulting in a disulfide and a hydrazine derivative (Fig. 1) (15). Diamide has an absorption maximum at a λ_max of 303 nm, whereas the hydrazine derivative is transparent at wavelengths greater than 230 nm (15). Therefore, the turnover of diamide could be followed spectrophotometrically by measuring the A₃₀₃ in filtered media. The correlation between growth recovery and diamide concentration in the medium shown in Fig. 2 suggested that the growth of cells is arrested until diamide is used up in the medium. That this is indeed the case is shown in Fig. 3B. Cells recovered growth shortly after the levels of diamide in the media dropped to near background levels.

Transcriptome and proteome analyses.

To characterize the changes in gene expression and protein synthesis, we used the transcriptome and proteome approach. For transcriptome experiments, exponentially growing cells of B. subtilis 168 (OD₅₄₀ of 0.4) were treated with 1 mM diamide. Prior to diamide treatment and 10, 30, and 50 min after the addition of diamide to the medium, cells were harvested and the RNA was isolated. The quality of RNA was analyzed by performing Northern blot analyses for selected genes (see Table 2). ³³P-labeled cDNA was obtained by reverse transcription with an oligonucleotide primer mix composed of specific primers (Sigma Genosys) for each of the 4,101 genes that encode mRNAs in B. subtilis. This radioactive cDNA mixture was then hybridized on DNA macroarrays of spots containing oligonucleotides corresponding to each of the 4,101 open reading frames of B. subtilis (Sigma Genosys). After computer-based comparison of control RNAs and RNAs of diamide-treated cells, induction and repression factors were calculated for each gene. Genes that (i) were in both experiments induced at least threefold at a given time point and that (ii) possessed a volume at least two standard deviations above the local background level on all replica spots on the arrays obtained upon diamide treatment at that given time point were defined as significantly induced for that time point. Significantly repressed genes were defined accordingly but had to be repressed threefold and to possess a volume of at least two standard deviations above the local background level on the arrays obtained immediately before diamide treatment. Lists of genes that are significantly induced and repressed by disulfide stress were prepared, and for each gene in this list quality flags concerning detectability and reproducability were assigned. Upon disulfide stress induction in B. subtilis, 350 genes were induced more than threefold and 578 genes were repressed more than threefold at at least one of the given time points of 10, 30, and 50 min after the addition of diamide in two independent experiments. At all three time points, as many as 141 genes were induced more than threefold, whereas 228 genes were repressed at all time points (Table 1).

TABLE 2.

Comparison of induction data obtained by DNA macroarray, Northern blot, and 2D-PAGE analyses

Gene	Induction ratio at:
	10 min			30 min			50 min
	Macroarray	Northern blot	2D-PAGE	Macroarray	Northern blot	2D-PAGE	Macroarray	Northern blot	2D-PAGE
clpE	46.8c	49.5	∞e	67.5c	144.6	∞	43.2c	67.4	∞
clpPa	11.9		4.7	17.8		6.3	10.7		6.7
cysK	8.3		5.6	15.9		5.1	16.4		5.5
greA	3.1		2.8	4.7		3.5	3.2		3.2
groES	2.6		2.8	1.4c		NDf	0.6c		1.6
gsiB	2.6	3.3		1.1c	4.1		2.0c	3.0
hagb	3.1		2.3	2.8		2.6	1.2		2.1
katA	32.3c,d		8.2	56.3c,d		13.5	14.9c,d		20.5
mrgAb	5.0c		19.2	61.1c		17.7	5.0c		10.8
trxA	9.7	9.4	4.6	27.4	18.9	11.3	26.0	24.9	13.6
ybaL	1.1		7.6	0.7c		3.6	1.1		4.7
ykuQ	0.5		5.6	0.3		3.5	0.3c		4.5
ylnD	2.3	2.6		0.8	0.5		0.5	0.3
yocJa	1.6c,d		4.7	38.4c,d		6.3	42.2c,d		6.7
yodC	6.5d		4.5	12.4d		7.0	11.9d		8.2
yuaE	15.9		∞	27.2		∞	47.8		∞
yvyD	3.0		2.7	2.3		2.8	4.5		3.5

^aClpP and YocJ were not separated by 2D-PAGE but could both be identified by MALDI-TOF-MS.

^bAll protein spots identified in the 2D-PAGE analyses were added.

^cDue to high deviation in both transcriptome experiments, a value closer to 1 was chosen and flagged as a “safe value.”

^dAt least one replica spot in one transcriptome experiment was near the detection limit or below it on the control array; this was flagged as the minimal induction value.

^e∞, No protein spot was detected on the control gel.

^fND, no protein spot was detected on the gel after diamide treatment.

TABLE 1.

Number of genes induced and repressed at least threefold in both transcriptome experiments at different time points after the addition of 1 mM diamide

Time (min) after addition of diamide	No. of:
	Induced genes	Repressed genes
10	196	308
30	236	480
50	290	469
All time points	141	228
At least one time point	350	578

For proteome analyses, cells of B. subtilis 168 were pulse-labeled for 5 min with l-[³⁵S]methionine immediately before and 10, 30, 50, and 125 min after the addition of 1 mM diamide to the exponentially growing culture at an OD₅₄₀ of 0.4. Proteins were extracted and analyzed by 2D-PAGE. Autoradiographs of dried gels were unwarped and analyzed with Delta2D Software (Decodon GmbH, Greifswald, Germany). Approximately 890 protein spots could be resolved by the software on our gels prior to diamide treatment. The massive change in the protein expression pattern was visualized by dual channel imaging (5) (Fig. 4). Identification of protein spots was performed by in-gel digestion and MALDI-TOF-MS from preparative nonradioactive gels and by comparison with the cytosolic protein map of B. subtilis (6). The 2D gels derived from samples at different time points were combined to generate a comprehensive overview of the change of the protein synthesis pattern under conditions of disulfide stress (Fig. 5). Our analysis showed that most vegetative proteins are massively downregulated during disulfide stress and regain their initial level of synthesis only after 125 min, when cells recovered exponential growth (Fig. 4 and 5). The synthesis of 26 major protein spots (possessing at least 0.2% of the overall blackness on the gel) was significantly upregulated during the stress compared to exponential growth before and 125 min after the addition of diamide (Fig. 4 and 5). Supplementary transcriptome and proteome data (i.e., a list of all significantly induced and repressed genes, DNA macroarray raw data, and the combination of all 2D gels) can be downloaded at http://microbio1.biologie.uni-greifswald.de:8080/institute/98. Induction ratios based on Northern, transcriptome, and proteome data of selected genes were compared and shown to be similar in trend (Table 2).

FIG. 4.

Diamide-induced changes in protein expression. (A to D) Dual-channel images of autoradiographs of 2D gels of l-[³⁵S]methionine-labeled intracellular proteins of B. subtilis 168 under control conditions (green) and 10 min (A), 30 min (B), 50 min (C), and 125 min (when the medium was cleared of diamide and exponential cell growth was restored) (D) after treatment with 1 mM diamide (red). Protein spots synthesized only under control conditions appear green; protein spots that were synthesized only upon treatment with diamide are red, whereas protein spots synthesized under both conditions are yellow (5). (E) Autoradiograph of 2D gel of l-[³⁵S]methionine-labeled intracellular proteins of B. subtilis 168 under control conditions. The protein spots that were induced more than twofold and possessed >0.2% of the overall intensity 10 min after the addition of diamide to the media are indicated by arrows (or circles, where missing) and either labeled with the corresponding gene or protein name or, if not identified, labeled with “DIA” and a number (for diamide induced).

FIG. 5.

Kinetics of protein synthesis in response to diamide stress. The protein synthesis pattern of exponentially growing cells (control conditions) was green and overlaid with the dewarped red protein synthesis pattern of the comparing gel at 10, 30, 50, and 125 min after the addition of 1 mM diamide. Protein spots synthesized only under control conditions appear green; protein spots synthesized only under conditions of disulfide stress appear red, while protein spots synthesized under both conditions appear yellow. Bar charts on the right show the normalized ratio of the intensity of the circled protein spots (in percent overall blackness). The circled proteins are ClpE (ATP-dependent Clp protease-like [class III stress gene]), chosen as a member of the class III heat shock regulon; CysK (cysteine synthetase A) of the cysteine biosynthesis family; TrxA (thioredoxin), an oxidative stress protein related to the homeostasis of the intracellular thiol-disulfide state; and PdhA (pyruvate dehydrogenase [E1 alpha subunit]), as a typical vegetative protein.

All major oxidative stress genes were induced by disulfide stress.

Several of the genes that are strongly induced upon diamide treatment are also known to be induced under hydrogen peroxide and paraquat stress. These include the genes encoding vegetative catalase, both alkylhydroperoxide reductase subunits, a DNA-binding protein, peptidyl methionylsulfoxide reductases, thioredoxin, thioredoxin reductase, and superoxide dismutase (Table 3). The regulator of the peroxide regulon perR was also induced by disulfide stress. The activity of the metalloprotein PerR seems to be modulated by redox-active cysteine residues (8, 12). Interestingly, it appears that all described members of the peroxide regulon, like perR itself, mrgA, katA, ahpC, and ahpF, were induced by disulfide stress except for the hemAXCDBL operon. This operon appears to be less induced by paraquat-induced oxidative stress (data not shown), which suggests an additional regulative mechanism for the heme biosynthesis operon, thereby discriminating hydrogen peroxide stress from other oxidative stresses.

TABLE 3.

Function and induction factors in the transcriptome experiments of selected genes induced by disulfide stress

Gene classification and name	Functiona	Induction factor at:
		10 min	30 min	50 min
Oxidative stress genes induced by disulfide stress
ahpC	Alkyl hydroperoxide reductase (small subunit)	10.6	10.0	6.0
ahpF	Alkyl hydroperoxide reductase (large subunit)/NADH dehydrogenase	5.2	9.3	5.8
katA	Vegetative catalase 1	32.3b,c	56.3b,c	14.9b,c
mrgA	Metalloregulation DNA-binding stress protein	15.0b	61.1b	5.0b
msrA	Peptidyl methionine sulfoxide reductase	5.0	11.5	21.4
perR	Transcriptional repressor of the peroxide regulon	8.8	9.9b	4.3
sodA	Superoxide dismutase	2.6	3.4	3.2
trxA	Thioredoxin	9.7	27.4	26.0
trxB	Thioredoxin reductase	8.7	13.5	17.9
yppQ	Unknown; similar to peptide methionine sulfoxide reductase	4.3	10.5	18.3
Class III heat stress genes induced by disulfide stress
clpC	Class III stress response-related ATPase	13.7	24.3	19.4
clpE	ATP-dependent Clp protease-like (class III stress gene)	46.8b	67.5b	43.2b
clpP	ATP-dependent Clp protease proteolytic subunit (class III heat shock protein)	11.9	17.8	10.7
ctsR	Transcriptional repressor of class III stress genes	22.2	28.5	23.1b
mcsA	Modulator of CtsR repression	10.5b	27.1	14.0b
mcsB	Modulator of CtsR repression	14.7	25.0	31.6
Class I heat stress genes induced by disulfide stress
dnaK	Class I heat shock protein (molecular chaperone)	4.1	3.4	2.2
groEL	Class I heat shock protein (chaperonin)	2.4	2.2	1.2
groES	Class I heat shock protein (chaperonin)	2.6	1.4b	0.6b
hrcA	Transcriptional repressor of class I heat shock genes	9.2	11.8	6.1
Genes related to cysteine biosynthesis induced by disulfide stress
cysK	Cysteine synthetase A	8.3	15.9	16.4
yjcI	Unknown; similar to cystathionine gamma-synthase	5.4	4.3	4.1
yjcJ	Unknown; similar to cystathionine beta-lyase	4.0	2.8	2.3
yrhA	Unknown; similar to cysteine synthase	19.9	35.2	45.0
yrhB	Unknown; similar to cystathionine gamma-synthase	20.5c	35.5c	44.8c
Genes related to resistance of toxic elements
arsB	Extrusion of arsenite	26.8b,c	36.4c	8.0c
arsC	Arsenate reductase	47.8b	53.5	19.7b
arsR	Arsenic resistance operon repressor	26.1b,c	40.1b,c	10.7b,c
yqcK	Unknown; similar to unknown proteins	35.0c	30.7c	12.5c
yvgW	Unknown; similar to heavy metal-transporting ATPase	57.2b,c	42.1b,c	20.0b,c
yvgX	Unknown; similar to heavy metal-transporting ATPase	50.9b,c	59.9b,c	61.2b,c
yvgY	Unknown; similar to mercuric transport protein	25.8b,c	58.1b,c	57.7b,c
yvgZ	Unknown; similar to unknown proteins	6.0	6.9	9.2

^aFunction according to the SubtiList database (http://genolist.pasteur.fr/SubtiList) (19).

^bDue to high deviation in both transcriptome experiments, a value closer to 1 was chosen and was flagged as a “safe value.”

^cAt least one replica spot in one transcriptome experiment near the detection limit or below it on the control array was flagged as the minimal induction value.

Comparison of our proteome data with previously obtained data from experiments with hydrogen peroxide or paraquat as oxidative stress inducers indicates a closer relationship between disulfide stress and paraquat-induced oxidative stress than between disulfide stress and hydrogen peroxide stress. For instance, genes such as trxA, trxB, and cysteine biosynthesis genes that are probably related to homeostasis of the intracellular thiol-disulfide state are more strongly induced by paraquat than by hydrogen peroxide (2).

Disulfide stress is an inducer of class III heat shock genes and other genes related to heat shock.

A group of genes most strongly induced by disulfide stress in both transcriptome and proteome experiments were the class III heat shock genes (clpC, clpP, and clpE) controlled by the global repressor CtsR (see reference 23 for a review) (Table 3). The gene clpC is located in an operon with ctsR and the modulators of CtsR repression mcsA and mcsB. McsA contains two putative zinc finger motifs, each consisting of two conserved CXXC motifs. In its native state, McsA binds CtsR, thereby preventing McsB from inactivating CtsR (16). Electrospray ionization-quadrupole TOF-MS experiments on His₆-tagged purified recombinant McsA showed that the CXXC motifs are easily oxidized, indicating that this protein might be the molecular interface for oxidative stress activation of class III heat shock proteins (data not shown). Protein stress caused by the formation of disulfide bonds of cysteine side chains, subsequently leading to misfolded proteins and toxic protein aggregates, might not be the only reason for induction of the specific heat stress response. Direct sensing of disulfide stress by McsA could also lead to the induction of the class III heat shock regulon. The similar induction pattern of all CtsR-regulated genes upon diamide treatment was confirmed by clustering the transcriptome data by using GeneSpring software (Silicon Genetics, Redwood City, Calif.). By subjecting the kinetic data to the changing method, the class III heat shock genes occurred in one distinct cluster, indicating a disulfide stress specific regulation by one common mechanism (data not shown).

We also determined that a number of class I heat shock genes, such as hrcA itself, dnaK, and groES, under the control of the global regulator HrcA were induced upon diamide treatment, albeit to a lesser extent, mainly 10 min after addition of diamide (Table 3).

Only very few of the class II heat shock genes that are controlled by the alternative sigma factor σ^B were significantly induced by disulfide stress. This may be due to transcription from additional σ^B-independent promoters, as could be shown by Northern blot analysis of trxA (data not shown).

Disulfide stress triggers the “stringent response.”

B. subtilis cells that were exposed to disulfide stress seemed to induce the stringent response. Both the proteomic and the transcriptomic patterns showed a massive downregulation of a large number of vegetative genes (Fig. 4 and 5). More than 80% of the recently described 141 negatively stringently controlled genes in B. subtilis (11), including genes for ribosomal proteins and housekeeping genes, are repressed at least threefold by disulfide stress, suggesting that disulfide stress could lead to a stalling of the ribosomes and, thus, to the release of the alarmone ppGpp. A similar stalling of the ribosome through oxidation of the CXXC motif of ribosomal protein L31 as an effect of disulfide stress in S. coelicolor was suggested by Paget et al. (20) and in E. coli oxidative stress leads to increased levels of ppGpp (26) and induction of the stringent response (7). Interestingly, rpmE, the gene encoding L31, is one of two negatively stringently controlled genes encoding ribosomal proteins that are not repressed significantly by disulfide stress in our study.

The stringent response does not only include genes that are strongly repressed but also genes that are RelA-dependently induced (11). The part of positively stringently controlled genes that is induced by disulfide stress is much smaller, and only 20% of all genes described by Eymann et al. were found to be induced. Positively stringently controlled genes related to amino acid biosynthesis were not induced except for genes involved in cysteine biosynthesis (Table 3 and Fig. 5). This suggests that free cysteines in B. subtilis could play a role as a low-molecular-weight thiol antioxidant similar to glutathione in other organisms (B. subtilis possesses no glutathione system). It is also noteworthy that two positively stringently controlled regulative genes of unknown function, yvyD and ytzE, were induced by disulfide stress as well.

Other findings.

Many genes encoding regulators or putative regulators seem to play a role in the disulfide stress response. A total of 17 genes that are known to be involved in regulation and initiation of RNA synthesis (according to the functional categories of SubtiList database [http://genolist.pasteur.fr/SubtiList] [19]) were induced by more than threefold 10 min after the addition of diamide to the medium. Fifty minutes after the addition of diamide, 24 regulatory genes were induced, indicating that there is not only a primary and one-dimensional cellular response but also a cascade of different secondary and even tertiary responses to disulfide stress. This became also apparent by the finding that the total numbers of genes that were significantly induced and repressed increased over time from 196 to 290 and from 308 to 469, respectively (Table 1). Among these regulators were, in addition to the already-mentioned regulators HrcA, CtsR, PerR, YtzE, and YvyD, also several other regulators whose regulons still need to be characterized.

Other genes that attracted our attention were genes involved in the detoxification of heavy metals and other toxic elements. The arsenic resistance operon consisting of the gene of the repressor arsR, the arsenite transporter arsB, arsenate reductase arsC, and yqcK are upregulated significantly at all time points. Also, the cadmium resistance gene yvgW and the three upstream genes, including its possible positive regulator yvgY (25), were among the genes most strongly induced by disulfide stress (Table 3). In proteins involved in resistance to toxic elements, reduced cysteines often play a major role in either binding and/or reducing the toxic elements and therefore are crucial for the activity of the detoxifying enzymes (18, 24).

A total of 137 of the genes that are induced significantly by disulfide stress at least at one of the time points chosen were of unknown function. This high number demonstrates that our knowledge of bacterial genomes is still far from being complete, but further comparison of expression profiles under different conditions could unravel the functions of some of these genes.

Conclusion.

It has previously been assumed that disulfide stress is just a subset of the oxidative stress response. Recent reports showing that the expression of whole regulons is controlled by redox-active cysteines have brought disulfide stress into focus (20, 28). This provides a preliminary indication that the disulfide stress response is different from the “classic” oxidative stress response induced by hydrogen peroxide and the superoxide stress induced by paraquat.

Here we report a genome-wide analysis of the influence of disulfide stress on the transcriptional and protein expression pattern in B. subtilis. We found 350 genes were induced and 578 genes were repressed upon diamide treatment. Thus, disulfide stress affected more than 20% of all 4,101 B. subtilis genes. The adaptation of B. subtilis to disulfide stress seems to be remarkably complex, and the response to disulfide stress appears to be a combination of different regulatory networks that involve genes from the oxidative stress response, the heat and protein stress response, the stringent response, and the heavy metal stress response. This suggests that redox-active cysteines might act as molecular switches in key positions of the regulation cascade of particular regulons, as suggested for PerR, the regulator of the peroxide regulon (8). Indeed, oxidative stress genes under control of PerR were among the strongest induced genes influenced by disulfide stress. The proposed role of redox-active cysteines in the ribosomal protein L31 in stalling ribosomes (20) is supported by the finding that disulfide stress induces the stringent response in B. subtilis. New candidates for regulators with possible redox-active cysteines are McsA, whose oxidation could lead to the high induction levels of the class III heat shock regulon observed after disulfide stress, and the regulator of the arsenic resistance operon ArsR.

These combined transcriptome and proteome analyses open up a broad and panoramic view of the response of the soil bacterium B. subtilis to the challenge of disulfide stress and let us anticipate the complexity of the regulatory mechanisms triggered directly or indirectly by a disturbance in the thiol-disulfide state within the cell.