A master regulator σB governs osmotic and oxidative response as well as differentiation via a network of sigma factors in Streptomyces coelicolor

Authors


Summary

The differentiating bacterium Streptomyces coelicolor harbours some 66 sigma factors, which support its complex life cycle. σB, a functional homologue of σS from Escherichia coli, controls both osmoprotection and differentiation in S. coelicolor A3(2). Microarray analysis revealed σB-dependent induction of more than 280 genes by 0.2 M KCl. These genes encode several sigma factors, oxidative defence proteins, chaperones, systems to provide osmolytes, cysteine, mycothiol, and gas vesicle. σB controlled induction of itself and its two paralogues (σL and σM) in a hierarchical order of σB→σL→σM, as revealed by S1 mapping and Western blot analyses. The phenotype of each sigma mutant suggested a sequential action in morphological differentiation; σB in forming aerial mycelium, σL in forming spores and σM for efficient sporulation. σB was also responsible for the increase in cysteine and mycothiol, the major thiol buffer in actinomycetes, upon osmotic shock, revealing an overlap between protections against osmotic and oxidative stresses. Proteins in sigB mutant were more oxidized (carbonylated) than the wild type. These results support a hypothesis that σB serves as a master regulator that triggers other related sigma factors in a cascade, and thus regulates differentiation and osmotic and oxidative response in S. coelicolor.

Introduction

Alternative sigma factors in bacteria co-ordinate gene expression in response to various environmental and endogenous signals. Current microbial genome sequence data reveal a variety of alternative sigma factors, ranging from one (Mycoplasma genitalium; Fraser et al., 1995) or two (Lactococcus lactis; Bolotin et al., 2001) sigma factors to about 66 sigma factors in Streptomyces coelicolor (Paget et al., 2002; Hahn et al., 2003). Most alternative sigma factors are related in sequence to σ70D) of Escherichia coli. Based on phylogenetic relatedness, σ70 family has been subdivided into several groups (Lonetto et al., 1994; Helmann, 2002; Gruber and Gross, 2003); group 1 for the σ70 orthologues, the essential housekeeping sigma factors, group 2 for non-essential close relatives of group 1 members, group 3 for more divergent but evolutionarily related sigma factors, group 4 for distantly related σ70 family members originally termed extracytoplasmic function (ECF) subfamily and group 5 for related proteins of TxeR subfamily found as positive regulators of toxin production in Clostridium spp.

Among these sigma factors, those that respond to diverse growth-limiting stresses and protect bacteria against further stresses are known as general stress response sigma factors. The best studied examples are σS (group 2) of E. coli representing γ-proteobacteria and σB (group 3) of Bacillus subtilis representing low G+C Gram-positive group of bacteria. σB in B. subtilis regulates the expression of more than 200 genes involved in heat, acid, ethanol, salt and oxidative stress resistance (Hecker and Volker, 1998; Petersohn et al., 2001; Price, 2002). Its homologue also regulates stress resistance in Listeria monocytogenes and Staphylococcus aureus (Becker et al., 1998; Kullik et al., 1998; Wiedmann et al., 1998), virulence in Bacillus anthracis (Fouet et al., 2000), adherent biofilm formation in Staphylococcus epidermidis and St. aureus (Rachid et al., 2000; Knobloch et al., 2001). In Mycobacterium tuberculosis, a high G+C Gram-positive bacterium, a σB-like factor contributes to its virulence (Chen et al., 2000).

The genome sequence of S. coelicolor A3(2) M145, containing 7825 protein-coding genes, revealed 10 group 3 sigma factors (Bentley et al., 2002; Hahn et al., 2003); σB (SCO0600), σL (SCO7278), σI (SCO3068), σN (SCO4034), σF (SCO4035), σH (SCO5243), σK (SCO6520), σM (SCO7314), σG (SCO7341) and σWhiG (SCO5621) in the order of similarity to σB. Homologues of these are also encoded in the genome of Streptomyces avermitilis (Ikeda et al., 2003). Of these, at least three (σB, σH, σI) are specifically induced by osmotic stress (Kormanec et al., 2000; Cho et al., 2001; Kelemen et al., 2001; Viollier et al., 2003a), suggesting that osmotic stress response of S. coelicolor is mediated by multiple σB-like sigma factors. They are regulated in diverse ways. σB is regulated at transcriptional level for its synthesis and post-translationally for its activity through interaction of its antisigma factor (RsbA) with an antianti-sigma factor (RsbV), involving phospho-relay mechanism (Lee et al., 2004a). σH is regulated at levels of transcription, translation start site selection, protein processing and possibly interaction with an antisigma factor (PrsH/UshX) (Sevcikova et al., 2001; Sevcikova and Kormanec, 2002; Viollier et al., 2003a,b). σI increases rapidly upon osmotic stress, most likely via increased transcription (Viollier et al., 2003a). σB, σH, σF and σWhiG have been also implicated in controlling proper differentiation (Chater et al., 1989; Potuckova et al., 1995; Cho et al., 2001; Sevcikova et al., 2001).

To investigate the genes regulated by σB, and hence its role in osmoprotection and differentiation, we characterized the global response of S. coelicolor to osmotic stress, using microarray analysis. From the pool of about 300 genes induced by 0.2 M KCl treatment, we explored the expression of several gene groups in further detail. The results reveal (i) a regulatory network (cascade) among sigma factors and (ii) the correlation between osmotic and oxidative responses, both of which are governed by σB.

Results

Identification of KCl-induced genes by microarray analysis

We analysed the transcriptome of the wild-type (J1501) and sigB null mutant (YD2108) using cDNA microarrays that contain 97% of predicted genes from S. coelicolor (7071 gene spots) on glass slides as described previously (Huang et al., 2001). To monitor changes in gene expression over time, mRNA profiles obtained at 5, 15, 30, 50 and 80 min after 0.2 M KCl treatment were compared with that of a time-zero control sample. We identified 287 gene signals that increased more than threefold at two or more time points after KCl treatment (Table S1). Among these, 89 genes appear to be induced from 26 operons, as judged from co-induction of neighbouring genes. We selected 93 genes induced more than threefold at three or more time points, and grouped them into four subclasses according to their induction profile (Fig. 1). Class I consists of 32 genes induced transiently with a peak around 30 min. They include genes for σB (SCO0600), a σB-paralogue σL (SCO7278), σHrdD, RsbV (an antianti-sigma factor for σB; Lee et al., 2004a), putative uptake systems for glycine betaine and glutamate, and gas vesicle components. Class II consists of 28 genes whose induction is rapid and sustained. They include genes for tellurite resistance, Fe–S assembly, cold shock proteins and translation control factors. Class III consists of 21 genes whose induction is slow (after 15 min) and/or peaks around 50 min. They include genes for catalases (catA, catA2), DpsA, bacterioferritin, sulphur assimilation, gas vesicle components and the principal sigma factor σHrdB. Class IV consists of 12 genes that are induced rapidly within 5 min with a transient decrease around 30 min. They include genes for ribosomal proteins and subunits for RNA polymerase. In the sigB null mutant, the fold induction of nearly all those genes were reduced by varying extents, suggesting that σB is a key regulator for gene induction after osmotic shock (Table S1). In contrast to large numbers of induced genes, only 27 genes were found to decrease by more than threefold upon KCl treatment. These encode mostly hypothetical proteins, putative membrane proteins, a cytochrome oxidase subunit and a decarboxylase (data not shown).

Figure 1.

Expression profiles of 93 genes prominently induced by KCl treatment. The induction patterns of 93 genes, enhanced more than threefold at three or more time points after treatment with 0.2 M KCl, were hierarchically clustered on the basis of their fold induction at five time points (5, 15, 30, 50 and 80 min; Eisen et al., 1998). The expression pattern was classified into four groups; transient (I), sustained (II), delayed (III), and oscillatory (IV) induction groups. The log2 fold induction (relative to the untreated control sample) of wild type (J1501) and sigB mutant (YD2108) was plotted. Individual genes of each group were listed on the right with open reading frame numbers along with known or predicted gene names and functions. Adjacent genes that might be co-transcribed are indicated by arrows.

KCl-induced sigma factors and their regulators

Among 287 genes prominently induced by KCl (Fig. 1, Table S1), many sigma factors and their known and predicted regulators were found. At least four sigma factor genes (sigB, sigL, sigM and hrdD) were dramatically induced (about nine- to 38-fold) in a σB-dependent manner (Fig. S1, Table S1). The phylogenic relationship among the osmotically induced sigma factors and their gene (operon) structure are shown in Fig. 2A and B.

Figure 2.

Osmotic induction of sigma factor genes and their dependence on σB.
A. Phylogenic relationship among σ70-type non-ECF sigma factors in S. coelicolor. Ten sigma factors were classified according to Helmann (2002).
B. Gene arrangements of six sigma factor genes (sigB, sigL, sigM, sigH, sigI and hrdD) that are induced by KCl. The genes for sigma factors and anti-sigma factors (known and putative) were shown with black and shaded arrows respectively. The positions for known promoters were indicated with bent arrows.
C. S1 mapping analysis of osmotically induced sigma factor genes and their dependence on σB. Induction of seven sigma factor genes (sigB, sigL, sigM, sigH, sigI, hrdD and hrdB), their regulators (rsbV encoding an antianti-sigma factor for σB, and SCO7277, a putative antisigma factor for σL) and catB gene previously shown to be σB-dependent, was determined by S1 mapping in the wild type (J1501) and sigB mutant (YD2108). RNAs were prepared from cells that were grown in YEME liquid medium at early exponential phase and treated with 0.2 M KCl for 10, 20, 30, 60 and 90 min. As a representative for constitutively expressed genes, sigBp2 transcripts were analysed in parallel.

To confirm microarray results, we performed S1 mapping analysis of the time-course of induction of these sigma factors and their putative regulators (Fig. 2C). The data revealed a close correlation between microarray and S1 analyses. As expected, sigBp1 transcripts were rapidly and dramatically induced up to more than 20-fold at 20 min, and decreased to near pre-stimulus level by 90 min. In the sigB mutant, sigBp1 transcripts were only marginally induced with delayed kinetics. Both sigL and sigM transcripts were induced more than 10-fold at maximum around 30 min, in a σB-dependent manner. Induction of hrdD was also dependent on σB as predicted from microarray data. The σB-dependent promoter (hrdDp3) was localized downstream of the previously reported σR-dependent promoter hrdDp2 (Paget et al., 2001), by about 19 nt difference between the transcription start sites. Induction of sigH (sigHp2) and sigI were less pronounced with delayed kinetics, and both were rather independent of σB. The hrdB gene encoding the principal sigma factor was prominently induced in a delayed fashion as demonstrated by microarray analysis (Fig. 1), and the induction was only marginally dependent on σB. The rsbV gene encoding an antianti-sigma factor for σB (Lee et al., 2004a) and the gene for a putative anti-sigma factor (SCO7277) for σL was also prominently induced in a σB-dependent manner. The catB gene encoding a differentiation-specific catalase, and previously reported to be σB-dependent (Cho et al., 2001), was found to be induced in a delayed fashion, implying an indirect control by σB The S1 mapping results suggest that upon osmotic upshift, activated (freed from its anti-sigma factor) σB increases the synthesis of itself as well as other related sigma factors (σL, σM and σHrdD), amplifying the positive regulatory cascade. Increase in antianti-sigma factor (RsbV for σB) fits with this strategy as well.

Increase in the level of σ B, σ L and σ M upon osmotic upshift

Among KCl-induced sigma factors, we further examined two σB-paralogues, σL and σM, as their gene induction appears to be directly dependent on σB. We tested whether the amount of σB, σL and σM proteins increased in accordance with the increase in transcripts upon osmotic upshift, using antisera raised against σB, σL or σM(Fig. 3). Consistent with transcript analyses, σB protein increased transiently with a peak at 30 min. It was not detected in sigB mutant as expected, but was still detected in sigL and sigM mutants with slightly shifted kinetics (Fig. 3). This suggests that σL and σM do not control σB synthesis, as much as σB controls them. σL was not detected in sigB and sigL mutants, whereas its transient increase was observed in sigM mutant. On the other hand, σM was not detected in any of the mutants. In the wild type, the increase in σM was slightly more delayed than σL. These results strongly support a hypothesis that σB is at the top of the regulatory cascade among σB-paralogues, followed by σL and then σM, in the hierarchical order. In a parallel experiment, CatB protein, previously known to be σB-dependent (Cho et al., 2001), was found to increase after 90 min following KCl treatment, and this increase was not detected in sigB, sigL and sigM mutants. Therefore, it is likely that the induction of catB gene is controlled indirectly by σB via σM, which controls catB either directly or via yet (an)other downstream regulator(s).

Figure 3.

Changes in the level of σB, σL and σM proteins after KCl treatment. The amount of each sigma factor after osmotic stress was monitored by Western blot analysis using antibodies raised against each purified sigma factor. Osmotic treatment of wild type and mutant cells (sigB, sigL and sigM) were done as described in Fig. 2C, with one more sampling at 120 min after KCl treatment. The specific band position for each sigma factor is indicated with an arrow. Induction of catalase B was analysed in parallel. A model for hierarchical regulation among these sigma factors is presented at right.

Disruption phenotype reflects hierarchy among σ B-like sigma factors

Considering σB is required for the formation of aerial mycelium, we investigated the contribution of other osmotically induced sigma factors to differentiation. Null mutations of sigL, sigM, sigH, sigI and hrdD were created in J1501 background using polymerase chain reaction (PCR)-targeted gene disruption system (Gust et al., 2003). As demonstrated in Fig. 4, we found that sigL null mutant showed white phenotype, forming fluffy aerial mycelium without further progress to sporulation on R2YE plates. Antibiotic production was also disturbed in sigL mutant with almost complete absence of actinorhodin. On the other hand, sigM mutant proceeded to sporulation, but with reduced number of spores than the wild-type by about 10-fold. We found that the phenotypes of sigH, sigI and hrdD mutants were indistinguishable from the wild-type (hrdD data not shown). When double mutations of each of sigL, sigM, sigH and sigI in sigB mutant background were created, we found that the bald phenotype from sigB mutation prevailed over the effect of other mutations (Fig. 4, lower row). Taken together, we propose that σL and σM, whose synthesis is osmotically induced in a σB-dependent way, also contribute to morphological differentiation, possibly with a sequential action of σB first (on aerial mycelium formation) and then σL (on sporulation) followed by σM. This coincides with the hierarchical order of regulation observed in the timing of induction by osmotic stress.

Figure 4.

Colony morphology of various sigma factor mutants. Single null mutants of sigB, sigL, sigM, sigH, or sigI genes as well as double mutants were created as described in the text. Mutant cells were grown on R2YE plates for 7 days and observed under a stereomicroscope.

Osmotic induction of cysteine biosynthetic gene cluster

One of the prominent KCl-induced gene clusters is that for cysteine biosynthesis. As shown in Fig. 5A, putative genes for the biosynthetic pathway from sulphate to cysteine and sulfonate transport were found to be induced strongly by KCl in microarray analysis (Table S1, Fig. 5A). Figure 5B illustrates a possible sulphur assimilation pathway for cysteine biosynthesis, similar to the one proposed in B. subtilis (Grundy and Henkin, 2002). Sulphate is predicted to be transported by an unknown sulphate permease. The cysN (SCO6097), cysD (SCO6098) and cysC (SCO6099) gene products are expected to encode enzymes responsible for activation of sulphate to 3′-phospho-adenosine-5′-phosphosulphate (PAPS). The cysH (SCO 6100) gene encodes a PAPS reductase. The cysI (SCO6102) and cysM (SCO2910) products are predicted to catalyse the last two steps of cysteine biosynthesis in S. coelicolor, the reduction of sulfite to sulfide and incorporation of sulfide into O-acetylserine synthesized by cysE gene product (SCO6103). In the absence of sulphate, S. coelicolor can utilize aliphatic sulfonates as a source of sulphur. This is mediated by the products of gene cluster ssuABC-SCO6093, which are located immediately downstream of cysEIHCDN genes. All these genes were significantly induced upon osmotic challenge with a maximal induction around 50 min after KCl treatment (Fig. 5A, Fig. S2). The two related genes for cysteine synthase, cysM (SCO2910) and SCO0992, and that for thiosulphate sulfotransferase (SCO4164; cysA2) were also induced with similar kinetics (Table S1).

Figure 5.

Osmotic induction of cysteine biosynthetic gene cluster and their final product.
A. Arrangement of cysteine biosynthetic gene cluster and their induction profile in microarrays. The gene cluster is composed of putative sulfonate transport genes (ssuABC-SCO6096) and sulphate-to-cysteine biosynthetic genes (cysN, D, C, H, SCO6101, cysI and E).
B. Illustration of possible pathway for cysteine biosynthesis. APS denotes adenosine 5′-phosphosulphate, and PAPS denotes 3′-phosphoadenosine 5′-phosphosulphate.
C. The changes in the amount of cysteine in wild type and sigB mutant upon KCl treatment. Amino acid analysis was performed as described in the text using HPLC. The level of cysteine was presented as molar percentage of total amino acids (mol percentage). Average values from three independent experiments were presented with error bars of standard deviations.

Triggered by these observations, the level of cysteine was monitored upon KCl treatment. Figure 5C demonstrates that cysteine increased gradually by about twofold within 2 h of osmotic shock. In sigB mutant, the unstressed level was lower than that of the wild type, and the KCl-induced increase was not observed. Therefore, as predicted from microarray results, cysteine synthesis increases in a σB-dependent way, with delayed induction pattern.

Increase in the biosynthesis of mycothiol by osmotic induction

Cysteine is a component of mycothiol, a cysteinyl disaccharide (1- d-myo-inosityl-2-(N-acetyl- l-cysteinyl) amino-2-deoxy-α- d-glucopyranoside) that is the principal low-molecular weight thiol compound found in many actinomycetes including Mycobacterium spp. (Newton et al., 1996). We therefore investigated the possibility that mycothiol synthesis increases upon osmotic stress. Four putative genes for mycothiol biosynthesis were found in S. coelicolor; mshA (SCO4204), mshB (SCO5126), mshC (SCO1663) and mshD (SCO1545) predicted to act on each step of the pathway as illustrated in Fig. 6A (Bornemann et al., 1997; Newton and Fahey, 2002). In the proposed pathway, MshA (N-acetylglucosamine transferase) catalyses the formation of GlcNAc-Ins (1- d-myo-inosityl-2-acetamido-2-deoxy-α- d-glucopyranoside), which is then deacetylated by MshB (deacetylase) to form GlcN-Ins (1- d-myo-inosityl-2-deoxy-α- d-glucopyranoside). GlcN-Ins is then ligated with cysteine by a ligase MshC (Sareen et al., 2002). The resulting Cys-GlcN-Ins (cysteine-glucosamine-inositol) is then acetylated to form mycothiol (MSH) by MshD (acetyltransferase) (Koledin et al., 2002).

Figure 6.

Osmotic induction of mycothiol biosynthetic genes and their final product.
A. S1 mapping analysis of mshA, B, C and D genes responsible for mycothiol biosynthesis. Mycothiol biosynthetic pathway was illustrated in the left along with the predicted responsible genes (Bornemann et al., 1997; Newton and Fahey, 2002). The same RNA samples used in Fig. 2C were analysed for msh transcripts.
B. Changes in mycothiol levels upon osmotic treatment in wild-type (grey bar) and sigB mutant (white bar). The amount of mycothiol was determined using an HPLC assay (Fahey and Newton, 1987) as described in the text. Average values from three individual experiments were presented in µmol per gram wet cells with error bars of standard deviations.

S1 mapping analysis showed that mshA, mshC and mshD genes were induced upon osmotic challenge, consistent with the microarray data (Fig. 6A, Fig. S2). The mshA gene was induced in a transient fashion, with a peak around 30 min, whereas mshC transcripts increased gradually up to 90 min. The mshD transcript increased less prominently with a sharp peak around 30 min. Almost no induction of mshA and mshC was observed in sigB mutant, suggesting that the induction was dependent on σB. The mshD gene was induced in sigB mutant with altered kinetics. Using a sensitive high-performance liquid chromatography (HPLC) assay for thiol compounds we monitored the change in the amount of mycothiol upon osmotic challenge. We found that mycothiol increased about 30% within an hour of osmotic challenge in the wild type, whereas the unstressed and induced levels were both lowered in sigB mutant (Fig. 6B).

σ B protects protein from oxidation

As osmotic stress induces many anti-oxidative defence proteins as well as small thiol compounds, it can be hypothesized that osmotic challenge elicits some oxidative damage to cellular components. As a marker for protein oxidation, we monitored protein carbonylation upon osmotic challenge in the wild type and sigB mutant, using antibody against derivatized carbonyl groups (Levine et al., 1990; Berlett and Stadtman, 1997). In Fig. 7, a representative result shows that the carbonylated proteins initially decreased upon osmotic treatment, but began to build up after an hour to the unstressed level. On the other hand, in sigB mutant the unstressed level of protein carbonylation is significantly higher than in the wild type. Long time exposure (over 90 min) to osmotic challenge resulted in significantly higher level of carbonylation in sigB mutant. These results strongly support a proposal that σB contributes a critical part to protect proteins from oxidation under ordinary culture conditions, as well as under hyper-osmotic conditions.

Figure 7.

Increase in oxidized proteins by sigB mutation. Oxidized (carbonylated) proteins in wild type and sigB mutant cells were detected by OxyBlot kit. The same samples prepared for Western blotting in Fig. 3 were analysed for carbonylated proteins following gel electrophoresis as described in the text.

Discussion

In this study, we presented evidences that σB is a master regulator for osmotic stress response in S. coelicolor, governing induction of more than 280 genes. Upon osmotic challenge it triggers the synthesis of at least two of its paralogues, σL and σM, which are also involved in the process of morphological differentiation. The sequential order of action in osmotic induction as well as in differentiation appears to be from σB to σL to σM. σB also contributes to protect cells against oxidative damage during ordinary growth as well as under osmotic stress.

Previously we catalogued possible σB regulon members based on consensus promoter sequence proposed from catB and sigB promoters; NNGNNT(N14-16)GGGTAC/T (Cho et al., 2000; 2001). A strict search through the complete S. coelicolor A3(2) genome database generated a list of 118 candidate σB-specific promoters (Lee et al., 2004b). In this study, we found over 280 genes that are prominently (over threefold) regulated by σB. Among 26 possible operons and 200 monocistronic genes presented in this study, 15 genes contain the above mentioned consensus sequence. Based on the difference in induction kinetics and the similarity among σB-like sigma factors in the promoter-recognition domains (domains 2.4 and 4.2), we propose that a large fraction of σB-dependent genes are not directly, nor solely, regulated by σB. σB can regulate these promoters indirectly by boosting the synthesis of other sigma factors or regulators, including the two σB-paralogues (σL and σM). The catB promoter that has previously been shown to be regulated by σB is most likely regulated via this sigma cascade. Taking those genes that are less prominently induced into account (for example, 796 genes that are induced more than twofold at two or more time points), we believe that more genes are positively regulated by σB directly or indirectly via cascades and networks of sigma factors and other regulators. When this expanded list of genes are considered, the salt-induced σB-dependent sigma factors consist of σH and σI (group 3 σB-paralogues), three group 2 sigma factors (σhrdA, σhrdC, σhrdD), seven ECF-type (group 4) sigma factors (σE, σR, σT, SCO3613, SCO4409, SCO4866, SCO7104), and the major sigma factor σhrdB (data not shown). The regulatory network will involve both the recognition of a single promoter by multiple paralogous sigma factors with shared recognition properties as suggested for ECF-type or σB-type sigma factors of Streptomycetes (Paget et al., 2001; Viollier et al., 2003a; Roth et al., 2004), and the recognition of a single gene with multiple promoters by different sigma factors.

Among prominent salt-induced genes (Table S1) are those predicted to participate in osmo-regulation; possible transport systems for osmoprotectants such as glycine betaine, proline, glutamate, fructose, maltose and oligopeptides, and biosynthetic genes for glycine betaine and trehalose. Genes for cell wall biogenesis (dihydrodipicolinate synthase) and membrane fluidity (cyclopropane-fatty acyl phospholipid synthase, fatty acid desaturase) could also increase tolerance against hyper-osmotic stress. In this respect, induction of many hydrolases and peptidases may also be considered to serve this function by providing osmolytes to counteract hyperosmostic stress. Even though the function of gas vesicle is not currently known, a strong salt-induction of the whole cluster of gas vesicle genes observed by microarray as well as S1 mapping (Fig. S3) suggest that gas vesicles may serve a function related with response against hyper-osmotic stress.

In addition to genes with possible osmoprotective function, we also observed significant induction of genes for anti-oxidative function. This anti-oxidative group includes genes for three catalases (catA, catB, catA2), a thioredoxin (SCO5438; trxA2), DpsA, two tellurite resistance proteins, a bacterioferritin, an Fe–S assembly system, cysteine and mycothiol biosynthesis, etc. Even though the probe (gene spot) for another thioredoxin gene trxC was absent in the microarray chips we used, we noticed induction of trxC (SCO0885) in a σB-dependent manner by S1 mapping (data not shown). The observation that the amount of cysteine increased during osmotic stress could be explained to provide anti-oxidative function in two ways, one to provide a substrate for mycothiol synthesis, and the other to replenish Fe–S clusters that are labile to oxidative attack. The former idea is supported by the observation that mycothiol synthesis increased upon osmotic treatment. The latter idea is supported by the result that the sufB1, sufB2, fdx, sufC, sufS and iscU gene cluster encoding Fe–S cluster assembly proteins increased significantly upon KCl treatment with similar delayed kinetics (Table S1).

The induction of nearly all the genes for osmo-regulation and chaperone synthesis is rapid, whereas that for anti-oxidative genes is usually slow and persistent. From these observations it can be hypothesized that the oxidative damage to cell components accumulates after hyper-osmotic shock, and various anti-oxidative functions are elicited as a secondary response to provide protective measures. The increase in oxidative damage after osmotic shock may result from an increase in reactive oxygen species or from an increased susceptibility of cell components, mainly proteins, to oxidation. Rapid increase in chaperones after osmotic shock supports a view that osmotic shock causes some deformation of cell components, and thus makes them more labile to oxidation. In plants, generation of oxidative stress by salinity has been reported in several cases (Hernandez et al., 2001; Mittova et al., 2002; 2004). In mammalian kidney cells, hyperosmolarity is a physiological correlate of heat shock, thought to create deformed proteins (Santos et al., 1998). In E. coli, the stationary sigma factor σS governs a shift from growth to stasis and some osmotic response (Hengge-Aronis, 2002; Nystrom, 2004). σS also protects oxidative damage resulting from the increased susceptibility of aberrant proteins produced at the onset of stasis (Dukan and Nystrom, 1998; Ballesteros et al., 2001). Our results demonstrate that a functionally similar role is played by σB in S. coelicolor. σB, the stationary sigma factor in liquid culture of S. coelicolor (Cho et al., 2001), governs a shift from mycelial growth to sporogenic differentiation, another form of starvation response on surface culture. To achieve this goal, it triggers a cascade of other sigma factors, allowing completion of differentiation process and protection against oxidative damage. There could be some common signal(s) or intracellular phenomena that are shared by osmotic and growth-limiting (stationary or starved) conditions. Our results broaden the common theme of correlation between growth limitation and oxidative damage in bacterial world, and a striking conservation of functional homologues controlling these apparently unrelated phenomena.

Experimental procedures

Bacterial strains and culture conditions

Streptomyces coelicolor A3(2) strains (J1501, M145, and their derivatives) were grown and maintained as described (Cho et al., 2000; Kieser et al., 2000). For liquid culture, YEME medium containing either 34% or 10.3% sucrose was inoculated with pregerminated spores (about 108−109 spores per 100 ml broth). For plate culture, 107 pregerminated spores or patches of mycelia were streaked on R2YE, NA, or minimal agar media. To facilitate harvest of aerial and sporulated mycelia, inoculums were spread on cellophane membrane on solid media. The growth rates and phases were determined as described (Cho and Roe, 1997). To apply osmotic stress in liquid culture, exponentially growing cells in YEME were treated with 0.2 M (final concentration) KCl for various lengths of time before harvest.

DNA manipulations

Restriction and modifying enzymes were used according to the manufacturer's recommendations (POSCOCHEM, Roche, NEB). Standard recombinant DNA methods were used. Ecoli DH5α, methylation-negative Ecoli ET12567 (MacNeil, 1988) and Scoelicolor A3(2) J1501 cells were used as hosts for various recombinant DNAs.

Purification of proteins

His-tagged σB protein was obtained as described previously (Cho et al., 2001). To prepare His-tagged σL protein, mutagenic forward (5′-GTGGGAGAGCATATGCAGACCG CC-3′, NdeI site underlined) and reverse (5′-GAGCTGGT CACCAACGCGGACC-3′) primers were used to amplify a PCR product of 1173 bp containing the entire sigL gene. To prepare His-tagged σM protein, mutagenic forward (5′-GAGCGGATCCATATGCTCATAG-3′, NdeI site underlined) and reverse (5′-GTGTCCCCTGGGATCCTTGTGTC-3′ BamHI site underlined) primers were used to amplify a 903 bp fragment containing the entire sigM gene. The PCR products containing sigL or sigM genes were cloned into pUC18, recovered with NdeI and BamHI digestion, and then cloned into pET-15b (Novagen) to yield pET2212 and pET2409 respectively. His-tagged proteins were overproduced in E. coli BL21 through IPTG induction and purified as recommended by the manufacturer (Novagen). Purified proteins were concentrated to about 0.3 mg ml−1 and dialysed against storage buffer (50 mM Tris-HCl, pH 7.8, 10 mM MgCl2, 0.1 M KCl, 0.1 mM EDTA, 1 mM DTT, 50% glycerol).

Preparation of antibodies and Western blot analysis

Polyclonal antibodies against σB, σL and σM were raised and prepared by conventional methods. For Western blot analysis, proteins resolved on 10% SDS polyacrylamide gel were transferred to nitrocellulose membrane (Schleicher and Schuell, BA-79) using Trans-Blot system (Bio-Rad) at 160 mA for 50 min. The membranes were washed three times in Tris-buffered saline with Triton X-100 (TBS-T; 20 mM Tris-HCl, pH 8.0, 10 mM NaCl, 0.1% Triton X-100) and blocked in TBS-T with 0.5% bovine serum albumin (BSA). The membrane was then incubated for 1–3 h with appropriate antisera (1:2000–1:10 000 dilution) in TBS-T with 0.5% BSA. Excess antibodies were removed by repeated washing with TBS-T. After 20 min incubation in TBS-T containing the secondary antibody (10−4 dilution of goat IgG against mouse IgGAM, Sigma) conjugated with horseradish peroxidase, the membrane was washed twice with TBS-T, and the signals were detected using ECL detection system (Amersham).

Analysis of cysteines and mycothiols

For amino acid analysis, free amino acids in TCA-soluble cell extracts were derivatized with PITC (phenylisothiocyanate) to form phenylthiohydantoin amino acids, which were then analysed by HPLC (Waters 990) through UV detection at 254 nm. Samples were resolved on a column (0.85 × 30 cm, PicoTag), with a linear gradient of 0–100% solvent B (60% acetonitrile) in solvent A (1.4 mM sodium acetate, 0.1% TEA, 6% acetonitrile, pH 6.3) at a constant flow rate of 1.0 ml min−1 for 25 min. Cysteine concentration was calculated using phenylthiohydantoin amino acids standards (Pierce) and represented as pmol per mg wet weight of cells. To detect thiol compounds, HPLC analysis of cell extracts labelled with monobromobimane (mBBr) was performed as described previously (Newton et al., 1996) with slight modifications. Cell pellets from 10 ml culture were resuspended in 0.5 ml of warm 50% acetonitrile in aqueous solution containing 2 mM mBBr (Fluka) in 20 mM Hepes-KOH, pH 7.6. The suspension was incubated in a 60°C water bath for 15 min, followed by acidification with 5 µl of 5 N methanesulfonic acid. The control samples were treated with 5 mM NEM prior to the extraction with 50% acetonitrile solution. For each sample, cell debris was cleared by centrifugation at 4°C, and the supernatant (10 µl) was injected into an HPLC column (Waters AtlantisTM) with an isocratic solution of 0.1% TFA and 40% methanol. The mBBR-derivatized mycothiol was detected by a fluorescence detector with excitation and emission at 380 and 480 nm respectively.

Preparation of RNA and S1 nuclease mapping

RNA isolation and S1 nuclease mapping analysis were performed as described previously (Cho et al., 2001). Probes used for mapping sigBp1, sigBp2, catBp and rsbVp transcripts were prepared as described (Cho et al., 2000; 2001; Lee et al., 2004a). The probes for sigL, sigM, sigH, sigI, hrdD, hrdB, SCO7277, mshA, mshB, mshC and mshD transcripts were prepared by PCR, encompassing 420–600 bp around the ATG start codon. The 5′ positions of forward and reverse primers are: −233 and +188 for sigL, −454 and +20 for sigM, −586 and +12 for sigH, −460 and +75 for sigI, −493 and −9 for hrdD, −497 and +1 for hrdB, −445 and +88 for SCO7277, −175 and +152 for mshA, −246 and +102 for mshB, −474 and +113 for mshC, −489 and +98 for mshD, relative to the ATG start codon. For microarray analysis, additional phenol/chloroform and chloroform extractions were done before the final precipitation of RNAs. The quality and quantity of RNA were estimated by UV absorbance and agarose gel electrophoresis.

Microarray analysis

DNA microarrays containing 7071 gene probes [about 97% of predicted genes of S. coelicolor A3(2) M145] were fabricated as described previously (Huang et al., 2001). Untreated RNA samples from wild type (J1501) and sigB mutant (YD2108) were labelled with Cy3-dCTP (green) and used as references, whereas RNA samples prepared from cells treated with 0.2 M KCl for various lengths of time were labelled with Cy5-dCTP (red) as described by Weaver et al. (2004). RNA (15 µg) and 5 µg of hexameric random primers (72% G+C) were denatured together in a total volume of 7 µl at 75°C for 15 min before snap cooling on ice. Cy3- or Cy5-dCTP (1 µl each; Amersham Pharmacia Biotech) was added to each reaction, together with a 7 µl cocktail containing Superscript II buffer (3 µl of a 5× solution), DTT (1.5 µl of a 0.1 M solution), dNTPs (1.5 µl of a mixture of 4 mM dATP, 4 mM dTTP, 10 mM dGTP and 0.2 mM dCTP) and Superscript II enzyme (1 µl of 200 U µl−1; Gibco BRL) to give a total volume of 15 µl. The mixtures were incubated for 10 min at room temperature before transfer to 42°C for 2–3 h. The reactions were stopped, and the labelled cDNA was purified through buffer exchange and concentration through Microcon-10 filters (Amicon). The Cy3- and Cy5-labelled cDNA samples were mixed together in SSC buffer containing SDS and polyA (DeRisi et al., 1997). The mixture was heated at 100°C for 2 min and applied to a microarray DNA chip. Hybridization was allowed at 63°C for 10–12 h, followed by washing as described previously (DeRisi et al., 1997). The hybridized chip was scanned using a GenePix 4000B scanner (Axon Instruments), and the data were analysed using Cluster and Treeview programs (Eisen et al., 1998).

Detection of oxidized proteins

Oxidized proteins containing carbonyl groups were detected using an Oxyblot kit (Chemicon). The protein sample (10 µg) was incubated with dinitrophenylhydrazine (DNPH) for 15 min, followed by neutralization with a solution containing glycerol and β-mercaptoethanol. These samples were electrophoresed on an 11% SDS-PAG and then transferred to a nitrocellulose membrane. After blocking, the membrane was incubated for 1 h with a rabbit antibody against DNPH (1:150), followed by incubation with anti-rabbit goat antibody (1:300) for 1 h at room temperature as recommended by the manufacturer. The signal was visualized through chemiluminescence with ECL detection kit (Amersham Pharmacia).

Gene disruption and complementation

The sigB mutant (YD2108) has been generated from J1501 cell through insertion of a thiostrepton resistance cassette (Cho et al., 2001). To create null mutation of other sigma factor genes, REDIRECT PCR-targeting method was employed (Gust et al., 2003), replacing each target gene with a disruption cassette. The cassette containing oriT and the aac(3)V gene conferring apramycin resistance was generated by PCR from 1.4 kb EcoRI-HindIII fragment from pIJ773 as a template using oligonucleotide primers that contain 39 nt gene-specific extensions at 5′ positions. Cosmid 5H1 (sigL), 5F8 (sigM), 27G11 (sigH), or E25 (sigI) containing each gene was introduced into E. coli BW25113 (Datsenko and Wanner, 2000) by electroporation along with gene-specific disruption cassettes. The entire target open reading frame was replaced with the oriT/aac(3)V cassette by recombination in E. coli. The resulting recombinant cosmid was introduced into non-methylating E. coli ET12567 and then to S. coelicolor J1501 via conjugation. Each sigma factor mutant was generated through double cross-over of the mutagenized cosmid with the chromosome, resulting in an apramycin-resistant and kanamycin-sensitive phenotype. The desired mutation was confirmed by PCR and Southern hybridization. In frame deletion mutants devoid of the marker cassette were also generated using FLP recombination system (Datsenko and Wanner, 2000; Gust et al., 2003). The cosmid that went through in frame deletion of cassettes by FLP recombinase was modified to replace kanamycin selection marker with an oriT/aadA cassette for spectinomycin resistance originated from pIJ778, in order to circumvent poor kanamycin selection on our culture plates. After conjugation, the final in frame deletion mutant was selected first by spectinomycin resistance and then by its subsequent loss resulting from double cross-over. The expected in frame deletion mutation in each target gene was confirmed by PCR and Southern hybridization. Complementation of the sigL and sigM in frame deletion mutants was done by introducing pSET152 plasmid (Bierman et al., 1992) containing PCR-amplified sigL (from 215 nt upstream to 41 nt downstream of the coding region) or sigM (from 204 nt upstream to 147 nt downstream of the coding region) genes at the BamHI site.

Acknowledgements

We are grateful to Y. J. Cho for technical assistance for microarray analysis and J. B. Seo for amino acid analysis in Korea Basic Science Institute. This work was supported by a KOSEF grant (R01-2004-000-10264) and an NRL grant for Molecular Microbiology lab from the Ministry of Science and Technology to J.-H. Roe, E.J. Lee, H.-S. Kim and J.-H. Park were recipients of BK21 fellowship for graduate students in Life Sciences at SNU.

Ancillary