Introduction

The secretion of heterologous proteins at high levels is an adverse condition for bacterial cells due to the imbalance between protein synthesis, translocation, and quality control, leading to the accumulation of misfolded proteins in the cell envelope. The secretion (envelope) stress is sensed by membrane-bound sensors and signal transduction systems, which are two-component systems (TCS) and extracytoplasmic function (ECF) sigma factors (Helmann 2002; Raivio and Silhavy 1999, 2001). In Bacillus subtilis, secretion stress is sensed at least by the CssRS TCS (a homologue of the Escherichia coli CpxRA). The CssRS TCS regulates the htrA and htrB genes, which encode the cell envelope-associated quality control proteases HtrA and HtrB (Darmon et al. 2002; Hyyryläinen et al. 2001; Noone et al. 2001). Overexpression of a secretory protein and the resulting misfolding activates the CssRS TCS, resulting in induction of the htrA and htrB genes (Darmon et al. 2002; Hyyryläinen et al. 2001). The HtrA and HtrB proteases degrade misfolded proteins and thereby rescue the cell—accumulation of misfolded proteins in the cell envelope is lethal (Hyyryläinen et al. 2001). The PrsA lipoprotein, a parvulin-type peptidyl-prolyl cis/trans isomerase, is anchored to the cell membrane and catalyses folding of secretory proteins (Vitikainen et al. 2001, 2004). The prsA3 mutation decreases the level of PrsA and increases misfolding of secretory proteins at the membrane/cell wall interface, increasing secretion stress and induction of the htrA and htrB genes (Hyyryläinen et al. 2001). PrsA is an essential protein in B. subtilis, suggesting that it also catalyses the folding of some essential protein in the cell envelope (Vitikainen et al. 2001).

It is important to characterise stress responses of industrial micro-organisms to obtain knowledge that can be used to optimise industrial production strains and processes. In this study we used DNA macroarrays to characterise secretion stress responses of B. subtilis. Bacterial cells were subjected to secretion stress by expressing a secretory heterologous α-amylase (AmyQ) at a high level in both a wild-type strain and the prsA3 mutant (severe secretion stress). We also characterised the CssRS regulon and genes that were differentially expressed when cells were depleted of the PrsA foldase. Severe secretion stress induced numerous genes, most significantly htrA, htrB, liaIHG (yvqIHG) and yqxL. The patterns of differentially expressed genes suggest roles for both functionally known and unknown proteins under stress conditions.

Materials and methods

Bacterial strains, plasmids, and growth media

The bacterial strains and plasmids used are listed Table 1. Strains that express AmyQ α-amylase harbour the plasmid pKTH3339 in which the amyQ gene is under the control of a xylose-inducible promoter (P xyn -amyQ) (Vitikainen et al. 2001). Bacteria were cultured in modified 2×L-broth (2% tryptone, 1% yeast extract and 1% NaCl). Erythromycin (Em) and chloramphenicol (Cm) were used at concentrations of 1 and 5 μg ml−1, respectively. Expression of P xyn -amyQ was induced in early exponential growth phase by adding either 0.2 or 0.02% xylose. P spac -prsA was induced at the beginning of growth to express PrsA either at a low sublethal level (24 μM IPTG) or at a near wild-type level (1 mM IPTG) (Vitikainen et al. 2001).

Table 1 Bacterial strains/plasmid
Full size table

Isolation of total RNA

Expression of P xyn -amyQ was induced with xylose in early exponential growth phase (Klett∼60), and the cells in 4 ml culture were harvested after 30 min of induction (cell density of about Klett 100). The cells were resuspended in 400 μl culture medium and mixed with a mixture containing 1.5 g glass beads (0.1 mm diameter; Sigma, St. Louis, Mo.), 50 μl 10% SDS, 50 μl sodium acetate pH 5.2, and 500 μl phenol/chloroform/isoamylalcohol (25:24:1). The mixture was briefly vortexed and then frozen in liquid nitrogen. Bacteria were allowed to thaw in a horizontal shaker at room temperature for 6 min with vigorous shaking. After centrifuging at 4°C for 2 min, the aqueous phase was withdrawn and extracted with chloroform. The aqueous phase was again separated by centrifugation and RNA was isolated using a High Pure RNA Isolation Kit (Roche, Mannheim, Germany) according to the manufacturer’s instructions.

Synthesis of 33P-labelled cDNA and hybridisation

For cDNA synthesis, 23 μl RNA (10 μg), 4 μl primer mix (specific primers for all B. subtilis genes, Sigma Genosys, Haverhill, UK), and 3 μl 10× hybridisation buffer (100 mM Tris-HCl pH 7.9, 10 mM EDTA, 2.5 M KCl) were mixed and heated to 95°C for 10 min. The temperature was slowly decreased to 42°C over 20 min. A solution containing 12 μl 5× first strand buffer (Sigma Genosys), 6 μl 0.1 M DTT, 2 μl dATP (10 mM), 2 μl dTTP (10 mM), 2 μl dGTP (10 mM), 4.5 μl [α-33P]-dCTP (10 μCi ml−1) and 1.5 μl Superscript II (Gibco BRL, Paisley, UK) reverse transcriptase (RTase) was mixed with the RNA solution. The RTase reaction mixture was incubated at 42°C for 90 min and then the reaction was stopped by adding 2 μl 0.5 M EDTA and 6 μl 3 M NaOH, and incubating at 65°C for 30 min and at 22°C for 15 min. The reaction mixture was neutralised by adding 20 μl 1 M Tris-HCl pH 8.0 and 6 μl HCl. Unincorporated labelled nucleotides were removed by using MicroSpin G-25 columns (Amersham, Piscataway, N.J.) according to the manufacturer’s instructions. The 33P-labelled cDNA was used to hybridise Panorama B. subtilis gene array filters (Sigma Genosys) according to the manufacturer’s instructions. The Panorama arrays contain duplicate spots of 4,107 open reading frames, representing all B. subtilis protein-coding genes. The array filters were exposed to phosphoimager screens to analyse the transcriptome profiles by phoshoimagery using a Fluorescent Image Analyzer FLA-2000 (Fujifilm).

Quantification of gene expression signals

The gene expression signals were quantified by using ArrayVision software (http://www.imagingresearch.com). Spot intensities were corrected by subtracting the background intensity and normalised by dividing with the mean spot intensity of all the genes. Experiments were carried out either two or three times with biologically independent samples. Microsoft Excel software was also used in data analysis. Genes with at least 2-fold differential expression, and signal-to-noise ratios greater than three in at least two independent experiments were considered induced. Reduced gene expression was considered significant if the difference in expression was at least 3-fold. Some smaller differences in gene expression are also listed in the tables if the genes belong to operons in which at least one gene fulfilled the criteria.

Results

DNA macroarray analysis of secretion stress response

B. subtilis cells were subjected to secretion stress by inducing the P xyn -amyQ gene construct on plasmid pKTH3339 with 0.2% xylose in the exponential growth phase (Klett 60). DNA macroarray analysis was carried out with exponentially growing cells to avoid the bacterial heterogeneity of later growth phases and consequent variation of data. Another set of DNA array experiments was carried out with a strain that additionally contained the prsA3 mutation (severe secretion stress).

In the prsA+ strain secreting AmyQ, the htrA and htrB (yvtA) genes, encoding extracytoplasmic quality-control proteases, were induced 4.3- and 5.8-fold, respectively, and yqxL, encoding a putative protein similar to CorA-type Mg2+ transporters, was induced 2.3-fold as compared to the non-secreting strain (Table 2). Four other genes (plsX, rpoB, rpsJ and yesO) were moderately induced (1.6- to 4.8-fold). These seven genes were also all upregulated in the prsA3 mutant secreting AmyQ (prsA3+AmyQ vs prsA3) (Table 2). In the prsA3 mutant the induction ratios of htrA, htrB and yqxL were clearly higher than in the prsA+ strain, at 12.3-fold, 15.2-fold and 13.8-fold, respectively (Table 2), suggesting that these genes respond to protein misfolding in the cell envelope. Interestingly, yvqI (renamed liaI in Mascher et al. 2004) and its downstream genes liaH (yvqH) and liaG (yvqG) were upregulated 2.1- to 10.3-fold (Table 3). Several of the induced genes encode components involved in protein secretion or general adaptation to atypical conditions. The csaA gene (3.5-fold) encodes a chaperone-like factor involved in early stages of protein secretion (Muller et al. 2000) and ffh (2.1-fold) encodes a signal recognition particle (SRP)-like component. The ylxM gene of the bicistronic ylxM-ffh operon, encoding a conserved functionally unknown protein, was also induced (2.5-fold). Some of the induced genes are involved in cell wall biosynthesis (Table 3). The dlt operon (1.6- to 2-fold) encodes the d-alanylation system of teichoic and lipoteichoic acids, and the pbpA (2.8-fold) gene encodes a penicillin-binding protein. The exuR and uxaB (alternative names yjmH and yjmI, respectively) genes, encoding a putative repressor of hexuronate utilization and a putative tagaturonate reductase, respectively, were strongly induced (about 5-fold) in a prsA3-dependent manner (Table 3). One of the induced genes was cssS (2-fold), which encodes the sensor histidine kinase of the CssRS TCS, a regulator of htrA and htrB. In the absence AmyQ secretion, the prsA3 mutation did not affect gene expression (prsA3 vs prsA+, data not shown).

Table 2 Genes induced by oversecretion of AmyQ α-amylase in both the prsA wild type and the prsA3 mutant
Full size table
Table 3 Genes induced by severe secretion stress in the prsA3 mutant
Full size table

AmyQ oversecretion alone did not reduce expression of any gene, but severe secretion stress (prsA3+AmyQ) reduced the expression of 23 genes at least 3-fold (Table 4). Interestingly, yacD, which encodes a PrsA-like protein, was strongly downregulated (0.22-fold) in the secretion-stressed prsA3 mutant (Table 4). Several multicistronic operons were also repressed. These include flhOP (0.15- to 0.22-fold), encoding flagellar proteins, lytABC (0.33- to 0.56-fold), encoding proteins involved in autolysis, and lonB (0.19-fold), encoding an ATP-dependent protease.

Table 4 Genes with reduced expression level in the prsA3 mutant oversecreting AmyQ
Full size table

The CssRS regulon and stress response to the absence of HtrA-type proteases

Inactivation of the CssRS TCS increases protein misfolding in the cell envelope (Hyyryläinen et al. 2001). The increased misfolding due to the absence of CssRS and Htr proteases may be sensed by other signal transduction mechanisms. In order to study this stress response and to identify genes belonging to the CssRS regulon, gene expression patterns of a cssS knockout mutant and its wild-type parent were compared by DNA array (cssS− vs cssS+) in the exponential phase of growth. In both strains, P xyn -amyQ was induced with 0.02% xylose and consequently AmyQ was synthesised and secreted at a low level. The only genes that were expressed at clearly lower levels in the cssS mutant were htrA and htrB (0.20-fold and 0.15-fold, respectively), suggesting that only these two genes are directly regulated by the CssRS TCS. Another prominent finding was the induction of the liaIHG (yvqIHG) genes (2.2- to 5.3-fold). Nine other functionally unknown genes were also moderately (about 2-fold) induced, including ydjF, which encodes a phage shock protein A homologue, and downstream genes of the operon (data not shown).

Sublethal depletion of PrsA foldase results in induction of genes involved in cell wall synthesis

We also used the DNA macroarray method to analyse changes in gene expression pattern in cells depleted of PrsA foldase to a sublethal level. These DNA arrays were carried out with cultures with a cell density of Klett 100 and the cells did not overproduce AmyQ. PrsA depletion might result in compensatory induction of a gene that encodes an essential component whose folding is dependent on PrsA (cascade effect) if such a gene is expressed from a regulatable promoter. This could also be an autoregulatory phenomenon. The above secretion stress results suggested that such feedback compensatory mechanisms exist in the cell. PrsA depletion may also cause a secretion stress response.

The CssRS-regulated quality-control protease gene htrB was induced 6.3-fold in PrsA-depleted cells (Table 5), suggesting that PrsA depletion causes accumulation of misfolded native proteins in the cell envelope. Several genes involved in cell wall synthesis were also moderately upregulated. These were dltBCE (1.7- to 2.3-fold), ddlA (1.9-fold) encoding the d-alanyl-d-alanine ligase A, and murF (1.8-fold), which probably forms an operon with ddlA. The maf (2.0-fold) mreB (1.5-fold) and mreD (1.5-fold) genes, which all belong to the mre operon and encode proteins involved in septum formation and cell-shape determination, respectively, and pbpX (2.7-fold), which encodes a penicillin-binding protein, were induced. Also induced were the σX-regulated genes, suggesting a stress response via this extracytoplasmic function sigma factor (Table 5).

Table 5 Genes induced in cells depleted of PrsA foldase
Full size table

Discussion

The secretion stress response in B. subtilis was studied by DNA macroarray analysis. Our approach was to induce secretion/folding stress by various means in order to identify components upregulated or downregulated by protein oversecretion and misfolding (the secretion stress regulon), and which may be involved in combatting the harmful effects of secretion stress. In addition to previously known components of the secretion stress regulon several new factors were identified.

High-level AmyQ secretion induced only seven genes in exponentially growing B. subtilis, suggesting that secretion stress was moderate under these conditions. The upregulated genes included htrA and htrB, consistent with previously published results (Hyyryläinen et al. 2001). The severe secretion stress in the prsA3 mutant caused a clearly stronger induction of htrA and htrB than mere AmyQ oversecretion alone, consistent with the fact that the degree of induction of these genes correlates with the degree of protein misfolding in the cell envelope (Hyyryläinen et al. 2001). A proteomic analysis has also revealed elevated levels of HtrA protease in the culture medium under secretion stress conditions (Antelmann et al. 2003). We also identified a third gene, yqxL, which responded to the secretion stress in a manner similar to that of the htr genes. The yqxL gene was weakly activated by mere AmyQ oversecretion (2.4-fold) but strongly by the severe secretion stress in the prsA3 mutant (13.9-fold). The YqxL protein belongs to the family of CorA-type Mg2+ transporters. In Salmonella typhimurium and E. coli, CorA is the primary influx system for extracellular magnesium (Kehres et al. 1998). In contrast to YqxL, the S. typhimurium and E. coli homologues are constitutively expressed. Some of the CorA-like proteins may have a function other than Mg2+ transport (Kehres et al. 1998). How does secretion stress and accumulation of misfolded proteins at the membrane-cell wall interface induce yqxL expression? Misfolded proteins could directly activate some still unknown signal transduction pathway that regulates yqxL, or indirectly affect the microenvironment at the membrane-cell wall interface and thereby affect activation of a regulatory mechanism controlling yqxL expression.

In addition to htrA, htrB and yqxL, severe secretion stress induced 17 genes or operons (at least 2-fold), among them the 10-fold induced liaI (yvqI) and downstream liaH (yvqH) and liaG (yvqG). Induction of lia (yvq) genes is mediated by the LiaRS (YvqCE) TCS that is encoded by the liaSR (yvqEC) genes in the same gene cluster as liaIHG (yvqIHG) (Kobayashi et al. 2001; Mascher et al. 2003; H.-L. Hyyryläinen et al. unpublished). Cell wall antibiotics that interfere with the lipid II cycle in the cytoplasmic membrane, and thus peptidoglycan synthesis, are strong inducers of LiaRS (YvqCE) (Mascher et al. 2004). We have shown that LiaRS (YvqCE) is also strongly activated (>100-fold induction ratio) by cationic antimicrobial peptides such LL-37 and protegrin-1 (Milla Pietiäinen et al., MS submitted). This study shows that secretion stress activates LiaRS (YvqCE), possibly due to the accumulation of misfolded proteins in the cell envelope and possibly causing perturbation in the membrane, although as an activator it is clearly weaker than cell wall antibiotics and cationic antimicrobial peptides. The roles of these proteins in the stress response remain unclear, but LiaH (YvqH) shows significant homology with the E. coli phage shock protein A (PspA).

DNA array analysis of the CssRS regulon suggested that this TCS is probably dedicated to regulating only the htrA and htrB genes. Fujita and his collaborators found that overexpression of cssR (yvqA) induced, in addition to the htr genes, two functionally unknown genes, ygxB and yjgD (Kobayashi et al. 2001). In our analysis, these two genes were not found to belong to the CssRS regulon. The inactivation of cssS and consequent absence of HtrA and HtrB proteases also caused stress, as evidenced by the induction of the stress-sensitive liaIHG (yvqIHG) genes via the LiaRS (YvqCE) TCS.

Several of the gene induction patterns suggest that cells combat the harmful effects of secretion/folding stress by compensatory up- or down-regulation of genes having a role in the affected functions. Components of the secretion apparatus are typically expressed from constitutive housekeeping genes. Our results, however, show that the CsaA molecular chaperone and the Ffh signal recognition particle-like component, which are involved in the early phase of protein secretion, are significantly induced to combat secretion stress. The upregulation of genes involved in cell wall synthesis/modification probably has a similar role. Among the induced operons was dlt, which encodes a system for the d-alanylation of teichoic acids and consequently modulates the charge density of the cell envelope. The dlt operon is regulated by the extracytoplasmic function sigma factor σX (Cao and Helmann 2004), suggesting that σX senses secretion/misfolding stress. There is a signal transduction pathway for sensing of misfolded proteins also in eukaryotic cells. The accumulation of misfolded proteins in the lumen of the endoplasmic reticulum (ER) activates the unfolded protein response, consequently upregulating the expression of several ER foldases (Patil and Walter 2001). Genes involved in many other cellular functions were also induced. Interestingly, the exuR (yjmH) and uxaB (yjmI) genes, which encode the ExuR repressor of the hexuronate utilization operon (exu locus) and the UxaB tagaturonate reductase, respectively, both involved in the utilization of galacturonate, were induced almost 5-fold. These genes are upregulated by galacturonate and repressed by glucose (Mekjian et al. 1999). The upregulation of exuR and uxaB expression was not caused by the xylose added to the culture medium to induce P xyn -amyQ, since in the absence of prsA3, mere xylose-induced amyQ expression did not result in the upregulation of these genes. It has been shown that there are two promoters that control gene expression in the exu locus, the σA-dependent exuP1 and the σE-dependent exuP2, and the latter promoter has been suggested to be responsible for the expression of exuR and uxaB (Mekjian et al. 1999). Our DNA array analysis, however, revealed that the yjmF and exuT genes of the exu locus, which are located upstream from exuR and uxaB, were expressed at significantly lower levels (>10-fold) than exuR and uxaB (data not shown). This suggests that a third promoter (exuP3) exists between exuT and exuR, which controls expression of downstream genes. Our transcriptome analysis suggests that posttranslocational protein misfolding induced exuR and uxaB from the predicted exuP3 by an unknown mechanism.

The reduced expression levels of numerous genes during severe secretion/folding stress can also be seen as an attempt by the cell to cope with the harmful effects of this stress. The expression of autolysins was decreased, possibly affecting the rate of turnover of the cell wall, together with increased expression of genes involved in cell wall synthesis. The yacD gene, which is the only prsA paralogue in B. subtilis, was strongly repressed, whereas prsA was expressed constitutively, suggesting that these parvulin-type folding factors have distinctly different roles in the cell wall.

The most prominent groups of genes induced by sublethal PrsA depletion were those involved in cell wall synthesis, including the ddl, dlt and maf-mre operons, and the σX regulon. Do these differentially expressed genes suggest a mechanism for the lethality of PrsA depletion? PrsA catalyses protein folding, and is probably required for the folding of some essential cell wall or membrane protein(s) (Vitikainen et al. 2001, 2004). A possible PrsA target could be the essential MreC cell-shape determining protein (Lee and Stewart 2003; Kobayashi et al. 2003). MreC is predicted to have one transmembrane segment at the N-terminus, and the large C-terminal domain is probably located at the membrane/cell wall interface (topology prediction at the TMHMM server http://www.cbs.dtu.dk/services/TMHMM) in the same compartment where PrsA is located. Furthermore, sublethal and lethal PrsA depletion cause similar changes in cell morphology as MreC depletion (Lee and Stewart 2003; Hyyryläinen et al. 2001).