Functional Analysis of Paralogous Thiol-disulfide Oxidoreductases in Bacillus subtilis *

The in vivo formation of disulfide bonds, which is critical for the stability and/or activity of many proteins, is catalyzed by thiol-disulfide oxidoreductases. In the present studies, we show that the Gram-positive eubacteriumBacillus subtilis contains three genes, denotedbdbA, bdbB, and bdbC, for thiol-disulfide oxidoreductases. Escherichia coli alkaline phosphatase, containing two disulfide bonds, was unstable when secreted by B. subtilis cells lacking BdbB or BdbC, and notably, the expression levels of bdbB and bdbC appeared to set a limit for the secretion of active alkaline phosphatase. Cells lacking BdbC also showed decreased stability of cell-associated forms of E. coli TEM-β-lactamase, containing one disulfide bond. In contrast, BdbA was not required for the stability of alkaline phosphatase or β-lactamase. Because BdbB and BdbC are typical membrane proteins, our findings suggest that they promote protein folding at the membrane-cell wall interface. Interestingly, pre-β-lactamase processing to its mature form was stimulated in cells lacking BdbC, suggesting that the unfolded form of this precursor is a preferred substrate for signal peptidase. Surprisingly, cells lacking BdbC did not develop competence for DNA uptake, indicating the involvement of disulfide bond-containing proteins in this process. Unlike E. coli and yeast, none of the thiol-disulfide oxidoreductases of B. subtilis was required for growth in the presence of reducing agents. In conclusion, our observations indicate that BdbB and BdbC have a general role in disulfide bond formation, whereas BdbA may be dedicated to a specific process.

Disulfide bonds are essential for the activity and/or stability of numerous eubacterial and eukaryotic proteins. Disulfide bonds can be formed spontaneously in vitro, but this process is much slower and less effective than the formation of disulfide bonds in vivo, where it is catalyzed by thiol-disulfide oxidoreductases (1,2). In Gram-negative bacteria, the formation of disulfide bonds occurs in the periplasm, which is a relatively oxidizing cellular compartment compared with the cytoplasm. In Escherichia coli six proteins involved in disulfide bond for-mation of periplasmic proteins have been identified: DsbA, DsbB, DsbC, DsbD, DsbE, and DsbG (see Refs. [3][4][5][6][7][8][9][10]. Current models (for review, see Refs. 11 and 12) propose that DsbA acts as the major oxidase in disulfide bond formation in the periplasm. The integral membrane protein DsbB is required to maintain the oxidized state of DsbA. In addition, DsbC acts as an isomerase that is needed for "proof-reading" of newly formed disulfide bonds. DsbC is kept in a reduced state by the integral membrane protein DsbD. The roles of DsbE and DsbG are less clear, but it has been suggested that DsbG is involved in maintaining a proper redox balance between the DsbA/B and DsbC/D systems (3) and that DsbE is involved in the folding of specific proteins such as c-type cytochromes (10,11). Strains lacking functional dsb genes display pleiotropic phenotypes, which include low levels of alkaline phosphatase (PhoA) 1 activity in the periplasm, sensitivity to reducing agents, and lack of motility (see Refs. 4 and 9).
In contrast to Gram-negative bacteria, very little is known about the catalysis of disulfide bond formation in proteins of Gram-positive bacteria. In fact, only three extracytoplasmic proteins with disulfide bonds were, thus far, identified in Bacillus subtilis, which is the paradigm for many studies in Gram-positive bacteria. First, the ComGC and ComGG proteins, which are essential for the uptake of DNA by competent cells, contain one intramolecular disulfide bond and one intermolecular disulfide bond, respectively (13). Second, the lantibiotic sublancin, encoded by the sunA gene, was shown to contain two disulfide bonds (14). In addition, disulfide bonds are formed in homologous proteins of B. subtilis with engineered cysteine residues such as certain neutral protease (NprE) mutants (15) and heterologous proteins such as PhoA of E. coli (16).
The only Bacillus protein that has, thus far, been invoked in the formation of disulfide bonds is Bdb from Bacillus brevis (17). This protein, which is predicted to have a cleavable signal peptide at its amino terminus, can complement a mutation in the E. coli dsbA gene. Because Bdb appeared to be translocated into the periplasm of E. coli, it was suggested that Bdb is translocated across the membrane of B. brevis and localized at the periphery of the cell envelope. The physiological role of Bdb is not known because the construction of a bdb mutant has not been reported.
A common feature of many thiol-disulfide oxidoreductase mutant strains, both of E. coli (see above) and Saccharomyces cerevisiae (18), is their sensitivity to reducing agents. Therefore, our previous attempts to identify thiol-disulfide oxidoreductases of B. subtilis were based on random approaches involving transposon mutagenesis and subsequent screens for * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  ␤-mercaptoethanol-sensitive mutants. These attempts resulted in the identification of three genes (i.e. ackA, resE, and yhdE) that are required for growth in the presence of reducing agents. Interestingly, the ackA gene, encoding acetate kinase (Gen-Bank accession number L17320), and the resE gene, encoding a histidine kinase (GenBank accession number 410142), turned out to be required for anaerobic growth but not for disulfide bond formation in secreted proteins such as PhoA of E. coli. The yhdE gene, which specifies a protein of unknown function (Gen-Bank accession number Y14082), was not involved in anaerobic growth or disulfide bond formation (data not shown). These unpublished observations suggested that B. subtilis either lacked thiol-disulfide oxidoreductases, that these enzymes were essential for growth, or that thiol-disulfide oxidoreductases were not required for growth on reducing agents.
To investigate further whether B. subtilis does contain thioldisulfide oxidoreductases that are actively involved in the formation of disulfide bonds in extracytoplasmic proteins we made use of the recently published genome sequence of this organism (19). Even though none of the predicted proteins of B. subtilis was annotated as a putative thiol-disulfide oxidoreductase, our data base searches revealed the presence of at least three genes for proteins with (limited) similarity to known thiol-disulfide oxidoreductases from other organisms. In the present studies, we document the identification of these genes, which we have named bdbA, bdbB, and bdbC (for Bacillus disulfide bond formation). We show that the BdbB and BdbC proteins are required for the activity and stability of disulfide bond-containing secretory reporter proteins. Table I lists the plasmids and strains used. TY medium contained Bacto tryptone (1%), Bacto yeast extract (0.5%), and NaCl (1%). Minimal medium (GCHE medium; Ref. 20) contained 1% glucose, 0.2% potassium Lglutamate, 100 mM potassium phosphate buffer (pH 7), 3 mM trisodium citrate, 3 mM MgSO 4 , 22 mg/liter ferric ammonium citrate, 0.1% casein hydrolysate, and 50 mg/liter L-tryptophane. Antibiotics were used in the following concentrations (in g/ml): chloramphenicol, 5; erythromycin, 1; kanamycin, 10; and ampicillin, 50. IPTG was used at 1 mM.

Plasmids, Bacterial Strains, and Growth Conditions-
DNA Techniques-Procedures for DNA purification, restriction, ligation, agarose gel electrophoresis, and transformation of competent E. coli DH5␣ cells were carried out as described by Sambrook et al. (21). Enzymes were from Roche Molecular Biochemicals (Germany). B. subtilis was transformed by adding DNA to cells growing in GCHE medium at the end of the exponential growth phase and continued incubation for 3-4 h at 37°C. The nucleotide sequences of primers used for PCR (5Ј-3Ј) are listed below; nucleotides identical to genomic template DNA are printed in capital letters, and restriction sites used for cloning are underlined.
To construct B. subtilis IbdbA, a fragment containing the ribosome binding site, the start codon and the first 180 nucleotides of the bdbA gene, but not the bdbA promoter(s), were amplified with the primers AB60bdbA (aaaagcttCGGAAAATAAGGAGTATTC) and AB61bdbA (aaggatccCAAGGAGGACAA CTTGTC). The amplified fragment was ligated into pMutin2, which resulted in plasmid pMI-bdbA. B. subtilis IbdbA was obtained by Campbell-type integration (single crossover) of plasmid pMI-bdbA into the chromosome of B. subtilis 168.
To construct B. subtilis ⌬bdbB, an internal fragment of the bdbB gene (191 nucleotides) was amplified by PCR with the oligonucleotides AB70bdbB (aaaagcttCCTATACCTATTATCTTAC) and AB71bdbB (aag-gatccTAGATACTCTACTTC). The amplified fragment was cloned into pMutin2, which resulted in plasmid pMI-bdbB. B. subtilis ⌬bdbB was obtained by Campbell-type integration of plasmid pMI-bdbB into the chromosome of B. subtilis 168.
To construct B. subtilis ⌬bdbC an internal fragment of the bdbC gene (202 nucleotides) was amplified by PCR with the oligonucleotides AB72bdbC (aaaagcttCTGTGCTGGTACCAGCG) and AB73bdbC (aag-gatccCGAGCACGGCACGCC). The amplified fragment was cloned into pMutin2, which resulted in plasmid pMI-bdbC. B. subtilis ⌬bdbC was obtained by Campbell-type integration of plasmid pMI-bdbB into the chromosome of B. subtilis 168. Correct integration of all plasmids into the chromosome of B. subtilis was verified by Southern hybridization.
Competence and Sporulation-Competence for DNA binding and uptake was determined by transformation with plasmid or chromosomal DNA (22). The efficiency of sporulation was determined by overnight growth in Schaeffer's medium (23), killing of cells with 0.1 volume of chloroform, and subsequent plating.
Enzyme Assays-The assay for alkaline phosphatase activity in growth media, and the calculation of PhoA units (per OD 600 ) were performed as described in Ref. 24. p-Nitrophenyl phosphate (Sigma) was used as the substrate.
To assay cellular ␤-galactosidase levels, overnight cultures were diluted 100-fold in fresh medium, and samples were taken at hourly intervals for optical density (OD) readings at 600 nm and ␤-galactosidase activity determination. The ␤-galactosidase assay and the calculation of ␤-galactosidase units (per OD 600 ) were performed as described in Ref. 25. 2-Nitrophenyl-␤-D-galactopyranoside (Sigma) was used as the substrate.
Western Blot Analysis-Western blotting was performed using a semidry system as described in Ref. 26. After separation by SDS-PAGE, proteins were transferred to polyvinylidene difluoride membranes (Roche Molecular Biochemicals). Proteins were visualized with specific antibodies and horseradish peroxidase anti-rabbit IgG conjugates using the ECL detection system of Amersham Pharmacia Biotech.

Identification of Three Putative
Thiol-disulfide Oxidoreductases-To identify putative thiol-disulfide oxidoreductases, data base searches were performed using the amino acid sequences of Bdb of B. brevis, the Dsb proteins of E. coli, and various thiol-disulfide oxidoreductases of yeast. Only two of these homology searches revealed proteins with significant sequence similarity to known thiol-disulfide oxidoreductases. First, a protein of 137 residues, specified by the yolI gene (located at 193.6 o on the B. subtilis chromosome; Ref. 19) showed sequence similarity to Bdb of B. brevis (43% identical residues and conservative replacements; Fig. 1A). By analogy to the bdb gene of B. brevis, we renamed the yolI gene of B. subtilis bdbA. Second, a protein of 148 residues, specified by the yolK gene, showed significant sequence similarity to the DsbB protein of E. coli (47% identical residues and conservative replacements; Fig. 1B). The yolK gene was renamed bdbB. The bdbA and bdbB genes are separated by only one gene of unknown function (i.e. yolJ). A third protein of 138 residues, specified by the yvgU gene (located at 293.6 o on the B. subtilis chromosome; see Ref. 20), was identified through homology searches with BdbB (51% identical residues and conservative replacements; Fig. 1B Notably, the spacing between Cys-104 and Cys-130 of DsbB of E. coli is substantially larger than the spacing between the corresponding cysteine residues of BdbB and BdbC. Like Bdb of B. brevis, BdbA is predicted to have a cleavable signal peptide but no membrane-spanning domains, suggesting that BdbA is translocated across the cytoplasmic membrane and retained in the cell wall or secreted into the growth medium. In contrast, algorithms of Sipos and von Heijne (29) predict that both BdbB and BdbC have four membrane-spanning domains, the amino and carboxyl termini being localized in the cytoplasm. This is in good agreement with the predicted membrane topology of DsbB from E. coli (28).
BdbA, BdbB, and BdbC Are Not Essential for Growth, Viability, and Resistance to Reducing Agents-To analyze the functions of BdbA, BdbB, and BdbC in B. subtilis, three mutant strains were constructed with the chromosomal integration plasmid pMutin2. In the first strain, denoted B. subtilis IbdbA, the coding sequence of the bdbA gene was left intact, but the promoter of this gene was replaced with the IPTG-inducible Pspac promoter present on pMutin2 ( Fig. 2A). In the second and third strains, denoted B. subtilis ⌬bdbB and B. subtilis ⌬bdbC, respectively, the coding sequences of bdbB and bdbC were disrupted by the integrated pMutin2 (Figs. 2, B and C). Irrespective of the growth medium or the presence of IPTG, B. subtilis IbdbA, ⌬bdbB, and ⌬bdbC showed growth rates similar to that of the parental strain B. subtilis 168 (data not shown), demonstrating that bdbA, bdbB, or bdbC is not essential for growth and viability of the cells, at least under the conditions used. Interestingly, B. subtilis ⌬bdbC was unable to develop competence for DNA uptake, whereas competence development was not affected in B. subtilis IbdbA and ⌬bdbB. None of these strains was affected in sporulation. 2 B. subtilis IbdbA, ⌬bdbB, and ⌬bdbC were tested for sensitivity to reducing agents by plating these mutant strains on TY-agar plates containing various concentrations of DTT, rang- ing between 5 and 25 mM. Significant inhibition of growth was observed at DTT concentrations higher than 10 mM, but, interestingly, none of the bdb mutant strains displayed an increased sensitivity for DTT compared with the parental strain (data not shown). This observation suggested that BdbA, BdbB, and BbC were not involved in disulfide bond formation or that B. subtilis lacked proteins with disulfide bonds required for growth and viability. In what follows, we show that the lack of DTT sensitivity of B. subtilis bdbB and bdbC mutants is, most likely, the result of the absence of essential proteins with disulfide bonds required for their function or overlapping specificities of BdbB and BdbC.
BdbB and BdbC Are Required for the Stability of Secretory Proteins Containing Disulfide Bonds-As a first approach to investigate the role of BdbA, BdbB, and BdbC in the formation of disulfide bonds in proteins that are exported from the cytoplasm and secreted into the growth medium, PhoA of E. coli was used. For two reasons PhoA is a particularly useful reporter protein for these studies. First, it contains two intramolecular disulfide bonds, which are essential for its activity and stability (30). Second, the formation of these disulfide bonds requires the activity of an oxidase such as DsbA of E. coli (see Refs. 4 and 9). Therefore, the IbdbA, ⌬bdbB, or ⌬bdbC mutation was introduced in a strain containing plasmid pPSPhoA5. This plasmid specifies PhoA of E. coli fused to the signal peptide and pro region of a lipase from Staphylococcus hyicus. When grown in TY medium, B. subtilis secretes this hybrid PhoA precursor protein efficiently into the medium where it is active, implying that its two disulfide bonds are formed correctly. 3 Next, PhoA activity was determined in the culture supernatants of B. subtilis IbdbA (grown in TY medium in the absence of IPTG), ⌬bdbB, ⌬bdbC, or the parental strain 168, all of which contained pPSPhoA5. Interestingly, compared with the parental strain, the PhoA activities in the growth media of B. subtilis ⌬bdbB and ⌬bdbC were reduced about 4-and 15-fold, respectively. In contrast, the PhoA activity in the medium of B. subtilis IbdbA was not affected (Table II). As shown by Western blotting (Fig. 3A), the reduced PhoA activities in the media of B. subtilis ⌬bdbB and ⌬bdbC were paralleled by the presence of about 3-and 6-fold reduced amounts of PhoA protein, respectively. In contrast, the secretion of the Bacillus amyloliquefaciens ␣-amylase AmyQ, which does not contain disulfide bonds (31), was not affected in these strains. This was demonstrated both by Western blotting (Fig. 3C) and ␣-amylase activity assays, which showed an average activity of 457 (Ϯ28) units/OD 600 in the different growth media. The reduced accumulation of the PhoA protein in the media of B. subtilis ⌬bdbB and ⌬bdbC was not caused by reduced rates of synthesis and translocation of PhoA, as evidenced by pulse labeling experiments in the presence or absence of sodium azide, an efficient inhibitor of the pre-protein translocase (32). These pulse labeling experiments (Fig. 4A) showed that within a labeling period of 2 min, these mutant strains synthesized amounts of PhoA similar to that of the parental strain. Furthermore, these exper-3 J. Meens and R. Freudl, personal communication.

FIG. 2. Construction of bdb mutant strains of B. subtilis.
Panel A, schematic presentation of the bdbA locus of B. subtilis IbdbA. By a single-crossover event (Campbell-type integration), the bdbA promoter region (PbdbA) was replaced with the Pspac promoter of the integrated plasmid pMutin2, which can be repressed by the product of the lacI gene. Simultaneously, the spoVG-lacZ reporter gene of pMutin2 was placed under the transcriptional control of the bdbA promoter region. The chromosomal fragment from the bdbA region, which was amplified by PCR and cloned into pMutin2, is indicated with black bars. Only the restriction sites relevant for the construction are shown (B, BamHI; H, HindIII); ori pBR322, replication functions of pBR322; Ap r , ampicillin resistance marker; Em r , erythromycin resistance marker; T 1 T 2 , transcriptional terminators on pMutin2; bdbAЈ, 3Ј-truncated bdbA gene. Panel B, schematic presentation of the bdbB locus of B. subtilis ⌬bdbB. The bdbB gene was disrupted by a Campbell-type integration of plasmid pMutin2. Simultaneously, the spoVG-lacZ reporter gene of pMutin2 was placed under the transcriptional control of the bdbB promoter region (PbdbB). The chromosomal fragment from the bdbB region, which was amplified by PCR and cloned into pMutin2, is indicated with black bars. bdbBЈ, 3Ј-truncated bdbB gene; ЈbdbB, 5Ј-truncated bdbB gene. SL, putative rho-independent terminator of transcription. Panel C, schematic presentation of the bdbC locus of B. subtilis ⌬bdbC. The bdbC gene was disrupted with pMutin2, using a strategy similar to the one described for the disruption of bdbB in panel B. PbdbC, promoter region of the bdbC gene; bdbCЈ, 3Ј-truncated bdbC gene; ЈbdbC, 5Ј-truncated bdbC gene.
iments showed that prepro-PhoA processing to pro-PhoA by signal peptidase depended on the activity of the preprotein translocase (Fig. 4A, ϩNaN 3 ) and was not impaired in B. subtilis ⌬bdbB and ⌬bdbC (Fig. 4A, no addition). The latter findings demonstrate that the reduced accumulation of PhoA in the medium was not caused by a general secretion defect of these strains. Taken together, these observations indicate that the reduced accumulation of PhoA in the medium of B. subtilis ⌬bdbB and ⌬bdbC strains was caused by a reduced stability of this protein which, most likely, is the result of the inefficient formation of disulfide bonds in the absence of BdbB or BdbC.
To examine possible roles of BdbA, BdbB, and BdbC in the secretion of a reporter protein of which the stability, but not the activity, depends on the formation of a single disulfide bond, the E. coli TEM-␤-lactamase (Bla) was used (33). In fact, this reporter protein was shown previously to be prone to proteolytic degradation in the growth medium of B. subtilis (34). For the present studies, the hybrid precursor pre(A13i)-␤-lactamase (pre-A13i-Bla), encoded by plasmid pGDL48, was used. This precursor, which contains a modified signal peptide derived from the B. subtilis YdjM protein, is processed by the type I signal peptidases of B. subtilis, albeit at a low rate (35,36). Consequently, pre-A13i-Bla and mature A13i-Bla are detectable in cells of B. subtilis which, in contrast to PhoA, offers the additional advantage that the effects of bdb mutations on the stability of this reporter protein in the membrane/cell wall environment can be analyzed. For these reasons, Western blotting experiments were performed to monitor the presence of A13i-Bla in cells of B. subtilis IbdbA, ⌬bdbB, ⌬bdbC, and the parental strain, all of which contained pGDL48. Notably, the cellular levels of A13i-Bla were only (mildy) reduced in the strain lacking a functional bdbC gene; compared with the parental strain, the level of pre-A13i-Bla was reduced about 2-fold, whereas the level of mature A13i-Bla was reduced approximately 5-fold (Fig. 3B). The cellular levels of A13i-Bla were not affected in B. subtilis IbdbA or ⌬bdbB (Fig. 3B). As shown by pulse-chase labeling (Fig. 4B), the synthesis of A13i-Bla in B. subtilis ⌬bdbC was not affected, demonstrating that the reduced levels of A13i-Bla in cells of this strain were caused by a reduced stability of the reporter protein. Most likely, this reflects a reduced efficiency of disulfide bond formation in A13i-Bla in the absence of functional BdbC.
Interestingly, the rate of pre-A13i-Bla processing to the mature form was increased significantly in the absence of BdbC; after a chase of 2 min, Ϸ55% (Ϯ3%) of the labeled A13i-Bla was mature in B. subtilis ⌬bdbC, whereas only Ϸ14% (Ϯ2%) was mature in the parental strain (Fig. 4B). This observation indicates that the reduced folding in the absence of BdbC makes pre-A13i-Bla a better substrate for the type I signal peptidases of B. subtilis. A similar but less pronounced effect was observed for processing of prepro-PhoA to pro-PhoA in the absence of BdbB or BdbC (Fig. 4A).
Medium-and Growth Phase-dependent Transcription of bdbA, bdbB, and bdbC-We have shown previously that the transcription of the secDF gene (37), the type I signal peptidase genes sipS and sipT (36,38), and the type II signal peptidase gene lsp (39) depends on the growth phase and medium. To test whether the transcription of the bdbA, bdbB, and bcbC genes is also growth phase-and medium-dependent, we made use of the transcriptional bdb-lacZ gene fusions present in B. subtilis IbdbA, ⌬bdbB, and ⌬bdbC, respectively (Fig. 2). These three strains were grown in minimal or TY medium, and samples  withdrawn at hourly intervals were assayed for ␤-galactosidase activity. Nearly constant low levels of ␤-galactosidase activity were observed during growth of all three strains in minimal medium ( Fig. 5; indicated with open symbols), suggesting that under these conditions bdbA, bdbB, and bdbC are expressed constitutively. In fact, the levels of ␤-galactosidase activity in B. subtilis ⌬bdbC were very low, being close to background levels ( Fig. 5; indicated with the symbol ‚). Completely different results were obtained when B. subtilis IbdbA, ⌬bdbB, and ⌬bdbC were grown in TY medium. In cells of B. subtilis IbdbA and ⌬bdbB, the levels of ␤-galactosidase activity increased during exponential growth, reaching a maximum in the transition phase between exponential and postexponential growth (t ϭ 0). The ␤-galactosidase levels decreased again in the postexponential growth phase ( Fig. 5; indicated with the symbols ࡗ and f, respectively). Compared with cells of B. subtilis IbdbA and ⌬bdbB grown in minimal medium, the ␤-galactosidase levels at t ϭ 0 were about 10-fold higher when these cells were grown in TY medium. In contrast, the ␤-galactosidase levels in cells of B. subtilis ⌬bdbC grown in TY medium were nearly constant during all growth phases ( Fig. 5; indicated with the symbol OE), and they were increased about 5-fold compared with cells grown in minimal medium. Taken together, these observations show that the expression levels of bdbA and bdbB depend both on the growth phase and the growth medium, whereas bdbC is constitutively expressed, the expression level depending on the growth medium.
Because the expression levels of bdbA, bdbB, and bdbC were significantly lower in cells grown in minimal medium compared with TY medium, the effects of minimal medium on the secretion of active PhoA were determined. To this purpose, B. subtilis IbdbA (grown in the absence of IPTG), ⌬bdbB, ⌬bdbC, or the parental strain 168, all of which contained pPSPhoA5, were grown overnight in minimal medium. As shown in Table III, the PhoA activities in the culture supernatants strongly depended on the presence of intact bdbB or bdbC genes; compared with the parental strain, the PhoA activity in the supernatant of B. subtilis ⌬bdbB was reduced about 4-fold, and the PhoA activity in the supernatant of B. subtilis ⌬bdbC cells was reduced about 10-fold. In contrast, the PhoA activity in the supernatant of BdbA-depleted cells was hardly affected. Most importantly, however, the PhoA activities in the culture super-natants of cells grown in minimal medium were reduced about 25-35-fold compared with the PhoA activities in culture supernatants of cells grown in TY medium (Table II). Similar PhoA activity levels were measured after growth in minimal medium supplemented with maltose instead of glucose (data not shown), showing that the reduced PhoA activities were not caused by carbon catabolite repression of PhoA synthesis. Taken together, these data indicate that the expression levels of the bdbB and bdbC genes set a limit to the secretion of active PhoA.

DISCUSSION
In the present studies, we have identified two genes of B. subtilis, denoted bdbB and bdbC, which are required for the stability of secreted proteins with disulfide bonds. Because BdbB and BdbC show similarity to DsbB of E. coli, it seems most likely that these two proteins are thiol-disulfide oxidoreductases, which catalyze the formation of disulfide bonds. Similar to DsbB of E. coli (7), BdbB and BdbC are required for the stability of PhoA, which is secreted into the growth medium of B. subtilis. The strongest defect in PhoA secretion was observed with the bdbC mutant strain. This mutant strain also showed reduced levels of translocated, yet cell-associated forms of a second reporter protein, A13i-Bla. Because BdbB and BdbC are predicted to be integral membrane proteins, our observations suggest that these proteins catalyze the formation of disulfide bonds in secretory proteins during or shortly after their translocation across the membrane. However, we can presently not exclude two alternative possibilities. First, BdbB and/or BdbC could be indirectly involved in the formation of disulfide bonds in secretory proteins by regeneration of an, as yet unknown, oxidase, as proposed for DsbB and DsbA of E. coli (40,41). Second, irrespective of their thiol-disulfide oxidoreductase activity, BdbB and/or BdbC could have chaperone-like activities that promote the folding of A13i-Bla and/or PhoA. For example, it has been shown previously that the endoplasmic reticular protein disulfide isomerase has chaperone-like activities in addition to its disulfide bond isomerase activity (see Refs. 42 and 43).
Interestingly, the bdbA gene, which specifies a homolog of the Bdb protein from B. brevis, was not required for the stability of PhoA or A13i-Bla. This observation is remarkable because the Bdb protein from B. brevis is, to date, the only protein derived from Bacillus species for which thiol-disulfide oxidoreductase activity has been demonstrated directly (17). Thus, it seems that BdbA of B. subtilis may be dedicated to the formation of disulfide bonds in a specific (subset of) protein(s). For example, because the B. subtilis bdbA gene is located immediately downstream of genes required for the synthesis and secretion of the lantibiotic sublancin, which contains two disulfide bonds (14), it is conceivable that BdbA is required for the formation or isomerization of disulfide bonds in sublancin.
Notably, bdbA, bdbB, and bdbC mutant strains of B. subtilis  are not sensitive to DTT, which indicates that this organism does not contain proteins with disulfide bonds that are essential for growth and viability or that the specificities of BdbA, BdbB, and BdbC overlap to some extent. This observation is consistent with our unpublished observation that genes for thiol-disulfide oxidoreductases of B. subtilis could not be identified through transposon mutagenesis and subsequent analysis of DTT-sensitive mutants. These findings are remarkable in view of the fact that many strains of E. coli and yeast, which carry mutations in genes for thiol-disulfide oxidoreductases, display increased sensitivity to reducing agents. The observation that the bdbC mutant strain is completely blocked in the development of competence is consistent with the fact that the ComGC and ComGG proteins of B. subtilis contain disulfide bonds, the formation of which is critical for DNA uptake by competent cells (13,44). Because the bdbB mutant strain was not impaired in competence development, it seems that BdbB, which is a paralog of BdbC, has no or only a very minor role in the formation of disulfide bonds in proteins required for DNA uptake. Thus it seems that the substrate specificities of BdbB and BdbC are not identical. Nevertheless, their substrate specificities do overlap to some extent because both proteins are required for the stability of secreted PhoA of E. coli.
We have reported previously that processing of pre-A13i-Bla was improved significantly by signal peptidase overproduction, both in B. subtilis (35) and E. coli (45). To explain this observation, we hypothesized that pre-A13i-Bla can fold into a conformation that prevents processing by the signal peptidase of E. coli and slows down processing by signal peptidases of B. subtilis. Thus, overproduction of signal peptidase would allow the processing of more pre-A13i-Bla molecules before their folding (45). Our present observation that the rate of pre-A13i-Bla processing was increased significantly in B. subtilis ⌬bdbC seems to confirm this hypothesis because the folding of A13i-Bla was impaired in this mutant.
The expression of the bdbA and bdbB genes follows a pattern that is reminiscent of the expression pattern of the secDF gene of B. subtilis (37). Like secDF, bdbA and bdbB are constitutively transcribed when cells are grown in minimal medium, whereas the transcription of these genes peaks in the transition phase between exponential and postexponential growth when cells are grown in TY medium. In contrast, genes for other known components of the protein secretion machinery of B. subtilis are transcribed in a different temporal-and mediumdependent manner. First, the transcription of the lsp gene for signal peptidase II is growth phase-dependent, irrespective of the growth medium, reaching a maximum in the exponential growth phase (39). Second, the transcription of the secA gene follows a pattern similar to that of lsp, but it peaks during the transition from exponential to postexponential growth (46). Third, the transcription of the sipS and sipT genes for the two major type I signal peptidases of B. subtilis is growth phasedependent and, as shown for sipS, depends on the growth medium only with respect to the transcription initiation site (36,38,47). Finally, the sipU, sipV, and sipW genes for three minor type I signal peptidases of B. subtilis are constitutively transcribed, irrespective of the growth medium (36,47). Thus, the third gene described in our present studies, bdbC, is transcribed in a way similar to that of sipU, sipV and sipW. The present observations suggest that the requirement for BdbA and BdbB peaks in the transition phase between exponential and postexponential growth when cells are grown in TY medium, whereas the requirement for BdbC is relatively low but constant under all growth conditions tested. Notably, the expression levels of bdbB and bdbC appear to set a limit for the secretion of active PhoA because PhoA activities in the supernatants of cells grown in minimal medium were significantly lower than the PhoA activities in supernatants of cells grown in TY medium.
In conclusion, our results show that BdbB and BdbC of B. subtilis are actively involved in the folding of certain secretory proteins, most likely by catalyzing the formation of disulfide bonds. Our ongoing research on the catalysis of disulfide bond formation in B. subtilis addresses three important questions that still remain to be answered: first, which are the natural substrates of BdbB and BdbC; second, are BdbB and BdbC directly or indirectly involved in the formation of disulfide bonds in PhoA and A13i-Bla; and third, do they act as oxidases or isomerases?