Bacillus subtilis Contains Four Closely Related Type I Signal Peptidases with Overlapping Substrate Specificities

Most biological membranes contain one or two type I signal peptidases for the removal of signal peptides from secretory precursor proteins. In this respect, the Gram-positive bacteriumBacillus subtilis seems to be exceptional, because it contains at least four chromosomally-encoded type I signal peptidases, denoted SipS, SipT, SipU, and SipV. Here, we report the identification of the sipT and sipV genes, and the functional characterization of SipT, SipU, and SipV. The four signal peptidases have similar substrate specificities, as they can all process the same β-lactamase precursor. Nevertheless, they seem to prefer different pre-proteins, as indicated by studies on the processing of the pre-α-amylase of Bacillus amyloliquefaciens in strains lacking SipS, SipT, SipU, or SipV. The sipU andsipV genes are constitutively transcribed at a low level, suggesting that they are required for processing of (pre-)proteins secreted during all growth phases. In contrast, the transcription ofsipS and sipT is temporally controlled, in concert with the expression of the genes for most secretory proteins, which suggests that SipS and SipT serve to increase the secretory capacity of B. subtilis. Taken together, our findings suggest that SipS, SipT, SipU, and SipV serve different functions during the exponential and post-exponential growth phase of B. subtilis.

Bacterial proteins that are exported from the cytoplasm through the general pathway for protein secretion are synthesized as precursors with an amino-terminal signal peptide. The signal peptide is required for the targeting of precursor proteins to the cytoplasmic membrane, and for the initiation of their translocation across this membrane. During, or shortly after the translocation process, most signal peptides are removed by type I signal peptidases (SPases), 1 which is a prerequisite for the release of secretory proteins from the extracytoplasmic side of the membrane (1)(2)(3)(4).
Homologous type I SPases have been identified in Grampositive and Gram-negative bacteria, the inner membrane of yeast mitochondria, and the endoplasmic reticular (ER) membranes of yeast and higher eukaryotes (5)(6)(7)(8). Despite the fact that considerable similarities can be observed between the known type I SPases when individual amino acid sequences are compared, only a few residues are strictly conserved in all known enzymes. These include serine and lysine residues, which are essential for enzymatic activity, possibly by forming a catalytic dyad (9 -12).
Based on computer-assisted analyses (5,6), 2 and studies on the membrane topology of type I SPases of Escherichia coli, yeast mitochondria and the canine ER (7,(13)(14)(15), it is predicted that the active sites of the known type I SPases are located either in the periplasm of Gram-negative bacteria, the cell wall of Gram-positive bacteria, the mitochondrial intermembrane space, or the lumen of the ER. Nevertheless, based on topological criteria, these enzymes can be divided into four distinct groups. SPases of the first and, apparently, largest group are type II membrane proteins with one amino-terminal membrane anchor. This group includes all known type I SPases from Gram-positive bacteria (6, 16 -19), cyanobacteria (Ref. 20; GenBank accessions D90899 and D90904); and the putative catalytic subunits of the ER type I SPases (8). Furthermore, at least one type I SPase from a Gram-negative bacterium (i.e. SipS of Bradyrhizobium japonicum; Ref. 21), and one from mitochondria (i.e. Imp1p; Ref. 22) belong to this group. Type I SPases of the second group, which have two amino-terminal membrane anchors, have been identified exclusively in Gramnegative bacteria (23)(24)(25). The type I SPase of Hemophilus influenzae (26) is the only known representative of the third group, having three putative amino-terminal membrane anchors. Finally, SPases of the fourth group seem to have one amino-terminal and one carboxyl-terminal membrane anchor. 2 Enzymes of the latter group have been identified in the Gramnegative bacterium Rhodobacter capsulatus (27) and yeast mitochondria (i.e. Imp2p; Ref. 28).
The SPase I of E. coli, also known as leader peptidase (Lep), is essential for cell viability (29), and SPase limitation results in the accumulation of precursors of exported proteins (30,31). Similarly, the type I SPases SpsB from Staphylococcus aureus (18), and Sec11p of the yeast ER membrane (32) are essential enzymes for cell viability. In contrast, the type I SPase SipS of Bacillus subtilis (SipS (Bsu)) is not essential for cell viability, and mutant B. subtilis strains with a disrupted sipS gene are still able to process secretory pre-proteins (6,33). This suggested the presence of at least one additional type I SPase in B. subtilis. Support for the latter hypothesis was first obtained through the identification of two genes for homologous, but non-identical SPases, denoted SipS (Bam) (17) and SipT (Bam) (also known as SipS2), in the closely related bacterium B. amyloliquefaciens (16). A few months later, genome sequencing analyses revealed the presence of two open reading frames, ycsB (34) and yhjF 3 from B. subtilis, the deduced amino acid sequences of which showed a high degree of similarity to that of SipS (Bsu). By analogy to other SPase-encoding genes of bacilli, we renamed the latter open reading frames sipU and sipV, respectively, although SPase activity of the corresponding proteins had not been demonstrated. SipS (Bsu) and SipS (Bam) appeared to be equivalent enzymes in B. subtilis and B. amyloliquefaciens: first, it was shown that the amino acid sequences of both enzymes are highly similar (91% identical residues and conservative replacements; Ref. 17); and, second, in both organisms the corresponding sipS genes were mapped immediately upstream of the rib operons for riboflavin biosynthesis (35). 4 In contrast, SipT from B. amyloliquefaciens showed a much lower degree of sequence similarity to SipU and SipV of B. subtilis (65 and 44% identical residues and conservative replacements, respectively), and the corresponding genes were mapped at different regions of the respective chromosomes (16,34). 3 The latter findings suggested that SipT from B. amyloliquefaciens is not the equivalent of SipU or SipV from B. subtilis and, consequently, it was conceivable that these organisms contain at least four chromosomal sip genes for type I SPases. The present studies were aimed at the verification of this hypothesis, and the functional characterization of the type I SPases of B. subtilis. We show that B. subtilis contains four closely related type I SPases which have similar, but non-identical substrate specificities. In addition, we show that the transcription of the sipT gene is temporally controlled, whereas the sipU and sipV genes are constitutively transcribed at a low level. Table I lists the plasmids and bacterial strains used. TY medium (tryptone/yeast extract) contained Bacto-tryptone (1%), Bacto-yeast extract (0.5%), and NaCl (1%). Minimal medium for B. subtilis was prepared as described in Ref. 41, and supplemented with glucose (0.5%), casamino acids (0.02%), tryptophan (20 mg/ml), histidine (20 mg/ml), methionine (20 mg/ml), tyrosine (20 mg/ml), adenine (20 mg/ml), uracil (20 mg/ml), nicotinic acid (0.4 mg/ml), riboflavin (0.4 mg/ml), and Fe-ammonium citrate (1.1 mg/ml). M9 media 1 and 2, used for the labeling of E. coli proteins with [ 35 S]me-  40. The sequences of DNA fragments, including PCRamplified fragments, were analyzed by the dideoxy chain termination method (45), using the T7 Sequencing Kit (Pharmacia, Uppsala, Sweden). [ 35 S]dATP (8 mCi/ml; Ͼ1000 Ci/mmol) from Amersham International. DNA and protein sequences were analyzed using version 6.7 of the PCGene Analysis Program (Intelligenetics Inc., Mountain View, CA). The BLASTP algorithm (46) was used for protein comparisons in GenBank. Correct integration of linearized DNA fragments, or plasmids in the chromosome of B. subtilis was verified by Southern hybridization. PCR under stringent conditions for the annealing of primers to template DNA was carried out with Vent DNA polymerase (New England Biolabs, Beverly, MA) as described in Ref. 11. When low-stringency conditions were required, the annealing temperature was lowered to 42°C, and the Supertaq DNA polymerase (Sphaero-Q, Leiden, the Netherlands) was used.

Plasmids, Bacterial Strains, and Media-
Pulse-Chase Protein Labeling, Immunoprecipitation, SDS-PAGE, and Fluorography-Pulse-chase labeling of E. coli and B. subtilis, immunoprecipitation, SDS-PAGE, and fluorography were performed as described in Refs. 42 and 43. 14 C-Methylated molecular weight markers were from Amersham International. Relative amounts of precursor and mature forms of secreted proteins were estimated by film scanning with an LKB ultroscan XL laser densitometer (LKB, Bromma, Sweden).
␤-Galactosidase Activity Assay-Overnight cultures were diluted 100-fold in fresh medium and samples were taken hourly for optical density (OD) readings at 600 nm and ␤-galactosidase activity determinations. The assay and the calculation of ␤-galactosidase units (expressed as units per OD 600 ) were carried out as described in Ref. 47.
Western Blot Analysis-Western blotting was performed as in Ref. 48. After separation by SDS-PAGE, proteins were transferred to Immobilon-polyvinylidene difluoride membranes (Millipore Corp., Bedford, MA). To assay the presence of the precursor and mature forms of B. amyloliquefaciens ␣-amylase in B. subtilis, cells were separated from the growth medium by centrifugation (5 min, 12,000 rpm, room temperature), and samples for SDS-PAGE were prepared as described in Ref. 42. The B. amyloliquefaciens ␣-amylase was visualized with specific antibodies and horseradish peroxidase anti-rabbit IgG conjugates (Amersham International). To monitor the presence of the SipV-Myc fusion protein in E. coli, samples for SDS-PAGE were prepared as in Ref. 11. SipV-Myc was visualized with specific monoclonal antibodies and horseradish peroxidase anti-mouse IgG conjugates (Amersham International).

Identification of the sipT Gene of B. subtilis-
To determine whether B. subtilis contains a sipT (Bam)-like gene, Southern hybridization experiments were performed. A 3.2-kb HindIII fragment of B. subtilis chromosomal DNA was found to hybridize weakly with the sipT (Bam) gene, but not with the sipS (Bsu), sipU (Bsu), or sipV (Bsu) genes, suggesting that B. subtilis contains a sipT (Bam)-like gene (data not shown). To identify the latter gene, a PCR was performed with the primers lba2-1 and lba2-2 ( Fig. 1), which correspond to sequences within the sipT (Bam) gene. Using chromosomal DNA of B. subtilis 8G5 sipS as a template, a 300-bp fragment was amplified. Sequence analysis of this fragment revealed the presence of an open reading frame, the deduced amino acid sequence of which showed a high degree of similarity to SipT (Bam) and, to a lesser extent, to SipS (Bsu), SipU (Bsu), and SipV (Bsu) (data not shown). The latter findings indicated that the 300-bp fragment was an internal fragment of a fourth sip gene of B. subtilis, possibly sipT.
The entire sipT (Bam)-like gene of B. subtilis was cloned in three successive PCR steps (schematically shown in Fig. 1, b-d). Sequence analysis showed that the upstream sequences of this gene contain the 3Ј end of the B. subtilis fruA gene for the enzyme II of the fructose-specific phosphoenolpyruvate phosphotransferase system (Fig. 1), which has been mapped at 126 degrees of the B. subtilis chromosome (49). The latter finding showed that the fourth sip gene of B. subtilis is indeed the sipT (Bsu) gene (GenBank accession U45883), because the sipT (Bam) gene is also preceded by fruA on the chromosome of B. amyloliquefaciens (16).
Like other sip genes from bacilli (6,16,17,34), the sipT (Bsu) gene (582 nucleotides) is preceded by a potential ribosomebinding site (GGAGG); it has a TTG start codon, and lacks upstream sequences with obvious similarity to the major classes of B. subtilis promoters. The deduced amino acid sequence of the SipT (Bsu) protein (193 residues), shows a high degree of similarity to that of the known type I SPases from B. subtilis and related bacilli (  subtilis as a template. The nucleotide sequences (5Ј-3Ј) of primers used for PCR are indicated below; nucleotides identical to genomic template DNA are printed in bold, and restriction sites used for cloning are underlined. After each PCR (a-d), amplified fragments were subcloned and sequenced. In step a, an internal, 300-bp fragment of sipT (Bsu) was amplified with primers lba2-1 (TGAAC-CGTACTTAGTGG) and lba2-2 (GATTGTCGCCCATGACG) under conditions of low stringency. In step b, a 600-bp fragment containing the 5Ј end of sipT (Bsu) was amplified by "inverse PCR" with primers lbt-3 (TTGAATTCACAAACAGCCTTTCTCCG) and lbt-4 (GAGAATTCG-GACCGGTTAAGGTTCCG), using a self-ligated, circular, 0.8-kb genomic EcoRI fragment as the template. In step c, a 250-bp fragment containing the 3Ј end of sipT (Bsu) was amplified from the B. subtilis genome by "walking PCR" using the primers lbt-4 (see above), lbt-8 (GATAGTCGACAAAGAAGAGAAACAACTG), and lbt-10 (GGAA-GTCGACATACGTACCTG GAATGG), in combination with a set of primers (X) with the following sequences: GGAAGATCTGAATTCATA-AAGGGAAGATG; GCGAATTCTTTTATCAGCGTTCTGGCT; TTTG-AATTCTACTTACTGTCACTCGTT; GATCGAATTCGATGGCGCTAC-TCTGGG; and GATCGAATTCATAAAGAACTAAACCTCGGTG. In step c, a first PCR was performed under conditions of low stringency with primers lbt-8 and X. This resulted in the amplification of a wide range of different fragments, which were used as template DNA in a second PCR under stringent conditions with primers lbt-10 and X. The products of the second PCR were used as template for a third PCR with primers lbt-4 and X. Finally, in step d, an 862-bp DNA fragment containing the complete sipT gene of B. subtilis 168 was amplified with primers lbt-12 (GATGGTCGACCTTTTAAGTATCGTGATCGG) and lbt-15 (CACGGTACCATGCATTGCATTGGTCGC). The sequence of sipT (Bsu) was determined from three independent isolates (GenBank accession U45883). Regions of sipT specifying the conserved domains (A-E), and relevant restriction sites are shown: EI, EcoRI; EV, EcoRV; Hi, HindII; Sy, StyI.
Because, after a chase of 10 min, no mature (A13i)-␤-lactamase could be detected in E. coli (pGDL131) cells containing the cloned sipV (Bsu) gene, pulse-chase labeling experiments were performed in which the chase was extended to 60 min. Under these conditions, approximately 90% of the (A13i)-␤lactamase was processed to the mature form in cells containing SipU (positive control) (Fig. 3B). In contrast, no mature (A13i)-␤-lactamase could be detected in E. coli (pGDL48; negative control), and E. coli (pGDL131; sipV (Bsu)) (Fig. 3B). To verify that the presence of pGDL131 resulted in the production of SipV in E. coli, an oligonucleotide specifying the human c-Myc epitope (EQKLISEEDLN, Ref. 50) was fused in-frame to the 3Ј end of sipV, resulting in pGDL132. As shown by Western blotting, the SipV-Myc protein was produced in E. coli cells containing pGDL132 (Fig. 3C). Pulse-chase labeling experiments revealed that pre-(A13i)-␤-lactamase was not processed in E. coli (pGDL132), like in E. coli (pGDL131) (data not shown). The latter findings indicate that SipV is produced in E. coli cells containing pGDL131, and that SipV is unable to cleave pre(A13i)-␤-lactamase in E. coli.
SipT, SipU, or SipV Are Not Essential for Cell Viability-It was previously shown that SipS is not essential for viability of B. subtilis, and cells lacking the sipS gene were still able to process secretory pre-proteins (33). To investigate whether SipT is essential for viability of B. subtilis, the following strategy was used: first, a plasmid-encoded copy of the corresponding gene was disrupted with a chloramphenicol resistance (Cm r ) marker. The resulting plasmid pHT100C, which is unable to replicate in B. subtilis, was linearized and, subsequently, used to transform competent B. subtilis 8G5 cells. As verified by Southern hybridization (data not shown), all chloramphenicol-resistant transformants (denoted B. subtilis 8G5 sipT-Cm) contained the disrupted sipT gene (schematically presented in Fig. 5A), showing that SipT is not required for cell viability.
Similarly, it was shown that SipU is not essential for cell viability by deleting a 197-bp EcoRI fragment from the chromosome of B. subtilis (schematically shown in Fig. 5A). The latter fragment contains the first 170 bp of the sipU gene specifying the conserved domains A (i.e. the membrane anchor) and B (containing the putative catalytic serine residue; Fig. 2). To this purpose, we used plasmid pINT34d, which consists of the chromosomal integration plasmid pORI280 (36) carrying a mutant copy of the sipU locus that lacks the 197-bp EcoRI fragment. Upon the Campbell-type integration of pINT34d into the sipU locus of the B. subtilis chromosome, and the subsequent selection of cells that had lost this plasmid from the chromosome, it was shown by PCR and Southern blotting that about 10% of the cells lacking pINT34d also lacked the 197-bp EcoRI fragment. This finding showed that SipU is not essential for cell viability. The resulting mutant strain was denoted B. subtilis 8G5 sipU.
To disrupt the chromosomal sipV gene, we first disrupted a plasmid-encoded copy of this gene with an erythromycin resistance (Em r ) marker. The resulting plasmid pV50E, which is unable to replicate in B. subtilis, was linearized and, subsequently, used to transform competent B. subtilis 8G5 cells. All erythromycin-resistant transformants (denoted B. subtilis 8G5 sipV-Em) contained the disrupted sipV gene (schematically presented in Fig. 5A), showing that, like SipS, SipT, and SipU, also SipV is not required for cell viability. Neither the disruption of the sipT or sipV genes, nor the removal of an essential part of the sipU gene, had a detectable influence on cell growth, the development of competence for DNA binding and uptake, or sporulation (data not shown).
Processing of ␣-Amylase Is Reduced in the Absence of SipT, and Improved in the Absence of SipS or SipU-Processing of the precursor of the B. amyloliquefaciens ␣-amylase AmyQ (previously also referred to as AmyE; Refs. 33 and 51) was recently shown to be improved in the absence of SipS, indicating that the production of SipS interferes with pre-AmyQ processing, and that this precursor could be a preferred substrate for other SPases, such as SipT, SipU, or SipV (33). To investigate the effects of the absence of SipT, SipU, or SipV on the processing of pre-AmyQ, B. subtilis 8G5 sipT-Cm, B. subtilis 8G5 sipU, and B. subtilis 8G5 sipV-Em were transformed with plasmid pKTH10. The latter plasmid contains the amyQ gene, and its presence in B. subtilis results in the secretion of ␣amylase at high levels (Ϯ1.3 g/liter; Refs. 37 and 51). First, we performed pulse-chase labeling experiments. The results showed that, compared with the parental strain 8G5, the rate of pre-AmyQ processing in the mutant lacking SipT was re-

FIG. 4. Processing of pre(A13i)-␤-lactamase in B. subtilis. A,
processing of pre(A13i)-␤-lactamase in B. subtilis 8G5 sipS harboring pGDL48 (no sip gene), pGDL41 (sipS (Bsu)), pGDL100 (sipT (Bsu)), pGDL121 (sipU (Bsu)), or pGDL131 (sipV (Bsu)) was analyzed by pulsechase labeling at 37°C and subsequent immunoprecipitation, SDS-PAGE, and fluorography. Cells were labeled with [ 35 S]methionine for 1 min prior to chase with excess of nonradioactive methionine. Samples were withdrawn at the times indicated. The loss of label in the time courses is due both to removal of the signal peptide and degradation of the mature ␤-lactamase in the medium (58). B, the kinetics of processing are plotted as the percentage of the total (A13i)-␤-lactamase (precursor ϩ mature) present in the precursor form at the time of sampling. Relative amounts of the precursor and mature forms of pre(A13i)-␤lactamase were determined by scanning of autoradiographs. E, pGDL48 (no sip gene); Ⅺ, pGDL41 (sipS (Bsu)); q, pGDL100 (sipT (Bsu)); f, pGDL121 (sipU (Bsu)); OE, pGDL131 (sipV (Bsu)).  8G5 sipT-Cm, B. subtilis 8G5 sipU, and B. subtilis 8G5 sipV-Em. The chromosomal sipT gene was disrupted with a Cm r marker by homologous recombination. To this purpose, B. subtilis 8G5 was transformed with the linearized plasmid pHT100C, which cannot replicate in B. subtilis, and which contains a mutant copy of sipT with a Cm r marker in the HindII site. Part of the sipU gene was removed from the chromosome of B. subtilis 8G5 using pINT34d, a derivative of the chromosomal integration plasmid pORI280 (36). In addition to an Em r marker and the E. coli lacZ gene, pINT34d carries a mutant copy of sipU, which was obtained by the excision of a 197-bp EcoRI fragment from a PCR-amplified genomic DNA fragment containing sipU and its flanking sequences. Since pINT34d is unable to replicate in B. subtilis, transformants (Em r and blue on plates with 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside) can only be obtained as a result of a Campbell-type integration into the homologous sipU sequences on the chromosome. Thus, the chromosome of transformants with pINT34d will contain both an intact, and a truncated copy of sipU, separated by sequences of pORI280. Upon growth for about 200 generations in the absence of Em, cells were selected that had spontaneously lost pORI280, together with one of the two copies of sipU (Em s and white on plates with 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside). Cells lacking the 197-bp EcoRI fragment were denoted B. subtilis 8G5 sipU. The chromosomal duced; after a chase of 4 min, about 30% of the labeled AmyQ was in the precursor form in the parental 8G5 strain whereas, under the same conditions, about 50% of the AmyQ was in the precursor form in B. subtilis 8G5 sipT-Cm (Fig. 5, B and C). In contrast, the rate of pre-AmyQ processing was increased in strains lacking either SipS or SipU; both in B. subtilis 8G5 sipS and B. subtilis 8G5 sipU only about 2% of the labeled AmyQ was present in the precursor form after a chase of 4 min. Processing of pre-AmyQ was hardly affected in B. subtilis 8G5 sipV-Em (Fig. 5, B and C).
To compare the effects of the absence of SipS, SipT, SipU, or SipV on the accumulation of pre-AmyQ, Western blotting experiments were performed with cells of B. subtilis 8G5 sipS (pKTH10), B. subtilis 8G5 sipT-Cm (pKTH10), B. subtilis 8G5 sipU (pKTH10), B. subtilis 8G5 sipV-Em (pKTH10), and the parental strain B. subtilis 8G5 (pKTH10), all grown overnight in TY medium. As previously shown for strains lacking SipS (33), compared with the parental strain 8G5, the absence of SipU resulted in a reduction of about 20% in the accumulation of pre-AmyQ (Fig. 5, D and E). In contrast, cells lacking SipT accumulated more pre-AmyQ (approximately 10%) than the parental strain, whereas the absence of SipV had no clear effect on the accumulation of pre-AmyQ (Fig. 5, D and E). Taken together, our findings indicate that pre-AmyQ is a preferred substrate for SipT, and that the presence of SipS or SipU interferes with efficient processing of this precursor. It is not clear whether pre-AmyQ is a substrate for SipV.
Distinct Regulation of sipT, sipU, and sipV Gene Expression at the Transcriptional Level-The transcription of the sipS (Bsu) gene is temporally regulated, sipS promoter activity being highest in the post-exponential growth phase (33). To examine whether this is also the case for the transcription of the sipT, sipU, and sipV genes, transcriptional sipT-lacZ, sipU-lacZ, and sipV-lacZ gene fusions were constructed, and introduced in the chromosome of B. subtilis 8G5 (schematically shown in Fig. 6A), using a similar strategy as described previously for a transcriptional sipS-lacZ gene fusion (33). This resulted in B. subtilis 8G5::pLGT207 (sipT-lacZ), B. subtilis 8G5::pLGU202 (sipU-lacZ), and B. subtilis 8G5::pLGV201 (sipV-lacZ), respectively. Next, these strains and B. subtilis 8G5::pGDE22 (sipS-lacZ; Ref. 33) were grown in TY and minimal medium, and samples withdrawn at hourly intervals were assayed for ␤-galactosidase activity.
In both media, nearly identical results were obtained with B. subtilis 8G5::pLGT207 (sipT-lacZ): the levels of ␤-galactosidase activity increased after the cells entered the transition state (t ϭ 0) between the exponential and the post-exponential growth phase (Fig. 6, B and C; indicated with the symbol f), and they continued to increase during the post-exponential growth phase, indicating that the promoter(s) of sipT became more active than in the exponential growth phase. Thus, the transcription of the sipT (Bsu) gene appears to be temporally controlled, similar to that of the sipS (Bsu) gene (Fig. 6, B and C; indicated with the symbol Ⅺ). In particular, in minimal medium the ␤-galactosidase levels observed in the strains with the sipS-lacZ or sipT-lacZ gene fusions were comparable (Fig.  6C). In TY medium, however, the sipS promoter activity appeared to be 1.5-2-fold higher than that of sipT (Fig. 6B).
The regulation of the transcription of the sipU and sipV genes seems to be completely different from that of sipS and sipT. When grown in TY or minimal medium, a nearly constant low level of ␤-galactosidase activity was observed in cells of B. subtilis 8G5::pLGU202 (sipU-lacZ) and B. subtilis 8G5::pLGV201 (sipV-lacZ), irrespective of the growth phase (Fig. 6, B and C; indicated with the symbol q and E, respectively). In fact, in both media the ␤-galactosidase levels of B. subtilis 8G5::pLGU202 (sipU-lacZ) were nearly equal to background ␤-galactosidase levels of control cells lacking a copy of lacZ (data not shown). Nevertheless, expression of the sipU gene was evident as colonies of B. subtilis 8G5::pLGU202 (sipU-lacZ) were blue on TY, or minimal plates with 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside; control cells lacking the lacZ gene, or containing a fusion between a non-transcribed gene and lacZ remained white (data not shown). These findings indicate that the activity of the sipU and sipV promoter(s) does not depend on the growth phase. DISCUSSION In those microorganims of which the genomes have been sequenced completely, at most two or three homologous type I SPases seem to be present. For example, the cyanobacterium Synechocystys contains two type I SPases (GenBank accessions D90899 and D90904), and the yeast Saccharomyces cerevisiae contains three of these enzymes (52). In the latter case, these enzymes are localized in two distinct membrane systems, the inner mitochondrial membrane (i.e. Imp1p and Imp2p; Refs. 22, and 28), and the ER membrane (i.e. the Sec11 protein; Ref. 32). Two homologues of bacterial type I SPases are also commonly found in the ER SPase complex of higher eukaryotes (8). 5 In contrast, for other genetically well characterized microorganisms, such as E. coli (GenBank accession ECOLI U00096), H. influenzae (26), and Methanococcus jannaschii (53), only one type I SPase seems to be sufficient, and type I SPases may even be completely absent from Mycoplasma genitalium (54). In our present studies we show that B. subtilis contains at least four chromosomally-encoded type I SPases (SipS, SipT, SipU, and SipV) involved in protein secretion. In addition, we have previously shown that certain strains of B. subtilis also contain plasmids (pTA1015/pTA1040) specifying related type I SPases (17). Thus, the composition of the protein secretion machinery of B. subtilis seems to be unique with respect to the high number of SPases.
A second remarkable property of the secretion machinery of B. subtilis concerns the high degree of similarity between the substrate specificities of SipS, SipT, SipU, SipV, SipP (pTA1015), and SipP (pTA1040). The conclusion that the substrate specificities of these six type I SPases are very similar is based on our present and previous (6,17) observations that all these enzymes are able to cleave the same substrate, pre(A13i)-␤-lactamase, albeit with different efficiencies, and under different conditions. By contrast, the type I SPases Imp1p and Imp2p in the inner mitochondrial membrane seem to have completely distinct substrate specificities (28).
Although similar, the substrate specificities of the four chromosomally-encoded type I SPases of B. subtilis are not identical, as our present results indicate that these enzymes have, at least in vivo, a different preference for the precursor of the ␣-amylase AmyQ of B. amyloliquefaciens. Pre-AmyQ processing was significantly reduced in strains lacking SipT, indicat-FIG. 6. Analysis of the expression of sipT, sipU, and sipV with transcriptional lacZ gene fusions. A, schematic presentation of the sipT, sipU, and sipV regions on the chromosomes of B. subtilis 8G5::pLGT207 (sipT-lacZ), B. subtilis 8G5::pLGU202 (sipU-lacZ), and B. subtilis 8G5::pLGV201 (sipV-lacZ), respectively. All three lacZ fusions were constructed with plasmid pLGW200 (38), a chromosomal integration plasmid for B. subtilis containing a promoterless spoVG-lacZ gene fusion. To construct a sipT-lacZ gene fusion, the 5Ј end of sipT, amplified by PCR with the primers lbt-9 (ATGAATTCAGCCCGGTTATCTCC) and lbt-12 (Fig. 1), was cloned in the multiple cloning site (MCS) upstream of the spoVG-lacZ fusion of pLGW200, resulting in pLGT207. Similarly, a sipU-lacZ gene fusion was constructed by cloning the 5Ј end of sipU, PCR-amplified with the primers lbu-3 (AGCTGTCGACATTGCCGGACAGGCC) and lbu-4 (AATAGGTACCGGAGGGAACCTCAACTTCG), in the MCS of pLGW200, resulting in pLGU202. To construct a sipV-lacZ gene fusion, the 5Ј end of sipV was PCR-amplified with the primers uni (GTAAAACGACGGCCAGT) and lbv-2 (TTGGAATTCGATTATCTCCAACGAC) from a pUC18-derived plasmid carrying the sipV gene. The amplified fragment was cloned in the MCS of pLGW200, resulting in pLGV201. The sip-lacZ gene fusions were introduced in the chromosome of B. subtilis 8G5 by Campbell-type integration. In B. subtilis 8G5::pLGT207, the transcription of lacZ is directed by the promoter(s) of sipT, in B. subtilis 8G5::pLGU202 by the promoter(s) of sipU, and in B. subtilis 8G5::pLGV201 by the promoter(s) of sipV. Only restriction sites relevant for the constructions are shown (EI, EcoRI; Hi, HindII; Sa, SalI; Sh, SphI; Su, StuI; Cl, ClaI), ori pBR322, replication functions of pBR322. B and C, time courses of the expression of sip-lacZ gene fusions were determined in cells growing in TY medium (B) or minimal medium (C) at 37°C. ␤-Galactosidase activities (in units per OD 600 ) were determined for B. subtilis 8G5::pGDE22 (Ⅺ, sipS-lacZ), B. subtilis 8G5::pLGT207 (f, sipT-lacZ), B. subtilis 8G5::pLGU202 (q, sipU-lacZ), and B. subtilis 8G5::pLGV201 (E, sipV-lacZ). Zero time (t ϭ 0) indicates the transition point between the exponential and the post-exponential growth phases.
ing that this precursor is a preferred substrate of SipT. In contrast, SipV did not seem to be involved in pre-AmyQ processing, whereas the presence of SipS and SipU interfered with efficient processing of this precursor. Taken together, these findings suggest that SipS and SipU somehow compete with SipT for binding of pre-AmyQ, and that SipT, but not SipS and SipU, can cleave this precursor efficiently. Similarly, preliminary data suggest that pro-OmpA of E. coli may be a preferred substrate of SipT, 6 whereas the presence of SipT seems to interfere with efficient secretion of levansucrase of B. subtilis. 2 What could be the advantage(s) for an organism, like B. subtilis, to acquire and maintain so many different SPaseencoding genes during its evolution? Our present observations indicate that multiple SPases may serve to guarantee a sufficient capacity for protein secretion under various conditions. First, we show that none of the four chromosomally-encoded type I SPases described in this article is, by itself, essential for cell growth and protein secretion. As SPase activity is essential for the viability of B. subtilis, 2 our present observations imply that the secretory precursor processing machinery of this organism is functionally redundant. Thus, B. subtilis can always avail of a "backup SPase," even in the case that a complete SPase-encoding gene would be lost. This may be of particular importance for B. subtilis and related bacilli, such as B. amyloliquefaciens, which secrete large amounts of proteins into the medium. Second, our present observations suggest that different chromosomally-encoded type I SPases of B. subtilis serve different functions in the exponential and post-exponential growth phases. For example, it seems likely that SipU and SipV are involved in the processing of secretory pre-proteins that are synthesized during all growth phases, because the corresponding genes are transcribed at a constitutive (low) level. In contrast, the transcription of the sipS and sipT genes is temporally regulated, the highest levels of transcription being observed in the post-exponential growth phase. The increase in the levels of transcription of sipS and sipT starts in the transition phase between the exponential and the postexponential growth phase and is, thus, concerted with the onset of the transcription of most secretory proteins of B. subtilis (55). In fact, in minimal medium, the transcription of both the sipS gene (33) and the sipT gene, but not the sipU and sipV genes, is controlled by the DegS-DegU two-component regulatory system, 2 which is also required for the transcription of several genes for secretory proteins (56). Therefore, it seems likely that SipS and SipT serve to increase the capacity for protein secretion in the post-exponential growth phase under conditions of increased synthesis of secretory proteins in B. subtilis. The latter hypothesis would be consistent with two of our previous findings: (a) the availability of SPase can be a limiting factor for the secretion of certain hybrid precursor proteins, which can be overcome by SPase overproduction (6,42); and (b) certain endogenous plasmids of B. subtilis contain SPase-encoding genes, suggesting that SPase can also be a limiting factor for protein secretion in a natural system (17). In addition, the special importance of SipS and SipT for protein secretion in B. subtilis is underscored by our recent 4 observation that only cells lacking both SipS and SipT were not viable, whereas all other sip gene mutations could be combined.
Finally, how many type I SPases does B. subtilis contain exactly? The systematic sequence analysis of the B. subtilis genome has been completed very recently, and it seems that there are no other genes for close homologues of SipS, SipT, SipU, and SipV. 7 However, our computer-assisted analyses revealed one additional gene (yqhE) for a potential type I SPase (SipW) that is more closely related to the type I SPases from archaea and the eukaryotic ER membrane than to bacterial type I SPases. The question whether SipW is actively involved in protein secretion remains to be answered.