A common feature in cells of prokaryotic and eukaryotic origin is the export of proteins from their site of synthesis, mostly the cytoplasm, to other destinations either inside or outside the cell. To achieve this, exported proteins are usually synthesized as precursors with an amino-terminal, transient "zip code" (signal peptide), which is recognized and deciphered by a cellular sorting and translocation machinery (318-320). Signal peptides consist of short stretches of amino acids which, after protein delivery to the correct subcellular compartment, are frequently removed by specialized signal peptidases. In general, a preprotein is first recognized by soluble targeting factors for its transport to the target membrane, where the protein becomes associated with a translocation machinery. Next, the polypeptide chain is transported through a proteinaceous channel. In most cases this transport process is driven by a translocation motor that binds and hydrolyzes nucleoside triphosphates. Finally, the signal peptide is removed, resulting in release of the mature protein from the translocase. If the protein is translocated in an unfolded conformation, the mature protein will fold into its native conformation shortly after release from the translocase. Notably, several integral membrane proteins retain their signal-like peptides and diffuse from the translocase laterally. These basic principles of protein transport across membranes apply to most eukaryotic and prokaryotic organisms (84, 216, 231, 249).

In eukaryotic cells, proteins can be transported to numerous destinations, such as the nucleus, the endoplasmic reticulum (ER), the Golgi apparatus, lysozomes, the plasma membrane, the cell wall, chloroplasts, mitochondria, peroxisomes, and the different membrane systems or compartments within the organelles mentioned. Furthermore, proteins can be secreted into the external environment of the cell. In contrast, in eubacterial and archaeal cells, protein sorting seems to be limited to a few compartments, such as the cytoplasmic membrane, the cell wall (gram-positive eubacteria and archaea), the periplasm, and the outer membrane (gram-negative eubacteria). In addition, eubacteria and archaea can secrete proteins directly into their growth medium. In order to do so, these unicellular organisms can exploit multiple pathways, such as the general secretory (Sec) pathway, the twin-arginine translocation (Tat) pathway, and ATP-binding cassette (ABC) transporters.

In the following sections of this review, the known signal peptide-dependent protein transport pathways, as they are present in the gram-positive eubacterium Bacillus subtilis, will be discussed, with a strong focus on the Sec pathway. Notably, the protein export machineries of the gram-negative eubacterium Escherichia coli and certain eukarya, such as the yeast Saccharomyces cerevisiae, have in many cases been characterized in more detail than those of B. subtilis. For matters of comparison, these machineries will also be discussed where appropriate. Because of the differences in the cell envelope structure, differences in the machineries for protein export in B. subtilis and E. coli were anticipated more than a decade ago (205). As described in this review, such differences do indeed exist, especially at the early and late stages of protein export, making detailed characterization of the underlying molecular mechanisms a fascinating scientific challenge.

B. subtilis and related Bacillus species are well known for their industrial use in the production of secreted proteins. These eubacterial species are particularly attractive for this purpose because they have a high capacity to secrete proteins into the growth medium and because of their nonpathogenicity. Moreover, good fermentation technologies exist for various bacilli (36, 40, 41, 263). Many proteins can be secreted to very high levels by B. subtilis, such as the -amylase AmyQ from Bacillus amyloliquefaciens (1 to 3 g/liter) (204), protein A from Staphylococcus aureus (>1 g/liter) (91), and human interleukin-3 (227). Although not precisely documented in the scientific literature, about 10-fold-higher secretion levels can be reached in optimized industrial fermentation systems using Bacillus amyloliquefaciens or Bacillus licheniformis strains. Unfortunately, the secretion of proteins of gram-negative eubacterial or eukaryotic origin by Bacillus species is often severely hampered due to several bottlenecks in the secretion pathway, such as poor targeting to the translocase, degradation of the secretory protein, and slow or incorrect folding. Therefore, it is not only of scientific but also of applied interest to define the so-called secretome of B. subtilis, which includes both the components of machineries for protein secretion and the native secreted proteins. In recent years, considerable progress has been made concerning the identification and characterization of host functions needed for protein secretion by B. subtilis. In particular, this progress was facilitated by the availability of the complete genome sequence of B. subtilis (149). Present research efforts on protein transport in B. subtilis are aimed first at obtaining a complete description of the secretome and second at identifying those secretome components that are limiting factors in secretion. This review will provide a first, largely genome-based survey of the secretome.

At first sight, protein transport in B. subtilis appears to be a relatively simple process, as its cell structure is considerably less complicated than that of eukaryotic cells. The cytoplasm is surrounded by the cytoplasmic membrane, which is covered by a thick layer (10 to 50 nm) of peptidoglycan-containing anionic polymers, such as teichoic and teichuronic acid. All proteins of B. subtilis lacking transport signals will be retained in the cytoplasm and fold, with or without the aid of chaperones, such as GroEL-GroES and DnaK-DnaJ-GrpE, into their native conformation (16, 90, 111, 113, 193). Other proteins contain membrane-spanning domains that are required for their insertion into the cytoplasmic membrane.

Most proteins that are completely transported across the cytoplasmic membrane are synthesized with an amino-terminal signal peptide. As B. subtilis, like other gram-positive eubacteria, lacks an outer membrane, many of these proteins are secreted directly into the growth medium. In most cases, these secreted proteins are enzymes involved in the hydrolysis of natural polymers, such as proteases, lipases, carbohydrases, DNases, and RNases. Such degradative enzymes are frequently synthesized as part of an adaptive response to changes in the environment, allowing the cell to benefit optimally from the available resources (95, 177, 263). Subsequently, specialized uptake systems in the cytoplasmic membrane internalize (partially) degraded substrates (14, 104). A second well-described class of secreted proteins, consisting of seven relatively small proteins, denoted PhrA to PhrK, are used to sense the cell density of the population, thereby regulating the onset of post-exponential-phase processes, such as competence development and sporulation (152, 208). These Phr proteins are, after their secretion and processing into small peptides, reimported to fulfill their inhibitory action on certain cytoplasmic phosphatases (206, 207, 209, 271). In contrast to the degradative enzymes and Phr proteins, most other exported proteins, involved in processes such as cell wall turnover, substrate binding, or protein secretion (217, 281), have to be retained at the membrane-cell wall interface to fulfill their function. To prevent the loss of these proteins, they can contain signals for their attachment to the membrane (lipid modifications) or the cell wall. Alternatively, some exported proteins have the potential to form pilin-like structures at the membrane-cell wall interface.

Strikingly, under conditions of nutrient starvation, the formation of two genuine internal compartments, as encountered in organelles of the eukaryotic cell, is induced. These compartments, which ultimately develop into an endospore, are confined by two membranes, the forespore inner and outer membranes. The forespore inner membrane confines the cytosol of the forespore, while the forespore outer membrane forms the initial barrier between the forespore and the cytosol of the mother cell (89, 279) (for details, see the section on sporulation-specific protein transport). Recent data indicate that certain proteins are specifically sorted from the cytosol of the mother cell or the forespore to the intermembrane space (IMS) between the two forespore membranes. The process of subcellular compartmentalization during sporulation in particular underscores the fact that, though simple at first glance, complex mechanisms for protein sorting have evolved in B. subtilis.

In the following sections, the amino-terminal cleavable signal peptides which are involved in the transport of ribosomally synthesized proteins in B. subtilis will be discussed. Furthermore, the different protein export routes and retention mechanisms which prevent the loss of certain exported proteins in the environment will be described.

Although the primary structures of different amino-terminal signal peptides show little similarity, three distinct domains can nevertheless be recognized (225, 316-319, 321). The amino-terminal N-domain of signal peptides contains at least one arginine or lysine residue, although this positively charged residue does not seem to be strictly required for protein export (51, 101). The positively charged N-domain has been suggested to interact with the translocation machinery (1) and negatively charged phospholipids in the lipid bilayer of the membrane during translocation (71). The H-domain, following the N-domain, is formed by a stretch of hydrophobic residues that seem to adopt an -helical conformation in the membrane (37). Helix-breaking glycine or proline residues are frequently present in the middle of this hydrophobic core. The latter residues might allow the signal peptide to form a hairpin-like structure that can insert into the membrane. In one model for signal peptide function, it was proposed that unlooping of this hairpin results in insertion of the complete signal peptide in the membrane (71) (Fig. 1). Helix-breaking residues found at the end of the H-domain, are thought to facilitate cleavage by a specific signal peptidase (SPase) (67, 202). The C-domain, following the H-domain, contains the cleavage site for SPase, which removes the signal peptide from the mature part of the secreted protein during or shortly after translocation. The mature part of the protein is thereby released from the membrane and can fold into its native conformation. Finally, the signal peptide is degraded by signal peptide peptidases (SPPases) and removed from the membrane (Fig. 1). Although different amino-terminal signal peptides tend to be quite similar in general structure, apparently small differences between individual signal peptides can cause cleavage by different SPases, export via different pathways, and transport to different destinations.

View larger version (25K): [in this window] [in a new window]   FIG. 1.   Model for signal peptide insertion into the cytoplasmic membrane and cleavage by SPase I. First, the positively charged N-domain of the signal peptide interacts with negatively charged phospholipids in the membrane, after which the H-domain integrates loopwise into the membrane. Next, the H-domain unloops, whereby the first part of the mature protein is pulled through the membrane. During or shortly after translocation by a translocation machinery (not shown), the signal peptide is cleaved by SPase I and subsequently degraded by SPPases. After its translocation across the membrane, the mature protein folds into its native conformation.

At present, four major classes of amino-terminal signal peptides can be distinguished on the basis of the SPase recognition sequence. The first class is composed of "typical" signal peptides which are present in preproteins that are cleaved by one of the various type I SPases of B. subtilis (289, 290, 292). Although most proteins having such a signal seem to be secreted into the extracellular environment, some of them are retained in the cell wall or sorted specifically to the IMS of endospores after membrane translocation via the Sec pathway. Notably, a subgroup of these signal peptides contain a so called twin-arginine motif (RR-motif), which might direct proteins into a distinct translocation pathway known as the Tat pathway.

The second major class of signal peptides is present in prelipoproteins, which are cleaved by the lipoprotein-specific (type II) SPase of B. subtilis (Lsp) (223, 291, 293). The major difference between signal peptides of lipoproteins and secretory proteins is the presence of a well-conserved lipobox in lipoprotein precursors. This lipobox contains an invariable cysteine residue that is lipid modified by the diacylglyceryl transferase prior to precursor cleavage by SPase II. After translocation across the cytoplasmic membrane, exported lipid-modified proteins remain anchored to the membrane by their amino-terminal lipid-modified cysteine residue (see the section on lipoprotein signal peptides for details). Notably, some signal peptides of lipoproteins contain a typical RR-motif. Consequently, the possibility exists that certain lipoproteins are exported via the Tat pathway rather than the Sec pathway.

The third major class is formed by signal peptides of prepilin-like proteins, which, in B. subtilis, are cleaved by the prepilin-specific SPase ComC (53). The recognition sequence for the prepilin SPase is, in contrast to that of secretory and lipoproteins, localized between the N- and H-domains, leaving the H-domain attached to the mature pilin after cleavage (53, 54, 157, 225).

Finally, the fourth major class of signal peptides is found on ribosomally synthesized bacteriocins and pheromones that are exported by ABC transporters (11, 203, 335). These signal peptides lack a hydrophobic H-domain and are removed from the mature protein by a subunit of the ABC transporter that is responsible for the export of a particular bacteriocin or pheromone or by specific SPases.

It is well established that B. subtilis can secrete certain proteins to high concentrations in the medium (184, 263). However, until recently it was difficult to estimate the number of exported proteins belonging to the secretome of B. subtilis. The completion of the B. subtilis genome sequencing project (149) and the availability of programs for the identification of signal peptides and transmembrane segments in large collections of protein sequences through worldwide web servers (194, 264) have now made it possible to predict the most likely location of all 4,107 annotated proteins (i.e., the proteome) of this organism. Computer-assisted studies have indicated that approximately 25% of the proteome of a given organism, such as B. subtilis, contains membrane sorting signals in the form of hydrophobic stretches of amino acids that can integrate in and span the membrane (35, 322; http://pedant.mips.biochem.mpg.de). Some of these putative membrane proteins contain amino-terminal signal peptides and may in fact be exported proteins, as indicated below.

To estimate the number of exported proteins, the amino termini of all annotated B. subtilis proteins in the SubtiList database (http://bioweb.pasteur.fr/GenoList/SubtiList) were used to predict amino-terminal signal peptides with the SignalP algorithm (194). This method incorporates a prediction of cleavage sites and a signal peptide/non-signal peptide prediction based on a combination of several artificial neural networks trained on the identification of signal peptides from gram-positive eubacteria. Next, all putative signal peptides were screened for the presence of a lipobox, RR-motif, or cleavage site for prepilin SPase. The numbers and features of each class of signal peptides are summarized in Fig. 2. It should be noted that polytopic membrane proteins, some of which can be cleaved by type I SPases (119, 288), were specifically excluded from the predictions, using the TopPred algorithm of Sipos and von Heijne (264). Furthermore, putative proteins with a single amino-terminal membrane-spanning domain, as encountered in certain type I SPases, might be falsely predicted to be secreted proteins. Finally, the neural networks of the SignalP algorithm, trained on data from gram-positive organisms, might not recognize some of the B. subtilis signal peptides with a more gram-negative or eukaryotic character.

View larger version (24K): [in this window] [in a new window]   FIG. 2.   Numbers and features of predicted amino-terminal signal peptides found in (putative) exported proteins of B. subtilis. To estimate the number of exported proteins, the first 60 residues of all annotated proteins of B. subtilis in the SubtiList database (http://bioweb.pasteur.fr/GenoList/SubtiList) were used to predict amino-terminal signal peptides with the SignalP algorithm for the prediction of signal peptides of gram-positive eubacteria (194). Next, to distinguish between potential secretory proteins and multispanning membrane proteins, putative membrane-spanning segments in protein sequences with a putative signal peptide were predicted with the TopPred2 algorithm (61, 264). All proteins containing additional hydrophobic domains (upper cutoff, 1.0; lower cutoff, 0.6; window size top, 11; window size bottom, 21) were regarded as membrane proteins, and their amino termini were excluded from the primary set of signal peptides. Finally, all putative signal peptides were screened for the presence of a lipobox, twin-arginine motif, or a cleavage site for the prepilin SPase. On the basis of SPase cleavage sites, predicted signal peptides were divided into four distinct classes: A, secretory (Sec-type) signal peptides and twin-arginine signal peptides; B, lipoprotein signal peptides; C, prepilin-like signal peptides; and D, bacteriocin and pheromone signal peptides. The number of predicted B. subtilis signal peptides of each class and the SPases responsible for their cleavage are indicated. Most signal peptides have a tripartite structure: a positively charged N-domain (N), containing lysine and/or arginine residues (indicated with +); a hydrophobic H-domain (H, indicated by a black box); and a C-domain (C) which specifies the cleavage site for SPase, as indicated by the scissors symbol. The average lengths of the complete signal peptide, N-domain, H-domain, and consensus SPase recognition sequences are indicated. Furthermore, helix-breaking residues, mostly glycine or proline (G/P) in the H-domain of certain signal peptides, are indicated. These residues are thought to facilitate loopwise membrane insertion and cleavage by SPase I, respectively (Fig. 1) (202). Finally, where appropriate, the most frequently occurring first amino acid (aa) of the mature protein (+1) is indicated.

Secretory (Sec-type) signal peptides. The signal peptide predictions resulted in the identification of 180 potential substrates for type I SPases. An RR-motif containing at least three residues of the R-R-X-#-# (where # is a hydrophobic residue) consensus sequence (18, 60) was found in 14 of these signals, suggesting that the corresponding preproteins are transported in a Sec-independent manner. The remaining 166 predicted "Sec-type" signal peptides (Table 1) had a length varying from 19 to 44 residues, with an average of 28 residues. These signal peptides contain on average two or three positively charged lysine (K) or arginine (R) residues in their N-domain, although some of the N-domains contain as many as 5 to 11 positively charged residues. The hydrophobic core (H-domain) has an average length of 19 residues, although a length of 17 or 18 residues seems to be preferred (Fig. 2 and 3). The C-domain of the predicted signal peptides carries a type I SPase cleavage site, with the consensus sequence A-X-A at position 3 to 1 relative to the SPase I cleavage site (Table 2). It is important to note that the C-domain must have an extended (-sheeted) structure for efficient interaction with the active site of type I SPases. Based on the crystal structure of the type I SPase of E. coli, the side chains of residues at the 1 and 3 positions are thought to be bound in two shallow hydrophobic substrate-binding pockets (S1 and S3) of the active site, whereas the side chain of the residue at position 2 is pointing outwards from the enzyme (202). It is presumably for this reason that residues tolerated at positions 3 and 1 of the signal peptide are generally small and uncharged, while almost all residues (except cysteine and proline) seem to be allowed at position 2 (Table 2). Nevertheless, a preference for serine (18%) at position 2 of the signal peptide seems to exist in B. subtilis. According to the predictions, an alanine residue is most abundant (27%) at position +1 of the mature protein, but all other residues, with the exception of cysteine and proline, seem to be allowed at this position (Table 2). The absence of proline at the +1 position is consistent with the observation that the SPase I of E. coli was inhibited by recombinant preproteins with proline at this position (12, 196). Finally, approximately 60% of the predicted signal peptides contain a helix-breaking residue (mostly glycine) in the middle of the H-domain, and about 50% contain a helix-breaking residue (proline or glycine) at position 7 to 4 relative to the predicted processing site for SPase I. 

View larger version (30K): [in this window] [in a new window]   FIG. 3.   Features of predicted secretory and lipoprotein signal peptides. (A) Length distribution of complete signal peptides (N-, H-, and C-domains). (B) Distribution of positively charged lysine or arginine residues in the N-domains of predicted signal peptides. (C) Length distribution of the hydrophobic H-domains in predicted signal peptides. Distributions are indicated as percentages of the total number of predicted secretory or lipoprotein signal peptides.

Twin-arginine signal peptides. Proteins containing a signal peptide with the RR-motif (R-R-X-#-#, where # is a hydrophobic residue) may be transported via the Tat pathway. Through a database search for the presence of this motif in amino-terminal protein sequences, a total number of 27 putative RR-signal peptides were identified. Putative SPase I cleavage sites are present in 14 of these predicted signal peptides. These cleavage sites show no striking differences from those of the predicted Sec-type signal peptides (Fig. 2 and Table 3). Notably, the RR-motif was also found in the signal peptides of five putative lipoproteins, suggesting that these proteins might also be substrates for the Tat pathway (Table 4). Moreover, eight additional signal peptide-like sequences with the RR-motif but lacking cleavage sites for SPase I or SPase II were identified. The corresponding proteins have the potential to remain attached to the membrane with an amino-terminal transmembrane domain (Table 3). Interestingly, some of these proteins even contain additional transmembrane segments. Thus, the possibility exists that certain membrane proteins are translocated via the Tat pathway. Altogether, the N-domains of predicted RR-signal peptides of B. subtilis have an average length of 13 amino acid residues and are twice as long as the N-domains of the typical (Sec-type) signals. Strikingly, no significant differences are observed between the H-domains of predicted Sec-type and RR-signal peptides of B. subtilis. In contrast, it has been suggested that the H-domains of RR-signal peptides of E. coli are, on average, longer and less hydrophobic than those of Sec-type signal peptides of this gram-negative organism (60). These observations may suggest either that a difference in the H-domains of Sec-type and RR-signal peptides is not important for translocation via the Tat pathway in B. subtilis or that some of the predicted RR-signal peptides do not direct proteins into the Tat pathway. For example, it is conceivable that the latter possibility could apply to WapA and WprA, the secretion of which was impaired by Ffh or SecA depletion (119). Finally, positively charged residues (arginine or lysine) in the C-domain, which can function as a so-called Sec avoidance signal that prevents interactions with Sec pathway components (23), are present in 8 of the 27 putative twin-arginine signal peptides.

Lipoprotein signal peptides. As the signal peptides of lipoproteins are, in general, shorter than those of secretory proteins, not all lipoproteins are recognized by the SignalP algorithm (194; our unpublished observations). In addition, some lipoproteins, such as CtaC (17) (Table 4), contain multiple membrane-spanning segments that were excluded from predictions of the signal peptides of secretory proteins mentioned above. Therefore, additional putative lipoprotein signal peptides were identified through similarity searches in the SubtiList database with signal peptides of known lipoproteins, using the Blast algorithm (4). Putative lipoprotein sorting signals identified by the latter method were combined with those identified by SignalP, resulting in a total number of 114 (Table 4). Signal peptides from lipoproteins differ in several respects from those of secretory signals. First, the structural features of lipoprotein signal peptides are more conserved than those of secretory signal peptides. This suggests that less variation in these peptides is allowed by the components involved in lipid modification and processing of lipoproteins. The C-domain contains a so-called lipobox with the consensus sequence L-(A/S)-(A/G)-C (Table 5), of which the invariable cysteine residue is the target for lipid modification and becomes the first residue of the mature lipoprotein after cleavage by SPase II. Second, both the N-domain (average of four residues) and the H-domain (average of 12 residues) seem, on average, to be shorter than the corresponding domains of signal peptides of nonlipoproteins (Fig. 2 and 3). Finally, helix-breaking residues are less abundant (27%) in the H-domain of lipoprotein signal peptides than in the corresponding regions of nonlipoprotein signal peptides. As transmembrane helices seem to require at least 14 hydrophobic residues to span the membrane (34, 46), the latter findings suggest that not all lipoprotein signal peptides can span the membrane completely. This implies that the active site of SPase II may be embedded in the cytoplasmic membrane, as was suggested for SPase I (202, 309). Alternatively, the N-domain of the signal peptide may not stay fixed at the cytoplasmic surface of the membrane during translocation. Strikingly, aspartic acid was absent from the +2 position of predicted mature lipoproteins. In mature lipoproteins of gram-negative eubacteria, an aspartic acid residue at the +2 position specifically prevents the sorting of these proteins to the outer membrane (167, 168, 221). The fact that aspartic acid at the +2 position is absent from Bacillus lipoproteins suggests that this sorting signal has evolved exclusively in gram-negative eubacteria. Nevertheless, it has to be noted that glycine, phenylalanine, or tryptophan residues can be found at the +2 position of various lipoproteins of B. subtilis (Table 5). Like aspartic acid, these residues prevent lipoprotein sorting to the outer membrane of E. coli (260). Thus, even though B. subtilis lacks an outer membrane, residues with a potential sorting function can be found at the +2 position of mature lipoproteins of B. subtilis. It is therefore conceivable that such residues are involved in the targeting of lipoproteins to specific membrane locations in B. subtilis. Nevertheless, it has to be emphasized that presently no experimental data are available to support this idea.

Type IV prepilin signal peptides. Only four proteins required for DNA binding and uptake during competence (ComGC, ComGE, ComGD, and ComGG), which are already known to contain type IV prepilin-like signal peptides (55), were identified in our SignalP and Blast homology searches for prepilin-like signal peptides in B. subtilis (Table 6). As the prepilin SPase (ComC) acts at the cytoplasmic side of the membrane (151, 157), the "C-domain" with the ComC cleavage site is localized between the N- and H-domains of the prepilin signal peptide (Fig. 2). After cleavage by ComC, the hydrophobic H-domain remains attached to the mature protein. Surprisingly, potential prepilin cleavage sites are also present in two known cold shock proteins, CspB and CspC (108), and one hypothetical protein, YqaF (Table 6), which have a (predicted) cytosolic localization as they lack an H-domain or transmembrane segments. Since the catalytic domain of ComC is located at the cytoplasmic side of the membrane, it is conceivable that cytoplasmic proteins containing a ComC cleavage site are substrates for this enzyme. However, it is presently not known whether CspB, CspC, or YqaF is processed.

Signal peptides of bacteriocins and pheromones. Bacteriocins and pheromones (other than PhrA to PhrK) with a cleavable amino-terminal signal peptide form a distinct group of exported proteins that are exported via ABC transporters. Signal peptides of this class of secreted proteins cannot be predicted by the regular algorithms for signal peptide prediction, such as SignalP (see previous sections), as they consist only of N- and C-domains and completely lack a hydrophobic H-domain (Fig. 2). In B. subtilis 168, the three known signal peptides of this type direct the secretion of the bacteriocins sublancin 168 (203) and subtilosin (336) and the pheromone ComX (159). Thus far, no other signal peptides with a similar structure have been identified in the amino acid sequences of B. subtilis proteins. Obviously, this does not exclude the possibility that less related signal peptides of this type do exist, particularly in view of the fact that at least 77 (putative) ABC transporters have been identified in B. subtilis (149).

Most proteins seem to be exported or inserted into the cytoplasmic membrane via the Sec pathway in B. subtilis. Nevertheless, several alternative export pathways seem to exist. First, the recently identified twin-arginine translocation (Tat) pathway seems to be present in B. subtilis, as judged from the identification of signal peptides with the RR-motif and conserved components of this pathway (see the section on Sec-independent protein transport). Protein secretion via this pathway was shown to be independent of Sec components in E. coli and plant chloroplasts. Possibly, this pathway has evolved specifically for the export of folded preproteins (65). Second, the assembly of extracellular prepilin-like structures depends on components which are, most likely, not involved in Sec-dependent protein secretion. Finally, at least three small prepeptides contain signal peptides lacking a hydrophobic domain. These prepeptides are transported across the membrane and cleaved by ABC transporters. The requirements for traveling via one of these export pathways are summarized in Fig. 4. In the following sections, each of these export pathways will be discussed in more detail.

View larger version (58K): [in this window] [in a new window]   FIG. 4.   Predicted protein transport pathways in B. subtilis. Based on the predictions of signal peptides and various retention signals, it is hypothesized that at least four different protein transport pathways exist in B. subtilis that can direct proteins to at least five different subcellular destinations. Ribosomally synthesized proteins can be sorted to various destinations depending on the presence (+SP) or absence (SP) of an amino-terminal signal peptide and specific retention signals such as lipid modification or cell wall-binding repeats (CWB). (A) Proteins devoid of a signal peptide remain in the cytoplasm. (B) Proteins with one or more membrane-spanning domains are inserted into the membrane either spontaneously (not shown), via the Sec pathway or, according to our predictions (Table 3), via the Tat pathway (+RR). (C) Proteins which have to be active at the extracytoplasmic side of the membrane can either be lipid-modified proteins (+lipobox) exported via the Sec or Tat pathways or prepilins exported by the Com system. (D) Proteins that need to be retained in the cell wall can be exported via the Sec or Tat pathway. In order to be retained, the mature part of these proteins contains cell wall-binding repeats (+CWB). (E) Proteins can be secreted into the medium via the Sec or Tat pathway or by ABC transporters. (F) Different mechanisms can be employed to transport proteins to the IMS of endospores.

The various components of the Sec-dependent secretion machinery can be divided into six groups: cytosolic chaperones, the translocation motor (SecA), components of the translocation channel (SecYEG and SecDF-YajC), SPases, SPPases, and, finally, folding factors that function at the trans side of the membrane. The main components of the secretion machinery of B. subtilis are depicted in Fig. 5 and listed in Table 7. In addition, the homologous and/or analogous components from E. coli, the archaeon Methanococcus jannaschii, and the eukaryon S. cerevisiae are listed in Table 7. As protein secretion has not been studied experimentally in archaea, the identification of components of the secretion machinery of M. jannaschii is based entirely upon data deduced from the genome sequence (44).

View larger version (40K): [in this window] [in a new window]   FIG. 5.   Main components of the Sec-dependent B. subtilis protein secretion machinery. The SRP complex consists of Ffh and the scRNA. See text for further details. C, carboxyl terminus; M, mature protein; N, amino terminus; R, ribosome; SP, signal peptide.

Secretion-dedicated chaperones. In B. subtilis, the only secretion-specific chaperone thus far identified is the Ffh protein (fifty-four homologue), a GTPase that is homologous to the 54-kDa subunit of the eukaryotic signal recognition particle (Srp54) (122). This protein forms a complex (denoted SRP) with the small cytoplasmic RNA (scRNA) that is functionally related to the eukaryotic 7S RNA (called scR1 RNA in S. cerevisiae) and the E. coli 4.5S RNA (186, 187). Recent data have shown that HBsu, a histone-like protein of B. subtilis, is also associated with the scRNA. Notably, HBsu was shown to bind to a region of scRNA that is not conserved in the 4.5S RNA of E. coli, suggesting that the E. coli SRP lacks an HBsu-like component (188, 333). The ternary ribonucleoprotein SRP complex of B. subtilis binds to the signal peptides of nascent chains emerging from the ribosome and is targeted to the membrane with the aid of the FtsY protein (also called Srb) (200). FtsY is a homologue of the eukaryotic SRP receptor -subunit (DP) that is essential for SRP-dependent protein secretion and cell viability, like the Ffh protein. In eukaryotic cells, SRP-dependent protein translocation occurs cotranslationally. The SRP of S. cerevisiae consists of a complex of seven subunits and the 7S RNA. Two of these subunits, Srp7 (Srp9 in mammalian cells) and Srp14, are responsible for a translation arrest as soon as the signal peptide emerges from the ribosome (262). The whole complex, consisting of the ribosome, nascent chain, and SRP, docks onto the SRP receptor (also termed docking protein), which consists of DP and DP. Next, SRP is released and polypeptide translation by the ribosome is resumed. Protein synthesis is likely to provide the driving force for cotranslational protein translocation across membranes (249). A similar SRP-mediated translation arrest probably does not occur in eubacteria. First, it was shown that E. coli SRP and FtsY do not arrest translation in a eukaryotic in vitro translocation assay (222). Second, eubacteria lack the SRP components that are responsible for translation arrest in eukaryotes. Moreover, a specific translation arrest may not even be required for cotranslational translocation in eubacteria, because the traffic distances are short and the protein translocation rates are high compared to the translation rate (for E. coli, estimated at approximately 10-fold) (225, 302).

View larger version (54K): [in this window] [in a new window]   FIG. 6.   Model for protein targeting to the B. subtilis Sec translocase. Precursor proteins with the most hydrophobic targeting signals are bound cotranslationally by the SRP complex. The complex formed by the nascent chain, ribosome, and SRP docks the preprotein at an unidentified site at the membrane with the aid of FtsY. The subsequent release of the nascent chain and ribosome from the SRP-FtsY complex appears to be preceded or accompanied by GTP binding to the Ffh protein in SRP and FtsY. Hydrolysis of GTP bound to both Ffh and FtsY mediates the dissociation and recycling of these targeting components. A subset of precursor proteins containing relatively fewer hydrophobic targeting signals interact posttranslationally with an unidentified chaperone, possibly CsaA, which targets the complex to the translocase. After their binding to the Sec translocase, precursors are translocated across the membrane through cycles of ATP-dependent insertion and deinsertion of SecA into the translocation channel (see text for details). This model is largely based upon the model for the E. coli targeting routes proposed by Valent et al. (302). C, carboxyl terminus; M, mature protein; N, amino terminus; R, ribosome; SP, signal peptide.

General chaperones. In addition to secretion-dedicated chaperones, chaperones with a general function in protein unfolding and folding might also function in protein translocation. Lill et al. (154) were the first to demonstrate that the so-called trigger factor (TF) is a cytosolic ribosome-bound protein that can maintain the translocation competence of the precursor form of the outer membrane protein OmpA in vitro. However, later studies indicated that depletion of TF did not block the export of proOmpA (109). Recently, it was shown that TF is a ribosome-bound peptidyl-prolyl cis/trans isomerase (PPIase, FK506 binding protein type [FKBP]) that interacts with both secretory and cytosolic proteins (117, 300, 301). Thus, the interaction with TF might represent a decision point for proteins to enter either the GroEL/ES folding pathway when these proteins are to remain cytosolic or to enter a targeting pathway to the translocase for secretory proteins. In addition, recent results indicate that secretory proteins are directed into the SecB-SecA-mediated posttranslational targeting pathway by means of their preferential recognition by TF (15). In B. subtilis, it was shown that TF, together with a second cytoplasmic PPIase (cyclophilin, also called PpiB), accounts for the entire PPIase activity in the cytoplasm (107). A direct involvement of TF or cyclophilin in protein secretion by B. subtilis has not been reported. However, it was recently shown that TF of the gram-positive eubacterium Streptococcus pyogenes is important for the secretion of the cysteine proteinase SCP (158). Interestingly, TF was required both for guiding SCP into the secretory pathway and for establishing an active conformation after translocation. This suggests that the cis-trans isomerization of certain peptidyl-prolyl bonds before translocation is important for the folding of the protein after translocation. However, an alternative explanation is that TF is involved in the secretion of an extracellular foldase that is required for folding of SCP at the trans side of the membrane.

The preprotein translocation machinery of the E. coli Sec pathway consists of at least seven proteins: SecA, which is the translocation motor, and the integral membrane proteins SecD, SecE, SecF, SecG, SecY, and YajC. Homologues of all of these components have been identified in B. subtilis. In the current model of preprotein translocation, which is based largely on results obtained in E. coli, several successive steps in the translocation of proteins are proposed (77, 79, 83, 304). First, SecA binds to acidic phospholipids and SecY (82, 110, 156, 161, 270) and is activated for recognition of SecB and the preprotein (110). Preprotein binding is followed by the binding and hydrolysis of ATP (155). The binding of ATP causes major conformational changes of SecA (72, 303), leading to a release of SecB (92, 93) and insertion of the carboxyl terminus of SecA into the membrane (85, 86, 94, 224). This membrane insertion, which occurs through the translocase complex (85, 87, 88, 224) promotes the translocation of a short fragment of the preprotein (251). Next, ATP is hydrolyzed by SecA, leading to release of the preprotein and deinsertion of SecA (85, 251). Once protein translocation is initiated by SecA, further translocation is driven by both repeated cycling of SecA through ATP binding and hydrolysis and the proton motive force (78, 100, 261).

The B. subtilis gene encoding SecA was initially identified as a gene called div, mutations in which affected cell division, sporulation, germination, protein secretion, autolysis, and the development of competence for DNA binding and uptake (236, 237). In fact, cloning and sequencing of the div gene, which is essential for cell viability, revealed that it encodes SecA (238). In addition, the B. subtilis secA gene was cloned independently, using hybridization with an E. coli secA probe (201). B. subtilis SecA can complement E. coli SecA mutants, provided that the protein is expressed at moderate levels (140). Notably, yeasts and archaea do not contain SecA homologues even though they contain a Sec-type protein-conducting channel (see below). In S. cerevisiae, the driving force for protein translocation across the ER membrane is generated either by the ribosome, in the case of cotranslational translocation, or by the ER-luminal Kar2 protein (an Hsp70 homologue), in the case of posttranslational translocation (249). The fact that SecA is absent from archaea, at least the ones for which the genome has been sequenced completely, suggests that these organisms use either another cytoplasmic ATPase, the SRP pathway, and/or protein synthesis as a force generator for protein translocation.

Notably, in addition to its role in preprotein translocation, SecA of B. subtilis could also fulfill the role of an export-specific chaperone, as suggested by Herbort et al. (116). Such a role for SecA of B. subtilis would be consistent with its recently documented low affinity for SecYEG (283). The latter observation implies that, at least in B. subtilis, the levels of cytosolic SecA are relatively high, which would facilitate early interactions with proteins destined for export.

A heterotrimeric complex of SecY, SecE, and SecG forms the main core of the translocation channel (42). This complex is found not only in eubacteria but also in eukaryotes (Sec61p complex) (106) and archaea (216). The SecY and SecE homologues in eukaryotes are called Sec61 and Sec61, respectively. SecG is not conserved, but this protein could be a functional analogue of the eukaryotic Sec61. Based on sequence similarity and the observation that archaea probably contain a Sec61 homologue rather than SecG, it was suggested that the archael translocase is more related to the eukaryotic Sec61p complex (216). During the last decade, homologues of SecY, SecE, and SecG have been identified in B. subtilis, either by complementation studies using E. coli sec mutants or by DNA sequencing (131, 185, 280, 312). SecY, SecE, and SecG of B. subtilis are membrane proteins, with 10, 1, and 2 membrane-spanning domains, respectively (131, 185, 312). Notably, B. subtilis SecE is considerably smaller than E. coli SecE and has only one membrane-spanning domain, while E. coli SecE has three such domains (250). Nevertheless, SecE of B. subtilis was able to complement the cold-sensitive and export-defective phenotype of an E. coli SecE mutant, showing that it is a true SecE homologue (131). Based on data from posttranslational protein transport experiments in the S. cerevisiae ER membrane, Plath et al. (213) postulated that SecE and signal peptides bind to the same or overlapping regions in SecY and that SecE functions as a surrogate signal peptide when the SecY channel is in its closed form in the absence of translocating protein. Upon the arrival of a signal peptide, it would displace SecE and thus open the SecY channel for transport. In contrast to SecY and SecE, SecG is not strictly required for preprotein translocation and cell viability. Nevertheless, it is required for efficient translocation, possibly by facilitating the movement of preproteins through the translocation channel in concert with the insertion and deinsertion cycles of SecA (165). The absence of SecG from E. coli causes a cold-sensitive growth phenotype, as frequently encountered in E. coli strains in which protein secretion via the Sec pathway is compromised (215). Similarly, disruption of the B. subtilis secG (yvaL) gene caused secretion defects that resulted in cold-sensitive growth (312), confirming the earlier conclusion by Bolhuis et al. (27) that protein translocation in B. subtilis is intrinsically cold sensitive, as it is in E. coli. Furthermore, the cold sensitivity of the B. subtilis secG mutant was exacerbated by overproduction of secretory preproteins. Interestingly, the growth and secretion defects of the B. subtilis secG mutant could be complemented by the expression of the E. coli secG gene even when secretory preproteins were overproduced. Finally, consistent with the role of SecG in E. coli, B. subtilis SecG stimulated the ATP-dependent in vitro translocation of the precursor pre-PhoB by the B. subtilis SecA-SecYE complex (283).

In addition to the genes encoding the SecYEG core elements of the translocase, a gene encoding the SecDF protein was also identified in B. subtilis (27). In contrast to the secD and secF genes identified in most other organisms, B. subtilis was shown to contain a natural gene fusion between the equivalents of secD and secF. Consequently, SecDF of B. subtilis is a molecular Siamese twin, with 12 putative transmembrane domains. Notably, SecDF shows both sequence similarity and structural similarity to secondary solute transporters. It was demonstrated that B. subtilis SecDF, which is not essential for cell viability, is merely required to maintain a high capacity for protein secretion (27). Unlike the SecD subunit of E. coli (166), the B. subtilis SecDF protein does not seem to be required for the release of a mature secretory protein from the membrane. It has been suggested that SecD and SecF of E. coli modulate the cycling of SecA (83, 86, 137). However, it was also noted that archaea, which contain separate SecD and SecF proteins, do not contain a SecA homologue (216). Therefore, it is conceivable that SecD-SecF has another function in protein translocation, such as assembly of the translocase (216) or clearing of the translocation channel from signal peptides or misfolded proteins (27). The latter idea would be consistent with the observation that SecDF shows structural similarity to secondary solute transporters.

For E. coli, it was shown that SecD and SecF form a heterotrimeric subcomplex with a third protein (denoted YajC) and that this complex constitutes a large "holoenzyme" with the SecYEG complex (83). YajC is specified by the first gene of the SecDF operon. A gene encoding a homologue of the E. coli YajC protein, denoted yrbF, was also identified on the B. subtilis genome (53% identical plus conservative residues), but its involvement in protein secretion has not been documented so far. This yrbF gene is located in the same chromosomal region as the secDF gene but, in contrast to E. coli, it is not cotranscribed with secDF (27). Disruption of the yajC gene of E. coli did not have a clear effect on protein export, but it was shown that overproduction of YajC suppresses the dominant-negative phenotype of the secY-d1 mutation, an internal in-frame deletion in the secY gene (287).

Finally, the Sec translocon in the ER of eukaryotic cells contains, in addition to the Sec61 core components, a number of other membrane proteins (338). These include the components of the Sec62/63 and Sec66/67 (also called Sec71/72) complexes. In addition, translocation complexes in the ER of mammalian cells contain the TRAM protein. The function of these proteins is not fully clear, but they may have functions analogous to those proposed for the SecD/F proteins (216).

SPases remove signal peptides from secretory preproteins when their C-domain emerges at the extracytoplasmatic side of the membrane. This reaction is a prerequisite for the release of the mature secretory protein from the membrane (64, 67). One of the most remarkable features of the B. subtilis protein secretion machinery is the presence of multiple, paralogous type I SPases. This in contrast to the situation in many eubacteria, archaea, and yeasts, in which one type I SPase seems to be sufficient for the processing of secretory preproteins (44, 67, 105, 269). For most eukaryotic species, however, the presence of two paralogous SPases appears to be characteristic (67). The largest numbers of known paralogous SPases appear to be present in the archaeon Archaeoglobus fulgidus, which contains four genes for type I SPases (139), and B. subtilis, in which seven sip genes for type I SPases have been identified so far. Five of the sip genes of B. subtilis (sipS, sipT, sipU, sipV, and sipW) are located on the chromosome (26, 289, 290, 307, 308); two additional sip genes (sipP) are located on plasmids that were identified in natto-producing strains of B. subtilis (171, 172). As was shown for E. coli (66, 306) and S. cerevisiae (25), SPase I activity in B. subtilis is essential for cell viability. Although all five chromosomally encoded SPases can process secretory preproteins, only SipS and SipT are of major importance for preprotein processing and viability, whereas SipU, SipV, and SipW have a minor role in protein secretion (290). Notably, SipS and SipT can be functionally replaced by the plasmid-encoded SPase SipP (292). The latter three SPases are therefore considered to be the "major" SPases, which have substrate specificities that differ at least partly from those of SipV, SipU, and SipW, the "minor" SPases (Fig. 7). These findings indicate that the minor SPases are specifically required for the processing of a subset of the 180 predicted secretory preproteins. Indeed, SipW seems to be specifically involved in the processing of pre-TasA and pre-YqxM, two preproteins that are encoded by genes flanking the sipW gene (276, 277, 278). Surprisingly, SipW shows high degrees of sequence similarity not only to certain SPases found in sporulating gram-positive eubacteria, but also to the SPases of archaea and the eukaryotic ER membrane. Together these SipW-like SPases form the subfamily of ER-type SPases. In contrast, all other known B. subtilis SPases are of the prokaryotic type (P-type). Such P-type SPases have thus far been found exclusively in eubacteria, mitochondria, and chloroplasts (290). As demonstrated by site-directed mutagenesis of various P-type SPases, including SipS of B. subtilis (298, 305), and by X-ray crystallography of the E. coli SPase I (202), the P-type SPases make use of a serine-lysine catalytic dyad. In all known eubacterial P-type SPases, the active-site serine residue is predicted to be localized at the extracytoplasmic membrane surface. In SipS, SipT, SipU, SipV, and SipP of B. subtilis, this serine residue is kept in position by a unique amino-terminal membrane anchor. In contrast to the latter SPases, SipW appears to have a carboxyl-terminal membrane anchor in addition to an amino-terminal membrane anchor that precedes its active-site serine residue (290). However, the major difference between P- and ER-type SPases is that the catalytic lysine residue of the P-type SPases is replaced with a histidine residue in the ER-type SPases (67, 290, 307). Recent studies have shown that conserved serine, histidine, and aspartic acid residues are critical for the activity of SipW and the ER SPase Sec11 of S. cerevisiae, indicating that ER-type SPases employ a Ser-His-Asp catalytic triad or, alternatively, a Ser-His catalytic dyad (293a, 311).

View larger version (62K): [in this window] [in a new window]   FIG. 7.   Type I and II SPases of B. subtilis. The type I SPases, responsible for the processing of the 180 predicted secretory preproteins, can be divided into two groups: the major SPases SipS (S), SipT (T), and SipP (P), which are important for cell viability, and the minor SPases SipU (U), SipV (V), and SipW (W), which are not important for cell viability under laboratory conditions (290, 292). SipP is encoded by plasmid-borne genes on the plasmids pTA1015 and pTA1040, which are present in certain natto-producing B. subtilis strains. All other SPases are chromosomally encoded. The transcription of the genes for the major SPases increases during the postexponential growth phase in concert with the genes for most secretory proteins, whereas the minor SPases are transcribed at a low level during all growth phases. SipW is the only ER-type SPase, showing a high degree of similarity to eukaryotic and archaeal SPases. In contrast to the prokaryotic (P-)type SPases of B. subtilis, which have one amino-terminal membrane anchor, SipW appears to have an additional carboxyl-terminal membrane anchor. Finally, B. subtilis contains only one gene (lsp) for a type II SPase (L) with four membrane anchors, which is required for the processing of the 114 predicted lipoproteins (291) (see text for details).

Even though it is well established that various type I SPases of B. subtilis have different substrate specificities, the molecular basis for these differences with respect to the composition of the C-domain of signal peptides and the structure of the active sites of the different SPases is presently unknown.

In contrast to the type I SPases, B. subtilis contains only one gene for a type II SPase (lsp) (223), which is specifically required for the processing of lipid-modified preproteins. All known SPases of this type are integral membrane proteins with four (putative) membrane-spanning segments, the amino and carboxyl termini having a predicted cytosolic localization (183, 223, 293). The potential active site of SPase II, formed by two aspartic acid residues, is located in close proximity to the extracytoplasmic surface of the membrane, similar to the active-site serine residue of type I SPases (293).

Interestingly, cells lacking SPase II are viable under standard laboratory conditions. This indicates that processing of lipoproteins by SPase II in B. subtilis is not strictly required for lipoprotein function, as at least one lipoprotein, PrsA, is essential for viability (144, 291). Although certain lipoproteins are required for the development of genetic competence, sporulation, and germination, these developmental processes were not detectably affected in the absence of SPase II. Cells lacking SPase II accumulated lipid-modified precursor and, surprisingly, mature-like forms of the lipoprotein PrsA, which is involved in the folding of secreted proteins. These forms of PrsA appeared to be reduced in activity, as the secretion of the B. amyloliquefaciens -amylase AmyQ, the folding of which is dependent on PrsA, was strongly impaired (291). It is presently not clear which proteases are responsible for the alternative processing of PrsA in the absence of SPase II. However, the involvement of type I SPases in this process appears to be highly unlikely (291). The cellular level of another lipoprotein, CtaC, is strongly reduced in the absence of SPase II, indicating that lipoprotein processing is important for the stability of certain proteins (17).

Diacylglyceryl modification of the cysteine residue at position +1 of the mature lipoprotein in lipoprotein precursors is catalyzed by the lipoprotein diacylglyceryl transferase (Lgt). This lipid modification is a prerequisite for processing of the lipoprotein precursor by SPase II (103, 147, 242, 243, 244, 294). Similar to the disruption of the lsp gene for SPase II, disruption of the lgt gene results in the accumulation of unprocessed (and unmodified) lipoproteins without affecting growth and cell viability (153, 291). Like SPase II, Lgt is required for the stability of several lipoproteins and the efficient secretion of AmyQ (17, 153). Interestingly, the processing rate of secreted nonlipoproteins was retarded in the absence of SPase II, which must be attributed to the malfunction of lipoproteins other than PrsA (291). Two candidate proteins that might be responsible for this effect are the putative lipoproteins SpoIIIJ and YqjG (Table 3), which show significant sequence similarity to the mitochondrial Oxa1 protein. The latter protein was shown to be required for export of the amino and carboxyl termini of the mitochondrially encoded precursor of cytochrome c oxidase subunit II (pre-CoxII) from the mitochondrial matrix to the intermembrane space (114, 115), proteolytic processing of pre-CoxII (13), and assembly of the cytochrome c oxidase and oligomycin-sensitive ATP synthase complexes (3, 33). As mitochondria lack Sec components, Oxa1p seems to represent a component of a specific protein export and/or assembly machinery that might be conserved in eukaryotic organelles and eubacteria (115, 235). Consistent with this hypothesis, it was recently shown that the E. coli homologue of SpoIIIJ/YqjG and Oxa1p, denoted YidC, is associated with the Sec translocase (256a).

Finally, processed lipoproteins of E. coli are further modified by aminoacylation of the diacylglyceryl-cysteine amino group (103, 147, 242, 243, 244, 294). The latter lipid modification step does not seem to be conserved in all eubacteria, as many organisms, including B. subtilis, lack an lnt gene for the lipoprotein N-acyltransferase (291).

After being cleaved off from the mature protein, signal peptides are rapidly degraded. In E. coli, the signal peptide of the major lipoprotein (Lpp; also called Braun's lipoprotein) was shown to be degraded by the membrane-bound protease IV (also called signal peptide peptidase, encoded by the sppA gene) (123, 125, 126). Nevertheless, in the absence of protease IV, significant levels of signal peptide degradation were observed (282), suggesting that other proteases can replace protease IV in this process. In addition to the protease IV, the cytoplasmic oligopeptidase A (OpdA) was shown to be involved in the degradation of the Lpp signal peptide (197, 198). Current models imply that protease IV cleaves the signal peptide in the membrane in two fragments (Fig. 1), which are further degraded in the cytoplasm by OpdA (198). Interestingly, independent of its effect on natural signal peptides, OpdA was also identified in suppressor screens using secretory proteins with defective signal peptides. In such screens, it was noticed that the prlC (protein localization) mutation, which turned out to be a mutation in the opdA gene, could suppress the export defect of certain LamB signal sequence mutants (57). At present, it is not clear how mutations in OpdA can lead to the observed prlC phenotype.

Protease IV is highly conserved in eubacteria and archaea. In contrast, OpdA appears to be absen