To access the private area of this site, please log in.

Username: 

Password:  

 

Chapter 64

Export of Proteins to the Cell Envelope in Escherichia coli

CHRISTOPHER K. MURPHY and JON BECKWITH

 

INTRODUCTION

A central process in the construction of a bacterial cell is the localization of proteins to different cellular compartments. In gram-negative bacteria such as Escherichia coli and Salmonella typhimurium, proteins must be exported to three major cell envelope compartments, the cytoplasmic membrane, the periplasmic space, and the outer membrane. Evidence has accumulated that the initial steps in the localization of proteins to these three compartments, the insertion and transfer across the cytoplasmic membrane, involve a common mechanism. Furthermore, this process appears to be conserved across bacterial species, with gram-positive bacteria and archaebacteria employing the same cellular apparatus for membrane localization and secretion (61, 73, 109). Strikingly, the core components of the export machinery in the cytoplasmic membrane of bacteria have their homologs in eukaryotic cells. From the yeast Saccharomyces cerevisiae to dog pancreas microsomes, the translocation apparatus in the membrane of the endoplasmic reticulum includes homologs of at least two of the components found in prokaryotes (61).

In addition to the localization of cell envelope proteins, many gram-negative bacteria must also export proteins beyond the outer membrane into the medium (128). In general, the normal export pathway is not sufficient to carry out this process. For export of proteins such as toxins, degradative enzymes, and others into the medium, bacteria have evolved a variety of specialized pathways, many of which do not even utilize the machinery involved in cell envelope protein export. We will not consider in this review the large literature on the mechanism of protein export into the medium.

The broad outlines of the steps in protein export are as follows. Proteins destined for the cell envelope are synthesized initially with export signals or signal sequences. In the case of periplasmic or outer membrane proteins, these signals are amino terminal and are cleaved from the protein during the export process. In the case of cytoplasmic membrane proteins, these signals are usually not cleaved. For this latter class of proteins the signals can be amino terminal or internal to the protein and can serve both as export signals and as membrane-spanning segments in the assembled protein. The export signals along with their attached proteins are recognized within the cell by an export apparatus. This recognition occurs while the proteins are still being translated, although translocation across the membrane is not entirely concomitant with translation. Rather, it appears that export of the protein may only begin after considerable extension of the polypeptide chain has taken place (132). Furthermore, under some conditions translocation of a protein can be observed after its complete translation and release from ribosomes (89). This posttranslational export is observed when protein export is made defective in some way, e.g., when the effectiveness of a signal sequence is reduced by mutation. In contrast, it is thought that protein secretion in eukaryotic cells is tightly coupled to translation.

The export apparatus, a collection of Sec proteins, has been defined mainly by genetic studies and the function of the components analyzed by a combination of genetic, biochemical, and in vitro secretion studies (13, 144, 182). Various genetic approaches described below have allowed the identification of six Sec proteins. These include the cytoplasmic protein SecB, the peripheral membrane protein SecA, and the integral cytoplasmic membrane proteins SecD, SecE, SecF, and SecY. An additional membrane protein, SecG, has been identified by its contribution to the efficiency of secretion in an in vitro system. The steps carried out by these proteins include recognition of proteins to be exported, targeting to the membrane export apparatus, and subsequent translocation across the membrane. Homologs of the proteins SecE and SecY are found in eukaryotic cells (61). Combined evidence from prokaryotes and eukaryotes suggests that these two sec gene products may be key to the passage of proteins through membranes.

During the export process, cleavable signal sequences are removed from their proteins via the action of leader peptidases. These cleavage enzymes are membrane proteins with their active sites localized to the periplasmic space. Two such enzymes have been detected in E. coli (37). One, LspA, is responsible for the removal of signal sequences from lipoproteins, proteins which are modified with glycerides (68, 163, 189, 191). The other leader peptidase, Lep, cleaves signal sequences from the remaining vast majority of cell envelope proteins (39, 193).

Several other proteins have been proposed to be part of the E. coli export machinery. We will discuss the basis for these proposals and the evidence for their role. We conclude in the Summary with a model for protein export reflecting the current state of knowledge.

 

THE SECRETION APPARATUS

 

Genetic Approaches to the Secretion Machinery in Bacteria

In Vivo Genetic Selections.

A variety of genetic approaches have been used to define the genes coding for components of the E. coli secretion machinery (Table 1). The first selection to yield mutations in such a gene involved the use of signal sequence mutants. Emr et al. (48) reasoned that it might be possible to restore protein export to a protein with a defective signal sequence by mutational alteration of a component of the secretion machinery. Selection for suppressors of a signal sequence mutation of the bacteriophage λ receptor yielded mutations in the secY gene. These mutations (termed prlA mutations) restored export to a variety of proteins with defective signal sequences. Subsequently, Ito et al. obtained, by localized mutagenesis, a temperature-sensitive mutation in the secY gene that showed a pleiotropic defect in protein secretion (71).

A second genetic selection asked for mutations that altered the localization of an exported protein to the bacterial cytoplasm from its normal location in the cell envelope. This was accomplished using gene fusions in which the cytoplasmic protein β-galactosidase was fused to exported proteins carrying their signal sequences (54, 81, 116). This attempt to force export of β-galactosidase across the cytoplasmic membrane resulted in the protein becoming embedded in the membrane, where it was unable to assume its active conformation. Mutations that interfered with the functioning of the signal sequence of the hybrid protein prevented this attempted export. As a result, β-galactosidase was restored to its cytoplasmic location, yielding an active enzyme and allowing cells to utilize lactose as a carbon source. Thus, selection for Lac⁺ derivatives of these fusion strains yielded mutants that were defective in export of the hybrid protein. While many of these mutations altered the hybrid protein’s signal sequence, others were found to be unlinked to the gene fusion and defined the secA, secB, and secD genes.

Finally, a detailed study of secA revealed a regulatory property of this gene which made another genetic screening procedure possible. Expression of the secA gene is regulated by the secretion needs of the cell (117). Mutations that block secretion result in up to a 10-fold increase in SecA synthesis. Thus, mutations in genes coding for components of the secretion machinery could be detected by screening for derivatives in which secA expression was increased. The secA gene was fused to the lacZ gene so that the level of β-galactosidase could be used to monitor increases in secA expression. These studies led to the identification of two additional genes, secE and secF (55, 139, 146).

Silhavy’s group also found mutations in the secA and secE genes through isolation of additional suppressors (prl mutations) of signal sequence mutants (153). Thus, the genetic studies led to the finding of six sec genes, secA, secB, secD, secE, secF, and secY. Three of these genes (secA, secE, and secY) were identified both by mutations that decreased export and by mutations that enhanced export of proteins with defective signal sequences.

A genetic selection also yielded mutations in the gene for the lipoprotein leader peptidase. This enzyme is sensitive to the antibiotic globomycin, which inhibits prolipoprotein processing and leads to cell death. Globomycin-resistant mutants, including one which is temperature sensitive for growth, map to the gene for this enzyme, lspA (188, 189). The gene was also identified by isolation of a plasmid clone that conferred globomycin resistance on E. coli (163).

"Reverse Genetics."

Verification for an in vivo role of two genes in secretion has been established by working backwards from a characterized protein to locate the corresponding gene. In the case of the major leader peptidase, the lep gene was first identified via isolation of a plasmid clone that expressed high levels of the protein (39). Subsequently, a null mutation in the lep gene was constructed in the presence of a cloned lep gene under the control of the arabinose promoter. Depletion of leader peptidase from cells led to cell death and to the accumulation of exported proteins with their signal peptide still attached (38). Despite the failure to cleave the signal sequence, the proteins were still exported, demonstrating that such cleavage is not necessary for protein translocation.

Starting with the amino acid sequence of SecG (see below), the secG gene was located on the E. coli chromosome and a null mutation in secG was constructed (110). This mutant was viable at 37°C and exhibited only a slight effect on secretion; at 20°C, however, the mutant grew very slowly and exhibited a significant export defect.

 

Function of the Sec Proteins

SecB.

Mutations in the secB gene have strong effects on the export of only a subset of cell envelope proteins, including maltose-binding protein, OmpA, and bacteriophage λ receptor (82). For these proteins, SecB is required for in vitro secretion into membrane vesicles as well (59, 83). Particularly as a result of studies on maltose-binding protein, the role of SecB is perhaps the best worked out of any of the components of the secretion machinery. SecB is required to maintain maltose-binding protein in an export-competent conformation (31, 58, 85, 181). In the absence of SecB in the cytoplasm, maltose-binding protein folds into what is perhaps its native conformation, exhibiting comparable resistance to added proteases.

SecB has been purified and appears to function as a tetramer (178). It binds, in general, to unfolded proteins in vitro (84, 164). However, it binds with considerably higher affinity to maltose-binding protein than to ribose-binding protein, the former being SecB dependent and the latter SecB independent for their export in vivo. Randall and coworkers have defined regions of maltose-binding protein that interact with SecB, and they suggest that initially the binding is to charged regions of maltose-binding protein (133, 164). Subsequently, hydrophobic regions of maltose-binding protein are exposed, possibly contributing to a tighter binding of SecB. A second suggested function of SecB is a role in targeting proteins to the secretion machinery in the membrane (40, 64, 103). While there is some evidence for binding of SecB to SecA (see below), no direct evidence for such targeting has yet been presented. Clearly, in the absence of SecB, even unfolded maltose-binding protein is exported considerably less efficiently than in its presence (32, 81, 85).

Curiously, alkaline phosphatase, which is normally quite independent of SecB, can, under some conditions, become SecB dependent for its export. At low temperature (87) or in prlA-mediated export of alkaline phosphatase missing its signal sequence (42), the absence of SecB severely limits alkaline phosphatase translocation. These two instances may be cases where alkaline phosphatase is folding in the cytoplasm more extensively because of the temperature or because of a longer residence time in the cytoplasm. Whereas the folding of alkaline phosphatase may be normally very slow, thus not necessitating a factor such as SecB, these different conditions may require such an activity.

While SecB is the only cytoplasmic chaperone found to be required for the export of several proteins, it may well be that certain of the proteins which are SecB independent utilize another chaperone. For instance, mutants in the genes for the chaperone complex GroEL-GroES have significant effects on β-lactamase export to the periplasm (87).

Cytoplasmic Membrane Components.

  SecA. Like SecB, SecA has been purified to homogeneity (28, 35, 78) and thus has been one of the most intensively studied Sec proteins biochemically. SecA is a multifunctional protein that exists as a homodimer (46) in two subcellular compartments, the cytoplasm and the cytoplasmic membrane (29). The protein contains domains that are responsible for binding and hydrolysis of ATP, interaction with RNA, interaction with other proteins, and ability to associate with the membrane (reviewed in reference 115). The central role of SecA in protein secretion is supported by the following: (i) it is capable of interaction with preproteins; (ii) it may interact with a number of components of the secretion apparatus; and (iii) it is essential both in vitro and in vivo. In addition to this evidence, prlD signal sequence suppressor mutations are alleles of the secA gene (8, 52, 142). However, the effect that these mutations have on the function of the SecA protein is at present unclear.

It is well established that SecA directly interacts with protein precursors destined for export from the cytoplasm. In vitro interaction of purified SecA with precursor protein occurs whether SecA is membrane bound or soluble (3, 36, 59, 93). This interaction is stimulated by the presence of ATP (79, 93), suggesting that the ATP-bound form of SecA is in a conformation favorable for preprotein binding (150). After binding to the preprotein, SecA hydrolyzes or releases ATP by an activity termed the SecA/lipid ATPase (150). The efficiency of this ATPase reaction is maximal in the presence of liposomes containing anionic phospholipids including cardiolipin and phosphatidylglycerol (93).

The interaction of SecA alone with model membranes has been thoroughly studied by biophysical and biochemical techniques. Insertion of SecA into membranes via an amino-terminal domain (29) occurs concomitantly with a partial unfolding of the protein (167), is inhibited by ATP, and is stimulated by anionic phospholipids (23). The importance of SecA interaction with the membrane in the secretion process in vivo is consistent with the finding that E. coli strains depleted of anionic phospholipids are defective for protein translocation (41). Maximal SecA/lipid ATPase activity is also dependent on the presence of preprotein (93). Specifically, the preprotein must be export competent (92) and possess a signal sequence for maximal effect (36). Mature protein or signal sequence alone fails to significantly stimulate SecA/lipid ATPase, although signal sequence alone can competitively inhibit the reaction in the presence of preprotein (36). Thus, SecA may recognize or bind both signal sequence and mature parts of the preprotein. The domain within SecA responsible for binding to preproteins has been defined by chemical cross-linking and corresponds to amino acid residues 267 to 340 (79).

Association of SecA with the cytoplasmic chaperone SecB has been demonstrated in vitro (59). However, although fairly specific, the affinity of SecA for SecB appears to be low. The affinity of their interaction is increased upon addition of preprotein, but it is not clear whether this is due to the ability of SecB to retain the preprotein in a state more able to interact with SecA, or to a proposed direct SecB-SecA interaction (59), both of which could lead to a SecA-SecB-preprotein ternary complex.

Taken together, the above in vitro results provide an insight into the early role of SecA in secretion in vivo. Upon binding ATP, SecA assumes a conformation, either in the cytoplasm or at the cytoplasmic membrane, that favors its interaction with preproteins. After binding preprotein (and perhaps SecB or other chaperones), SecA hydrolyzes and/or releases ATP and undergoes a conformational change that enhances its (and the preprotein’s) association with the cytoplasmic membrane. In this way, SecA presumably aids in the insertion of the preprotein into the bilayer, thereby facilitating precursor interaction with other membrane-bound components of the secretion apparatus. The SecA-interactive components of the cytoplasmic membrane have been tentatively identified as SecY and SecE, but may include other proteins (see below).

SecA exhibits further ATPase activity upon binding, along with the preprotein, to cytoplasmic membrane components of the secretion apparatus. This ATPase activity, which may not be distinct from that of SecA/lipid ATPase, appears to be required for translocation of the preprotein across the cytoplasmic membrane and has thus been termed "translocation ATPase" (36, 92). Dissection of SecA ATPase activities by directed mutagenesis has revealed that SecA possesses two ATP-binding sites, one low- and one high-affinity site (102, 107). Both are required for functional complementation in vitro and in vivo (107). However, the assignment of either of these sites to allosteric versus catalytic roles in the secretion process has not been achieved.

Translocation ATPase activity and protein export are both inhibited by azide (118); mutations within secA can be isolated that confer resistance to azide (53, 118). It has been proposed that translocation ATPase is the result of successive cycles of SecA binding and release of the translocating preprotein (115, 119, 147). In this scheme, ATP may bind to SecA after SecA preprotein inserts into the membrane and complexes with membrane components. ATP binding results in release of the preprotein and partial deinsertion of SecA from the membrane. After conformational change, SecA, now in a form able to bind preprotein, binds to the next segment of the preprotein to be translocated. ATP hydrolysis occurs, and SecA again inserts into the membrane, along with this preprotein segment. This mechanism of SecA action is further supported by experiments suggesting that preproteins at various stages of translocation interact with SecA (and SecY; see below) (74).

SecA also binds to RNA. As described above, SecA expression is derepressed up to 10- to 20-fold by the effects of sec mutations (excluding secB mutations) and other blocks to protein secretion (117, 140). Down-regulation occurs by binding of SecA protein to its mRNA (43, 148), thereby reducing SecA expression at the level of translation. In wild-type E. coli, derepression of SecA expression may allow cells to cope with increased burdens on the secretion apparatus, for instance during periods of accelerated growth (115).

  SecY. SecY is a large protein integral to the cytoplasmic membrane and is essential for protein export in vitro (111) (for an alternative finding, see reference 180). Because SecY spans the membrane 10 times (4), it has been proposed that this protein may form, either by itself or in combination with other components of the secretion apparatus, a pore through which translocating proteins pass (4, 74, 177). Biochemical, physiological, and genetic studies suggest that SecY may interact with several components of the secretion apparatus. Various studies support the view that SecY is at least part of the receptor, or site to which SecA binds with high affinity to the cytoplasmic membrane. SecA has been shown to shield SecY from proteolysis in vitro (59). Defects caused in vitro by a temperature-sensitive mutation of SecY can be overcome by the addition of excess SecA protein (50). Antibody specific for cytosolic epitopes of SecY can block high-affinity binding of SecA (59, 92) and preproteins (177) to membrane vesicles. Translocating preproteins can be cross-linked to SecY (74).

SecY interaction with the two other cytoplasmic membrane components of the secretion apparatus, SecE and SecG (described below), is inferred from the following findings. (i) SecY is copurified with SecE and SecG from the cytoplasmic membrane (26, 27). (ii) Overexpression of SecY in cells requires that SecE also be overproduced (the converse is not true; 99). The domain of SecE required for SecY overexpression has been mapped and includes the same minimal portion of SecE able to complement a deletion of the secE gene in cells (see below) (112, 145). The physiological relevance of this overexpression phenomenon is unclear. (iii) Reconstitution of in vitro translocation with SecA and preprotein requires inclusion of membrane vesicles containing SecY and SecE (26, 27); addition of SecG to this reconstituted system results in even more efficient translocation (113). (iv) Studies by Silhavy and coworkers have provided a genetic basis for specific interactions between SecE and SecY. Certain prl alleles of secY and secE, when expressed in the same cell, result in synthetic lethality (14, 120). These prl mutations alter residues within SecY’s membrane-spanning segments 7 and 10 and SecE’s membrane-spanning segment 3, suggesting that these membrane-spanning segments interact in the proposed secretion complex. In other elaborate experiments involving the titration of Sec components with mutant forms of SecY and SecE and with mutant preproteins, the Silhavy group has also put forth a model invoking sequential interactions between signal sequences, SecY, SecE, and SecA (14).

SecY is likely to be physically adjacent to preproteins at late stages of translocation (74). Initial interactions of the signal sequences of preproteins with SecY are less well understood. At first glance, the occurrence of prlA alleles in secY appears to support the proposal that SecY and signal sequences directly interact (48). However, this suppression is not allele specific as one would expect for such a direct interaction. Rather, a single prlA suppressor can allow for the export of preproteins with a wide variety of signal sequence mutations (9, 47, 48, 120, 152). In fact, proteins completely lacking a signal sequence can be exported in strains harboring the prlA4 allele of secY (42). This lack of specificity of interaction between SecY and signal sequences, coupled with the fact that prlA mutations affect specific regions of SecY, has led Silhavy and coworkers to propose that SecY actually provides a "proofreading" function that can recognize good signal sequences and allow preproteins connected to them to be exported. The prlA mutations alter or remove this proofreading function (present in membrane-spanning segment 7 of SecY) so that proteins bearing altered signal sequences are allowed to be productively exported. This type of mechanism may also explain the export of proteins lacking signal sequences as the absence of proofreading by SecY of "cryptic" signal sequences present in the mature part of the protein (42). Alternatively, interaction between SecY and the signal sequence may not be required at all for translocation to take place.

  SecE and SecG. In addition to the SecY protein, two other cytoplasmic membrane proteins, SecE and SecG, are presumed to be a part of the core translocation apparatus. SecE, as well as being indispensable in in vitro translocation reactions (162), is essential in vivo for viability and for protein secretion (146). SecE spans the cytoplasmic membrane three times (146). Mutations that alter amino acid residues in membrane-spanning segment 3, the prlG1 and prlG2 alleles of secE, allow proteins bearing mutant signal sequences to be exported, suggesting that SecE may interact with the signal sequences of preproteins (153). Alternatively, since prlG1 exhibits a synthetic lethal phenotype when combined with the prlA4 allele of SecY, it has been proposed that this membrane-spanning segment indirectly affects signal sequence interaction with membrane-spanning segment 7 of SecY, thus altering the latter’s inherent proofreading function (120).

A detailed mutational analysis of SecE has revealed that a minimal form of it, which contains membrane-spanning segment 3, the carboxyl-terminal periplasmic domain, and a portion of its second cytoplasmic loop, can complement for SecE function in vivo (145). SecE homologs in other bacteria and eukaryotes found to date, in fact, contain only domains corresponding to these three portions of SecE (61, 109). Membrane-spanning segment 3 can be functionally replaced with a heterologous membrane-spanning segment, and the sole essential part of the protein resides in its second cytoplasmic loop region (109). Amino acids within this essential cytoplasmic domain of SecE are conserved across species of eubacteria, pointing to their important role in protein export (61, 109). Possible functions for SecE in secretion include contribution to formation of a "translocation pore" or recruitment of other factors or proteins required for secretion to the translocation site via interactions in the cytoplasm (109).

SecG, originally termed band 1 or p12, was identified as a protein that copurifies with the SecY/SecE complex obtained from cytoplasmic membrane preparations (26, 27, 113). The sequence of the secG gene suggests that SecG is a cytoplasmic membrane protein that may span the membrane two or three times (113). In an in vitro membrane vesicle system reconstituted with SecE and SecY, SecG stimulates in vitro translocation about 20-fold (113). This major effect is surprising in light of its nonessentiality in vivo at 37°C (110).

  SecD and SecF. SecD and SecF are cytoplasmic membrane proteins that, unlike SecY, SecE, and SecG, have large soluble domains located in the periplasmic space (55, 125). Various studies on SecD and SecF in vivo have provided clues as to their function in secretion. As discussed above, no prl mutations have been isolated in the secDF locus, despite their providing a large target for mutagenesis (13). Co-overexpression of SecD and SecF, however, does result in the suppression of signal sequence defects and even allows for more efficient secretion of proteins in otherwise wild-type cells (123). Deletion of the genes encoding SecD and SecF from the chromosome, or depletion of the proteins in cells, results in a severe secretion defect, although such strains are viable under certain conditions (123) and the secretion defect is not as severe as that seen when SecE is depleted from cells (C. K. Murphy and B. Traxler, submitted for publication). Similar defects in secretion are manifested whether SecD or SecF or both are inactivated in cells, indicating that they may function together in a complex (123). Whether such a proposed complex is part of the SecY/SecE complex remains unclear. We estimate cellular levels of SecD and SecF to be about 10-fold lower than the predicted levels of SecE and SecY, and copurification of SecD and SecF with SecY and SecE has not been detected (125). (Mizushima and coworkers have estimated different levels of production of SecD and SecF [101].) These results, taken together, imply that SecD and SecF function catalytically, but do not rule out the possibility that they form a transient or nonstoichiometric complex with SecE and SecY (123).

Two different mechanisms for the function of SecD and SecF have been proposed based on in vivo studies. Mizushima and colleagues have demonstrated that the release of mature secreted proteins is inhibited when spheroplasts competent for translocation are preincubated with antibody that binds to periplasmic determinants of SecD (100). The mature (signal sequence cleaved) species of maltose-binding protein in these studies was found to be sensitive to externally added protease, whereas in untreated spheroplasts, the maltose-binding protein attained a soluble, protease-resistant form. These results suggest that SecD may aid in protein folding and/or release from the cytoplasmic membrane or secretion apparatus. (It should be noted that such protease-sensitive translocation intermediates have been identified in other sec mutants and under other conditions that impair secretion [124, 166]). This is consistent with genetic evidence showing that SecD and SecF function at a step after SecY (15). An alternative proposal is that SecD and SecF affect the energized state of the membrane. Arkowitz and Wickner have detected a defect, albeit modest, in electrochemical gradient (Δμ) maintenance in cells depleted of SecD and SecF (7). It is unclear at present as to whether this result is due to a secondary effect of the depletion, since the cells were depleted to the extent that cell growth was severely inhibited. Thus, a clear connection between SecD and SecF and Δμ will require further analysis.

Conclusive demonstration of the importance of SecD and SecF in in vitro translocation assays has not been achieved (101), although effects can be seen under certain conditions depending on the preprotein examined (7). The discrepancy between in vivo and in vitro results may be attributable to several characteristics of both the SecD and SecF proteins and the in vitro assays. The in vitro translocation system is inefficient catalytically as compared to secretion in vivo (2). Any stimulatory effect caused by the presence of SecD and SecF may not be observable in in vitro systems. In addition, SecD and SecF may provide some function in secretion that cannot be assayed for in the present in vitro translocation systems. These systems are able only to detect translocation of preprotein into vesicles, as assayed by signal peptide cleavage and acquisition of resistance to externally added protease. Such assays would presumably not detect a step or steps that occur after translocation, such as folding or release from the membrane translocation site (123). Translocation into membrane vesicles in the absence of SecD and SecF is defective for certain proteins (7). These vesicles were also defective for maintenance of Δμ. However, because the vesicles were derived from severely depleted cells, it is unclear whether the effect seen was due to a primary defect (see above).

 

Leader Peptidases

Of the two leader peptidases of E. coli, LspA, which processes prolipoproteins (137, 187), and Lep, which acts on most cell envelope proteins with cleavable signal sequences, the latter has been the more thoroughly studied (37). Lep, which has been the subject of intensive topological (143, 185), structure-function (16), and substrate specificity (10, 80, 149) analyses, is anchored to the cytoplasmic membrane by two membrane-spanning segments (108, 185) and contains its active site in the periplasm (16, 155, 165). Lep activity is not affected by either chemicals (193) or mutations that inhibit the activity of other known proteases (reviewed in reference 37). Mutagenic analysis of Lep suggests that Lep and Lep homologs make up a novel class of serine proteases. Although signal sequences overall contain no consensus amino acid sequence, they do have certain features that are critical to their being recognized as substrate for signal peptidase. Signal sequences that are cleaved efficiently contain small amino acids at positions –1 and –3 and a helix-breaking residue at position –6 (149). After their cleavage from the mature protein, signal sequences are broken down to amino acids by various proteases. One cytoplasmic membrane peptidase, protease IV encoded by the sspA gene, and the cytoplasmic enzyme oligopeptidase A have been suggested to be implicated in this process (67, 114, 156). However, it is at present unclear whether specific proteases are dedicated to signal peptide degradation, or whether, alternatively, more than one general protease carry out this function in the cell.

 

Is There a Bacterial Equivalent to Mammalian SRP?

Eukaryotic cells contain a complex of proteins and 7S RNA known as the signal recognition particle (SRP; 57, 175). This SRP complex recognizes the signal sequence of nascent preproteins through its SRP54 protein component (62, 97, 192) and maintains preproteins in an export-competent conformation prior to their translocation (136). In E. coli, homologs of SRP54 and 7S RNA, Ffh (P48) and 4.5S RNA, respectively, have been identified, raising the possibility that these components perform functions similar to their mammalian counterparts (60, 126, 127, 135, 138). In vitro studies have demonstrated that E. coli Ffh or 4.5S RNA can substitute for SRP54 or 7S RNA, respectively, in assays designed to test for enzymatic properties of mammalian SRP (12, 96, 138). In physiological studies, depletion of 4.5S RNA or induction of a dominant lethal form of 4.5S RNA in E. coli results in accumulation of pre β-lactamase, but not other precursor proteins, and results in an induction of the heat shock response (126, 138). Depletion of Ffh in E. coli also results in accumulation of pre-β-lactamase, with only a modest effect on precursor accumulation of other preproteins (122).

Although these results hint at a role for Ffh and 4.5S RNA in secretion in E. coli, more recent evidence raises questions about this role. For instance, Pedersen and colleagues have shown that the early primary effect of depletion of 4.5S RNA is a global reduction of peptide elongation rate in the cell (72). Furthermore, signal sequence processing was not different either in 4.5S RNA-depleted or -replete cells. Defects in processing were only manifested after cells had been depleted for extended periods of time. These findings raise the possibility that secretion defects seen in previous 4.5S RNA depletion experiments could be the secondary effects of interference with protein translation. In addition, other studies have demonstrated that cells do not appear to require an Ffh-4.5S RNA complex for viability (186). The role of Ffh protein in the cell remains to be elucidated. One concern in trying to verify a role in secretion for a particular gene product is that it is often only β-lactamase that shows a severe export defect in mutant backgrounds or in depletion experiments. It may be that β-lactamase is particularly sensitive to disruptions of cellular physiology (such as induction of the heat shock response) and, therefore, is not a good probe to use for assaying export defects.

 

ENERGY REQUIREMENTS FOR SECRETION

Clearly, considerable energy must be expended to translocate an amphiphilic preprotein across the hydrophobic cytoplasmic membrane bilayer. Several potential sources of energy may contribute to protein export. As discussed above, ATP hydrolysis by SecA is one of the energy sources required for protein translocation to occur (30). In vitro experiments suggest that ATP hydrolysis is at least needed for steps subsequent to preprotein-SecA insertion into the cytoplasmic membrane (147, 159). In the presence of nonhydrolyzable ATP analogs, this insertion step takes place efficiently, resulting in translocation of about 20 amino acid residues of the preprotein and signal sequence cleavage, but further translocation does not take place (160). Several in vitro studies also indicate that SecA hydrolyzes ATP during subsequent steps of translocation (see above). The importance of ATP in the secretion process can also be demonstrated in vivo. Protein export in cells is blocked in the presence of azide, probably because the initial steps, which result in SecA ATPase activity, are inhibited; azide-resistant mutants of E. coli, in fact, map almost exclusively to the secA gene (see above).

The membrane potential (

) has also been shown to be crucial for protein secretion both in vivo and in vitro. In cells exposed to CCCP (carbonyl cyanide m-chlorophenylhydrazone), is dissipated and protein secretion is blocked. The role of in translocation has been examined in vitro. Although SecA- and ATP-driven translocation of preprotein into vesicles can take place in the absence of (159), translocation efficiency is greatly stimulated by the establishment of (56). If translocation can take place in the absence of , then what role does this form of energy provide to the process? One proposal is that the membrane potential (ΔΨ) component of directly contributes to mechanistic movement by acting on charged parts of the preprotein in an electrophoretic fashion. However, Kato et al. have demonstrated that a model preprotein, devoid of charged residues, is efficiently translocated into membrane vesicles in a -dependent manner (77). A second proposal holds that influences the direction of protein translocation (147). In support of this, reversal of in vesicles has been shown to result in reverse translocation of preproteins engaged in the secretion apparatus (45). Thus, ATP hydrolysis by SecA releases the translocating preprotein, and at this stage, affects either the preprotein or translocation apparatus so that passage of a portion of the preprotein through the cytoplasmic membrane towards the periplasm occurs. Whether affects the preprotein or a component(s) of the secretion apparatus remains one of the central questions in the protein secretion field.

 

SIGNAL SEQUENCES

Signal sequences are, in general, essential for the export of the proteins to which they are attached. For instance, oligonucleotide mutagenesis has been used to remove the signal sequences of β-lactamase, alkaline phosphatase, and maltose-binding protein, resulting in severe effects on export (20, 42). Nevertheless, even with its signal sequence deleted, 1% of alkaline phosphatase is still found in the periplasmic space. Furthermore, substantial amounts of a Rhodobacter sphaeroides cytochrome can be exported in constructs where the protein is missing its signal sequence (22). Also, under certain conditions, relatively efficient export of E. coli proteins can take place in the absence of a signal sequence. Specifically, prlA mutations in the secY gene can suppress the effects of complete deletion of the signal sequence of alkaline phosphatase and restore export of up to 40% of the protein (42). This export is relatively efficient in the total amount of protein translocated, but inefficient in its kinetics. The alkaline phosphatase passes through the membrane posttranslationally and at a rate which is orders of magnitude slower than that of the wild-type protein.

Signal sequences, while not sharing sequence homology, do have several features in common (169, 170, 173). In E. coli, they range in length from 18 to approximately 30 amino acids. They have one or more basic amino acids near their amino terminus, a central hydrophobic core of seven or more amino acids, and a hydrophilic carboxy terminus containing the sequence motif recognized by leader peptidase. The roles of these subdomains have been established by mutational analysis. Mutations in the hydrophobic core that lower its hydrophobicity can result in defects in export. For instance, replacement of a hydrophobic amino acid with a charged amino acid such as arginine can reduce export by as much as 100-fold (106). In contrast, complete removal of the basic amino acids at the amino terminus of the sequence has a much weaker effect on export (69, 130, 168). Even these effects can be mitigated by increasing the length of the hydrophobic core (63, 129). Mutations in the amino acids that constitute the leader peptidase recognition site at the carboxy terminus of the signal sequence can block cleavage, but do not interfere with the export process (1, 10, 11, 51, 158). These findings point to the hydrophobic region of signal sequences as central to their function in recognition by the export apparatus and translocation across the membrane.

Numerous lines of evidence indicate that there is no common recognition motif in the hydrophobic core of signal sequences. Searches for sequence homology have revealed none. Substitution of random genomic sequences in the yeast Saccharomyces cerevisiae suggests that many hydrophobic sequences can replace a natural signal sequence (76). Replacement of the signal sequence of alkaline phosphatase with polyleucine sequences can result in even more efficient export of the protein than in the wild type (44, 141). Both genetic and in vitro experiments indicate that, in addition to its hydrophobic nature, α-helical structure is a required feature of the hydrophobic core (24, 25, 49).

The function of the signal sequence remains obscure, in part because so many different roles have been proposed for it. Furthermore, the requirement of a signal sequence for export can be dramatically relieved in the prl mutants described above. One proposal is that the signal sequence retards folding of exportable proteins, thus maintaining them in a secretion-competent conformation for a longer time than would be the case otherwise. Studies in vitro show that this retardation of folding does occur (88, 95, 121). According to another proposal, the signal sequence is recognized by what is thought to be the first Sec component to be encountered by the nascent polypeptide chain, SecB (179). However, both in vivo and in vitro evidence shows that a signal sequence is not necessary for interaction of SecB with a protein (42, 134). The hydrophobic nature of the signal sequence has led to the suggestion that it functions by inserting into the lipid bilayer (17, 18, 70). Such insertion does in fact occur in model in vitro systems (24, 33, 65, 66, 75, 105, 176). Other proposals suggest that the signal sequence is recognized by the SecA or the SecY protein (3, 93, 120). Finally, in vitro studies show that signal sequences stimulate the formation of an ion channel in the membrane (151).

 

ROLE OF THE MATURE SEQUENCE OF EXPORTED PROTEINS IN TRANSLOCATION

We have already referred to the findings that export competence and protein folding work at cross-purposes. A direct correlation exists between the folding of maltose-binding protein and its ability to be exported. Furthermore, mutants defective in the folding of maltose- and ribose-binding proteins have enhanced ability to export these proteins (32, 34, 94, 161). Signal sequence mutants of these proteins can be suppressed by the folding mutants. These results are explained by the fact that signal sequence mutants slow down the export process and, as a result, enhance the inhibitory effect of the folding of the protein. Slowing down the folding pathway by mutation gives more opportunity for the protein to be engaged by the export apparatus and translocated across the membrane.

Thus, it is probably inherent in cell envelope proteins that their normal folding pathway is slower than that of cytoplasmic proteins. This property could be due to (i) the above-mentioned effects of signal sequences on folding, (ii) features inherent in the normal folding pathway, or (iii) "chaperones" such as SecB that inhibit folding.

If signal sequences are sufficient to retard folding of a protein long enough to allow export, does that mean that attaching a signal sequence to any protein (e.g., a normally cytoplasmic protein) will allow its export? Unfortunately, there has not been enough systematic study of this question to give a clear answer. Reports of export of cytoplasmic proteins have, in general, not been quantitative enough to draw strong conclusions (98, 157). Our own experience with β-galactosidase and a subunit of tryptophan synthetase suggests that cytoplasmic proteins are not necessarily readily exported (90) (G. Jander and J. Beckwith, unpublished results). The efficiency with which a cytoplasmic protein is exported may also be related to the length of the polypeptide chain (19, 90). It seems likely that the folding of the proteins in their normal environment, the cytoplasm, competes with the export process. If this is so, it suggests that a signal sequence may require special features of its attached protein in order to retard its folding. These features may simply be the speed of folding intrinsic to the protein itself. Attempts to detect by sequence inspection any features (e.g., hydrophobicity) that distinguish exported from cytoplasmic proteins have failed.

The nature of the amino terminus of the mature sequence of exported proteins is also critical for the export process. The presence of too many basic amino acids in this region results in a block in protein translocation (21, 91, 131, 154, 190). The important property appears to be the distribution of charges around the signal sequence. The more positive charges at the amino terminus of the signal sequence, the less the effect of the basic amino acids at the beginning of the mature sequence (131). Studies by von Heijne show that the effects of a string of charged amino acids decline as the cluster is placed at increasing distances from the signal sequence (in this case of a membrane protein) (5). It has been proposed that the appropriate charge distribution around the signal sequence can properly orient it in the membrane so that export can take place (171). The nature of the electrochemical gradient across the lipid bilayer (negative inside, positive outside), as well as the loop model, which pictures the amino terminus of the signal sequence apposed to the cytoplasmic face of the membrane and the carboxy terminus appearing on the other side of the membrane, are consistent with this proposal. In fact, dissipation of the electrochemical gradient can mitigate the effects of an incorrect charge balance (6). Similar effects of charge distribution have been noted around membrane-spanning segments of membrane proteins (21, 172, 174). Studies with altered phospholipids raise the possibility that the effect of charge is dependent on the nature of the lipid itself (acidic, neutral, or basic) (86). However, Andersson and von Heijne have suggested that interaction with the membrane electrochemical potential (negative inside, positive outside) is responsible for the effects of charged amino acids (6).

 

CYTOPLASMIC MEMBRANE PROTEINS AND THE SECRETORY PATHWAY

Growing evidence suggests that integral proteins of the cytoplasmic membrane may share the same pathway for their localization as exported proteins. The best studied of these systems, the Lep leader peptidase, is inhibited in its membrane assembly by mutations in secA and secY (184). The Lep protein contains two membrane-spanning segments. Initial experiments suggested that the more complex membrane protein, MalF, which spans the membrane eight times, does not require the Sec machinery for its membrane localization (104). However, recent experiments with a strain that permits strict depletion of the SecE protein suggest that MalF also depends on the Sec proteins (B. Traxler and C. K. Murphy, in preparation). Since the membrane-spanning segments that act as export signals in cytoplasmic membrane proteins consist of considerably longer hydrophobic stretches than signal sequences, they may have greater affinity for the export machinery. If so, effects on membrane protein insertion could be difficult to detect except with sec mutations that have particularly strong effects. Early experiments indicate that the M13 coat protein can assemble into the membrane in the absence of the Sec proteins (183). Testing membrane incorporation under more stringent conditions of depletion of the secretion machinery might yet show an effect on even this protein.

 

IN VITRO VERSUS IN VIVO STUDIES

For the most part, the in vivo and in vitro studies have cohered very nicely. Most of the genes and gene products identified in vivo are important for the functioning of the in vitro system. However, certain disparities between the two usually complementary approaches do exist. For instance, SecD and SecF appear to play a role in protein export in vivo (123), but no clear-cut effects have been seen in vitro (7, 101). In contrast, SecG provides a big boost to the in vitro system at 37°C (113), but in vivo, its absence has a barely detectable effect on export (110). These results indicate that the in vitro system will require further development in order to more closely mimic the in vivo situation. Conversely, the bacterial cell may have ways of compensating for defects in some of its components by expressing factors that can substitute for them.

 

SUMMARY

While many aspects of the export of proteins to the bacterial cell envelope have been well characterized, others remain obscure. The dispensability of certain of the system’s components simplifies analysis (especially in vitro analysis) of those that are essential. On the other hand, the fact that several of the components are nonessential complicates a complete definition of the process. The absence of a signal sequence can be partially compensated for by mutations within the mature portion of the polypeptide chain or by mutations (prl mutations) in various components of the secretion machinery. The absence of SecB does not fully prevent export of a protein such as maltose-binding protein and can also be compensated for by mutations in the mature sequence or by mutations in the heat shock regulatory system.

Nevertheless, the broad outlines of the export process are emerging. We propose the following working model (see Fig. 1). Exported proteins are made with amino-terminal (or internal, for cytoplasmic membrane proteins) signal sequences. An important function for the signal sequence early in the export pathway is to interact with the mature sequence of a protein and retard its folding. The SecB protein (or perhaps other, as yet undefined chaperones) can also interact with mature sequences of certain proteins with the same result. With other proteins (e.g., alkaline phosphatase), the export may be so rapid and the folding slow enough in the cytoplasm that chaperones such as SecB are not necessary. The retarded folding of proteins may allow the signal sequence and the mature sequence to interact with the membrane and the secretion apparatus contained therein. The signal sequence may facilitate this interaction by inserting via its hydrophobic core into the lipid bilayer.

Contact between the signal sequence and the SecA/Y/E (and G?) complex may trigger a conformational change in the apparatus which allows the mature sequence to penetrate a pore in this protein complex. Initial interaction of the preprotein with the nucleotide-bound form of SecA may actually result in "activation" of the secretion complex. Activation may be an essential part of the translocation process if an active secretion complex minus preprotein were lethal to the cell (i.e., if an open "pore" were formed by the secretion complex). These interactions stimulate the ATPase activity of SecA, which provides energy for the translocation process. Proton motive force may provide directionality to signal sequence insertion or protein translocation. Cleavage of the signal sequence by Lep (or LspA for lipoproteins) takes place during the translocation. An alternative model for protein translocation suggests that the mature sequence passes into an interface between the Sec machinery and the lipid bilayer.

The roles of SecD and SecF in protein translocation are at present unclear. They may form a nonstoichiometric complex with SecE and SecY, perhaps organizing several secretion complexes at particular points in the cytoplasmic membrane, or making such secretion complexes more efficient. Alternatively, SecD and SecF may assist in folding or release of newly translocated proteins.

How can prl mutants be explained in this general scheme? In prlA, prlG, and perhaps prlD mutants, the secretion apparatus may be conformationally altered so that it no longer requires interaction with the signal sequence. The secretion apparatus in such mutants is thus less stringent, perhaps allowing any unfolded protein to be accepted by the Sec protein complex.

Despite recent major advances, several questions remain unanswered in the field of protein secretion. Do other chaperones or components involved in secretion exist that have thus far been undetectable by biochemical or genetic means? What are the exact roles of the cytoplasmic membrane components? What is the structure of the translocation complex? How does ATP hydrolysis energize protein translocation? The answers to these questions will undoubtedly increase our understanding of the basic process of translocation of proteins through membranes.

ACKNOWLEDGMENTS

This work was supported by a grant from the National Institutes of Health. J.B. is an American Cancer Society Research Professor.