Introduction
Protein secretion is required for numerous aspects of the bacterial life cycle, including organelle biogenesis, nutrient acquisition, and virulence-factor expression.
Gram-negative bacteria face a special challenge in this regard, as secreted proteins must cross the periplasm and the outer membrane (OM) in addition to the cytoplasmic or inner membrane (IM). Secretion is defined in this review as any process in which a protein crosses the OM barrier. Thus, in addition to export of proteins to the extracellular medium, secretion includes export of proteins that remain anchored to the cell surface or are assembled into cell-surface organelles.
Secretion across the OM represents a different problem from export of unfolded proteins across the IM. Proteins acquire structure in the periplasm, including formation of disulfide bonds, and may completely fold before crossing the OM. In addition, ATP or other sources of energy are not known to be present at the OM.
Export systems must therefore be self-energized or have mechanisms for harnessing the energy available at the IM. Gram-negative bacteria have evolved a number of pathways for protein secretion that can be organized into six overarching groups.
Four of these pathways represent terminal branches of the general secretory pathway (GSP), exporting proteins with cleavable aminoterminal signal sequences that require the Sec system for translocation across the IM (see [1] for a recent Sec review).
The other pathways are Sec-independent and capable of exporting substrates directly from the cytoplasm outside the bacteria. At least three pathways share secretion mechanisms with, and probably evolved from, organelle biogenesis systems, and one pathway is dedicated to the assembly of surface structures.
This review will present a brief overview of each secretion pathway, focusing on secretion mechanisms and will highlight important advances over the past year. Readers are referred to recent reviews for in-depth discussions of each pathway [2–8].
Autotransporters and proteins requiring single accessory factors
The autotransporter secretion pathway is a terminal branch of the GSP that exports proteins with diverse functionalities, including proteases, toxins, adhesins and invasins [2]. The prototypical member of the autotransporter family is the Neisseria gonorrhoeae IgA1 protease [9].
This pathway has also been termed type IV secretion. However, the type IV nomenclature has also been claimed for the much more complex DNA- and protein-exporting system described below. To avoid confusion, this review will use the more descriptive autotransporter designation. Autotransporters do not require accessory factors for transit from the periplasm to the bacterial surface (Figure 1).
Figure 1 | Models for the autotransporter and single accessory pathways. (a) Both pathways are branches of the GSP, and proteins cross the IM via the Sec system, followed by cleavage of their amino-terminal signal sequence in the periplasm by signal peptidase. (b) For autotransporters, the carboxy-terminal or β domain of the protein inserts into the OM to form a β-barrel secretion channel through which the passenger domain passes to the cell surface. The passenger domain is released into the extracellular medium by proteolysis. (c) In pathways requiring a single accessory factor, a separate protein forms the β -barrel OM channel, which may be gated.
A typical autotransporter contains three domains: an aminoterminal signal sequence for secretion across the IM by the Sec system, an internal passenger or functional domain, and a carboxy-terminal β -domain [2].
The b-domain inserts into the OM to form what is predicted to be a β -barrel pore structure, similar to the bacterial porins, through which the passenger domain passes to the cell surface. A linker region connecting the passenger and b-domains is also essential for export and may guide the passenger region through the β -domain channel [10].
Secretion across the OM does not appear to require input of external energy. Once secreted, the passenger domain is either retained on the bacterial surface or released into the environment by proteolysis. The nature of the passenger domain does not appear to be important, as foreign proteins can be substituted for this domain and successfully secreted [11,12]. Work on secretion of foreign proteins indicates that formation of disulfide bonds by the passenger domain hinders export to the cell surface, pointing to a requirement for unfolded or partially folded substrates [11,12].
However, this is contradicted by recent research showing export of a functional, disulfide-bonded antibody fragment fused to the IgA b-domain [13•]. Much attention has been focused on the structure of the b-domain. Computer modeling and gel-mobility studies support the formation of a β -barrel structure [2,10]. The b-domain of the Bordetella pertussis autotransporter BrkA was recently shown to form pores in experimental lipid bilayers [14]. It is unknown whether a single b-domain forms the channel or if oligomerization is required. A separate pathway has been identified that requires a single accessory factor for secretion across the OM.
Examples include export of B. pertussis filamentous hemagglutinin and Serratia marcescens hemolysin by FhaC and ShlB, respectively [15,16]. The accessory protein is thought to form a b-barrel pore in the OM for secretion of the substrate to the cell surface (Figure 1). Foreign proteins can be secreted by this pathway, but acquisition of disulfide bonds appears to block export to the cell surface [17]. Recent work demonstrated that both FhaC and ShlB form channels in experimental lipid bilayers [15,16]. The channels may be gated and evidence suggests that ShlB acts as a monomer. This single accessory pathway appears to be functionally similar to the autotransporters, except that separate proteins encode the passenger and b-domains. However, there is no sequence homology between the two systems and the accessory factors appear instead to be related to a group of OM proteins that share homology with a protein-translocating porin found in the outer envelope of chloroplasts [18]. Thus, the autotransporter and single accessory pathways probably evolved independently and may, in fact, employ completely different mechanisms.
Chaperone/usher pathway
The chaperone/usher pathway is a branch of the GSP dedicated to the assembly and secretion of a broad range of adhesive virulence structures on the Gram-negative bacterial surface [3]. Secretion across the OM by this pathway requires only two components: a periplasmic chaperone and an OM protein termed an usher (Figure 2).
Model for biogenesis of P pili by the chaperone/usher pathway. (a) Pilus subunits cross the IM via the Sec system, followed by cleavage of their amino-terminal signal sequence. The periplasmic chaperone PapD binds to each subunit via a conserved carboxy-terminal subunit motif (white box), allowing proper subunit folding and preventing premature subunit–subunit interactions. (b) The crystal structure of the PapD–PapK chaperone–subunit complex. The chaperone (green) consists of two Ig folds. The subunit (blue) consists of a single Ig fold that lacks the usual seventh b strand, resulting in exposure of its hydrophobic core. The G1 b strand of PapD binds to the conserved carboxy-terminal motif of PapK, donating its hydrophobic residues to complete the structure of the subunit in a mechanism termed donor strand complementation. (c) Chaperone–subunit complexes are targeted to the OM usher for assembly into pili and secretion across the OM. Subunit–subunit interactions are thought to take place by interaction of conserved amino-terminal (black box) and carboxyterminal (white box) motifs. The amino-terminal motif of one subunit may complete the structure of the preceding subunit in a mechanism termed donor-strand exchange to build the pilus fiber. The usher channel is only able to allow passage of a linear fiber of folded subunits, forcing the pilus rod to adopt its final helical conformation at the cell surface. The location of Pap subunits in the pilus is indicated.
The prototypical organelles assembled by this pathway are P and type 1 pili, expressed by uropathogenic Escherichia coli [19,20].
These pili are composite structures consisting of a thin, flexible tip fibrillum connected to a rigid, helical rod (Figure 2) [21]. Following export across the IM via the Sec pathway, pilus subunits interact with the periplasmic chaperone via a conserved carboxy-terminal motif present on each of the pilus subunits. The chaperone facilitates release of pilus subunits into the periplasm, guides their proper folding, and caps subunit-interactive surfaces to prevent premature interactions in the periplasm [22]. Next, periplasmic chaperone–subunit complexes are targeted to the OM usher [23,24]. Interaction with the usher triggers chaperone dissociation from the subunit by an unknown mechanism, exposing subunit-assembly surfaces and allowing incorporation of the subunit into the pilus fiber. The usher provides a translocation channel through the OM for secretion of the pilus. The P pilus usher, PapC, has been imaged by electron microscopy [3] and revealed as ring-shaped oligomeric structures containing central pores 2–3 nm in diameter. Pore formation was confirmed by reconstitution of purified usher protein into liposomes [3]. Interestingly, OM components of the unrelated type II and type III secretion systems also form ring-shaped oligomeric complexes with central pores (see below). A 2–3 nm usher channel would be sufficient to allow passage of folded subunits assembled into a linear fiber, such as the pilus tip fibrillum, but not the helical pilus rod. Therefore, the pilus rod would be constrained to a linear fiber of folded subunits while traversing through the usher. It would only be able to coil into its final helical conformation once it reached the cell surface (Figure 2). The chaperone/usher pathway does not appear to require input of external energy for secretion of pili across the OM [25] and coiling of the rod outside the cell may facilitate the outward translocation of pili. Recently, crystal structures of FimC–FimH and PapD–PapK (corresponding to chaperone-subunit complexes from type 1 and P pili, respectively) have been solved (Figure 2) [26••,27••]. The chaperone consists of two immunoglobulin-like (Ig) domains oriented in an L-shape. The FimH adhesin also consists of two domains: a receptor-binding domain and a pilin domain. The pilin domain of FimH and the single domain of the PapK subunit have Ig folds that lack the seventh (carboxy-terminal) b-strand present in canonical Ig folds. This results in a deep groove along the surface of the subunit and exposure of its hydrophobic core. In the chaperone–subunit complexes, the chaperone donates its G1 b-strand and a portion of its F1-G1 loop to complete the Ig fold of the subunit by occupying the groove (Figure 2). This donor-strand complementation interaction stabilizes the subunit by shielding its hydrophobic core. Indeed, alternating hydrophobic residues of the G1 strand become an integral part of the hydrophobic core of the subunit. Part of the groove is formed by the conserved carboxy-terminal subunit motif, which has also been implicated in subunit–subunit interactions [28]. Thus, the donor-strand complementation interaction caps an interactive surface and prevents premature pilus formation in the periplasm. Indeed, the folding, stabilization, and capping of the subunit may be part of a single dynamic process that ensures subunit interactive surfaces are not accessible until pilus biogenesis at the usher. The crystal structures suggest a model for pilus assembly. Subunits also have an amino-terminal extension with a conserved alternating hydrophobic motif that participates in subunit–subunit interactions [28]. This motif is similar to the alternating hydrophobic motif present in the chaperone G1 b-strand. The amino-terminal extension projects away from the subunit, where it would be free to interact with another subunit [27••]. During pilus biogenesis, the amino-terminal extension of one subunit may displace the chaperone G1 b-strand from its neighboring subunit in a mechanism termed donor-strand exchange. The mature pilus would thus consist of an array of Ig domains, each of which contributes a strand to the fold of the preceding subunit to produce a highly stable organelle.
Type I secretion
Type I or ATP-binding cassette (ABC) protein exporters are employed by a wide range of Gram-negative bacteria for secretion of toxins, proteases and lipases [4]. Secretion of a-hemolysin by E. coli represents the prototypical type I exporter [29]. Related ABC systems are conserved from prokaryotes to eukaryotes and export a variety of toxic compounds and antibiotics.
The type I pathway is Sec independent and secretes proteins directly from the cytoplasm across the OM without a periplasmic intermediate. Substrates of this pathway lack a cleavable amino-terminal signal sequence, instead, they possess a carboxy-terminal ~60 amino acid secretion signal [4]. The type I export apparatus consists of three proteins: an IM ABC exporter, an IM-anchored protein that spans the periplasm, termed a membrane fusion protein (MFP), and an OM protein (OMP) (Figure 3).
Model for secretion by the type I pathway. Type I substrates do not posses an amino-terminal signal sequence recognized by the Sec system, instead they possess a carboxy-terminal signal sequence that is not cleaved. During secretion, the periplasmic MFP interacts with both the IM ABC exporter and the OM channel-forming protein (OMP) to allow secretion to the extracellular medium without a periplasmic intermediate. Both the OMP and MFP are thought to assemble as trimers. ATPase activity by the ABC protein may energize substrate release into the medium.
TolC, the OMP for hemolysin export, assembles as a trimeric complex in the OM and is predicted to consist of a porin-like b-barrel membrane domain with a carboxy-terminal hydrophilic region that extends into the periplasm [30]. However, recent sequence analysis indicates that TolC and other OMPs are unrelated to porins [31•]. The OMP presumably functions as the OM secretion channel. Indeed, pore-forming activity by TolC oligomers has been demonstrated in experimental lipid bilayers [32]. The periplasmic MFP component also assembles as a trimer and interacts with both the OMP and ABC exporter [29,33]. MFPs typically contain a hydrophobic amino terminus believed to span the IM or are anchored in the IM by lipid modification of the amino terminus [31•]. The bulk of the MFP is thought to extend across the periplasm to contact the OMP and/or the OM itself. The MFP is thought to facilitate substrate secretion without a periplasmic intermediate by forming a closed bridge or channel across the periplasm or by fusing the inner and outer membranes, allowing direct contact of the ABC exporter and OMP channel. Experimental evidence has led to two related models for secretion by the type I pathway. Work on E. coli hemolysin secretion [29] indicates that the ABC exporter and MFP associate before substrate binding. Substrate binding to this complex then triggers contact of the MFP with the OMP. This bridging is transient, collapsing after export of the substrate. ATP hydrolysis by the ABC exporter drives substrate release outside the cell and is not required for substrate binding or assembly of the complex.
Work on S. Marcescens hemoprotein and Erwinia chrysanthemi metalloprotease secretion points towards a slightly different order of events [33]. In this model, the ABC exporter and MFP do not associate before substrate binding. Instead, an ordered set of associations takes place in which the substrate first binds to the ABC exporter, which then triggers binding of the MFP. This complex in turn interacts with the OMP, allowing secretion of the substrate. Future work is needed to resolve these two models or to determine if they reflect a variation among individual type I mechanisms.