Type V Secretion: the Autotransporter and Two-Partner Secretion Pathways

Introduction

The autotransporter and two-partner secretion (TPS) pathways are simple secretion pathways that are commonly grouped together under the rubric “type V” secretion (30). Both pathways are widely distributed in Gram-negative bacteria. Classical autotransporters are single polypeptides that consist of an N-terminal extracellular domain (“passenger domain”) that varies in size from ~20 kD to ~400 kD and a ~30 kD C-terminal β barrel domain (“β domain”) that resides in the outer membrane (OM) (50; Fig. 1A). Some passenger domains undergo a proteolytic maturation process that releases them from the cell surface. Passenger domains mediate a wide variety of effector functions, many of which are associated with virulence (see chapter by Henderson and Nataro). After autotransporters are translocated across the inner membrane (IM) via the Sec pathway, the ββdomain integrates into the OM and facilitates the secretion of the passenger domain by an unknown mechanism. For many years it was thought that the β domain forms a channel through which the covalently linked passenger domain is extruded (whence the name “autotransporter”), but recent work has challenged this view and has strongly suggested that another factor promotes the translocation reaction (7). In addition to classical autotransporters, a distinct subfamily of “trimeric autotransporters” has been identified (23). These proteins have the same domain structure as classical autotransporters, but their β domains contain only ~80 amino acids (108, 125). Theββ domains trimerize to form a single β barrel and the passenger domains (which can approach 400 kD in length) form an intertwined three-stranded structure (Fig. 1B). All trimeric autotransporters studied to date remain cell-associated and function as adhesins. The TPS pathway is similar to the autotransporter pathway in that a single virulence factor is transported into the extracellular space through the action of a dedicated OM component (2, 115, 133, 145). In the TPS pathway, however, the secreted polypeptide (“TpsA” protein or “exoprotein”) and its OM partner (“TpsB” protein or “transporter”) are coordinately expressed but not covalently linked (Fig. 1C).

Fig. 1. Schematic illustration of components of type V secretion pathways.

Classical autotransporters (A) are single polypeptides consisting of a large N-terminal extracellular domain (“passenger domain”) and a ~30 kD β barrel domain (‘β domain”) that resides in the OM. Trimeric autotransporters (B) are homooligomers consisting of three intertwined passenger domains and three ~8 kD β domains that form a single β barrel. In two-partner secretion (C), a large secreted polypeptide (“TpsA” or “exoprotein”) and its ~60 kD transporter (“TpsB”) are encoded as separate polypeptides within a single operon. The passenger domains of classical autotransporters and TpsA molecules (dark blue) are predominantly α helical, while the passenger domains of trimeric autotransporters (light blue) contain distinct β solenoid structural elements. The β domains of classical and trimeric autotransporters (green) are structurally closely related. TpsB proteins, however, consist of a ~20 kD N-terminal periplasmic domain (brown) and a very different β barrel (orange).

Despite the many parallels among the classical autotransporter, trimeric autotransporter and TPS pathways that are described below, the evolutionary relationship among these pathways is not entirely clear. In part because the classical and trimeric autotransporter pathways appear to share a similar secretion mechanism, it is likely that the two types of autotransporters arose from a single common ancestor. In contrast, there are fundamental differences in the mechanism of secretion via the autotransporter and TPS pathways.

Furthermore, autotransporter β domains and TpsB proteins belong to two distinct classes of OM proteins based on sequence (147) and structure. For these reasons the autotransporter and TPS pathway probably represent convergent solutions to the secretion of large polypeptides, and it is unlikely that TPS components were derived from an ancestral autotransporter in which the two domains became “unlinked” (or vice-versa).

In this chapter, I review our current understanding of the biogenesis of components of type V pathways and the mechanism of secretion through each pathway. Although I will describe studies on autotransporters and TPS components that are produced by a variety of different Gram-negative organisms, it is likely that insights obtained from one system are generally applicable to other systems. Moreover, many studies on type V secretion have utilized E. coli autotransporters or TPS components as model proteins or have involved the use of laboratory strains of E. coli as convenient hosts for type V components derived from other organisms.

The structure of polypeptides secreted by type V pathways

The passenger domains of autotransporters and TpsA proteins are not only variable in length but also extremely diverse in sequence. Even closely related passenger domains, such as the members of the E. coli Serine protease autotransporters of Enterobacteriaceae (“SPATE”) family, are only 30-50% identical. In fact, the sequences of members of different families of passenger domains are often difficult to align using standard tools such as BLAST. TpsA proteins all contain a conserved ~250 residue N-terminal segment called the TPS domain (61) that plays a specialized role in secretion (see below), but more C-terminally located segments likewise often share limited homology.

X-ray crystallographic analysis and in silico predictions indicate that the vast majority of both classical autotransporter passenger domains and TpsA proteins fold into an elongated solenoid structure known as a β helix despite their sequence diversity (67, 71). While the passenger domains of some autotransporters such as the Bordetella pertussis pertactin protein are entirely α helical, the passenger domains of other autotransporters such as the E. coli Hbp and Helicobacter pylori VacA proteins also contain small globular domains (37, 41 101; Fig. 2, top). In some cases, the virulence function of an autotransporter is encoded in the loops that connect the rungs of the β helix. Pertactin, for example, contains RGD motifs that are thought to mediate its adhesive activity. In contrast, the protease activity of proteins such as Hbp and other members of the SPATE family are associated with globular domains. The passenger domains of classical autotransporters are likely secreted as monomers, and in general do not appear to oligomerize (57, 95, 120). VacA and the Chlamydia trachomatis PmpD protein, however, clearly form higher order structures (28, 84, 128). Single-layered and double-layered rings consisting of as many as 12-14 subunits of VacA have been observed. The oligomerization of VacA is required for its function as a pore-forming protein (85). Only ~300 residue N-terminal fragments of TpsA proteins (which encompass the TPS domain) have been crystallized, but as predicted form α helical structures (21, 143, 149). Like classical autotransporter passenger domains, some TpsA proteins such as the B. pertussis FHA protein are also predicted to contain globular domains (70).

Fig. 2. Crystal structures of type V secretion pathway components.

Top, crystal structures of secreted polypeptides. The structures of the passenger domains of the classical autotransporters pertactin (Prn) and Hbp are shown (37, 101). The structures of the collagen binding domain of the trimeric autotransporter YadA (residues 26-241) and the TPS domain of the exoprotein FHA (residues 72-375) are also shown (21, 96). β strands are colored yellow and β helices are colored red except in the case of YadA, where individual subunits are colored blue, orange and yellow. Bottom, crystal structures of OM components. Structures of the β domains of the classical autotransporters NalP and EspP and the trimeric autotransporter Hia are shown (3, 91, 100). The structure of the TpsB protein FhaC is also shown (22). β strands are colored yellow and β helices are colored red except in the case of Hia, where individual subunits are colored blue, orange and yellow.

Many trimeric autotransporters also contain modules whose structure is reminiscent of the β helices formed by classical autotransporters. The crystal structure of the ~200 residue collagen binding domain of the Yersinia enterocolitica YadA protein reveals a unique left-handed β roll that is held together by extensive hydrophobic interactions between the three subunits (96; Fig. 2, top). A β roll is a solenoid-like structure in which two β strands from each subunit contribute to a single superhelical turn. Many trimeric autotransporters are predicted to contain at least one β roll module, and some (especially those that contain large passenger domains) are predicted to have multiple β roll modules (71, 96). β roll structures differ in their curvature and twist from α-helical structures and therefore may have evolved to create distinct binding specificities (77). The β roll modules are connected to the β domain and to each other by spacers that form coiled coils (78, 83). Proteins like YadA that contain a single β roll module have a lollipop appearance in electron micrographs (56), while proteins that have a larger number of modules presumably resemble beads on a string. Many of the colled-coils of trimeric autotransporters are unusual in that they contain polar residues (instead of hydrophobic residues) in predicted core positions (131). An analysis of a coiled coil segment from the Salmonella protein SadA shows that these residues form a highly ordered network of polar interactions with an anion at the center (49). As discussed below, the specific properties of these coiled coils may help to maintain the passenger domain in a secretion-competent conformation in the periplasm and may even help to drive the secretion reaction. A few trimeric autotransporters such as the Haemophilus influenzae Hia protein do not contain any β solenoid motifs, but instead contain β prism modules that mediate the adhesive activity of the protein (92, 148). Like β roll modules, β prism modules are formed through the intertwining of β strands from all three subunits. Finally, some trimeric autotransporters contain a mixture of β roll and β prism segments (130). As suggested by the extensive intertwining of the three strands, trimerization is essential for the folding and stability of trimeric autotransporters (24).

The prevalence of solenoid-like elements in polypeptides secreted through type V pathways is striking because they are not found in other secreted bacterial virulence factors. There are several possible explanations for this structural conservation. As discussed below, the common architecture of passenger domains and TpsA proteins might arise at least in part from constraints imposed on the secretion of large polypeptides from the periplasmic space. Consistent with this possibility, globular polypeptide domains that are fused to passenger domains and TpsA proteins are sometimes secreted inefficiently (47, 74, 112). Second, an elongated shape (which is created by long coiled-coil segments as well as β solenoid structures in the case of trimeric autotransporters) may allow the effector domains of passenger domains and TpsA proteins that remain anchored to the cell surface to traverse lipopolysaccharide and capsular layers and interact with host factors. In addition, a solenoid shape creates diverse interfaces that can be used to mediate either monovalent or multivalent interactions (77). Finally, because repetitive structures are thought to accelerate evolution (86), a solenoid-like architecture may have been exploited to promote rapid adaptation to different environments.

The structure of OM components of type V secretion pathways

The β domains of autotransporters are more closely related in sequence than the passenger domains and can be identified by simple search algorithms. The β domains of classical autotransporters are members of a single family of protein domains (PFAM03797) and the β domains of trimeric autotransporters are members of a second unrelated family (PFAM03895). Nevertheless, both types of β domain show considerable sequence diversity and an absence of universal sequence motifs.

Despite the sequence diversity both within and between these two families, the overall architecture of autotransporter β domains appears to be highly conserved. The structure of the β domains of the two classical autotransporters that have been crystallized to date are remarkably similar. The β domains of the Neisseria meningitidis NalP and E. coli O157:H7 EspP proteins (which are only ~15% identical) fold into nearly superimposable 12-stranded β barrels (3, 100; Fig. 2, bottom). In both cases the β barrel forms a hydrophilic channel that is ~10 A in diameter. Interestingly, a polypeptide segment that is C-terminal to the passenger domain cleavage site is embedded inside both β barrels in an α-helical conformation. The α-helical segment completely traverses the NalP β barrel but penetrates only about halfway into the EspP pore. This disparity is most likely due to a difference in the mechanism of passenger domain cleavage. While the NalP passenger domain is cleaved in the extracellular space, the EspP passenger domain is cleaved in the middle of the barrel and only a small segment of the protein remains inside the barrel following passenger domain release (see below). Perhaps more remarkably, the crystal structure of the C-terminal 107 residues of Hia reveals a single 12-stranded β barrel comprised of four β strands from each subunit that is very similar to the β barrels formed by classical autotransporters (91; Fig. 2, bottom). The diameter of the Hia pore is only slightly larger than those of NalP and EspP, and the β strands lie at nearly the same angle relative to the axis of the barrel. While the NalP β barrel is traversed by one α-helical segment, the Hia β barrel is traversed by three, one contributed by each protomer. The α-helical segments are required for the stability of the β barrel and initiate the coiled coil region that connects the β barrel to the β prism modules in the passenger domain.

The conservation of the structure of autotransporter β domains suggests that a specific fold is required for their function. The lack of sequence conservation, however, suggests that there are few if any “catalytic” residues that are essential for the translocation reaction. Consistent with the hypothesis that the function of the β domain is conserved, the C-terminal ~90 amino acids of several different trimeric autotransporters promote the secretion of YadA (1). Furthermore, the last ~300 residues of Hia promote the secretion of a passenger domain derived from a classical autotransporter (123). Although the β domains of distinct families of autotransporters diverge in sequence, there is considerable homology within families. Thus β domains and passenger domains may have coevolved to optimize secretion.

TpsB proteins differ dramatically from autotransporter β domains in both sequence and structure. Sequence analysis shows that TpsB molecules are members of the BamA superfamily (also known as the Omp85 or YaeT superfamily) (42). Members of this family have been found in bacterial, chloroplast and mitochondrial OMs and have been implicated in either protein translocation or membrane protein integration reactions (Fig. 3). All members of the family contain one or more N-terminal polypeptide transport-associated (POTRA) domains that are tethered to the membrane by a C-terminal integral membrane domain. POTRA domains are ~75 residue modules that have been postulated to mediate a chaperone-like function (113), and structural analysis of the POTRA domains of the E. coli BamA protein has led to the proposal that they promote the folding of β barrel proteins through a process known as β augmentation (73, 76). The crystal structure of the FHA transporter (FhaC) reveals two POTRA domains and a 16-stranded monomeric β barrel that forms a ~16 A pore (22; Fig. 2, bottom). The pore is occluded, however, by both an N-terminal β helix and an extracellular loop (L6). Cryo-EM studies indicate that another TpsB protein, H. influenzae HMW1B, also forms an occluded channel (80). Presumably at least the loop would need to be displaced if TpsA proteins are secreted through the TpsB pore. In this regard it is interesting to note that a sequence motif (motif 3) that is conserved throughout the BamA superfamily (94) is located in L6. A highly conserved tetrad, VRGY, is located at the extremity of L6 and reaches into the periplasm. A second conserved motif (motif 4) includes residues located on the inner face of the β barrel along which L6 is positioned. These conserved regions may regulate the displacement of L6 from the β barrel of TpsB proteins and other members of the BamA superfamily in response to a substrate (see below).

Fig. 3. Schematic illustration of members of the BamA superfamily.

Members of the BamA superfamily are found in the bacterial, mitochondrial and chloroplast OMs. Some members of the family have been implicated in membrane protein integration reactions, whereas others promote protein translocation. All family members have between one and five N-terminal POTRA domains and a ~40 kD membrane-embedded β barrel domain. Two sequence motifs (designated motifs 3 and 4; see ref. 94) are conserved throughout the BamA superfamily. TpsB proteins also contain an N-terminal αhelix that is connected to the rest of the protein by an unstructured linker (light blue).

Initial steps in the biogenesis of polypeptides secreted through type V pathways

Autotransporters and TPS pathway components are translocated across the IM via the Sec pathway (9, 10, 44, 102, 119) and therefore must remain in a largely unfolded conformation in the cytoplasm. Despite their large size, most autotransporters and TpsA proteins appear to be targeted to the Sec machinery post-translationally. Whereas especially hydrophobic signal peptides are often recognized by the signal recognition particle (SRP) and targeted to the IM cotranslationally (79), the signal peptides of many autotransporters are not predicted to be SRP substrates. Curiously, some autotransporters (including many of those produced by E. coli) and TpsA proteins (most notably those that have putative adhesive functions) contain unusual signal peptides that exceed 50 amino acids in length. These signal peptides consist of two domains, a C-terminal segment that resembles a typical Sec signal peptide and a unique ~25 residue N-terminal extension that contains a highly conserved sequence motif. Although it was originally proposed that the unusual signal peptides promotes cotranslational targeting (50, 119), several studies have shown that the presence of the N-terminal extension actually inhibits SRP recognition and promotes post-translational targeting (19, 31, 102). In fact, analysis of the 55 residue signal peptide of EspP led to the discovery that the protein can remain translocation-competent in the cytoplasm much longer than many moderately sized proteins (102). The ability of autotransporters and TPS exoproteins to remain translocation-competent may reflect an inherent tendency to fold slowly (67, 69). Post-translational targeting may also be promoted by an interaction with DnaK, which has been shown to maintain OM proteins in a prolonged translocation-competent state and has recently been implicated in the secretion of two Shigella autotransporters (65, 104). The presence of the unusual signal peptide extension in both autotransporters and TpsA proteins might be viewed as evidence of a common evolutionary ancestry (51), but the discovery of a B. pertussis intimin-like protein that also harbors the signal peptide extension (124) suggests that it has been shuffled between unrelated protein families.

The great majority of autotransporters, TpsA and TpsB proteins are translocated across the IM as unmodified polypeptides. A family of E. coli autotransporters that function as adhesins (AIDA-I, Ag43, and TibA) and a family of HMW1-like exoproteins, however, are glycosylated in the cytoplasm. The passenger domains of the three E. coli proteins are ~30% identical and share imperfect ~19 residue repeats (75). Two homologous heptosyltransferases (TibC and AAH) which are encoded directly upstream of TibA and AIDA-I have been shown to attach multiple O-linked heptose moieties to both autotransporters (6, 81, 82, 93). Ag43 is also naturally glycosylated, and AAH and TibC heptosylate the protein in laboratory strains of E. coli (118). Neither TibC nor AAH contains a signal peptide, and there is direct evidence that AIDA-I is glycosylated prior to its translocation across the IM (15). Glycosylation is required for the binding of all three proteins to mammalian cells and increases the stability of AIDA-I (6, 36, 118). Unlike the autotransporters, HMW1 is subject to N-linked glycosylation. The carbohydrate addition requires phosphoglucomutase and HMW1C, a cytoplasmic glycosyltransferase that is encoded immediately downstream from hmw1-hmw1B (45). While HMW1 is glycosylated at multiple asparagine residues that reside in the universal NX(S/T) consensus sequence, the protein is modified with mono-hexose or di-hexose sugars instead of the N-acetylated sugars that are generally found in N-linked glycans (46). Glycosylation of HMW1 appears to protect it from premature degradation. An E. coli adhesin called EtpA that is >30% identical to HMW1 and that is likewise encoded directly upstream from a putative glycosyltransferase has also been shown to be glycosylated (40).

A few classical autotransporters including NalP, the B. pertussis SphB1 protein (which is homologous to NalP) and the H. pylori AlpA protein contain a lipobox motif [LA(A/G)|C] straddling the signal peptide cleavage site (27, 97, 135). This motif targets proteins to an acyltransferase that attaches a lipid to the cysteine residue on the periplasmic side of the IM (43). Most lipoproteins are released from the IM and then targeted to the OM by the Lol system (see chapter by Tokuda). SphB1 mediates the proteolytic processing of the FHA exoprotein (see below), but it must be tethered to the OM by the N-terminal lipid moiety in order to perform its function (27). It is unclear, however, whether the lipid is transferred across the OM or remains anchored to the inner leaflet of the OM. While the acyl group presumably plays a role in the biogenesis of NalP, the first ~30 residues are absent from the cleaved passenger domain and the significance of lipidation for protein function is unknown.

Transit through the periplasm

It is generally believed that OM proteins are released from the Sec machinery into the periplasm before they reach their final destination, although other models for the transport of proteins from the IM to the OM have been envisioned (110). While autotransporters and TpsB proteins therefore likely reside transiently in the periplasm, the existence of periplasmic intermediates of TpsA proteins has been more controversial. To explain the observation that an ~80 kDa N-terminal fragment of FHA (Fha44) is trapped inside cells when the synthesis of FhaC is delayed, it was proposed that TpsA proteins are transported across the IM and OM in an obligately coupled fashion and are never released into the periplasm (47). Without ruling out the possibility that translocation across the IM and OM are coupled under certain conditions, however, a more recent study demonstrated that the E. coli O157:H7 exoprotein OtpA can reside briefly in the periplasm before it crosses the OM (20).

Several lines of evidence indicate that there are considerable constraints on the folding of TpsA proteins in the periplasm. Experiments in which the synthesis of OtpA and its cognate transporter (OtpB) were uncoupled showed that the exoprotein can fold rapidly into a secretion-incompetent conformation (20). Moreover, effective secretion of TpsA proteins appears to be incompatible with the formation of even small elements of tertiary structure. A naturally occurring disulfide-bonded loop at the C-terminus of HMW1 acts as a stop-transfer sequence that anchors the protein to the OM, and a version of OtpA that was engineered to contain a similar C-terminal disulfide bond is secreted only very inefficiently (12, 20). Indeed the dimensions of the FhaC pore suggest that it can only secrete proteins that lack significant tertiary structure.

While the folding of passenger domains in the periplasm is also constrained, the autotransporter secretion mechanism tolerates (and conceivably even requires) a more significant degree of passenger domain folding. The presence of naturally-occurring or artificial disulfide bonds that create small elements of tertiary structure has no discernable effect on passenger domains secretion, and the results of one study suggest that at least one passenger domain may undergo substantial folding in the periplasm (8, 66, 120). Furthermore, a chimeric passenger domain comprised of the cholera toxin B subunit (CtxB) fused to EspP is secreted efficiently despite folding of the CtxB moiety in the periplasm (120). Some folded structures, however, are clearly not secreted efficiently. The introduction of unnatural disulfide bonds that connect distant segments of pertactin or Hbp appears to create a structure that jams the transport channel (66, 68). A chimeric alkaline phosphatase-Hia passenger domain is secreted only in the presence of reducing agents, and a maltose binding protein (MBP) moiety fused to the passenger domain of the E. coli autotransporter AIDA-I is secreted inefficiently unless a mutation that slows the folding of MBP is introduced (112, 123). It is not entirely clear in these cases, however, if the heterologous polypeptide domain interferes with translocation per se or with folding of the protein before it reaches the OM. In this regard it is important to note that even small insertions in the passenger domain can seriously impair its secretion (11, 34). This observation suggests that minor perturbations of folding can easily create a translocation-incompetent conformation.

Available evidence indicates that autotransporter passenger domains and TpsA proteins are maintained in a secretion-competent conformation in the periplasm by two distinct mechanisms. First, periplasmic chaperones likely prevent both types of polypeptides from misfolding. Consistent with this notion, the chaperones DegP, SurA, and Skp interact with passenger domains and TPS exoproteins in vivo and in vitro, and their depletion impairs secretion through type V pathways (4, 59, 111, 141). More surprisingly, the long signal peptides associated with a subset of autotransporters and TPS exoproteins also appear to play a role in promoting secretion-competence. Replacement of the native EspP signal peptide with a generic signal peptide has no effect on the transport of the protein across the IM but strongly impairs the translocation of the passenger domain across the OM (129). Curiously, the EspP signal peptide also transiently jams the Sec machinery. Based on these results, a model was proposed in which the EspP signal peptide dissociates from the Sec complex (and becomes susceptible to cleavage) relatively slowly and thereby prevents passenger domain misfolding by serving as a transient membrane anchor. The observation that truncation of the EspP passenger domain obviates the need for a long signal peptide suggested that the tendency of the full-length protein to misfold in the periplasm is related to its size or overall structure (129). Consistent with this hypothesis, long signal peptides are not randomly distributed, but instead are clustered within specific families or subsets of autotransporters and TpsA proteins. Extended signal peptides are very common among trimeric autotransporters that have long passenger domains (>1000 amino acids), but are not found among those that have short passenger domains. Likewise, long signal peptides are found on all members of the SPATE family, the AIDA-I/TibA/Ag43 family of E. coli adhesins, the YapH family of large putative Yersinia autotransporters, and TpsA proteins that have been implicated in adhesion. Although the sequences of the secreted polypeptides within these clusters are not always closely related, it seems plausible that a similarity in sequence, size or function might lead to the use of similar folding pathways and impose the need for a long signal peptide to offset an elevated tendency to misfold.

There is good evidence that important aspects of the assembly of the C-terminus of autotransporters are catalyzed in the periplasm at the same time that the folding of the passenger domain is restricted. Although the compartment in which trimeric autotransporters oligomerize has not been investigated, studies on a prototypical trimeric OM protein, E. coli TolC, suggest that they trimerize slowly in the periplasm (144). In addition, an analysis of EspP biogenesis has demonstrated that the α-helical segment that traverses the β barrel is incorporated into the pore of the β domain (which presumably is folded into a “proto-barrel”) in the periplasmic space (58). The β domain and the embedded polypeptide are then integrated into the OM as a pre-formed unit. This assembly reaction is essential for the progression of autotransporter biogenesis because deletion of a small portion of the α-helical segment leads to the rapid degradation of the β barrel and blocks its integration into the OM. Furthermore, as discussed below, the discovery that the C-terminus of an autotransporter assembles in the periplasm has important implications for the mechanism of passenger domain secretion. Although molecular chaperones such as SurA and Skp are required for the efficient biogenesis of OM proteins in general (122), their role in the assembly of autotransporter β domains has not been investigated.

Current models of the mechanism of protein secretion through type V pathways

The hypothesis that autotransporters represent a self-contained secretion system originated in the late 1980s and stems largely from the observation that deletion of the β domain of the prototypical autotransporter, the Neisseria gonorrhoeae IgA protease, abolishes passenger domain secretion (103). Based on this result and the finding that the β domain resides in the OM, it was proposed that the passenger domain is secreted through the pore formed by the β domain to which it is covalently linked. Both the observation that N-terminal truncations do not affect passenger domain translocation (which implies that the N terminus lacks a signal that targets it to the transport channel) and direct tests of secretion directionality have demonstrated that secretion is initiated at the C-terminus (59, 68, 74, 87, 129). Thus, translocation of the passenger domain (or passenger domains in the case of trimeric autotransporters) through a single β domain would presumably involve the formation of a C-terminal hairpin followed by the progressive extrusion of more N-terminally located segments into the extracellular milieu (“self-transport model”, Fig. 4A). More recently, the finding that a C-terminal 45 kDa fragment of the IgA protease forms oligomeric ring-like structures and that folded single chain Fv fragments fused to the IgA protease passenger domain are secreted efficiently led to an alternative hypothesis in which passenger domains are secreted through a large central channel formed through the assembly of multiple β domain protomers (138, 139). While the possibility that the IgA protease forms oligomeric channels cannot be excluded, the fact that the β domains that have been crystallized form typical β barrels with a hydrophobic exterior strongly suggests that a hydrophilic channel (which presumably is required for passenger domain secretion) could not be formed through their oligomerization. Furthermore, biochemical analysis of several different E. coli autotransporters indicates that they are almost certainly monomeric (57, 95, 120).

Fig. 4. Models of the mechanism of autotransporter and two-partner secretion.

Top, models of the mechanism of autotransporter secretion. In the self-transport (or “hairpin”) model (A), the passenger domain is secreted by the covalently linked β domain. The C-terminus of the passenger domain first forms a hairpin that is embedded inside the β domain pore. Subsequently, segments of the passenger domain are threaded through the β domain pore in a C-to-N terminal direction until the entire polypeptide reaches the extracellular milieu. In the facilitated transport model (B), the passenger domain is extruded across the OM in a C-to-N terminal direction by the Bam complex or another external transporter by an unknown mechanism. In this model, the primary role of the β domain is to target the passenger domain to the appropriate transport factor. In both of these models a segment of the passenger domain is incorporated into the β domain prior to its integration into the OM. Although only a classical autotransporter is depicted, it is likely that trimeric autotransporters utilize a similar secretion mechanism. Bottom, model of the mechanism of two-partner secretion (C). In this model, an interaction between the TPS domain of TpsA and the TpsB POTRA domains gates open the pore of the TpsB β barrel by catalyzing a conformational change in loop 6 (L6). TpsA is then threaded through the pore in an N-to C-terminal fashion. After the C-terminus of the protein is transported through the pore the N-terminus dissociates from the POTRA domains and is secreted. The highly conserved VRGY motif in L6 is colored green. It should be noted that in all of these models the vectorial folding of the passenger domain or TpsA protein in the extracellular space drives translocation by preventing the secreted polypeptide from sliding back through the translocation channel.

Despite its attractiveness, the “autotransporter” hypothesis has been strongly challenged by a variety of experimental results and theoretical considerations. In light of recent structural studies, it is difficult to reconcile the idea of protein secretion through the β domain pore with the observation that passenger domains that have tertiary structure are secreted efficiently. The pore of the NalP and EspP β domains is sufficiently wide to facilitate the transport of a fully extended polypeptide in a hairpin configuration or a single β helix, but is too narrow to accommodate the passage of naturally occurring disulfide-bonded loops (let alone folded CtxB domains). Molecular dynamics simulations show that the NalP β barrel is extremely stable and is unlikely to expand spontaneously to permit the transport of folded segments (72). Although it is conceivable that disulfide-bonded segments unfold prior to their secretion and then refold in the extracellular space, a variety of experimental observations indicate that this scenario is very unlikely (7). Furthermore, the three α-helical segments that are embedded inside the Hia β barrel are packed extremely tightly, and it is difficult to imagine polypeptide movement through the pore given the tightness of the fit. Beyond these fundamental inconsistencies, studies on EspP biogenesis strongly suggest that the C-terminus of the passenger domain is embedded inside the β domain in a linear configuration (instead of the hairpin configuration that would be expected if the β domain were a translocation channel) prior to the integration of the β domain into the OM (58). Finally, sequence analysis, structural studies and site-directed mutagenesis have failed to identify key catalytic residues in either the β domain or the C-terminus of the passenger domain. β domains lack the conserved sequence motifs that are found in TpsB proteins, and the interior of the NalP, EspP and Hia β barrels differ radically in composition and charge. Mutation of a variety of conserved residues that point into the lumen of the β barrel has no effect on the secretion of the EspP passenger domain (29). In addition, the insertion of short linkers near the C-terminus of the EspP passenger domain has only a subtle effect on translocation despite the fact that they would likely disrupt any hairpin structure that is encoded in the polypeptide sequence (58).

In light of these results, an alternative model has been proposed in which the passenger domain(s) is transported across the OM by an external transporter (Fig. 4B). The factor that would most likely mediate the translocation reaction is the so-called Bam complex. This complex consists of the aforementioned integral OM protein BamA and several lipoproteins (designated BamB-E in E. coli) (121, 139, 146). Although its function is not well understood, the Bam complex has been clearly implicated in the integration of β barrel proteins into the OM and has been shown to be required for autotransporter biogenesis (63, 109, 114, 140). As mentioned above, however, members of the BamA superfamily (most notably TpsB proteins and the chloroplast Toc75 protein) have also been associated with protein translocation reactions (54). Consistent with the idea that the Bam complex plays an important role in passenger domain secretion, recent photocrosslinking experiments that used a transiently stalled EspP translocation intermediate showed that the passenger domain interacts with BamA during its transit across the OM (59). Furthermore, liposome swelling assays have provided evidence that the N. meningitidis Bam complex can form pores that are large enough to transport at least partially folded polypeptides (107). Based partly on these observations, it has been proposed that the Bam complex catalyzes the integration of autotransporter β barrels into the OM and the translocation of autotransporter passenger domains across the OM in a concerted fashion (58, 59). In this model, the β domain is essential for autotransporter secretion because it targets the passenger domain to the Bam complex, not because it functions as a translocase itself. Presumably the incorporation of a short polypeptide segment inside the β domain pore is required for effective targeting.

While the “facilitated transport” model is consistent with available evidence and provides a reasonable alternative to the autotransporter hypothesis, the mechanism by which the Bam complex would facilitate a combined translocation/insertion reaction (or another transporter would simply promote passenger domain secretion) is unclear. At the present time, no model is completely satisfying. Assuming that the structure of BamA resembles that of the homologous FhaC protein, It is difficult to imagine how β barrel proteins would be integrated into the OM through the pore formed by a BamA monomer, especially if they undergo significant folding in the periplasm. It seems more likely that the Bam complex acts as a “chaperone” that promotes spontaneous insertion of the β domain into the OM. In that case, perhaps only the passenger domain passes through the BamA pore. The mechanism of passenger domain secretion might then resemble the mechanism by which TpsA proteins are secreted through TpsB transporters (see below). It is unclear, though, how the passenger domain would enter and ultimately be released from the interior of the BamA β barrel. Alternatively, multiple subunits of the Bam complex might contribute to the formation of a novel translocation channel that can be gated laterally. Another possibility is that the passenger domain passes through the β domain pore as originally proposed, but that the Bam complex is required to expand the β domain pore and hold it in an “open” conformation. It is difficult to imagine, however, how the aqueous nature of the translocation channel or the stability of the β barrel of trimeric autotransporters would be maintained in this scenario. In all probability, the role of the Bam complex in autotransporter biogenesis will only emerge from a much better understanding of its overall function.

At least in some respects, the mechanism of protein secretion through the TPS pathway appears to be rather distinct from the mechanism of autotransporter secretion. Because the two components of the TPS pathway are independent polypeptides, the exoprotein must be targeted to the transporter. The targeting function is mediated by the N-terminal TPS domain, which is both necessary and sufficient for TpsA secretion (44, 55, 105, 116). Studies on FHA indicate that the TPS domain encodes an extremely extended signal that targets the protein to FhaC in an unfolded conformation (55), The TPS domain is recognized by the TpsB POTRA domains, and the FhaC crystal structure suggests that this interaction brings the N-terminus of the TpsA protein into proximity to the tip of loop L6 (and the VRGY motif) that protrudes into the periplasm (22, 55, 127). It has been proposed that the TpsA protein is subsequently transported through the pore of the TpsB β barrel (22, 90). Consistent with the idea that TpsB functions as a secretion channel, formation of a C-terminal disulfide bond in HMW1 not only creates a stop-transfer element, but also blocks the secretion of additional HMW1 molecules when HMW1B is limiting (12). In addition, both FhaC and HMW1B have been shown to have a conductance that is similar to that of porin monomers (33, 60, 126). Loop L6 in FhaC seems to represent a critical mobile element because it becomes accessible to protease digestion only during active secretion (48). Furthermore, deletion of L6 abolishes secretion and reduces the channel activity of FhaC (22). The data suggest a model in which secretion of the TpsA protein through the TpsB β barrel pore is initiated by conformational changes that lead to the expulsion of L6 from the barrel (22; Fig. 4C). Work on FHA showing that the C-terminus of the protein is distal to the cell surface has led to the proposal that the N-terminus of TpsA proteins remains bound to the TpsB POTRA domains while the rest of the protein is threaded through the TpsB pore in an N-to C-terminal direction (88). Presumably the N-terminus is released from the POTRA domains and is transported across the cell surface at a late stage of the secretion reaction. This model implies that most of the TpsA protein is secreted in a hairpin conformation.

Ironically, the greatest challenge to this elegant model arises from the fact that it is not easily reconciled with our current understanding of autotransporter secretion. If the original autotransporter hypothesis is correct and the Bam complex is only an OM protein insertase, then the homology between BamA and TpsB proteins and the conservation of sequence motifs that appear to play an important role in secretion becomes difficult to explain. Alternatively, if the Bam complex facilitates the secretion of passenger domains but does not transport them through the BamA pore, then it becomes difficult to explain why some members of the BamA superfamily use the β barrel as a translocation channel while others do not. Clearly, further work is needed to resolve these conceptual puzzles.

Regardless of the exact mechanism of secretion, a great deal of energy must be required to drive polypeptides that exceed 100 kD across the OM. Given that the periplasm is devoid of ATP, however, the source of this energy is unclear. Because autotransporters are presumably released into the periplasm en route to the OM and there is no obvious activation or gating of the Bam complex by an IM protein, it is unlikely that the membrane potential maintained across the IM is harnessed to drive passenger domain secretion. Likewise, the finding that the translocation of TpsA proteins across the OM can be uncoupled from their translocation across the IM implies that the Sec machinery does not provide the energy for secretion.

Based partly on the observation that the passenger domains of classical autotransporters fold slowly in vitro, it has been proposed that secretion through type V pathways is driven by the diffusion of an unfolded polypeptide (or polypeptides in the case of trimeric autotransporters) across the OM followed by its vectorial folding in the extracellular space (67; Fig. 4). Presumably the folding of the secreted polypeptide prevents it from sliding back into the periplasmic space. As a corollary, it has also been proposed that β solenoid structures have been conserved in type V pathways because their folding properties potentiate secretion in the absence of ATP. Several lines of evidence have provided support for this general model. First, the observation that the translocation of the EspP passenger domain is stalled near the site of a linker insertion that also perturbs folding suggests a direct connection between passenger domain folding and secretion (59). Furthermore, equilibrium denaturation titrations have revealed the presence of a stable ~20-25 kD core near the C-terminus of two weakly related passenger domains that might serve as a template for the folding of the rest of the protein (67, 106). It is also noteworthy that the unusual properties of the coiled coil regions of trimeric autotransporters tend to hinder coil formation. Attaching these segments to folded domains effectively raises their local concentration and promotes assembly (49). Thus it is possible that stable coiled coils do not form until their assembly is triggered by the folding of adjacent domains in the extracellular environment. The hypothesis that folding drives secretion, however, does not account for the observation that polypeptide segments that fold in the periplasm (e.g., CtxB) can be secreted efficiently (87, 120). This hypothesis also does not account for the secretion of a C-terminally deleted passenger domain of the B. pertussis autotransporter BrkA that does not fold properly or the single misfolded passenger domain of YadA and Hia heterotrimers formed through the co-expression of full-length and truncated protomers (24, 98). Furthermore, if secretion is linked to vectorial folding, then it is unclear how passenger domains can be secreted in a C-to-N terminal direction while TpsA proteins are secreted in an N-to-C-terminal direction unless the folding of α-helical proteins can be initiated at either end.

Proteolytic processing and maturation of proteins secreted via type V pathways

While some autotransporters (including all trimeric autotransporters) and TpsA proteins remain intact, others undergo proteolytic processing. Proteolytic processing of autotransporters releases the passenger domain from the β domain and, as discussed below, serves several different purposes. At least in the case of EspP, processing occurs after the passenger domain is translocated across the OM (120). Some TpsA proteins are synthesized as proproteins that are processed either during or after their secretion. As discussed below, at least in the case of FHA the pro domain is thought to promote the folding of the remainder of the protein. Interestingly, some TpsA proteins that are not subject to proteolytic cleavage undergo a different sort of maturation in which they are activated by their cognate TpsB transporters.

Passenger domains are cleaved from the cell surface by a remarkable variety of distinct mechanisms. In most cases, proteolysis occurs in the extracellular space and a small segment of the polypeptide that has been transferred across the OM remains associated with the β domain. The Shigella IcsA protein is cleaved by a dedicated protease called IcsP that is related to OmpT and other members of the so-called omptin family (35, 117; Fig. 5A). The passenger domains of several N. meningitidis autotransporters including App, Aus I, MspA and the IgA protease are cleaved by NalP, which encodes a serine protease (132, 135, 136; Fig. 5B). The passenger domains of App and the IgA protease are further processed in an autocatalytic reaction that releases a small C-terminal “α-fragment”, but it is not known whether the reaction is intramolecular or intermolecular. NalP is also processed in two steps; one step involves an autocatalytic reaction and the other involves cleavage by an unidentified protease (135). The passenger domains of the H. influenzae Hap and B. pertussis SphB1 proteins are cleaved in an intermolecular autocatalytic reaction by an endogenous serine protease (Fig. 5C; 25, 38, 52). AIDA-I likewise undergoes autoprocessing, but the cleavage reaction appears to be intramolecular and involves acidic residues located near the C-terminus of the passenger domain (Fig. 5D; 16). Interestingly, the SPATE proteins are not processed by their endogenous serine protease activities. Instead, the passenger domains of these proteins (as well as several B. pertussis autotransporters including pertactin and BrkA) are cleaved in a unique reaction that occurs inside the pore of the β domain (Fig. 5E). This reaction involves a nucleophilic attack on the polypeptide backbone by an invariant asparagine residue located on the N-terminal side of the cleavage junction and is similar to self-cleavage reactions mediated by inteins, the EscU component of the E. coli type III secretion machine, and eukaryotic viral capsid proteins (29). The mechanism by which VacA, PmpD, and other autotransporters including the S. marcescens PrtS/Ssp protein and the B. henselae Cfa and Arp proteins are processed is still unknown, so additional cleavage reactions might yet be discovered.

Fig. 5. Mechanisms of passenger domain and exoprotein proteolytic processing.

Top, mechanisms of passenger domain cleavage. (A) The passenger domain of IcsA (blue) is cleaved by a dedicated protease called IcsP (pink). (B) The passenger domains of the IgA protease (IgaP) and App are cleaved by NalP and are further processed in an autocatalytic reaction to release a small C-terminal α-fragment. It is not known whether the latter reaction is intramolecular or intermolecular. (C) The passenger domain of Hap is cleaved in an intermolecular autoproteolytic reaction. (D) AIDA-I is cleaved near the C-terminus of the passenger domain in an intramolecular reaction that involves acidic residues. (E) The passenger domains of the SPATEs and the pertactin (Prn) family of B. pertussis autotransporters are released in an intra-barrel autocatalytic reaction. Bottom, mechanisms of exoprotein processing. (F) The C-terminal ~1200 residues of FHA are removed from the protein by the autotransporter SphB1. It is not known whether the C-terminal pro domain is cleaved prior to its translocation across the OM, but it appears to be rapidly degraded. (G) The N-terminal ~370 residues of HMW1 are also released in a proteolytic reaction. The identity of the protease is not known, but the pro domain is probably cleaved after its secretion. HMW1 is anchored to the OM by a C-terminal disulfide bond.

Clearly the passenger domains of proteins like the SPATEs and VacA that perform their function only after internalization by host cells must be released from the cell surface. Likewise the proteolytic processing of autotransporters such as the IgA protease presumably increases their effectiveness by allowing the passenger domain to diffuse away from the bacterial cell. In some cases, however, release of the passenger domain serves a highly specialized purpose. IcsA, for example, recruits an actin tail that facilitates the movement of S. flexneri in the cytoplasm of host cells and between cells. This function requires that the protein be localized to a single pole. Elimination of passenger domain cleavage by site-directed mutagenesis or disruption of the icsP gene leads to the accumulation of a small amount of the protein in non-polar regions and to the formation of aberrant actin tails (32, 35). These observations suggest that IcsP serves to release the passenger domains of mislocalized IcsA molecules. In the case of SphB1, whose passenger domain remains anchored to the OM by an N-terminal lipid, proteolytic processing may allow the passenger domain to adopt an appropriate orientation. In contrast, it is unclear why some passenger domains that function as adhesins are cleaved. Hap is cleaved in an intermolecular reaction that is concentration-dependent, and because the protein is ordinarily produced at a low level the passenger domain is released from the cell surface slowly (39). Slow cleavage might promote bacterial spread. Alternatively, the production of soluble passenger domain molecules might titrate circulating antibodies or enable proteins like Hap, which encode a proteolytic activity, to mediate a second non-adhesive function. The proteolytic processing of adhesins such as AIDA-I, Ag43 and pertactin is especially enigmatic because the cleaved passenger domains remain noncovalently associated with the cell surface (5, 13, 17). Furthermore, elimination of AIDA-I cleavage does not affect its function (14).

At least in some cases, the pro domain of TpsA proteins appears to promote proper folding of the mature domain. FHA is synthesized as a precursor (designated FhaB) containing a ~1200 residue C-terminal pro domain that represents nearly one-third of the protein. The observation that the pro domain is required for the efficient secretion of full-length FHA but not short N-terminal fragments suggests that it functions as an intramolecular chaperone that prevents misfolding or degradation of the protein in the periplasm (105). Consistent with this hypothesis, truncated versions of FhaB that contain only a portion of the pro domain are secreted efficiently but show a marked loss of function in in vitro adhesion assays (88). Because the effector function of mature FHA is mediated by the last ~500 residues, the results suggest that the pro domain influences the final conformation of this globular segment of the protein (89). Assuming that translocation occurs in an N-to C-terminal direction, it is likely that the pro domain is cleaved either at a late stage of exoprotein secretion or following secretion (Fig. 5F). Because FHA release from the cell surface requires cleavage by SphB1 or an alternative protease that cleaves at a slightly different location, the pro domain also appears to anchor the protein to the OM (25, 88). Even after processing, only a fraction of the FHA is released from the cell surface, but this release is essential for effective colonization in an animal model (26). The release of FHA, like the release of autotransporter adhesins, may enhance bacterial dissemination. HMW1 is also synthesized as a proprotein, but the pro domain (which is less than 400 residues) is located at the N terminus and encompasses the TPS domain (44). The function of the pro domain is unclear, and proteolytic processing of the protein is not required for either secretion or adhesive activity. Given that the free pro domain is observed in the extracellular milieu, cleavage most likely occurs after the N terminus is secreted (Fig. 5G).

Unlike the TpsA adhesins that undergo proteolytic maturation, cytolysin/hemolysins such as the Serratia marcescens ShlA protein are activated during secretion in a non-proteolytic process (115). Activation does not appear to involve covalent modification of the protein, but requires both a physical interaction with ShlB (which can be reproduced by mixing the inactive form of ShlA, designated ShlA*, and ShlB in vitro) and the presence of phosphatidlyethanolamine (53, 99). During the activation process ShlA* undergoes a conformational change (142). The available evidence suggests a model in which an interaction with ShlB alters the conformation of ShlA* and allows phosphatidylethanolamine to stabilize the protein it its active conformation. The observation that a sequence motif in the TPS domain that is required for secretion is also required for activation suggests that the two processes are coupled (116). Presumably the cytolysins are synthesized in an inactive form to prevent them from killing their hosts prior to their released into the extracellular milieu.

Polar localization of autotransporters

While the polar localization of IcsA has been clearly established and is obviously required for its function, there is evidence that other autotransporters may also be distributed in the OM in a non-random fashion. A recent study showed that several different autotransporters including AIDA-I, BrkA and the S. flexneri SepA protein exhibit a unipolar localization both in their native environment and in E. coli (62). NalP was shown to have a unipolar distribution in E. coli but was localized to multiple discrete foci in N. meningiditis, which is a spherical organism that lacks distinct poles.

Several lines of evidence suggest that positional information is encoded in the passenger domain of IcsA. Experiments in which fragments of IcsA were fused to GFP indicate that residues 1-104 and 507-620 are each sufficient for polar localization in enterobacteria and Vibrio cholera (18). In the same study the synthesis of untagged IcsA was observed to interfere with the polar localization of IcsA-GFP. Taken together, the results suggest that the polar positioning of IcsA is mediated by a limiting cellular factor that is evolutionarily conserved. The observation that neither deletion of the signal peptide nor inactivation of SecA or SecY affects the localization of IcsA, however, provided evidence that the secretion machinery is not involved in polar targeting (9, 18). Moreover, although IcsA is found in septa (which correspond to future poles) in filamentous cells, localization does not require the activity of known cell division factors (64). Recent work has suggested that DnaK may play a role in polar positioning (65). Although the mechanism and biological significance of the localization of other autotransporters is unknown, the polar localization of a SepA-GFP fusion protein also appears to be independent of secretion (62).

Concluding remarks

Although a great deal of new information about autotransporters and TPS exoproteins and transporters has emerged through the use of biochemical, structural and theoretical methods in recent years, the mechanistic details of protein secretion through type V pathways are still somewhat enigmatic. Recent results have strongly challenged the original proposal that autotransporters are self-contained protein translocation systems and have suggested that exogenous factors, most likely the Bam complex, play a key role in the secretion process. Far from solving the problem of autotransporter secretion, the involvement of exogenous factors raises many new questions. Given the biochemical and biophysical properties of βbarrel proteins, it is difficult to envision a protein conducting channel that would possess the lateral gating properties of the Sec complex (134) that are presumably required to release passenger domains after their translocation is complete. It seems likely that the passenger domains of classical and trimeric autotransporters are secreted by a similar mechanism, but this notion needs to be proven experimentally. The mechanism of secretion via the TPS pathway is likewise not entirely clear. A variety of results suggest that TpsA proteins pass through the pore formed by the TpsB β barrel, but this model raises questions about the similarity of TpsB proteins to BamA, which at least so far has only been shown to promote membrane protein integration. While type V secretion mechanisms might ultimately resemble well-established protein translocation paradigms, these and other paradoxes suggest that at least some of the properties of type V secretion might turn out to be rather novel.