Looks can be deceiving: recent insights into the mechanism of protein secretion by the autotransporter pathway

Summary

Autotransporters are a large superfamily of cell surface proteins produced by Gram-negative bacteria that consist of an N-terminal extracellular domain (‘passenger domain’) and a C-terminal β-barrel domain that resides in the outer membrane (OM). Although it was originally proposed that the passenger domain is translocated across the OM through a channel formed exclusively by the covalently linked β-barrel domain, this idea has been strongly challenged by a variety of observations. Recent experimental results have suggested a new model in which both the translocation of the passenger domain and the membrane integration of the β-barrel domain are facilitated by the Bam complex, a highly conserved heteroligomer that plays a general role in OM protein assembly. Other factors, including periplasmic chaperones and inner membrane proteins, have also recently been implicated in the biogenesis of at least some members of the autotransporter superfamily. New results have raised intriguing questions about the energetics of the secretion reaction and the relationship between the assembly of autotransporters and the assembly of other classes of OM proteins. Concomitantly, new mechanistic and structural insights have expanded the utility of the autotransporter pathway for the surface display of heterologous peptides and proteins of interest.

Introduction

The autotransporter pathway is one of the simplest of the diverse secretion systems used by Gram-negative bacteria to deliver proteins into the extracellular environment. Classical autotransporters are single polypeptides that contain a signal peptide that directs transport across the inner membrane (IM) via the Sec pathway and two major domains, an N-terminal extracellular domain (‘passenger domain’) and a C-terminal domain that resides in the outer membrane (OM). Passenger domains often function as hydrolases, cytotoxins, or adhesins, or have other activities associated with virulence (Henderson and Nataro, 2001). Following their transport across the OM, many passenger domains are released from the cell surface by a proteolytic cleavage (Dautin and Bernstein, 2007). Passenger domains are highly diverse in size and sequence, but often exceed 100 kD in length. X-ray crystallographic analysis and in silico studies indicate that the vast majority of passenger domains fold into a repetitive structure known as a β-helix despite their sequence diversity (Emsley et al., 1996; Otto et al., 2005; Junker et al., 2006; Gangwer et al., 2007; Heras et al., 2014). The C-terminal domains are typically ∼30 kD in length and are also diverse in sequence, but contain short conserved sequence motifs (Celik et al., 2012; Leyton et al., 2014). Like most polypeptides that are fully integrated into the OM, these domains fold into a closed β-sheet known as a β-barrel. All of the C-terminal domains that have been crystallized to date form nearly superimposable 12-stranded β-barrels (Oomen et al., 2004; Barnard et al., 2007; van den Berg, 2010; Tajima et al., 2010; Zhai et al., 2011). The two major domains are connected by a short ‘linker’, part of which forms an α-helical segment that is embedded inside the β-barrel domain (Drobnak et al., 2015). In a recent survey, hidden Markov models identified > 1500 genes that likely encode autotransporters in the genomes of Proteobacteria, Chlamydiales and Fusobacteria and confirmed that the autotransporter pathway is very widespread (Celik et al., 2012).

The classical autotransporter pathway has been designated the ‘type Va’ pathway In the numerical classification of bacterial secretion pathways, and it is generally grouped together with several other seemingly related secretion pathways under the umbrella of ‘type V’ secretion (Leo et al., 2012; Fig. 1). In the type Vb or two-partner secretion (TPS) pathway, a single β-helical ‘exoprotein’ that resembles an autotransporter passenger domain (TpsA) is secreted by a dedicated OM transporter (TpsB) (Jacob-Dubuisson et al., 2001; Clantin et al., 2004; Yeo et al., 2007). The two partners are co-ordinately expressed, but not covalently linked. Furthermore, the TpsB proteins differ from autotransporter C-terminal domains in that they are members of the ‘Omp85’ superfamily (Gentle et al., 2005; Heinz and Lithgow, 2014). These proteins have 16-stranded β-barrels plus a variable number of periplasmic ‘POTRA (polypeptide transport-associated) domains’ that are thought to mediate protein– protein interactions (Clantin et al., 2007; Gruss et al., 2013; Noinaj et al., 2013). In the type Vc or ‘trimeric autotransporter’ pathway, three identical polypeptide chains contribute to the formation of a trimeric passenger domain and a single 12-stranded β-barrel that is structurally nearly identical to the β-barrels of classical autotransporters (Linke et al., 2006; Meng et al., 2006). The passenger domain contains one or more globular ‘head’ domains (often ‘β-roll’ or ‘β-prism’ structures) that are connected to each other and the β-barrel by coiled-coil ‘stalk’ regions. The C-terminus of the trimeric passenger domain is continuous with three α-helices that traverse the β-barrel. The recently described type Vd pathway appears to be a chimera of the type Va and type Vb pathways (Salacha et al., 2010). In this pathway, a passenger domain that is homologous to the patatin family of lipases is covalently linked to a C-terminal domain that contains features reminiscent of the type Vb transporters. Finally, in the type Ve or intimin/invasin pathway the order of domains is reversed: members of this family contain an autotransporter-like β-barrel domain at the N-terminus and a passenger domain comprised of immunoglobulin-like repeats at the C-terminus (Hamburger et al., 1999; Fairman et al., 2012; Leo et al., 2014).

Fig. 1.

Illustration of the type V secretion pathways.

A. In the type V secretion pathways, structurally related 12-stranded β-barrel domains (red) or larger β-barrels with associated POTRA domains (green) facilitate the secretion of a polypeptide (passenger domain or exoprotein) that typically has a β-helical (blue), mixed coiled-coil/β-roll/β-prism (purple) or globular (brown) structure. In most cases, the β-barrel domains and passenger domains are covalently linked, but they are separate polypeptides in the type Vb pathway. The type Vc pathway is unique in that both the β-barrel domain and the passenger domain are formed through the assembly of three identical subunits. The passenger domain is located at the N-terminus of the polypeptide in the type Va, Vc and Vd pathways, but is located at the C-terminus in the type Ve pathway.

B. Crystal structures of polypeptides associated with the type Va, Vb, Vc and Ve pathways are shown. These include the pertactin (Prn) passenger domain (Emsley et al., 1996; PDB ID: iDAB), the NalP β-barrel domain (Oomen et al., 2004; PDB ID: 1UYO), a fragment of the HMW1 exoprotein (Yeo et al., 2007; PBD ID: 2ODL), FhaC (Clantin et al., 2007; PDB ID: 4QKY), a fragment of the EibD passenger domain (Leo et al., 2011; PDB ID: 2XQH), the C-terminus of Hia (Meng et al., 2006; PDB ID: 2GR7), the invasin passenger domain (Hamburger et al., 1999; PDB ID: 1CWV) and the intimin β-barrel domain (Fairman et al., 2012; PDB ID: 4E1S). The helix inside the FhaC β-barrel was generated from a neighboring asymmetric unit in the crystal lattice. Structures of polypeptides associated with the type Vd pathway are not currently available.

It should be emphasized that much remains to be learned about the mechanism(s) by which polypeptides are secreted by the type V secretion pathways (especially the type Vd-e pathways, which have been the subject of only a few studies to date). Nevertheless, the striking structural similarities (and presumably evolutionary relationships) among these pathways suggest that some aspects of the assembly process are conserved. In this review I will focus on recent studies that have advanced our understanding of the biogenesis of classical autotransporters, but I will also allude to studies on other type V pathways that have yielded potentially relevant insights.

A changing view of the autotransporter secretion mechanism

It was originally proposed that classical autotransporters are completely autonomous secretion systems in which the substrate and the transporter are covalently linked. The idea that the C-terminus forms a translocation channel in the OM was first published in a classic study on the founding member of the autotransporter superfamily (the Neisseria gonorrhoeae IgA protease) and was supported by the observation that C-terminal deletions abolish the translocation of N-terminal segments (Pohlner et al., 1987). In the quarter century since the self-transport hypothesis was proposed, a variety of observations have been reported that are consistent with the idea that the C-terminus facilitates passenger domain translocation. Most notably, X-ray crystallography not only confirmed that the C-terminus forms a cylindrical β-barrel structure, but also demonstrated the presence of a polypeptide segment embedded in the β-barrel that, at least at first glance, seems to reflect the termination stage of the translocation reaction (Oomen et al., 2004; Barnard et al., 2007; van den Berg, 2010; Tajima et al., 2010). The recent finding that the mutation of conserved β-barrel domain residues that are required for efficient folding or membrane integration impairs passenger domain translocation also suggests that the β-domain plays a role in translocation (Pavlova et al., 2013; Leyton et al., 2014). Furthermore, the observation that the replacement of the β-barrel domain with the β-barrel of an unrelated OM protein abolishes translocation suggests that the β-barrel domain does not simply target the passenger domain to an unlinked transporter (Sauri et al., 2011). Taken together with compelling evidence that translocation proceeds in a C-to-N-terminal direction (Ieva and Bernstein, 2009; Junker et al., 2009), it is easy to envision a scenario in which a hairpin formed at the C-terminus of the passenger domain is inserted into the β-barrel and maintained while N-terminal segments are gradually threaded through the pore (Fig. 2).

Fig. 2.

Early model for autotransporter secretion. It was proposed over 25 years ago that the passenger domain of an autotransporter (blue) is secreted through a channel formed by the covalently linked β-barrel domain (red), which functions as a completely autonomous transporter (Pohlner et al., 1987). In this model, the β-barrel domain first inserts into the OM and folds into a stable structure (step i). Subsequently the C-terminus of the passenger domain forms a hairpin inside the pore of the β-barrel domain (step ii). The passenger domain is then gradually threaded through the pore in a C-to N-terminal fashion in an unfolded conformation (step iii) until the entire polypeptide is secreted (step iv). Translocation is driven at least in part by the folding of the passenger domain in the extracellular space. Finally, the passenger domain of many autotransporters is released by a proteolytic cleavage (step v). In some versions of this model, the passenger domain forms a hairpin inside the β-barrel prior to its insertion into the OM.

Despite the attractiveness of the self-transport hypothesis, this elegant model has been challenged in the last decade by several observations. Perhaps most significantly, crystallographic analysis has shown that the β-barrel pore is ∼ 10 Å in diameter and is therefore only wide enough to accommodate a single α-helix or a hairpin in an extended conformation (Oomen et al., 2004; Barnard et al., 2007; van den Berg, 2010; Tajima et al., 2010; Zhai et al., 2011). There is considerable evidence, however, that polypeptides that have tertiary structure are secreted by the autotransporter pathway. At least some ∼ 10–20 kD polypeptides whose structure is stabilized by calcium ions or the formation of disulfide bonds in the periplasm (e.g., a fragment of the Bordetella pertussis Rtx toxin, the cholera toxin B subunit, single chain Fv fragments) are secreted effectively when they are fused to autotransporters (Veiga et al., 2004; Skillman et al., 2005; Kang’ethe and Bernstein, 2013a). The growth of cells in minimal medium appears to increase secretion efficiency, possibly because misfolding of the fusion proteins is reduced under slow growth conditions. While it has not been proven that these polypeptides are fully folded prior to their translocation across the OM, a systematic analysis of the secretion of peptides that vary in length and structural complexity indicated that the active pore is ∼ 17–20 Å in diameter (Sauri et al., 2012). Furthermore, a subset of native passenger domains have been shown to undergo disulfide bonding in the periplasm, including a Chlamydia autotransporter that is highly enriched in cysteine residues (Skillman et al., 2005; Swanson et al., 2009; Leyton et al., 2011). In this regard it should be noted that native type Ve passenger domains that are associated with equally narrow β-barrels also contain a disulfide bond (Hamburger et al., 1999). Finally, several lines of evidence strongly suggest that the linker that is embedded inside the β-barrel forms an α-helix before the completion of translocation and imply that the active transport channel contains at least an α-helical segment and an extended polypeptide (Ieva et al., 2008; Peterson et al., 2010). Taken together, the available data imply that the β-barrel domain would have to be in an expanded or open conformation during translocation. Molecular dynamics simulations, however, have confirmed that fully folded autotransporter β-barrels are rigid structures (Khalid and Sansom, 2006; Tian and Bernstein, 2010).

The apparent incompatibility between the size of the autotransporter β-barrel pore and the structure of secreted polypeptides suggests that the mechanism of autotransporter translocation is more complicated or even fundamentally different than originally envisioned. Indeed the results of recent studies on autotransporter biogenesis that have analyzed assembly intermediates are consistent with this possibility. Most notably, site-specific photo-cross-linking experiments performed with stalled translocation intermediates of the Escherichia coli autotransporter EspP have shown that passenger domain residues that lie near the site of stalling are in close proximity to BamA, a member of the Omp85 superfamily (Ieva and Bernstein, 2009). BamA is a highly conserved component of the barrel assembly machine (Bam) complex, a heterooligomer that catalyzes the integration of β-barrel proteins into the OM (Voulhoux et al., 2003; Wu et al., 2005; Hagan et al., 2010) and that is required for autotransporter assembly in vivo (Voulhoux et al., 2003; Jain and Goldberg, 2007; Rossiter et al., 2011). The similarity of BamA to the TpsB proteins that mediate type Vb secretion raises the intriguing possibility that it plays a role in the passenger domain translocation reaction. Additional photo-cross-linking studies have shown that the EspP β-barrel domain is not only in contact with the Bam complex during the entire translocation reaction, but also undergoes a positional or conformational shift (Ieva et al., 2011). Chemical cross-linking experiments have indicated that Hbp is likewise in proximity to the Bam complex when the translocation of the passenger domain stalls and have shown that the β-barrel domain is incompletely folded at this stage (Sauri et al., 2009). These are key observations because they strongly suggest that the assembly of the β-barrel and the translocation of the passenger domain are co-ordinated processes and challenge the self-transport model, which holds that the β-barrel is completely assembled before translocation begins.

The experiments described earlier together with other efforts to dissect autotransporter assembly into a sequence of discrete steps have led to a revised model for the mechanism of passenger domain translocation (Pavlova et al., 2013). Based on the observation that the EspP linker region is refractory to proteolysis and chemical modification at an early stage of assembly and is required for the membrane integration of the β-barrel domain (Ieva et al., 2008), the model posits that the β-barrel domain begins to fold in the periplasm and incorporates the C-terminus of the passenger domain in a hairpin conformation (Fig. 3, step i). Cross-linking data suggest that the β-barrel domain interacts with the molecular chaperone Skp at this stage (Ieva et al., 2011; Pavlova et al., 2013). Skp is a heterotrimer with jellyfish-like projections that define a cavity large enough to accommodate at least a ∼ 20 kD polypeptide (Korndörfer et al., 2004; Walton and Sousa, 2004; Walton et al., 2009). Subsequently the autotransporter is targeted to the OM, where BamA presumably recognizes a conserved C-terminal motif found in most OM proteins (Robert et al., 2006). At this stage crosslinks between specific residues of the EspP β-barrel domain and BamB and BamD are first detected (Fig. 3, step ii). (Pavlova et al., 2013). The stereospecificity of the cross-linking is consistent with the idea that the β-barrel domain is in a single conformation and suggests that the membrane integration of the β-barrel involves the concerted action of multiple Bam complex subunits. The interaction between the β-barrel domain and the Bam complex precedes the initiation of translocation, which requires the β-barrel to reach a specific (but poorly defined) stage of assembly (Fig. 3, step iii) (Pavlova et al., 2013). Based on evidence that the BamA β-barrel has a unique ability to open laterally that first emerged from structural studies (Noinaj et al., 2013; 2014), it is tempting to speculate that an open form of BamA functions as a component of the passenger domain translocation channel. Indeed the hypothesis that the autotransporter β-barrel domain and BamA transiently form a relatively large hybrid channel would account for both the cross-linking of stalled translocation intermediates to BamA and the secretion of folded polypeptides. At the very least, the Bam complex likely maintains the autotransporter β-barrel in an expanded conformation at this stage. Translocation then appears to proceed by a progressive transfer of passenger domain segments from the molecular chaperone SurA, which binds to the first POTRA domain of BamA (Bennion et al., 2010), to the POTRA domains and then to the transport channel (Pavlova et al., 2013) (Fig. 3, steps iii–iv). Indeed exoproteins appear to follow a similar path from the TpsB POTRA domains to the β-barrel pore during type Vb secretion (Baud et al., 2014). The passenger domain is released by proteolytic cleavage following the completion of translocation (Fig. 3, step v), and in a final step, the β-barrel domain dissociates from the Bam complex (Ieva et al., 2011; Fig. 3, step vi).

Fig. 3.

Revised model for autotransporter secretion. Evidence derived from a variety of recent experiments plus the discovery that the membrane insertion of β-barrel proteins is catalyzed by the Bam complex has led to a new model for autotransporter biogenesis (Pavlova et al., 2013). In this model, the β-barrel domain (red) begins to fold in the periplasm and incorporates the C-terminus of the passenger domain (blue) in a hairpin conformation (step i). At this stage, the β-barrel domain interacts with the molecular chaperone Skp. Subsequently, the β-barrel domain interacts with BamA, BamB and BamD in a stereospecific fashion (step ii). Passenger domain translocation is initiated only after the β-barrel reaches a specific stage of assembly (step iii). Translocation involves the progressive movement of the passenger domain from the molecular chaperone SurA to the BamA POTRA domains to the transport channel (step iv). During the translocation reaction the BamA β-barrel may be in an open conformation and may form part of the transport channel. As in the original model, translocation is driven at least in part by the folding of the passenger domain in the extracellular space. After the entire passenger domain is secreted (step v) the β-barrel domain folds completely and is released from the Bam complex (step vi). The passenger domain of EspP and related autotransporters is released in an unusual intrabarrel cleavage reaction that appears to occur prior to the release of the β-barrel domain from the Bam complex (Dautin and Bernstein, 2007; Dautin et al., 2007; Ieva et al., 2011), but in other cases proteolytic maturation likely occurs after the β-barrel domain is fully assembled. In Escherichia coli the Bam complex consists of five subunits, but BamC and BamE have been omitted for clarity.

Although many aspects of this model remain to be tested or confirmed, it appears to be more consistent with the available evidence than the self-transport model, which, after all, was proposed long before the Bam complex was discovered. At present, it is not entirely clear how the Bam complex might promote the transport of a polypeptide across the OM, while simultaneously promoting the assembly of the covalently linked β-barrel. In this regard it should be noted that two recent studies have yielded striking evidence that the E. coli Bam complex catalyzes the insertion of a small lipoprotein called RcsF into the β-barrel of OM proteins during their assembly (Cho et al., 2014; Konovalova et al., 2014). While it is not known if the RcsF threading reaction is mechanistically related to passenger domain translocation, these studies nevertheless underscore the versatility of the Bam complex and its ability to act on multiple polypeptides (or independent polypeptide domains) in a co-ordinated fashion.

Role of additional factors in autotransporter assembly

A recent study showed that the E. coli autotransporter Ag43 can be assembled into proteoliposomes containing only the Bam complex after its translocation through the Sec machinery of spheroplasts (Norell et al., 2014). Moreover, another study showed that the purified Bam complex and SurA are both necessary and sufficient to mediate the assembly of EspP in vitro (Roman-Hernandez et al., 2014). Although the efficiency of the in vitro reaction (which is only ∼ 10–20%) appears to be limited by the ability of the substrate to remain assembly competent, the results do not rule out the possibility that additional factors participate in autotransporter assembly in vivo. Indeed an unidentified ∼ 30 kD protein that may interact with the passenger domain during translocation has been detected in cross-linking experiments (Ieva and Bernstein, 2009). In addition, TamA, a member of the Omp85 superfamily, and TamB, a large IM-anchored protein that interacts with TamA, have been implicated in the biogenesis of the Citrobacter rodentium autotransporter p1121 and Ag43 (Selkrig et al., 2012). When TamA is reconstituted into a planar lipid environment and mixed with TamB, a conformational change in the POTRA domains can be observed in the presence of Ag43 (Shen et al., 2014). Although the exact function of TamA/TamB is unclear, these proteins do not play a detectable role in the biogenesis of EspP and Hbp (Sauri et al., 2009; Kang’ethe and Bernstein, 2013a) and therefore appear to facilitate the assembly of only a subset of autotransporters.

Presumably autotransporters must remain in an assembly-competent conformation during their passage through the periplasm. Several periplasmic chaperones that prevent misfolding or play other general roles in envelope protein biogenesis including DegP, FkpA, Skp and SurA have been shown to interact with autotransporters in vivo and/or in vitro (Ieva and Bernstein, 2009; Ruiz-Perez et al., 2009; 2010; Ieva et al., 2011). There is also evidence that the elimination of specific periplasmic chaperones impairs autotransporter assembly in vivo, at least under some experimental conditions (Purdy et al., 2007; Ruiz-Perez et al., 2009). Interestingly, a subset of classical autotransporters, trimeric autotransporters and TpsA proteins contain unusually long signal peptides that play an important role in maintaining the translocation-competence of the passenger domain in the periplasm (Szabady et al., 2005). The effect of the long signal peptide appears to be mediated at least in part through an interaction with YidC, an IM assembly factor that is associated with the Sec machinery (Jong et al., 2010). While no other factors that maintain the translocation-competence of the passenger domains of classical autoransporters have been identified, the efficient assembly of a subset of trimeric autotransporters requires the activity of a trimeric IM lipoprotein that is encoded in the same operon (Grin et al., 2014).

Energetics of autotransporter secretion

Like the transport of any polypeptide across a membrane barrier, the translocation of passenger domains across the OM must either be thermodynamically favorable or (as is often the case) driven by an input of energy. The periplasm is devoid of adenosine triphosphate (ATP), however, and there is no electrochemical gradient across the OM that can be harnessed to drive translocation. It is conceivable that an IM protein that interacts with passenger domains or with the OM transport machinery derives energy from cytoplasmic ATP or the membrane potential across the IM to drive the translocation reaction. The observation that the Bam complex and SurA are sufficient to promote passenger domain translocation into proteoliposomes, however, appears to be inconsistent with this hypothesis and implies that autotransporter assembly does not require an exogenous energy source (Roman-Hernandez et al., 2014).

To account for the energetics of autotransporter secretion, it was proposed over 20 years ago that small segments of the passenger domain passively diffuse across the OM and then fold sequentially. Folding would trap the passenger domain in the extracellular space and provide the driving force for the translocation reaction (Klauser et al., 1992). Indeed this hypothesis is extremely appealing given that virtually all passenger domains have a modular β-helical structure that might fold in a stepwise fashion. Even the passenger domain of a Pseudomonas autotransporter that forms an atypical globular structure contains secondary structure elements that are predicted to fold sequentially (van den Berg, 2010). Furthermore, there is now significant evidence in support of the ‘vectorial folding’ model. Studies that have analyzed the refolding of passenger domains in vitro or the effect of mutations on passenger domain secretion in vivo have provided evidence that the folding of a conserved ∼ 20–25 kD C-terminal passenger domain ‘stable core’ segment plays an important role in driving the secretion reaction (Velarde and Nataro, 2004; Junker et al., 2006; Renn and Clark, 2008; Peterson et al., 2010; Soprova et al., 2010). The results of one of these studies (Peterson et al., 2010) show that mutations that impair the folding of the C-terminal segment in the extracellular space do not affect its exposure on the cell surface, but strongly delay the secretion of the remainder of the passenger domain. In the case of an autotransporter that lacks the C-terminal stable core, an N-terminal segment appears to destabilize the entire passenger domain to prevent it from folding prematurely in the periplasm (Besingi et al., 2013). Atomic force microscopy experiments have also shown that a β-helical TPS exoprotein unfolds in a stepwise, hierarchical process (Alsteens et al., 2013). Finally, the results of kinetic simulations are consistent with the idea that passenger domain translocation is driven by the free energy of folding in the extracellular environment (Drobnak et al., 2015).

Despite the attractiveness of the vectorial folding hypothesis, several observations have suggested that the energy for autotransporter secretion is not derived exclusively from stepwise passenger domain folding. Multiplepoint mutations introduced into the middle of the EspP passenger domain have been shown to destabilize the protein, but impair translocation only modestly (Kang’ethe and Bernstein, 2013b). Furthermore, when the native EspP passenger domain was replaced with the intrinsically disordered receptor domain (RD) of the B. pertussis CyaA toxin, the heterologous polypeptide was secreted as rapidly and efficiently as the native passenger domain (Kang’ethe and Bernstein, 2013a). Interestingly, the RD is exceptionally acidic, and the conversion of multiple acidic residues to neutral or basic residues stalled translocation. This observation, together with the discovery that native passenger domains are generally acidic, raises the possibility that charge interactions and/or the Donnan potential across the OM (Stock et al., 1977) play a role in driving the translocation reaction. Finally, the secretion of small folded polypeptide domains by the autotransporter pathway that has been suggested by several studies would clearly require an alternative energy source. It is conceivable that residual energy generated during the integration of the β-barrel domain into the OM (perhaps the same energy that drives the exposure of the extracellular loops of all β-barrel proteins) can be harnessed for passenger domain translocation. While this hypothetical energy source may not drive the translocation of the entire 675 residue RD or folded polypeptides, it may nevertheless play an important role in the translocation of C-terminal segments of native passenger domains.

Recent advances in autodisplay

Autotransporters have long been used as delivery systems to display peptides and proteins of interest on the cell surface (Jose and Meyer, 2007). The use of this so-called ‘autodisplay’ technology has been exploited for library screening, industrial applications such as biocatalysis and bioremediation, and vaccine development. In many cases N-terminally truncated derivatives of autotransporters have been used for autodisplay, and it was recently shown that the β-barrel domain plus a short linker segment are sufficient to present non-native polypeptides on the cell surface (Sevastsyanovich et al., 2012). Although most autodisplay work has been conducted in E. coli, a few research groups have validated the use of autotransporter-based surface display in alternative organisms (Tozakidis et al., 2015). Recent studies have extended the basic concept of autodisplay and shown that polypeptides fused to autotransporters can also be presented on the surface of OM vesicles and bacterial ghosts (Daleke-Schermerhorn et al., 2014; Hjelm et al., 2015). Taken together, the numerous examples of autodisplay that have appeared in the literature demonstrate that a remarkable diversity of polypeptides can be secreted effectively by the autotransporter pathway. Hydrolases, esterases, metallothioneins, fluorescent proteins, albumin-binding domains derived from protein G and anticalins (engineered proteins built on a β-barrel scaffold) are among the many different proteins that have been displayed on the cell surface (Jose and Meyer, 2007; Binder et al., 2010; Sevastsyanovich et al., 2012; Fleetwood et al., 2014). These studies have reinforced the idea that secretion is heavily influenced by the folded status of the passenger domain. Nevertheless, it seems likely that at least a few of the many different polypeptides that are secreted by autodisplay fold partially in the periplasm and that those polypeptides that remain unstructured in the periplasm fold at very different rates in the extracellular environment. Thus, the existence of the autodisplay phenomenon provides yet another reason to suspect that autotransporters are not autonomous secretion systems that rely solely on vectorial folding to drive protein translocation.

One particularly notable recent advance is the development of a structurally informed platform for autodisplay. This approach is based on the use of crystal structures to locate peripheral segments of passenger domains that do not contribute to the formation of the β-helical core. At least in the case of Hbp and closely related autotransporters, these segments are dispensable for secretion (Jong et al., 2012). Interestingly, the secretion of chimeras in which up to three of these segments have been replaced with heterologous polypeptides simultaneously has been observed (Jong et al., 2012; 2014). This work suggests that it may be feasible to expand the utility of autodisplay as further insights into the structure and biogenesis of autotransporters emerge and to use autodisplay technology to develop multivalent vaccines.

Concluding remarks

Despite several important recent advances, the mechanism of autotransporter secretion remains poorly understood. To obtain a more complete picture of autotransporter assembly, it will almost certainly be necessary to elucidate the function of the Bam complex. The fact that a broad range of polypeptides can be secreted by autodisplay or are natural substrates of type V pathways – all of which rely directly or indirectly on the function of BamA or BamA-like proteins – suggests that the Omp85 family of transporters is highly versatile. In addition to identifying similarities and differences between the autotransporter pathway and other type V secretion pathways, it will be interesting to determine the relationship between the assembly of autotransporters and the assembly of other types of OM proteins. Indeed it is conceivable that the β-strands of many OM proteins enter the BamA β-barrel sequentially and insert into the OM through the lateral gate, as recently proposed (Noinaj et al., 2013; 2014), whereas autotransporter β-barrel domains partially fold in the periplasm and insert into the OM by a different mechanism that involves interactions with BamB and BamD. In any case, it is striking that the extracellular loops that connect the β-strands of bacterial β-barrel proteins are usually short. Loop sizes do not appear to exceed ∼ 75 residues even in cases where a large amount of polypeptide must be exposed on the cell surface (Noinaj et al., 2012). Furthermore, insertions of only 30–50 residues are tolerated in the short loops (< 20 residues) of the E. coli porin PhoE (Janssen and Tommassen, 1994). These observations suggest a mechanistic constraint that confines large secreted polypeptides to discrete N-or C-terminal domains. With respect to the energetics of secretion, it should be noted that structural studies on the chaperone/usher and type VIII pathways have suggested that transporters can drive translocation across the OM by defining a low-energy pathway or using an entropy-based diffusion mechanism (Geibel et al., 2013; Goyal et al., 2014). While these studies may not be directly relevant to the autotransporter pathway, they illustrate the diversity of energy sources that can be tapped in the absence of ATP.