Structural comparison of the transport units of type V secretion systems

Introduction

The secretion of proteins is crucial not only for pathogenic gram-negative bacteria, but also a wide variety of enzymes are secreted for utilization of exogenous nutrients. During evolution, different secretion mechanisms developed in order to traverse the inner and outer membrane barriers; they include the type I–VIII secretion systems (T1SS–T8SS) and the chaperone-usher (CU) pathway (Desvaux et al., 2009). Among these, T1SS, T3SS, T4SS, T6SS, and T7SS are composed of protein complexes spanning both the inner and the outer membrane, promoting secretion in a single step (Rego et al., 2010). In contrast, T2SS, T5SS, T8SS, and the CU pathway are dependent on the Sec or the Tat translocon to cross the inner membrane. The final translocation is mediated by specific secretion mechanisms resulting either in the release into the extracellular space or the surface display of effector proteins (Dautin and Bernstein, 2007).

Among the two-step secretion systems, the T5SS is probably the simplest process for the export of adhesins, enzymes, toxins, and other virulence factors with varying sizes and structures (Henderson et al., 2004). Typically, members of T5SS contain a signal peptide at the N-terminus mediating Sec-dependent transport across the inner membrane and a passenger domain that exerts biological activity in the extracellular space. A linker domain connects the passenger and the β-domain, which forms a β-barrel with a hydrophilic pore in the outer membrane. Different subgroups were identified with respect to special characteristics, such as polypeptide-transport-associated (POTRA) domains, oligomerization state, and domain arrangement.

Subgroup Va is defined as the group of ‘classical’ autotransporters and includes the monomeric autotransporters containing all information required for translocation across both membranes in a single polypeptide chain (Jose et al., 1995; Desvaux et al., 2004). The two-partner-secretion (TPS) system is assigned to subgroup Vb. In this subgroup, the passenger and the transporter are expressed as two different proteins referred to as TpsA and TpsB (Clantin et al., 2007). Here, the transporter TpsB contains two POTRA domains for recognition of its corresponding passenger TpsA. A third subgroup (Vc) covers the trimeric autotransporters, which are composed of three identical subunits (Cotter et al., 2005). Each subunit contributes four β-strands to the resulting 12-stranded β-barrel and thus deploying three passenger domains in the extracellular space. In 2010, Salacha et al. discovered a new type of autotransporter protein in Pseudomonas aeruginosa, termed PlpD. This protein shares the common features of classical autotransporters (Va) and TpsB proteins (Vb) in one polypeptide chain. In PlpD, the passenger domain is connected to the transport domain by only one POTRA domain. Therefore, this protein was assigned to the new subgroup Vd, termed fused TPS systems (Leo et al., 2012). Recently, the crystal structures of two adhesins, intimin of Escherichia coli and invasin of Yersinia spp., were described (Fairman et al., 2012). These structures revealed architectures similar to those of classical autotransporters of subgroup Va but in reverse order. Here, the N-terminal part of the protein chain forms a β-barrel domain and the C-terminal part harbors the passenger domain. Consequently, intimin and invasin were assigned to a new subgroup Ve, comprising inverted autotransporters (Leo et al., 2012).

Despite extensive biochemical and structural data, the exact mechanism of the outer membrane transport of T5SS remains unclear. Different pathways have been discussed, including the hairpin model (Henderson et al., 2004) and the BamA/Omp85 model (Voulhoux et al., 2003; Robert et al., 2006), which is based on an evolutionary ancient ensemble of periplasmic and outer membrane proteins, e.g., the Bam complex, found to have homologues even in mitochondria (Tommassen, 2010).

The type V secretion mechanism

In T5SS, the translocation of passenger domains across both membranes of the gram-negative envelope is mediated in a two-step mechanism (Rego et al., 2010). To cross the inner membrane, the signal peptide of the protein chain is recognized by the Sec translocon (Figure 1, step 1) (Desvaux et al., 2004) and translocated therefore in an unfolded form. During translocation, the signal peptidase cleaves off the signal peptide. According to the BamA/Omp85 model (Voulhoux et al., 2003), the autotransporter protein (Figure 1, step 2), devoid of the signal peptide, adopts a stable intermediate in the periplasm by interaction with periplasmic chaperones, such as Skp and SurA, and the outer membrane Omp85/Bam complex (Brandon and Goldberg, 2001; Skillman et al., 2005; Ieva and Bernstein, 2009; Ruiz-Perez et al., 2009; Leyton et al., 2012). The β-domain integrates into the outer membrane in some way, forming a β-barrel with a hydrophilic pore occupied by a linker domain through which the passenger domain is apparently translocated to the cell surface (Figure 1, step 3).

Figure 1

Schematic overview of type V secretion mechanism.

Brown: Sec translocon; purple: signal peptidase; yellow: signal peptide; green: β-barrel; blue: passenger domain; red: linker domain; violet: periplasmic chaperones; orange: Omp85. For simplicity, we omitted the ribosome as point of initial protein synthesis of the autotransporter. On the basis of unpublished data (JJ), one can hypothesize that some sort of co-translation mechanism in the Sec-mediated translocation of autotransporters across the inner membrane might exist. The peptide chain is transported across the inner membrane by the Sec translocon (step 1). In the periplasm, the intermediate adopts a stable conformation while interacting with periplasmic chaperones and BamA (step 2), and the passenger domain is finally translocated across the outer membrane (step 3).

Structural and functional diversity of T5SS passengers

Passengers of T5SS are involved in major processes of gram-negative bacteria, such as virulence, bacterial motility, or nutrient acquisition and function as adhesins, enzymes, or toxins (Henderson et al., 2004; Leo et al., 2012; Jacob-Dubuisson et al., 2013). Despite varying in size with a maximum up to 500 kDa, most of the T5aSS and T5bSS passengers form extended β-helices after translocation, such as the passengers of pertactin, EspP, Hbp, VacA, and FHA. One exception is the passenger domain of EstA (Va), which does not contain β-helices (van den Berg, 2010). Each passenger undergoes one of three possible post-translocational treatments: (i) it can either remain covalently linked to the transport unit, (ii) or can be cleaved but remain attached to the cell surface, or (iii) the passenger is released into the extracellular space after cleavage. Trimeric autotransporters (Vc), e.g., Hia and YadD, contain α-helical passenger domains, which remain covalently linked after translocation across the outer membrane and function as adhesins (Cotter et al., 2005). For PlpD, which is a member of the recently defined subgroup of fused autotransporters (Vd), no structures are available. However, the passenger of the patatin-like protein D (PlpD) was predicted to fold similarly to the potato patatin Pat17 containing α-helices, β-sheets, and connecting loops (Rydel et al., 2003; Salacha et al., 2010). The passengers of T5eSS, e.g., intimin and invasin, are released into the extracellular space and fold into several Big (bacterial immunoglobulin-like) domains and a C-type lectin-like domain at the C-terminus. Intimin and invasin mediate adhesion or the rearrangement of the host cytoskeleton (Hamburger et al., 1999; Batchelor et al., 2000; Luo et al., 2000).

In summary, a structural comparison of only the passenger domains cannot be used to differentiate between the subgroups of T5SS. Thus, it is necessary to consider the characteristics of the whole autotransporter.

The architecture of type V transport units

In contrast to the passengers translocated by T5SS, the transport units share high structural similarity (Henderson et al., 1998; Thanassi and Hultgren, 2000; Meng et al., 2006; Leyton et al., 2012). In recent years, several crystal structures from all T5SS subgroups, except Vd, were solved (Table 1). All structures of NalP (Va), FhaC (Vb), Hia (Vc), and intimin (Ve, Figure 2) compared here (Figures 2 and 3) consist only of the transport unit, except for FhaC (Oomen et al., 2004; Meng et al., 2006; Clantin et al., 2007; Fairman et al., 2012). For NalP and intimin, only the linker and the β-barrel domains were cloned, expressed, and crystallized. The structure of Hia was solved for a construct where the N-terminal part was deleted.

Figure 2

Overview of transport units of type V secretion systems.

Green: β-barrel; blue: POTRA domains; orange: periplasmic α-helix; red/magenta/brown: α-helices and unstructured peptide chains traversing the pore of the β-barrel. (A) Lateral view of crystal structures of the transport units in ribbon presentation. (B) Top view from periplasmic side of crystal structures of the transport units in ribbon presentation.

Figure 3

Overview of transport units of type V secretion systems with surface presentation of domains traversing the β-barrel pore.

Green: β-barrel; blue: POTRA domains; orange: periplasmic α-helix; red/magenta/brown: α-helices and unstructured peptide chains traversing the pore of the β-barrel. (A) Lateral view of crystal structures of the transport units as ribbon diagrams with surface presentation of domains traversing the pore. (B) Top view from the periplasmic side of crystal structures of the transport units as ribbon diagrams with surface presentation of domains traversing the pore.

All type V secretion transport units localize to the outer membrane of gram-negative bacteria. These are composed of a linker and an amphipatic β-barrel domain, which is a common feature of proteins spanning the outer membrane of gram-negative bacteria, mitochondria, and chloroplasts (Schulz, 2002). The amphipatic β-barrel does not appear as a structural component in other proteins, with the peculiar exception of intracellular fatty acid-binding proteins in animal cells (Smathers and Petersen, 2011). The autotransporters β-barrels consist of 12 antiparallel β-strands in the case of NalP, Hia, and intimin (Table 1, Figure 2), and 16 strands for FhaC. The β-strands of the β-barrels are connected by short turns on the periplasmic side and larger extracellular loops. In the case of Hia, the connecting loops have the same length on both sides of the β-barrel (Figure 2). The central pores of the β-barrels are hydrophilic and occupied by the α-helical linker in the case of NalP, FhaC, and Hia. For intimin, the linker domain traverses the pore in an extended conformation (Fairman et al., 2012). The diameters of these pore domains range between 10 and 20 Å, where Hia harbors the widest pore with three α-helical domains in contrast to NalP and FhaC with only one (Figure 2). The surface representation illustrates the dimensions of the linker domain relative to the pore (Figure 3). For NalP, the linker domain is located in the center of the pore, resulting in dense packing of this domain close to the wall of the β-barrel (Figure 3B). The linker domain of FhaC traverses the pore in a diagonal manner owing to interactions of charged clusters of amino acids in the pore and corresponding amino acids in the linker domain, creating an asymmetric channel (Clantin et al., 2007). The pore of Hia is fully occupied by the three α-helical linker domains (Figure 3). The linker domain of intimin strongly interacts with one side of the pore, which also results in an asymmetric architecture similar to the one observed for FhaC.

As shown by Struyvé et al. (1991), outer membrane proteins of gram-negative bacteria share a common C-terminal consensus sequence containing a phenylalanine or tryptophan at -1 and a hydrophobic residue at positions -3, -5, -7, and -9 (Table 2), which is recognized by Omp85 (Robert et al., 2006). Moreover, it was demonstrated for PhoE, an E. coli porin, that the last C-terminal amino acid in the signature sequence (a phenylalanine) is essential for the correct assembly of β-barrels in the outer membrane. Compared with the five conserved amino acids of the consensus signature sequences postulated by Struvyé et al. (1991), NalP, FhaC, and intimin contain all five conserved amino acids in the last β-strand of their β-barrel (Table 2). The same is true for each subunit of the homotrimeric Hia. Loveless and Saier (1997) extended this signature sequence to the last 18 amino acids of the last β-strand by sequence alignment of different autotransporters. For each position, several amino acids were identified, resulting in a considerable variation of the signature sequence except for the terminal phenylalanine or tryptophan (Table 2). The highest agreement with the consensus signature sequence stated by Loveless and Saier (1997) shows the homotrimeric autotransporter Hia, which matches 13 amino acids, followed by FhaC with 12, NalP with 10, and finally intimin with 9 conserved amino acids (Table 2). These results strongly support the thesis that autotransporters are integrated in the outer membrane by a common machinery, which is also responsible for the integration of other outer membrane proteins.

Despite the high similarities of the overall architecture, the transport units contain special features that differ between the subgroups, such as POTRA domains, oligomerization state, and domain arrangement. For example, the TpsB protein FhaC (Vb) contains two POTRA domains (Figure 2, blue subdomains), which are responsible for the recognition of their dedicated passenger protein, filamentous hemagglutinin (FHA) (Clantin et al., 2007). Each POTRA domain consists of three β-strands and one short α-helix. Intimin (Ve) contains a periplasmic α-helix, which is located between the linker and the β-barrel domain (Figure 2, orange subdomains) (Fairman et al., 2012). The exchange of this helix for a short glycine-glycine linker had no influence on the heat modifiability and proteinase K sensitivity, indicating a properly folded β-barrel and an unaffected translocation of the passenger (Fairman et al., 2012). This helix could possibly function as a plug of the pore from the periplasmic side that avoids leakage of the cells. In contrast to members of all other subgroups, Hia (Vc) translocates its high identical passenger domains within only one β-barrel (Figure 2).

Next, we were interested if it is possible to distinguish between transport units of different subgroups of T5SS by a structural comparison. Therefore, the global root mean square deviation (RMSD) was calculated, which is a relative measurement for describing overall structural similarity. For that, the transport unit structures of each subgroup were compared to a reference structure of the same subgroup (Table 3). Among the monomeric autotransporters, EspP, a serine protease autotransporter (SPATE) from E. coli, showed the highest (RMSD of 1.20 Å) and BrkA from Bordetella pertussis the lowest (RMSD of 2.57 Å) structural similarity to NalP with respect to the C_α atoms. Regarding the function of EspP, which like NalP is assigned to the SPATEs, the high structural similarity is not surprising. The structures of YadA and Hia, which both function as adhesins, are very similar over a length of 273 C_α atoms (RMSD of 1.81 Å). The global RMSD for invasin compared with intimin is very low (0.69 Å), indicating a nearly identical structure over the whole length of the protein. Comparing structures of different subgroups, Hia and intimin to NalP, the global RMSDs of intimin (2.25 Å) and Hia (2.63 Å) are in the same range as for BrkA (Va) (2.57 Å). For FhaC, the structural identity is significantly lower (RMSD of 3.63 Å) owing to the different numbers of β-strands.

In consequence, the structural comparison of the transport units is not sufficient to differentiate between the subgroups of T5SS. For example, NalP and intimin share a higher structural similarity compared with NalP and BrkA, but belong to different subgroups, Va and Ve, respectively. Thus, a comprehensive analysis of the whole protein is required for the correct classification of each autotransporter.

In recent years, considerable experimental data were obtained in terms of the biogenesis of the monomeric autotransporter EspP from E. coli (Ieva and Bernstein, 2009; Ieva et al., 2011; Pavlova et al., 2013). In these studies, several conserved amino acids were analyzed by mutational and crosslinking experiments and by bioinformatic comparison to other autotransporters. To analyze if other autotransporters contain identical amino acids at the same position in the spatial arrangement as in EspP, a structural alignment of EspP with NalP, FhaC, Hia, and intimin was performed. Indeed, four of these already investigated amino acids of EspP (Pavlova et al., 2013) were also present in NalP and FhaC, including the C-terminal phenylalanine of the conserved signature sequence (see above; Figure 4, highlighted in blue). As depicted in Figure 4, G₁ and G₂ correspond to the glycine residues at the left and the right side of each β-barrel, to which the distance measurements listed in Table 4 refer. For EspP, heat modifiability tests revealed that mutation of G2(1066) or G2(1081) slightly interferes with the stability of the β-barrel (Pavlova et al., 2013). In consequence, the exposure of the passenger domain on the cell surface and the integration of the β-barrel into the outer membrane were delayed, as shown by crosslinking and protease accessibility tests. However, the release of the passenger domain was unaffected. The midbarrel residue Y(1125) was shown to interact with the periplasmic chaperone Skp and lipopolysaccharides from the outer membrane (Ieva et al., 2011). The role of G₁(1207) in the biogenesis of EspP has not yet been determined, but this residue seems to be important for the general translocation mechanism as it is highly conserved among autotransporters (see superfamily entry pfam03797 in the Conserved Domain Database, http://www.ncbi.nlm.nih.gov/cdd).

Regarding the spatial arrangement, the distances between the conserved residues in EspP and NalP are very similar to each other (Table 4). Owing to the 16-stranded β-barrel of FhaC, the distances of G₂(249) and G₁(359) to F(554) and Y(287) to G₂(249) differ from those observed in EspP and NalP, which are composed of 12-stranded β-barrels. Nevertheless, all of these amino acid side chains in each structure are exposed to the environment and thus are accessible for interactions in a comparable spatial arrangement (Figure 4). Regardless of the subgroup, these conserved amino acids could presumably play the same role in the biogenesis of T5SS.

Figure 4

Conserved amino acids among the transport units of EspP, NalP, and FhaC.

Green: transport units in ribbon presentation; blue: conserved amino acids in stick presentation. For a better overview, the POTRA domains of FhaC are hidden. The conserved amino acids are labeled with one-letter code and position number. G₁ and G₂ correspond to the distance measurements in Table 4.

Translocation across the outer membrane

In the past decade, different models for translocation of passenger domains across the outer membrane of gram-negative bacteria were postulated (Veiga et al., 2002; Oomen et al., 2004; Kostakioti and Stathopoulos, 2006). Most biochemical and structural data obtained by numerous investigations support two of these models: (i) the hairpin model and (ii) the Omp85 model (Figure 5).

Figure 5

Postulated mechanisms of type V passenger translocation across the outer membrane.

Green: β-barrel; blue: passenger domain; red: linker domain; orange: Omp85 complex; violet: periplasmic chaperones. According to the hairpin model (A), the linker domain forms a hairpin-like structure in the pore of the β-barrel (step A1) and the passenger domain is pulled through this pore (step A2). Finally, the passenger domain adopts its fully folded conformation at the cell surface (step A3). In the Omp85 model (B), the autotransporter interacts with BamA (step B1) and is subsequently integrated into the outer membrane in an ‘open’ conformation (step B2), allowing transport of partially folded passengers. After translocation, across the outer membrane, the passenger domain obtains its active conformation in the extracellular space (step B3).

Alternative model

Because neither the hairpin model nor the Omp85 model can describe a translocation mechanism consistent with all data derived from structural and biochemical analyses (Bernstein, 2007; Leyton et al., 2012), a synopsis of both appears to be worth looking at and has been proposed before (Benz and Schmidt, 2011; Ieva et al., 2011; Jose et al., 2012; Grijpstra et al., 2013; Pavlova et al., 2013).

The alternative model is a combination of both the hairpin and the Omp85 model (Figure 6), supported by recently obtained evidence from crosslinking and mutagenesis experiments, heat modifiability analysis, and protease accessibility tests with the monomeric autotransporter EspP (Ieva et al., 2011; Pavlova et al., 2013). As described below, we extend this model by the fact that a recently discovered translocation and assembly module (TAM) is involved in the translocation of autotransporters (Selkrig et al., 2012).

Figure 6

Alternative model mechanism of type V passenger translocation across the outer membrane.

Green: β-barrel; blue: passenger domain; red: linker domain; light orange: Omp85; yellow: BamB; dark orange: BamD; violet: Skp; purple: SurA; light gray: TamA; dark gray: TamB; P: POTRA domains. The autotransporter interacts with the periplasmic chaperone Skp and forms a stable intermediate with the linker domain incorporated in the pore of the prefolded β-barrel (step 1). This intermediate is targeted to the outer membrane interacting with TAM and the Bam complex (step 2). The prefolded β-barrel is integrated in the outer membrane while the passenger domain interacts with SurA and the POTRA domains of BamA (step 3), resulting in a stepwise transfer across the outer membrane (step 4). Finally, the passenger domain adopts its fully folded conformation at the cell surface (step 5).

After translocation across the inner membrane, the autotransporter partially folds in the periplasm, interacting with the periplasmic chaperone Skp (Figure 6, step 1). At this stage, the linker domain is incorporated inside an incompletely folded β-barrel, forming a hairpin-like structure (Ieva et al., 2008). Subsequently, the prefolded autotransporter is targeted to the outer membrane interacting with the Bam complex in a unique orientation (Figure 6, step 2, highlighted in yellow/orange) (Ieva et al., 2011; Pavlova et al., 2013). As the midbarrel residues are accessible to crosslinking and the hairpin formed by the linker domain is not detectable at the cell surface at this stage, the β-barrel does not seem to be fully integrated into the outer membrane (Pavlova et al., 2013). Recently, another secretion complex, the TAM, was discovered in different proteobacteria (Selkrig et al., 2012). This module consists of TamA, an Omp85-family protein located in the outer membrane, which interacts with TamB in the inner membrane, presumably through the POTRA domains of TamA. It was shown for Citrobacter rodentium that mutants lacking TamA or TamB do not express the putative autotransporter protein p1121. Furthermore, double mutants of E. coli lacking both TamA and TamB were not able to secrete the recombinant autotransporter adhesins Ag43 and EhaA. Therefore, cell-cell aggregation was abolished. Additionally, the precursor form of Ag43 accumulated in the periplasm (Selkrig et al., 2012). Moreover, crosslinking experiments with EspP in E. coli revealed that during the transient delay of passenger translocation, the passenger domain of EspP was crosslinked to an unidentified cellular protein with a molecular weight of 30–40 kDa (Ieva et al., 2011), which might correspond to TamA. Consequently, the TAM is assumed to interact with the passenger domain in the periplasm, providing an essential step in the translocation process presumably by stabilizing the periplasmic intermediate (Figure 6, step 2, highlighted gray). Probably, owing to a conformational change or repositioning of this intermediate, the β-barrel integration into the outer membrane is completed and the hairpin is exposed at the cell surface, leading to the initiation of passenger domain translocation (Figure 6, step 3). In this way, the TAM would dissociate from the passenger domain, while a stretch of 80 amino acids of the passenger domain, as shown by crosslinking experiments, interacts with several POTRA domains of BamA/Omp85 and the chaperone SurA in the periplasm (Pavlova et al., 2013). Thus, in the alternative model, the passenger is transferred in a stepwise process through a pore composed of the autotransporter in an ‘open conformation’ mediated by the Bam complex. During another repositioning of the β-barrel (Figure 6, step 4), the passenger domain is fully translocated and the assembly of the β-barrel is completed, resulting in the dissociation of the Bam complex and the surface exposure of the passenger domain (Figure 6, step 5).

Concluding remarks

In the last decade, more and more insight into the T5SS has been gained by structural, bioinformatic, and biochemical experiments. Comparing the crystal structures of transport units, which represent the state after translocation, the overall architecture of the type V transport units share high similarity also between the different subgroups, indicating a common translocation mechanism through the outer membrane. Different model mechanisms were discussed. Among those, the hairpin and the Omp85 model are supported by some but not all experimental results and clearly show the dependency of autotransporter proteins on additional factors, such as the Bam complex. Thus, the term ‘autotransporter’ is now redundant.

An alternative model was proposed, which is a combination of both the hairpin and the Omp85 model. Here, the fact that passengers contain disulfide bonds or modified amino acids, the interaction of the autotransporter with periplasmic chaperones and the Bam complex, and the formation of an α-helical structure in the pore before insertion of the β-barrel in the outer membrane are taken into account. We extended this model by a novel TAM, which was discovered recently and shown to be involved in the translocation of autotransporters across the outer membrane (Selkrig et al., 2012). To determine the exact function of the TAM and its interaction partner, further investigations are obviously required. Nevertheless, our understanding of the type V secretion mechanism improved during the last decade, resulting in a more and more complete picture.