Gram-negative bacteria use the type II secretion system to transport a large number of secreted proteins from the periplasmic space into the extracellular environment. Many of the secreted proteins are major virulence factors in plants and animals. The components of the type II secretion system are located in both the inner and outer membranes where they assemble into a multi-protein, cell-envelope spanning, complex. This review discusses recent progress, particularly newly published structures obtained by X-ray crystallography and electron microscopy that have increased our understanding of how the type II secretion apparatus functions and the role that individual proteins play in this complex system.
Extracellular secretion in gram-negative bacteria requires complex transport systems to move secreted proteins from the cytoplasm into the extracellular environment. Several highly specialized pathways have evolved for this purpose, including the type II secretion (T2S) system. Secretion via the T2S pathway occurs in two distinct steps. First, the proteins to be secreted are expressed with signal peptides that target them to either the Sec or Tat machinery for transport across the cytoplasmic membrane into the periplasm (Voulhoux et al., 2001; de Keyzer et al., 2003; Palmer et al., 2005). This is followed by removal of the signal peptides and recognition and translocation of the fully folded proteins across the outer membrane by components of the T2S system. The T2S pathway was first discovered in Klebsiella oxytoca, where it was found to be required for secretion of the starch-hydrolyzing lipoprotein, pullulanase (d'Enfert et al., 1987). Since then, Vibrio cholerae, Shiga-toxin producing Escherichia coli, Legionella pneumophila, Yersinia enterocolitica, Pseudomonas aeruginosa, Burkholderia pseudomallei, Erwinia chrysanthemi and Xanthomonas campestris have been shown to make use of this pathway for secretion and targeted delivery of toxins, proteases, cellulases and lipases (Dow et al., 1989; Filloux et al., 1990; Reeves et al., 1993; Sandkvist et al., 1997; DeShazer et al., 1999; Lathem et al., 2002; Iwobi et al., 2003; Rossier et al., 2004). Functional T2S pathways are also likely present in many additional species, as sequencing of bacterial genomes has identified T2S-like genes in numerous species (Cianciotto, 2005).
The number of genes identified as being essential for T2S is species-dependent and varies from 12 to 15. The homologous genes and their subsequent proteins have been designated in most systems by the letters A–O and S. P. aeruginosa is an exception, however, as the letters P–Z and A have been used instead. To indicate homology, in this review we will use the term T2S followed by the letter(s) of the gene product. For example, T2S:ER will be used to indicate that P. aeruginosa XcpR is the homolog to the T2S family protein E. For clarity with respect to previously published reviews, the general term T2S is equivalent to Gsp (Desvaux et al., 2004).
Based on data from many different experimental approaches, including subcellular localization and protein-protein interactions between individual components of the T2S system, it has been proposed that the T2S proteins assemble to form a complex that spans the entire gram-negative cell envelope. This multi-protein assembly includes a component in the cytoplasm, an inner membrane subcomplex that reaches into the periplasmic compartment and a secretion pore in the outer membrane (Fig. 1). The exact configuration of this apparatus and the mechanism, or order, of its assembly in the cell envelope is still unclear. In this review, we will highlight the very recent advances, particularly in structural data from X-ray crystallography and electron microscopy, which have increased our understanding of the complex machinery that makes up the T2S system. For broader reviews of T2S, see references Russel (1998), Sandkvist (2001a), Filloux (2004), and for reviews that highlight the involvement of T2S in pathogenesis and other bacterial processes, see references Sandkvist (2001b) and Cianciotto (2005).
The inner membrane complex
A functional T2S transport system requires the presence of T2S:ER, a cytoplasmic protein that is associated with the cytoplasmic membrane via the membrane proteins T2S:LY and T2S:MZ. T2S:ER proteins are members of a large family of secretion nucleoside triphosphates, which are thought to power bacterial protein secretion through NTP hydrolysis (Planet et al., 2001). T2S:ER proteins are multi-domain proteins (Fig. 2a), and while they all contain domains N1 to C2, domain N0 is only present in a few homologs including XpsE and XcpR in Xanthomonas campestris and Pseudomonas aeruginosa, respectively. All T2S:ER proteins have typical ATP-binding sites, called Walker A boxes, commonly present in ATPases, ATP synthases and kinases. Recently it was demonstrated that purified EpsE, the T2S:ER family member from Vibrio cholerae, is a Mg2+-dependent ATPase (Camberg & Sandkvist, 2005). A mutation introduced in the Walker A ATP-binding motif reduces the specific ATPase activity in vitro by several fold and results in a protein that is unable to support secretion in vivo (Sandkvist et al., 1995; Camberg & Sandkvist, 2005). The rate of ATP hydrolysis by purified EpsE is similar to those observed for other secretion ATPases, which range from approximately 0.5 to 15 nmol−1 min−1 mg−1 protein.
Interestingly, EpsE displays cooperative ATPase activity, and purified preparations of EpsE contain a small fraction of oligomers, specifically hexamers, with increased specific activity (Camberg & Sandkvist, 2005). This may suggest that the functional form of T2S:ER proteins in vivo is hexameric. Several members of the type IV secretion ATPase family, such as HP0525 from Helicobacter pylori and the type IV pilus retraction ATPase, PilT, have been characterized as hexameric ATPases (Yeo et al., 2000; Forest et al., 2004). The X-ray structure of an N-terminally truncated form of EpsE, containing the N2 domain and the complete C-terminal domain with its three subdomains C1, CM and C2, was also recently solved (Figs 1 and 2a) (Robien et al., 2003). The truncated EpsE crystallized as anti-parallel helical filaments, however a hexameric ring model was proposed, as 8 of the 10 closest structural homologs assemble into multi-subunit rings. Structural data of EpsE also uncovered a metal ion tetrahedrally coordinated by a tetracysteine motif, protruding from the rest of the protein in an extended hairpin-like loop. Additional biochemical analysis revealed that EpsE incorporates 1 mol of zinc per mol of EpsE (Camberg & Sandkvist, 2005).
T2S:ER is dependent on the bitopic cytoplasmic membrane protein T2S:LY for membrane association (Sandkvist et al., 1995; Ball et al., 1999; Py et al., 1999; Possot et al., 2000). T2S:LY consists of a large cytoplasmic domain, a single transmembrane helix and a smaller periplasmic domain. The cytoplasmic domain binds to and localizes T2S:ER to the cytoplasmic membrane, whereas either the transmembrane helix or the periplasmic domain, or both, interact with T2S:MZ (Sandkvist et al., 2000). This interaction results in mutual protection of T2S:LY and T2S:MZ against proteolysis (Michel et al., 1998; Sandkvist et al., 1999), as well as localization of T2S:LY to the cell poles of V. cholerae (Scott et al., 2001). The recently published X-ray structure of the cytoplasmic domain of the V. cholerae TS2:LY homolog, cyto-EpsL, unexpectedly revealed structural homology with the actin-like ATPase super family (Fig. 1) (Abendroth et al., 2004a). Cyto-EpsL consists of three β-sheet-rich domains, with domains I and III of cyto-EspL corresponding to the conserved domains 1A and 2A of the actin-like ATPases. Cyto-EpsL does not contain the non-conserved 1B and 2B domains of the actin-like proteins, and as a consequence, cyto-EpsL is incapable of forming actin-like filaments. Domain II of cyto-EspL has only been observed in one other member of the actin-like ATPase super family: the cell division protein FtsA, where it likely participates in intramolecular interactions that result in recruitment of other proteins essential for cell division to the bacterial midcell (Rico et al., 2004). Structural and sequence analysis suggested that domain II of cyto-EpsL is also involved in protein-protein interactions (Abendroth et al., 2004a). The absence of domains 1B and 2B of the actin-like proteins, as well as the presence of the flexible domain II, makes cyto-EpsL unique among known members of the actin-like ATPase super family. Additionally, unlike the vast majority of the actin-like proteins, neither the sequence nor the X-ray crystal structure of cyto-EpsL reveal a nucleotide-binding site (Abendroth et al., 2004a). Instead, EspL likely participates indirectly in ATP hydrolysis through its interaction with the T2S ATPase EpsE.
Mutational studies have shown that the interface between EpsE and EpsL is formed by the N-terminal 90 residues of EpsE (Sandkvist et al., 1995) and a region of the cytoplasmic portion of EpsL that has been designated as domains II and III in the X-ray structure of cyto-EpsL (Sandkvist et al., 2000; Abendroth et al., 2004a). The first structural view of these interacting partners completes the structural picture of EpsE (which could not be done in the earlier X-ray structure of the N-terminally truncated EpsE) and confirms that domains II and III of cyto-EspL are involved in the interaction with the N1 subdomain of EpsE, which is folded into a compact α+β structure composed of three α-helices and β-strands each (Fig. 1) (Abendroth et al., 2005). One of the N1 α-helices appears to be wedged in a cleft between domains II and III of cyto-EspL. Given that this α-helix only partially fills the groove between the cyto-EpsL domains II and III, it is possible that other residues of EpsE or another member of the T2S machinery, such as T2S:FS, may also bind in this pronounced cleft.
The X-ray structure of subdomains N0 (helical region) and N1 (core domain) of XpsE from X. campestris has very recently been determined in two crystal forms (Fig. 2b) (Chen et al., 2005). In both forms, subdomain N1 adopts the same α+β fold as seen for N1 of EpsE in complex with the cytoplasmic domain of EpsL (Fig. 2c) (Abendroth et al., 2005) and shows some structural homologies with a NifU-like domain (PDB ID: 1veh). Of note, another member of the NifU-like family, IscU, is known to bind to and stimulate the activity of an ATPase (Silberg et al., 2001; Chen et al., 2005); an intriguing finding that may suggest that the N1 subdomain similarly regulates the activity of the T2S:ER ATPase. Consistent with this suggestion is the finding that removal of the N-terminal 90 residues of EpsE results in increased ATPase activity (Camberg & Sandkvist, unpublished data). In the two crystal forms of N0–N1 of XpsE, the 62-residue helical N0 subdomain adopts two vastly different conformations (Fig. 2b), but in spite of this conformational variability, the very N-terminal 36 residues appear indispensable for the functioning of the T2S system in X. campestris. Chen et al. hypothesized that these residues participate in binding to XpsL and that the two conformations might represent the bound and unbound states of the XpsE N-terminus; a hypothesis that would be consistent with the finding that EpsE may exist in two different forms in V. cholerae (Sandkvist et al., 1995). However, the structural superposition of the N1 subdomain of XpsE onto that of EpsE in the EpsE–EpsL complex leads to severe sterical clashes between subdomain N0 of XpsE in both open and closed conformation and the cytoplasmic domain of EpsL (Fig. 2d). This could mean that the N0 domain of XpsE adopts yet another conformation when bound to XpsL. Alternatively, the observed sterical clashes may support the suggestion of Chen et al. that XpsE and EpsE belong to two slightly distinct families of T2S:ER proteins and differ in the presence of domain N0, which may result in different T2S:LY binding properties.
The three T2S components T2S:MZ, T2S:LY and T2S:ER form a stable complex at the cytoplasmic membrane. T2S:MZ is a small inner membrane protein with a short cytoplasmic segment, one transmembrane helix and a periplasmic C-domain. The V. cholerae homolog, EpsM, inherently localizes to the cell pole, independently of other Eps proteins (Scott et al., 2001). It is also capable of directing EpsL to the bacterial cell pole. Besides localization and stabilization of T2S:LY, little has been reported about the role of T2S:MZ in the complex. All functional T2S systems contain a T2S:MZ homolog that is essential for secretion, however. The sequence homology between the T2S:MZ family members is concentrated in the predicted transmembrane helix and two blocks of conserved residues present in the periplasmic region of the polypeptide chain (Abendroth et al., 2004b). The short, N-terminal, portion of the protein shows very little homology with other family members.
The X-ray structure of a major portion of the periplasmic domain of EpsM has recently been solved (Fig. 1) (Abendroth et al., 2004b). Full-length EpsM and the soluble periplasmic domain are capable of forming dimers in solution, and two monomers of periplasmic EpsM are present in each asymmetric unit of the crystal. In the model of the dimer, subunit-subunit contact occurs mostly through interactions between C-terminal residues 123 through 135 in each EpsM monomer. Residues that make up the dimer interface are conserved only among the closest homologs, however.
Interestingly, when two EpsM monomers come together a 10 Å deep cleft that contains a hydrophobic bottom and a hydrophilic rim is formed at the interface (Abendroth et al., 2004b). The majority of hydrophobic residues in the bottom of the cleft are highly conserved between T2S:MZ family members, whereas the hydrophilic residues that line the rim show very little conservation. The periplasmic domain of EpsM represents a ferredoxin-like fold with a unique αββ–αββ permutation, and like other ferredoxin-like proteins which tend to bind substrates at the β-strand side (Russell et al., 1998), the EpsM dimer interface might also form a ligand-binding domain. Interestingly, additional electron density was found in the cleft between the subunits (Abendroth et al., 2004b). The chemical identity of the small molecular ligand could not be determined by X-ray crystallography. The shape of the additional density and possible molecular interactions are compatible with a short peptide, however (J. Abendroth & W. G. J. Hol, unpublished data).
A third cytoplasmic membrane protein, T2S:FS, has also been implicated in the cytoplasmic platform of the T2S machinery. Co-immunoprecipiation of OutE, OutL and OutF as well as co-purification of XcpRE, XcpSF and XcpYL with histidine-tagged XcpZM by metal affinity chromatography have demonstrated interactions between the T2S components ER, LY, MZ and FS (Py et al., 2001; Robert et al., 2005b). Similar interactions have also recently been demonstrated for the two T2S:ER and T2S:FS homologs BfpD and BfpE, which are required for biogenesis of the bundle-forming pilus of enteropathogenic Escherichia coli (Crowther et al., 2004). Furthermore, the ATPase activity of BfpD was found to be slightly stimulated by a fragment of the N-terminal cytoplasmic domain of BfpE (Crowther et al., 2005).
Sequence analysis of the T2S components has revealed homology with several components involved in type IV pilus biogenesis, suggesting that the T2S apparatus may function similarly to the type IV pilus machinery (Nunn, 1999; Burrows, 2005). The T2S:GT, HU, IV, JW and KX proteins display some similarity to the subunits of type IV pili and, like the pilin subunits, are translated with a specialized leader peptide required for export across the cytoplasmic membrane. Furthermore, T2S:OA shares a high degree of identity with the prepilin peptidase (PilD). PilD is a novel bifunctional aspartic acid protease that cleaves the leader peptides of type IV pilin subunits and N-methylates the newly generated N-terminus (Nunn & Lory, 1993; LaPointe & Taylor, 2000). T2S:OA performs the same function for the proteins T2S:GT, HU, IV, JW and KX in the T2S system. In some species no gene encoding T2S:OA is linked to the rest of the T2S genes; in these cases, PilD acts on both the precursor forms of type IV pilin subunits and the related proteins of the T2S system (Nunn & Lory, 1992).
The notable similarity with type IV pilus components has led to T2S:GT, HU, IV, JW and KX proteins being termed pseudopilins. It has recently been shown that certain T2S pseudopilins are indeed capable of forming a pilus-like, helical, fiber that spans the cell envelope and is exposed on the cell surface upon overexpression (Fig. 3) (Sauvonnet et al., 2000; Durand et al., 2003; Köhler et al., 2004). In agreement with the finding that T2S:GT is expressed at much higher levels than the other pseudopilins, immunogold-labeling and overexpression studies show that the pilus-like structure may be comprised of only T2S:GT subunits (Nunn & Lory, 1993; Durand et al., 2003; Vignon et al., 2003). Although the formation of the surface exposed polymer interferes with secretion, possibly by occluding the channel in the outer membrane, it suggests that the pseudopilin does assemble into a pilus-like structure that may span the periplasmic compartment and participate in protein secretion across the outer membrane (Possot et al., 2000; Hu et al., 2002; Durand et al., 2003).
The recently published crystal structure of a truncated form of T2S:GT, PulG from Klebsiella oxytoca, confirms the suspected structural parallels between T2S:GT proteins and type IV pilins (Fig. 1) (Köhler et al., 2004). Overall, PulG and the type IV pilins display a similar structure made up of an N-terminal hydrophobic α-helix followed by four β-strands arranged in a globular domain. PulG lacks, however, a highly variable loop region containing a disulphide bond found in all type IV pilins. Electron microscopy and modeling predict that PulG most likely assembles into a left-handed helical pilus with the α-helix of each subunit packing into the core of the pilus structure (Fig. 3). This helix in the type IV pilin monomer is proposed to be critically involved in subunit–subunit interactions that facilitate pilus formation. When the pilus is not assembled, the same region of the protein would anchor the pilin subunits in the membrane.
Pseudopilins and type IV pilins have similar overall structures and share some sequence similarity; however, there are key differences between the two groups of proteins. PulG appears to be more compactly folded than type IV pilin proteins (Köhler et al., 2004). Additionally, it is proposed that at least two of the type IV pilus proteins pack into right-handed helices (Parge et al., 1995; Craig et al., 2003), rather than the left-handed helix observed for PulG by electron microscopy (Köhler et al., 2004). As further demonstration of the differences between these two groups of proteins, Köhler (2004) showed using chimeric proteins generated between PulG and the type IV pilins PilE or PilA that the type IV pilins can be assembled into pili by the T2S system only when fused to the first 17 or 21 amino acids of the N-terminal helical domain of PulG. Even then, efficiency of the assembly was reduced. Therefore, despite their N-terminal sequence and structural similarity, this region of the pseudopilins and type IV pilins appears to not be functionally interchangeable.
Although interactions between the pseudopilins have been demonstrated, the function of the minor pilins T2S:HU, IV, JW and KX is not known (Nunn & Lory, 1993; Hu et al., 2002; Kuo et al., 2005). While all of them are essential for T2S mediated transport of proteins across the outer membrane, only T2S:IV appears to be required for the formation of the surface exposed pilus when the major pseudopilin T2S:GT is overexpressed (Sauvonnet et al., 2000). T2S:IV may be necessary for initiation of T2S:GT assembly or it may perhaps anchor the T2S:GT pilus to the cell envelope, as no pilus-like structure is visualized on the cell surface in its absence (Sauvonnet et al., 2000; Vignon et al., 2003; Durand et al., 2005). T2S:KX, on the other hand, may possibly terminate pilus assembly as overexpression of PulK from K. oxytoca has been shown to abolish the formation of the PulG pseudopilus (Vignon et al., 2003), and overproduction of the P. aeruginosa protein XcpX interfered with the assembly of XcpT, the T2S:GT homolog, into a pseudopilus (Durand et al., 2005). Conversely, the absence of XcpX resulted in a dramatic increase in the length and number of pseudopili observed. These findings collectively suggest that T2S:KX controls the length and possibly also the number of assembled pili. T2S:GT and T2S:KX likely interact directly, as XcpX can be cross-linked with XcpT and also appears to render XcpT more susceptible to proteolytic degradation when the two proteins are incubated together in vitro (Durand et al., 2005). As T2S:KX is an atypical pseudopilin in that it is longer than T2S:GT, HU, IV or JW, its incorporation into the growing filament could terminate pseudopilus elongation by disrupting packing of the helix or perhaps by interfering with recognition of next subunit extending the fiber (Parge et al., 1995). Given that XcpX also appears to destabilize XcpT in vitro, it is also possible that the presence of XcpX acts as a regulator or signal for disassembly of the pseudopilus subunits (Durand et al., 2005). It is likely that regulation of pilus assembly termination or disassembly is kinetically and stoichiometrically well controlled and may be, in part, governed by the affinity of T2S:KX for T2S:GT. Interestingly, T2S:KX differs from the other pseudopilins and pilins in that it lacks the highly conserved glutamate residue found at position +5; a residue that has been proposed to aid in the helical assembly of type IV pilins, as mutant pilins with E+5 substitutions are incapable of self-assembly (Pasloske et al., 1989; Strom & Lory, 1991; Parge et al., 1995; Craig et al., 2003). The lack of E+5 in T2S:KX may reduce the affinity for T2S:GT to prevent the premature termination of pilus assembly. In agreement with this suggestion, the T+5E substitution in XcpX displayed a stronger negative effect on XcpT assembly, as its overexpression completely prevented the formation of surface exposed XcpT pili (Durand et al., 2005). In contrast, a similar mutation in the homologous PulK had no deleterious consequence on secretion or pilus assembly (Vignon et al., 2003).
Reports that connect the pseudopilins to other components of the T2S system are fewer and may be due to their limited, and perhaps transient, interactions. Nevertheless, interactions between the major pseudopilin XpsG and the outer membrane protein XpsD and the T2S:CP-like XpsN in Xanthomonas campestris have been revealed by chemical cross-linking (Lee et al., 2005). Furthermore, the minor T2S:JW pseudopilin OutJ from Erwinia chrysanthemi was recently reported to interact with subdomains of the outer membrane protein OutD, the inner membrane protein OutL and the pilin-like protein OutI in a yeast-two hybrid system (Douet et al., 2004). These interactions await confirmation in E. chrysanthemi.
The outer membrane complex
T2S:CP interacts with T2S:LY and T2S:MZ present in the inner membrane and T2S:DQ in the outer membrane (Possot et al., 1999; Sauvonnet et al., 2000; Gerard-Vincent et al., 2002; Robert et al., 2002; Lee et al., 2004). Specifically, a region of the periplasmic domain near the membrane-spanning helix has been shown to contribute to interactions with the T2S:LY–T2S:MZ subcomplex (Gerard-Vincent et al., 2002; Lee et al., 2004), and the C-terminal region has been shown in many species to be important for interaction with T2S:DQ (Bleves et al., 1999; Possot et al., 1999). These interactions allude to a role for T2S:CP in linking the two membranes, where it may act as a energy transducer between the ATP-binding T2S:ER in the cytoplasm and the secretion pore in the outer membrane. As mentioned earlier, cross-linking between the T2S:CP-like protein, XpsN, in Xanthomonas campestris and pseudopilin protein XpsG has also recently been shown (Lee et al., 2005), suggesting that T2S:CP may play a direct role during pseudopilus assembly or activity as well.
Although T2S systems are generally well conserved, the components of the system have been found to be highly species-specific. T2S:CP and T2S:DQ, in particular, cannot be replaced by homologs even from closely related species (de Groot et al., 1991; Lindeberg et al., 1996). This suggests that these two proteins may be able to determine specificity for the substrates transported by the T2S system and/or perhaps for the assembly of the secretion complex. Domain swapping between closely and distantly related T2S:CP, as well as T2S:DQ, homologs have identified several regions that may contribute to the species-specificity. The results from the analysis of the T2S:CP and T2S:DQ chimeras are at times difficult to interpret, however, most likely because these components interact with more than one other protein. For instance, at the same time as T2S:CP may interact with both T2S:LY and T2S:MZ in the cytoplasmic membrane and T2S:DQ in the outer membrane, there is also the possibility that it interacts with T2S:GT and the proteins to be secreted. Sorting out the function of particular domains and establishing their contribution to species-specificity by swapping domains may therefore be challenging. For example, the PDZ, or coiled–coiled (CC region), motif in T2S:CP has, on the one hand, been suggested to participate in protein-protein interactions to promote T2S:CP homo-oligomerization (Gerard-Vincent et al., 2002), and on the other, to directly participate in the recognition of secreted proteins and confer secretion specificity to a subgroup of secreted proteins (Bouley et al., 2001). In the latter study, it was suggested that there are two sets of secreted proteins in Erwinia that are recognized and secreted by different mechanisms; one set is directly interacting with the PDZ domain of OutC and the other is not (Bouley et al., 2001). It is also possible, however, that the PDZ domain participates indirectly by promoting oligomerization and presentation of a recognition motif located elsewhere in T2S:CP. A subset of the secreted proteins may recognize this motif in the monomer, and as T2S:CP protein oligomerizes, the affinity increases. For the weaker interacting proteins in the other subgroup of secreted proteins, oligomerization may be required for binding.
T2S:DQ proteins are members of the secretin family of outer membrane proteins required for T2S, type IV pilus biogenesis, type III secretion, and filamentous phage extrusion. Several secretins including two T2S:DQ family members, as well as pIV, which is required for f1 phage secretion in Escherichia coli, have been purified and well characterized. They form stable, heat- and detergent-resistant ring-like assemblies of 12–14 subunits with internal pores that range from 50 to 100 Å in diameter that are occluded with a centrally located plug (Bitter et al., 1998; Nouwen et al., 2000; Opalka et al., 2003; Chami et al., 2005). Precisely how opening and closing the pore is regulated and what components are involved is not known. It has been hypothesized, however, that the very C-terminal region of T2S:CP may control the gating, as deletion or modification of this region results in increased levels of secreted proteins released to the extracellular environment (Robert et al., 2005a). Additionally, the suggestion that docking of secreted proteins may displace the centrally located plug, thus creating an uninterrupted channel through which the proteins can be secreted, has also been discussed (Chami et al., 2005).
Sequence comparisons between T2S:DQ family members have shown that the N-terminal domain is not well conserved, while the C-terminal domain is highly conserved (Genin & Boucher, 1994; Peabody et al., 2003). The conserved C-terminus is also protease-resistant and contains several putative amphipathic transmembrane β-strands (Bitter et al., 1998). Deletion and limited proteolysis studies as well as circular dichroism spectral analysis, nano-gold labeling and cryo-electron microscopy have strengthened the suggestion that the C-terminal and N-terminal regions are two structurally separate domains; the C-terminal domain is necessary for outer membrane insertion and multimerization, and the N-terminus may be free to interact with other T2S components and exoproteins in the periplasmic space (Chen et al., 1996; Shevchik et al., 1997; Bouley et al., 2001; Chami et al., 2005). 3D reconstruction of PulD revealed a three stacked ring structure with a cup and saucer appearance that is similar to the structure of pIV (Opalka et al., 2003; Chami et al., 2005), but for unknown reasons, quite different from the EM structure of PilQ, the T2S:DQ-like secretin involved in type IV pilus biogenesis in Neisseria meningitides (Collins et al., 2005). When modeled into an outer membrane environment, the PulD saucer is anchored in the outer membrane with some exposure to the cell surface, while the majority of the cup formed by the N-terminus is exposed to the periplasmic compartment. The chamber of the cup is likely the site of insertion of secreted proteins prior to their outer membrane translocation (Chami et al., 2005).
Exactly how T2S:DQ inserts into the outer membrane and assembles into a oligomeric ring is not known, however, the general outer membrane assembly factor Omp85 (YaeT) may be one of the factors that participates in these processes (Voulhoux et al., 2003). In addition, specific T2S components such as the lipoprotein T2S:S and the ATPase T2S:A in complex with T2S:B play important roles in promoting outer membrane insertion and stabilization of the T2S:DQ oligomer in some species (Shevchik et al., 1997; Condemine & Shevchik, 2000; Schoenhofen et al., 2005). To date, genes encoding these proteins have not been found in all T2S systems, however. This does not rule out the existence of other, non-homologous, proteins that could function in a similar capacity in systems where these proteins are not found.
It is assumed that T2S:DQ acts as the export channel for T2S proteins, and although exoprotein secretion through this channel has not been directly visualized yet, several secreted proteins have been shown to interact with the T2S:DQ homolog in Erwinia chrysanthemi (Shevchik et al., 1997). Furthermore, accumulation of assembled type IV pili in the periplasm of Neisseria gonorrhoeae in the absence of PilQ suggests that PilQ forms the outer membrane conduit through which type IV pili are transported to the cell surface (Wolfgang et al., 2000). Additional support for secretins acting as transport channels comes from studies showing that actively transported phage can block pIV-mediated uptake of oligosaccharides into E. coli cells (Marciano et al., 2001).
The N. meningitides PilQ has recently been shown to interact directly with the type IV pilus (Collins et al., 2005). Following separation by SDS-PAGE, purified type IV pili and full-length and truncated forms of PilQ proteins were subjected to far-Western analysis. Interaction of the type IV pilus protein with full-length, as well as with N- and C-terminal fragments of PilQ was observed. Transmission electron microscopy on preparations of purified pili incubated with PilQ oligomers revealed that the PilQ oligomer localized to only one end of the pilus, resulting in the formation of a ‘mallet-type’ structure (Fig. 4). A view of a section through the center of the pilin/secretin complex reveals that, in contrast to the previous structure of the PilQ oligomer alone, the interior of the PilQ/pilus complex is inaccessible to stains and therefore apparently filled; suggesting that the PilQ oligomer is capable of dynamic behavior previously predicted, but unobserved (Fig. 4).
Concluding remarks and future directions
The individual components that make up the T2S system are largely well defined. The challenge for future years is to reveal how these proteins come together to create a large, highly organized, multiprotein complex capable of secreting a diverse array of fully folded exoproteins across the outer membrane of gram-negative bacteria. Recent structural characterizations noted above are an important first step toward completing our picture of how the T2S system assembles and functions, however numerous interactions between T2S proteins still need to be uncovered to better understand these processes. Particularly exciting is the work showing the predicted, and unpredicted, similarities between T2S and other cellular systems. For instance, as EpsL and FtsA show unexpected structural similarities, FtsA's role in cell division will perhaps give us more clues to the function of EpsL and possibly other members of the T2S complex. The two different structures of the N-terminal portion of XpsE is another potentially very interesting finding that may suggest that T2S:ER cycles between two functional forms with different ATPase activity.
Another multiprotein secretion system, the type III secretion needle complex, has been successfully extracted from bacterial membranes (Kubori et al., 1998; Marlovits et al., 2004); an important achievement that, if repeatable with the T2S system, would greatly add to our knowledge of T2S. Whether it is ultimately possible to purify the entire T2S complex or whether we will have to continue to express, purify and crystallize domains of the proteins and, like a puzzle, piece them together remains an open question. A promising avenue seems to be to separate the T2S system into larger sub-assemblies for the studies of inner membrane, pseudopilus and outer membrane subcomplexes. This has been initiated in recent years, but much remains to be done to obtain higher resolution and to discover how these subcomplexes interact with one another.
Many additional challenges will have to be overcome to characterize this dual membrane complex, not the least of which is the possibility that the secretion apparatus is a dynamic structure constantly assembling and disassembling. The mechanism and energy requirement for the function of the putative periplasmic-spanning pilus-like piston in the extrusion of proteins through the outer membrane needs to be addressed. The type IV pilus system contains two ATPases with opposing functions, one for polymerization of pilin and the other for pilus retraction or disassembly, while the T2S system may only contain one ATPase. T2S:A proteins, such as ExeA, may fulfil the energy requirement for the retraction function, as ExeA has been shown to exhibit ATPase activity (Schoenhofen et al., 2005). T2S:A is not present in every T2S system, however. Filloux and colleagues have put forth another hypothesis that suggests that incorporation of the minor pseudopilin T2S:KX could drive pilus retraction by inducing pilus disassembly (Durand et al., 2005). Incorporation of T2S:KX into the growing pilus would not only bring the T2S:GT polymerization to a halt, but also initiate pilus disassembly, because binding of T2S:KX results in a conformational change in T2S:GT that makes it more susceptible to proteolytic degradation (Durand et al., 2005). Understanding how proteins to be secreted are sorted from other resident periplasmic proteins and determining how the peptidoglycan layer is negotiated during assembly of the T2S machinery and/or during transport are also of great importance. Elucidation of the atomic resolution architecture and dynamics of the T2S system in a variety of functional states is an immense challenge, albeit important goal, which in the future may lead to the identification of novel ways to block secretion of virulence factors in a large number of pathogens.