Decoding the roles of pilotins and accessory proteins in secretin escort services


  • Jason Koo,

    1. Program in Molecular Structure and Function, The Hospital for Sick Children, Toronto, ON, Canada
    2. Department of Biochemistry, University of Toronto, Toronto, ON, Canada
    Search for more papers by this author
  • Lori L. Burrows,

    Corresponding author
    1. Department of Biochemistry and Biomedical Sciences, McMaster University, Hamilton, ON, Canada
    • Program in Molecular Structure and Function, The Hospital for Sick Children, Toronto, ON, Canada
    Search for more papers by this author
  • P. Lynne Howell

    Corresponding author
    1. Department of Biochemistry, University of Toronto, Toronto, ON, Canada
    • Program in Molecular Structure and Function, The Hospital for Sick Children, Toronto, ON, Canada
    Search for more papers by this author

Correspondence: Lori L. Burrows, 4H18 Health Sciences Centre, 1200 Main St. West, Hamilton, ON L8N3Z5, Canada. Tel.: +1 905 525 9140; fax: +1 905 522 9033; e-mail:

P. Lynne Howell, Room 3507A, The Hospital for Sick Children, 555 University Ave., Toronto, ON M5G1X8, Canada. Tel.: +1 416 813 5378; fax: +1 416 813 5379; e-mail:


Secretins are channels that allow translocation of macromolecules across the outer membranes of Gram-negative bacteria. Virulence, natural competence, and motility are among the functions mediated by these large oligomeric protein assemblies. Filamentous phage also uses secretins to exit their bacterial host without causing cell lysis. However, the secretin is only a part of a larger membrane-spanning complex, and additional proteins are often required for its formation. A class of outer membrane lipoproteins called pilotins has been implicated in secretin assembly and/or localization. Additional accessory proteins may also be involved in secretin stability. Significant progress has recently been made toward deciphering the complex interactions required for functional secretin assembly. To allow for easier comparison between different systems, we have classified the secretins into five different classes based on their requirements for proteins involved in their assembly, localization, and stability. An overview of pilotin and accessory protein structures, functions, and characterized modes of interaction with the secretin is presented.


Secretion of molecules and macromolecules requires stringent control of membrane channel gating to maintain cell integrity. In Gram-negative bacteria, secretion involves crossing two membrane barriers. Protein trafficking through the inner membrane is largely mediated by the Sec or Tat systems, as reviewed recently by Facey & Kuhn (2010) and Robinson et al. (2011), respectively. In the Sec system, SecB recognizes the nascent preprotein destined for secretion from the cytoplasm and delivers it to SecA, which in turn propels the preprotein through the SecYEG pore into the periplasm. The specific number of proteins involved in Tat-mediated translocation is variable in Gram-negative bacteria, but TatA and TatC comprise the minimal functional unit. Outer membrane channels are more diverse and can be subdivided into three broad groups: monomer or multimeric β-barrel porins; α-helical multimeric barrels; and other protein assemblies for which there is currently no structural data. A recent review of the various membrane channel types has been published by Karuppiah et al. (2011). In short, a β-barrel formed by a single protein places a significant limitation on the size of the molecule that can be secreted. Larger channels formed through protein multimerization have thus evolved to allow passage of larger substrates.

Bacteria and bacteriophages use multiple systems to move macromolecules across the outer membrane without causing the cell to rupture. Secretins are common components of many of these transport systems, including type IVa and IVb pili (T4aP, T4bP), type II secretion (T2S), type III secretion (T3S), and DNA uptake and filamentous bacteriophage extrusion systems. The other types of secretion systems use alternative strategies to pass through the outer membrane that do not contain the conserved ‘secretin domain’. Many of these secretory nanomachines are of therapeutic interest owing to their roles in export of virulence factors during bacterial infection. Secretins are homo-multimeric complexes that form a gated channel in the outer membrane that open to allow passage of folded proteins, assembled multi-protein complexes, and DNA. Efforts to determine the structures of secretins by X-ray crystallography and electron microscopy were recently reviewed by Korotkov et al. (2011). The protein that multimerizes to form the secretin is typically comprised of two parts: a conserved C-terminal region containing the ‘secretin domain’ that is embedded into the outer membrane and a variable, system-specific N-terminal region. Both of these regions may interact with other components of the system as well as with the substrates to be secreted or internalized. The N-terminal region contains several different types of subdomains: (1) a N0 domain that resembles the TonB-dependent signaling receptor that may allow signal transduction between the inner membrane and outer membrane components of the system during secretion or uptake (Larsen et al., 1999; Brillet et al., 2007); (2) up to three heterogeneous nuclear ribonucleoprotein K homology-like domains that may fulfill the DNA binding role of competence systems (Tarry et al., 2011); and (3) additional elements that have yet to be structurally characterized. Despite the similarities in the overall architecture of the proteins forming secretins, the mechanisms that control secretin assembly vary both between and within systems. This review provides an overview of the differences in the assembly requirements of secretins. Particular focus will be given to the variability in the structure and function of pilotins and accessory proteins and their role in secretin stabilization, localization and/or assembly, their mode of interaction with the secretin-forming protein, and the effect(s) that the absence of the pilotin or accessory protein has on the secretin.


Proteins involved in secretin assembly are diverse in structure, functional role, and genomic context. These differences may reflect the evolutionary divergence from an ancestral secretin by recruitment of a specific set of proteins to optimize the system for a particular function. Generally, there are two classes of ancillary proteins: (1) pilotins and (2) accessory proteins. Localization and/or assembly of secretins is the proposed function of pilotins (Table 1). Pilotins have a type II N-terminal signal sequence followed by a conserved cysteine, which allows the protein to be lipidated and transferred to the inner leaflet of the outer membrane by the Lol system (Okuda & Tokuda, 2010). Binding of the pilotin to the secretin subunit followed by their co-transport to the outer membrane by the Lol system has been proposed to be the mechanism by which pilotins aid in secretin localization (Okon et al., 2008; Collin et al., 2011). Accessory proteins can stabilize the secretin itself, the secretin subunits prior to membrane insertion or are co-dependent with the secretin for mutual stability. Accessory proteins are membrane-associated and contain periplasmic regions that are thought to interact directly with the secretin. Systems may contain either a pilotin, an accessory protein(s), or both. Conservation of particular genes across a system does not necessary correlate with similar function, as significant differences have been documented between bacterial species.

Table 1. Secretin subunits and their cognate pilotins and accessory proteins in the secretion, pilus and filamentous phage systems have been tabulated along with their known effect(s) on secretin formation
SystemSpeciesSecretin subunitSecretin classPilotinPilotin classInteraction site on secretinEffect on secretin if absentAccessory proteinInteraction site on secretinEffect on secretin if absentReferences
  1. BMRB codes for MxiM bound to MxiD are 15503, 15497, 7407 and 15504.

  2. PDB codes for have been listed in parentheses where available.

  3. Pilotin classes: 1, α-helices forming TPRs; 2, predominantly β-strand; 3, α-helical non-TPRs.

  4. Secretin classes: 1, auto-assemble and can self-localize to the outer membrane; 2, auto-assemble but can not reach the outer membrane by themselves; 3, auto-assemble but inefficiently localize to the outer membrane by themselves; 4, are able to target to the outer membrane but not auto-assemble; 5, are unable to assemble or target to the outer membrane.

  5. IM, inner membrane; ND, interaction site or localization not determined; OM, outer membrane; TPR, tetratricopeptide repeat.

  6. a

    Expression of either MxiM or MxiJ stabilizes the secretin.

T2SA. hydrophilaExeD3    ExeA/BNDIM, monomerAst et al. (2002)
T2SA. salmonicida & Vibrio sp.GspD3    GspA/BNDND, reduced multimerStrozen et al. (2011)
T2SE. chrysanthemiOutD3OutS3 (predicted)C-terminalOM, less Multimer   Shevchik et al. (1997); Shevchik & Condemine (1998); Condemine & Shevchik (2000)
T2SK. oxytocaPulD2PulS3 (predicted)C-terminalIM, Multimer   Nouwen et al. (1999)
T2SP. aeruginosaHxcQ1       Viarre et al. (2009)
T3SS. flexneriMxiD3MxiM (PDB: 1Y9T, 1Y9L, 2WJ1)2C-terminalDegraded without MxiJaMxiJNDDegraded without MxiMaSchuch & Maurelli (2001); Lario et al. (2005); Okon et al. (2008)
T3SS. typhimuriumInvG2InvH3 (predicted)C-terminalIM, Multimer   Crago & Koronakis (1998)
T3SY. enterocoliticaYscC2YscW2 (predicted)Not C-terminalIM, less Multimer   Burghout et al. (2004); Ross & Plano (2011)
T4aPM. xanthusPilQNDTgl1 (predicted)NDND, Monomer   Nudleman et al. (2006)
T4aPNeisseria sp.PilQ4PilW (PDB: 2VQ2)1NDOM, MonomerPilP (PDB: 2IVW), FimVMiddle, NDVariable effect, reduced monomer and multimerDrake et al. (1997); Carbonnelle et al. (2005, 2006); Balasingham et al. (2007); Trindade et al. (2008); Szeto et al. (2011); Wehbi et al. (2011)
T4aPP. aeruginosaPilQ5PilF (PDB: 2HO1, 2FI7)1NDIM, Monomer   Koo et al. (2008)
T4bPE. coliBfpB4    BfpGN-terminusOM, monomerSchmidt et al. (2001); Daniel et al. (2006)
T4bPV. choleraeTcpC4    TcpQNDDegradedBose & Taylor (2005)
Filamentous bacteriophage f1 pIVND    pI, pXINDNDFeng et al. (1999); Haigh & Webster (1999)

Pilotins that have been identified and characterized to date are listed in Table 1. Although many systems have identifiable pilotin orthologues, they are either absent or have yet to be identified in others. Competence systems, filamentous phage, and T4bP each lacks pilotins. Most T2S systems characterized to date have pilotins, except for Pseudomonas aeruginosa Hxc and Xcp, Escherichia coli Gsp, Aeromonas hydrophila Exe, and Vibrio cholerae Eps. Immediate differences can be found between the remaining pilotin-containing T2S, T3S, and T4aP systems by comparing the genomic organization of the genes encoding the secretin subunit and the pilotin. The gene encoding the secretin subunit is typically clustered with other genes that encode a variable number of proteins involved in system assembly. The T2S pilotins Erwinia chrysanthemi outS and Klebsiella oxytoca pulS as well as the T3S pilotins Salmonella typhimurium invH, Shigella flexneri mxiM and Yersinia enterocolitica yscW are each encoded with other components of their respective assembly systems. In contrast, the T4aP pilotins Pseudomonas pilF (pilW), Neisseria pilW, and Myxococcus xanthus tgl are located elsewhere in the genome and surrounded by non-T4aP genes.

While pilotins fulfill similar roles in localizing and/or assembling the secretin, the structure of specific pilotins can vary significantly. Based on the available structural data or on sequence-based predictions, we divided pilotins into one of three different classes: Class 1 pilotins are composed entirely of α-helical tetratricopeptide repeats (TPRs) and are roughly double the size of other pilotins. Class 2 pilotins are comprised predominantly of β-strands, while Class 3 pilotins are predominantly α-helical non-TPR proteins. The structure of pilotins clearly divides the secretion and pilus systems.

Sequence identity among T4aP pilotins PilF, PilW, and Tgl is poor, ranging from 13% to 25%. However, the structures of PilF and PilW that have been determined by X-ray crystallography (Koo et al., 2008; Trindade et al., 2008) show that they have a common protein fold. PilF and PilW are each composed of six TPRs with a nearly identical tertiary fold (Fig. 1a). Structural alignments show that a rigid body motion between the individual TPR elements causes the observed structural differences between the proteins. These differences reflect either inherent dynamic protein motion or artifacts caused by protein packing during crystallization. While the structure of Tgl has yet to be solved, it is of the same length as its counterparts and is predicted by TPRpred (Karpenahalli et al., 2007) to contain six TPRs with high confidence (per protein P-value of 4.8E−39 and a 100% probability of having TPR structure). Canonical TPRs are formed by 34 amino acid residue repeats that fold into a pair of α-helices interlocked by a pattern of large and small side chains as defined by the TPR consensus sequence (D'Andrea & Regan, 2003). Like many other protein repeats, TPRs are generally found to be involved in protein–protein interactions (Andrade et al., 2001) and therefore support the hypothesis that the pilotin interacts directly with the secretin subunit. Pilotins in T4aP systems appear to be absolutely required for secretin assembly, and the TPRs may act as a scaffold for this process. However, low sequence identity resulting in very different surface properties of PilF, PilW, and Tgl prevents large functionally conserved surfaces from being identified (Fig. 1b) and likely reflects evolution from a common protein fold into three highly specialized pilotin–secretin interaction interfaces.

Figure 1.

Known and predicted structures of Class 1–3 pilotins. (a) Structures of Class 1 pilotin PilF (PDB: 2HO1), Class 1 pilotin PilW (PDB: 2VQ2), Class 2 pilotin MxiM (PDB: 1Y9L), and putative Class 3 pilotin GspS (PDB: 3SOL). PilF (in blue) and PilW (in red) have been structural aligned. The aligned regions, TPRs 1–4, have been colored in lighter shades. The arrow shows the direction of rigid body motion that relates the two structures. (b) Residues conserved between PilF, PilW, and Tgl have been mapped in blue on the surface representation of PilF; (c) The secondary structures of ExsB, MxiM, YscW, GspS, InvH, OutS, and PulS are depicted based on predictions or solved structures. The mature protein is shown without the cleaved type II signal sequence. α-Helices are represented by red rods and β-strands by green arrows. Protein structure figures were generated with PyMOL (Schrödinger, 2010) and secondary structures prediction schematics generated with Jalview (Waterhouse et al., 2009).

Pilotins in T2S and T3S are about half the size of those found in T4aP systems and are not predicted to contain TPRs. Only one structure of a secretion system pilotin has been solved to date: the S. flexneri T3S pilotin, MxiM (Lario et al., 2005). The structures of E. coli T2S GspS (PDB: 3SOL), an orthologue of InvH, OutS, and PulS, and P. aeruginosa T3S ExsB (Izore et al., 2011), an orthologue of YscW, have also been recently determined but have yet to be functionally characterized. These structures, paired with secondary structure predictions using JPRED (Cole et al., 2008), suggest they represent two different groups, one predominantly comprised of β-strands (Class 2) and the other of α-helical (Class 3) (Fig. 1a). With the exception of InvH, the T2S and T3S systems appear to contain pilotins of Class 3 and Class 2, respectively. The β-strand Class 2 pilotins include MxiM and Y. enterocolitica T2S YscW. MxiM is composed of 10 β-strands that fold into an incomplete β-barrel to enclose a channel ~ 8 Å across (Fig. 1a) (Lario et al., 2005). Two helices within the series of β-strands effectively occlude the pore from one side. Additional density within the pore was suggestive of a bound lipid tail and led to a proposed mechanism for MxiM-mediated outer membrane insertion of the secretin through membrane disruption. Despite sharing only 4% identity with MxiM, YscW is predicted to have a similar arrangement of secondary structure elements (Fig. 1c). Tertiary structure predictions using Phyre2 (Kelley & Sternberg, 2009) produces a model with high confidence for YscW based on its putative orthologue ExsB. Unlike MxiM, ExsB contains fewer but longer β-strands that form a β-sandwich without a central pore (Izore et al., 2011). In place of the long α-helix that was found to block the MxiM pore, YscW only contains a α-helical turn. These results suggest either that the bound lipid in MxiM is an artifact of the crystallization process, which required detergents to be present, or that the lipid disruption mode of secretin insertion into membranes is not universally used by Class 2 pilotins.

Class 3 pilotins InvH, OutS, and PulS are predicted to be similar in size to the β-strand pilotins and to be predominantly α-helical, although they lack predicted TPRs (Fig. 1c). Structural data for this group are limited to the crystal structure of E. coli T2S GspS (PDB ID: 3SOL), an orthologue of the Class 3 pilotins that has not been functionally characterized. While the sequence identity among GspS, OutS, and PulS ranges from 30% to 36%, the sequence identity of InvH to OutS, PulS, and GspS is only 3%, 12%, and 14%, respectively. The structure of GspS is a four α-helix bundle, as is predicted for OutS and PulS (Fig. 1c). One face of GspS forms a distinct groove that could provide a convenient binding surface for an interacting partner. InvH is predicted to contain shorter α-helices and a large central region without regular secondary structure. Tertiary structure predictions by Phyre2 (Kelley & Sternberg, 2009) produces high confidence models (100%) for OutS and PulS based on GspS. As InvH is significantly different from the others at the sequence level, models can only be generated for a fragment of the protein at confidence levels of 47.3% or lower, and are not templated on GspS.

Accessory proteins

Accessory proteins that have been functionally characterized in secretin-containing systems are listed in Table 1. Accessory proteins are not always present in a particular system, nor are their functions always the same. Many accessory proteins appear to be involved in stability of the secretin or of the secretin subunit prior to assembly. Accessory proteins that have been reported to influence secretin formation include ExeA/B in A. hydrophila; GspA/B in Vibrio species and Aeromonas salmonicida; OutB in E. chrysanthemi; MxiJ in S. flexneri; PilP in Neisseria meningitidis and P. aeruginosa; FimV in P. aeruginosa; pI/pXI in filamentous phage; BfpG in E. coli; and TcpQ in V. cholerae.

In T2S, GspA/B in Vibrio species and A. salmonicida (ExeA/B in A. hydrophila) has been found to be important for expression of the secretin. However, the protein pair is not universally present – or has yet to be identified – in all T2S systems (Strozen et al., 2011). GspA spans the inner membrane and has domains in both the cytoplasm and the periplasm (Schoenhofen et al., 1998; Howard et al., 2006). A surprisingly similar arrangement and orientation is predicted for the filamentous phage accessory protein, pI, which raises the possibility that the two could be evolutionarily related. The cytoplasmic domains of GspA and pI are both predicted to contain a nucleoside-binding domain and can be modeled with high confidence by Phyre2 (Kelley & Sternberg, 2009) (99.9% or greater). ATP hydrolysis by these domains is necessary for both secretion and phage assembly (Russel, 1995; Schoenhofen et al., 2005), suggesting they may be involved in priming the secretin for activity. The periplasmic portion of GspA, but not pI, is predicted to contain a three-helix-bundle-type peptidoglycan (PG)-binding domain that is well modeled by Phyre2 (Kelley & Sternberg, 2009).

Despite the resemblance of pI to GspA, the similarity is not maintained in the second accessory component in these systems, pXI and GspB, respectively. GspB is encoded separately from GspA, while pXI is formed by an alternate translation start site within the pI transcript and plays a different role (Haigh & Webster, 1999). The Erwinia Out system contains a GspB homolog, OutB, but oddly, lacks a GspA equivalent. Phyre2 (Kelley & Sternberg, 2009) is able to generate only partial models of ExeB, GspB, OutB, and pXI and all are of low confidence. Secondary structure predictions also show significant variations between the proteins.

MxiJ is an accessory protein involved in S. flexneri T3S secretin formation (Schuch & Maurelli, 2001). A structure of MxiJ is not available but it can be well modeled on its homologs, S. typhimurium PrgH and E. coli EscJ. PrgH and EscJ are integral proteins involved in T3S and are thought to form 24-membered rings in the inner membrane (Yip et al., 2005; Schraidt & Marlovits, 2011). While the MxiJ homolog is a common component of T3S systems, the consequences of mutating this protein are inconsistent across T3S systems. The presence of either MxiJ or the pilotin, MxiM, is sufficient for secretin assembly (Schuch & Maurelli, 2001). In the absence of YscJ in Y. enterocolitica, the secretin formed by YscC appears normal (Diepold et al., 2010). However, without E. coli EscJ or P. aeruginosa PscJ, secretion is abolished, although whether this is attributable to a malformed secretin has not been demonstrated (Ogino et al., 2006; Burns et al., 2008). To date, these systems have not been shown to require a MxiM-like pilotin.

Structures of T4bP accessory proteins TcpQ and BfpG have yet to be determined, but in both cases Phyre2 (Kelley & Sternberg, 2009) predicts the C-terminal half of the protein to adopt a VirB7-like fold. VirB7, together with VirB9 and VirB10, is involved in forming the outer membrane pore in type IV secretion systems and resembles the N0 domain found in secretins (Souza et al., 2011) although none of the Vir proteins contains a ‘secretin domain’. The presence of an N0-like domain in this non-secretin protein family suggests that Gram-negative bacteria have adopted a common protein fold to allow communication between components of membrane-spanning systems.

PilP is an inner membrane component of the T4aP system whose role in secretin assembly has yet to be fully determined, as several inconsistencies appear in the published data (Drake et al., 1997; Carbonnelle et al., 2006; Balasingham et al., 2007; Tammam et al., 2011). Structures of PilP fragments from P. aeruginosa (PDB: 2LC4) and N. meningitidis (Golovanov et al., 2006) have been solved by NMR and adopt an identical fold, but do not share the same surface properties. The N-terminus appears to lack regular secondary structure, while the C-terminus forms a β-sandwich. The C-terminus of the PilP orthologue in E. coli T2S, GspC, has been shown to interact directly with the N0 domain of the secretin subunit (PDB: 3OSS). An additional inner membrane protein, FimV, has been shown to affect the function of T4aP and T2S in P. aeruginosa (Semmler et al., 2000; Coil & Anne, 2010; Michel et al., 2011; Wehbi et al., 2011). Mutation of fimV reduces both secretin and secretin monomer levels (Wehbi et al., 2011). No structure of FimV is yet available but the periplasmic N-terminus encodes a LysM-type PG-binding motif, while the highly acidic cytoplasmic C-terminus contains putative TPR motifs. Deletion of the LysM domain alone impairs secretin formation, suggesting PG interactions are important for assembly.

Secretin variations

Outer membrane secretins are formed by multimerization of 12–15 molecules of a single protein into ring-like structures (Korotkov et al., 2011). The subunit in each system that forms the secretin is listed in Table 1. Secretins characterized to date can be divided into five classes: (1) self-assembling and self-membrane targeting; (2) self-assembling but not self-membrane targeting; (3) self-assembling but inefficiently self-membrane targeting; (4) self-membrane targeting but not self-assembling and (5) not self-assembling and not self-membrane targeting (Fig. 2). Little correlation is readily apparent between the classes of secretins and the systems to which they belong.

Figure 2.

Schematic representation of the five secretin classes with respect to phenotypes that arise from the absence of their pilotin and/or accessory protein(s).

Class 1 secretins have only recently been identified. The single Class 1 secretin that has been characterized to date is HxcQ from the P. aeruginosa T2S. This class of secretins is unique as they are themselves lipoproteins that are directly targeted to the outer membrane by the Lol pathway, where they auto-assemble (Viarre et al., 2009). In contrast (see below), all other secretins require additional proteins for stability, localization and/or assembly. The genomic organization of Class 1 secretins also differs from the typical Gsp-type T2S systems, where the secretin gene is usually paired with, and immediately downstream of a gspC orthologue (encoding GspC, PulC, OutC, XcpP, EtpC, or XpsN). The equivalent gene in the Hxc system, hxcP, is instead located at the opposite end of the gene cluster from the secretin gene. As GspC and the secretin in T2S are hypothesized to be responsible for co-recognition of multiple substrates (Bouley et al., 2001; Gerard-Vincent et al., 2002; Douzi et al., 2011), the uncoupling of hxcP and hxcQ may reflect the fact that Hxc secretes only the alkaline phosphatase, LapA, under phosphate limiting conditions (Ball et al., 2002).

Secretins in Class 2 are able to assemble independently but need their pilotins to localize correctly to the outer membrane. Examples of this class include InvG, PulD, and YscC. In the absence of their cognate pilotins, InvH and PulS, the amounts of monomeric InvG and PulD are decreased in the cell (Hardie et al., 1996; Crago & Koronakis, 1998). In contrast, the amounts of pilotin YscW and secretin subunit YscC were found to be inversely correlated (Burghout et al., 2004). Furthermore, a dominant-negative effect on secretion was observed when mistargeted YscW was expressed in the wild-type background (Burghout et al., 2004). Oligomers, corresponding to the assembled secretin, were shown to localize to the inner membranes in all three systems (Crago & Koronakis, 1998; Burghout et al., 2004; Guilvout et al., 2006). Assembly of secretins in the inner membrane by PulD has been shown to have a toxic effect through the induction of the phage shock response and to partially dissipate the transmembrane electrochemical potential, implying that this secretin is incompletely gated (Guilvout et al., 2006). These results lead to the hypothesis that the pilotin binds the secretin subunit to allow their co-localization to the outer membrane prior to self-assembly, thereby preventing premature formation at the inner membrane that would be deleterious to cellular integrity.

Class 3 secretins, like their Class 2 counterparts, self-assemble but require assistance for efficient outer membrane targeting. Secretins that fall into this class are from T2S systems that rely on accessory proteins (Table 1) for full functionality. In the absence of the accessory protein GspA in A. salmonicida, Vibrio cholerae, Vibrio parahaemolyticus, and Vibrio vulnificus, GspD is still able to form multimers but less efficiently than wild-type (Strozen et al., 2011). In contrast, ExeD multimers in A. hydrophila were not observed in an exeA/B mutant unless ExeD was overexpressed (Ast et al., 2002). While multimer localization in the cell was not determined in any of these accessory protein mutants, the fact that secretion was measurable suggests that at least some functional secretins were present in the outer membrane. Despite the high sequence identity between GspA in A. salmonicida, V. cholerae, V. vulnificus, and V. parahaemolyticus and ExeA in A. hydrophila, only secretion by A. hydrophila and A. salmonicida was greatly reduced or abolished in the absence of the accessory proteins, which suggests they are more strongly required in Aeromonas (Ast et al., 2002; Strozen et al., 2011).

The E. chrysanthemi Out system shares some similarity with Gsp/Exe but has an additional level of complexity. In this system, the GspB homolog, OutB, is present but a GspA homolog is absent. Mutation of the putative accessory protein outB, like the double mutation of exeA/B in A. hydrophila, can be overcome by overexpression of OutD and the two proteins are mutually stabilizing (Condemine & Shevchik, 2000). However, unlike Gsp and Exe, a pilotin, OutS, is required for secretion (Condemine et al., 1992), as well as stability and efficient outer membrane localization of OutD (Shevchik et al., 1997; Shevchik & Condemine, 1998). Shigella flexneri MxiD similarly requires both a pilotin, MxiM, and accessory protein, MxiJ. However, MxiJ has no sequence similarity to GspB. Expression of either MxiM or MxiJ prevents MxiD from degradation (Schuch & Maurelli, 2001).

Secretins in Class 4 are able to reach the outer membrane but are unable to form stable assemblies in the absence of their accessory proteins. BfpB from E. coli T4bP falls into this category, as multimers of BfpB cannot form without BfpG (Schmidt et al., 2001). Despite being part of a T4bP system, TcpC in V. cholerae behaves differently. TcpC and its accessory protein, TcpQ, are mutually stabilizing, and each is completely degraded in the absence of the other (Bose & Taylor, 2005). Another example of a Class 4 secretin is PilQ from N. meningitidis T4aP. In the absence of PilW, PilQ remains monomeric in the outer membrane – or does not form stable multimers – and does not support T4P activity (Carbonnelle et al., 2005). The inner membrane protein PilP has been reported to affect PilQ stability in Neisseria, but published results are inconsistent (Drake et al., 1997; Carbonnelle et al., 2005, 2006; Balasingham et al., 2007).

Pilotins are required for both proper localization and assembly of Class 5 secretins. PilQ in P. aeruginosa, unlike its homolog in N. meningitidis, is retained in the inner membrane without the PilF pilotin (Koo et al., 2008). Untethering of PilF from the membrane by mutation of its lipidation site causes PilQ assembly in both membranes and shows that secretin assembly mediated by PilF is a separate function from localization.

Interactions involved in secretin formation

Given the variation in the requirements for secretin assembly, the mode of interaction between pilotins and accessory proteins with their cognate secretin has been the focus of much study. Biophysical techniques and functional characterization of mutants have begun to pinpoint the region(s) of the secretin subunit involved and the stoichiometry of the interaction. The majority of pilotins have been found to interact with the C-terminus of the secretin subunit, whereas accessory proteins bind in the N-terminal region.

Protein chimeras between secretin C-termini and several different proteins have been used to show an interaction between the secretin and pilotin. Attachment of the C-terminal 65 amino acids of PulD or 43 amino acids of InvG to the filamentous phage protein pIV rendered the chimeras dependent on the pilotins, PulS and InvH, respectively, for phage assembly and allowed the chimera–pilotin complex to be co-immunoprecipitated (Daefler et al., 1997; Daefler & Russel, 1998). Attachment of the 65 amino acid C-terminus of PulD to maltose binding protein or the 62 amino acid C-terminus of OutD to the secreted protein PelD rendered both dependent on their pilotins, PulS and OutS, respectively, for stability (Daefler et al., 1997; Shevchik & Condemine, 1998). The same region of OutD was also demonstrated to be required for OutS-mediated stability of OutD (Shevchik et al., 1997) and to bind OutS by far-western blotting (Shevchik & Condemine, 1998). Interestingly, the 65 amino acid C-terminus of PulD could be further divided by function into two regions: the C-terminal 25 amino acids are required for outer membrane targeting by PulS, while the region 25–65 amino acids upstream from the C-terminus are important for stability mediated by PulS (Daefler et al., 1997). Subsequent biophysical characterization has shown PulS binds with high affinity directly to the C-terminal 28 amino acids of PulD (Nickerson et al., 2011).

Structural methods have also been applied to look at secretin–pilotin interactions. The original cryo-electron microscopy model of the PulD secretin in complex with the pilotin PulS showed the 12-fold symmetrical complex to form a funnel-like cylinder with 12 peripheral spokes emanating from the central region (Nouwen et al., 1999) (Fig. 3a). Limited proteolysis of the PulD–PulS complex showed that PulS forms a part of the spoke (Chami et al., 2005). The mode of binding between PulD and PulS suggests that the C-terminus of the secretin is located at or near the inner leaflet of the outer membrane that was defined by the location of the spoke. Yeast two-hybrid interaction (Schuch & Maurelli, 2001) and isothermal calorimetry (Lario et al., 2005) studies established that the C-terminal 46 amino acid tail of MxiD interacts with MxiM. Subsequent NMR studies have revealed the atomic level details of the C-terminal 18 amino acids of MxiD binding to MxiM (Okon et al., 2008). The MxiD C-terminus was shown to undergo a transition from a disordered to α-helical state on binding to MxiM (Fig. 3b). A similar transition was also observed on binding of PulD by PulS (Nickerson et al., 2011). The binding of the Class 2 and 3 pilotins described above to the C-termini of their respective secretins subunits strongly suggests a 1 : 1 stoichiometry. Whether this same mode of binding is also used by Class 1 pilotins remains to be determined, but some differences are evident: (1) the cryo-electron microscopy reconstruction of the PilQ secretin from N. meningitidis showed fourfold symmetry with much weaker 12-fold symmetry and lack of peripheral spokes (Collins et al., 2001, 2003, 2004) (Fig. 3c); and (2) sequence alignments show that PilQ in T4aP lacks the C-terminal tail found in the above examples (Daefler et al., 1997; Korotkov et al., 2011). A different mode of binding is, however, not unprecedented. Deletion of the C-terminal 96 amino acids of YscC, corresponding to the expected binding region of the pilotin, YscW, did not prevent the outer membrane targeting or assembly of the secretin (Burghout et al., 2004).

Figure 3.

Structures of secretins in complex with their pilotin or accessory protein. (a) Cryo-electron microscopy reconstruction of PulD in complex with pilotin PulS. Figure from (Nouwen et al., 1999) was reproduced with permission. Copyright 1999 National Academy of Sciences, USA. (b) NMR structure of MxiD with pilotin MxiM (PDB: 2WJ1) (Okon et al., 2008). MxiM is colored in green and the bound C-terminal peptide of MxiD is in blue. (c) 3D reconstruction of cryonegative-stained PilQ, purple, PilQ in complex with PilP, in green, and the PilQ–PilP complex labeled with nanogold, green and yellow, respectively. Figure from (Balasingham et al., 2007) was reproduced with permission.

Compared to studies of pilotin–secretin interactions, less is know about how the accessory proteins and secretins associate. Filamentous phage pI and pIV were shown to interact both in vivo and when co-expressed in isolation from the other phage proteins using crosslinking approaches (Feng et al., 1999). An interaction between BfpB and BfpG was also demonstrated by crosslinking and affinity purification (Daniel et al., 2006). Yeast two-hybrid studies further refined the binding site to the N-terminal third of BfpB (Daniel et al., 2006). While PilP does not consistently affect PilQ stability or assembly, an interaction between the two proteins has been demonstrated. Far-westerns and cryo-electron microscopy show PilP binds a central region of PilQ (Fig. 3c) (Balasingham et al., 2007). Significant structural rearrangements in the ‘cap’ and ‘arms’ regions were visible in the PilP–PilQ secretin complex compared to the PilQ secretin complex alone. Nanogold labeling showed that PilP was localized to the displaced regions of the secretin; the stoichiometry could not be determined as several different surfaces were labeled.


Our knowledge of the ways in which secretins and pilotins/accessory proteins interact has grown significantly through the implementation of innovative functional assays and the advances in protein structure determination. Over time, the increasing diversity of mechanisms by which secretins are formed has become evident. While bacteria have a general secretion pathway for the majority of exoproteins, additional systems have evolved to specialize in and accommodate very specific functions: T4P production, the T3S needle-like injectosome, DNA uptake, and secretion of specialized proteins in response to environmental stimuli. Presumably, these systems are costly to maintain in the genome but have been retained to enable survival in niche environments. The fact that filamentous phage also use secretins to extrude from their bacterial hosts certainly prompts speculation about the degree of co-evolution between the host and pathogen.

A significant impediment to studying the in vivo interactions within these large membrane-spanning complexes has been the technical barriers to extraction of intact protein complexes from the membrane environment. However, the increasing body of research in membrane proteins and membrane protein complexes shows this is clearly no longer a deterrent. Continued research will undoubtedly lead to the development of novel methods to work with membrane proteins that will allow us to better understand the interactions between secretins and the proteins required for their formation.

Future perspectives

Despite the accumulation of a significant amount of data on secretin–pilotin and accessory protein interactions to date, many outstanding questions remain. While the Lol system is likely responsible for trafficking a pilotin–secretin subunit complex to the outer membrane, the process by which the secretin is assembled is unknown. Does the pilotin itself disrupt the membrane to assemble the secretin or are other cellular factors involved? Lipid binding by MxiM suggests a membrane disruption model but not all pilotins have been found to bind lipids nor do they share the same protein fold. While MixM and many other pilotins bind the secretin subunit C-terminus, this is not always the case. The variations in secretin structure raise the issues of where the pilotin and secretin interact, and the stoichiometry of the interaction. Accessory proteins have been demonstrated to be critical in stabilizing secretins but the mechanism by which this occurs remains unknown. The stimuli that trigger secretin opening to enable passage of substrates and the role that secretins play in mediating substrate specificity also need to be determined. Structural data at the atomic level that show any of the interactions required for secretin formation, channel dynamics, and substrate recognition would be of tremendous value not only to aid our understanding of secretin assembly but also of how large membrane-spanning complexes in general assemble and function.

Note added in proof

Following the acceptance of this manuscript, the structure of K. oxytoca PulS was published by Tosi et al. 2011. As predicted, PulS is a Class 3 pilotin that, like E. coli GspS, contains a distinct groove formed by helix α1 flanked by helices α3 and α4. Mutation of the groove has shown it to be critical for PulS function.


The authors would like to thank Dr. Lili Sampaleanu and Ms. Stephanie Tammam for fruitful discussions. Work in the Howell and Burrows laboratory on type IV pilus assembly is supported by grant MOP 93585 from the Canadian Institutes of Health Research (CIHR). J.K. is the recipient of a Canada Graduate scholarship from CIHR. P.L.H. and L.L.B. are recipients of a Canada Research Chair and CIHR New Investigator award, respectively.