Type IVa pili are bacterial nanomachines required for colonization of surfaces, but little is known about the organization of proteins in this system. The Pseudomonas aeruginosa pilMNOPQ operon encodes five key members of the transenvelope complex facilitating pilus function. While PilQ forms the outer membrane secretin pore, the functions of the inner membrane-associated proteins PilM/N/O/P are less well defined. Structural characterization of a stable C-terminal fragment of PilP (PilPΔ71) by NMR revealed a modified β-sandwich fold, similar to that of Neisseria meningitidis PilP, although complementation experiments showed that the two proteins are not interchangeable likely due to divergent surface properties. PilP is an inner membrane putative lipoprotein, but mutagenesis of the putative lipobox had no effect on the localization and function of PilP. A larger fragment, PilPΔ18-6His, co-purified with a PilNΔ44/PilOΔ51 heterodimer as a stable complex that eluted from a size exclusion chromatography column as a single peak with a molecular weight equivalent to two heterotrimers with 1:1:1 stoichiometry. Although PilO forms both homodimers and PilN–PilO heterodimers, PilPΔ18-6His did not interact stably with PilOΔ51 alone. Together these data demonstrate that PilN/PilO/PilP interact directly to form a stable heterotrimeric complex, explaining the dispensability of PilP's lipid anchor for localization and function.
Type 4 pili (T4P) are expressed by a wide variety of Gram-negative bacteria, and have also been identified in Gram-positive bacteria (Pelicic, 2008). T4P play an important role in many processes including twitching motility, phage infection, DNA uptake, attachment to surfaces and biofilm maturation (Bradley, 1974a,b; Burrows, 2005). Rapid pilin polymerization and depolymerization make these long thin flexible filaments incredibly dynamic. As they generate retraction forces in excess of 100 pN (Maier et al., 2002) the T4P apparatus is the most powerful molecular machine identified to date. Most importantly, T4P are essential virulence factors for many bacteria (Hahn, 1997; Wolfgang et al., 1998a,b; Soto and Hultgren, 1999; Heiniger et al., 2010).
Although mechanistic details are few, a picture of the organization of the T4P assembly machinery is beginning to emerge from biochemical, genetic, structural and bioinformatic analyses (Hobbs and Mattick, 1993; Ayers et al., 2009). PilQ, encoded by the highly conserved pilMNOPQ operon (Pelicic, 2008) is a member of the secretin family and forms the outer membrane pore through which the pilus is extruded. PilM is a cytoplasmic protein with predicted structural homology to the actin-like cytoplasmic domain of GspL from the evolutionarily related type II secretion system (T2SS), while structural evidence (Sampaleanu et al., 2009) suggests that the bitopic inner membrane proteins PilN and PilO are similar to GspL and GspM respectively. The periplasmic domains of PilN and PilO form a stable heterodimer (Sampaleanu et al., 2009), and those of GspL and GspM have been shown to interact with each other (Sandkvist et al., 1999; Possot et al., 2000; Lee et al., 2001; Robert et al., 2005a). PilP is predicted to be an inner membrane lipoprotein (Drake et al., 1997) with no obvious homologue in the T2SS. PilP contains a highly conserved lipobox (L15A16G17C18) in its signal peptide (Inouye et al., 1977; Hayashi and Wu, 1990). Sequence analysis revealed that all PilP proteins have a lipidation sequence, and PilP from Nesseria gonorrhoeae has been shown by radiolabelling experiments to be lipidated (Drake et al., 1997). Based on co-fractionation data and the protein stability phenotypes of mutant strains, PilM, PilN, PilO and PilP were predicted to form an inner membrane subcomplex potentially involved in alignment of the secretin with the pilus assembly machinery (Ayers et al., 2009).
Although the T2SS has no PilP homologue, the corresponding component may be GspC, a protein that interacts with GspM and GspL (Gerard-Vincent et al., 2002; Robert et al., 2005a), and with the GspD secretin. GspC and its homologues are the only proteins known to interact with both inner and outer membrane components of the T2SS (Bleves et al., 1999; Possot et al., 1999). Both GspC and PilP-like proteins have been implicated in control of secretin gating (Lee et al., 2000; 2005; Robert et al., 2005b; Balasingham et al., 2007) leading to their classification as ‘secretin-dynamic associated’ (SDA) proteins (Ayers et al., 2010). In Neisseria meningitidis, PilP has been experimentally demonstrated to interact with PilQ (Balasingham et al., 2007), and in this work we show that Pseudomonas aeruginosa PilP interacts with the PilN/PilO heterodimer; this finding – in combination with bioinformatics information – suggests that PilP is functionally equivalent to the homology region of GspC.
Here we also report the NMR structure of the C-terminal domain of P. aeruginosa PilP and show that it is structurally homologous to N. meningitidis PilP (PilPNm), although the overall sequence identity is low (28%). Despite the striking structural similarity between the two proteins, PilPNm is unable to complement a P. aeruginosa pilP mutant, indicating that there are important species-specific differences that require consideration when making functional inferences. We show that the cysteine of the highly conserved lipobox of PilP is not essential for its function in vivo, likely due to the formation of a stable heterotrimeric complex with the periplasmic domains of PilN/PilO. In contrast, PilP did not form a stable complex with a PilO homodimer, confirming that the PilN/PilO/PilP heterotrimer is most likely to be the functional entity. Finally, we propose a model in which PilP is the linker component, aligning the inner and outer membrane subcomplexes of the T4P export machinery.
Structure determination of PilPΔ71
To gain insight into the structure and function of the protein at the molecular level, we pursued the structure determination of P. aeruginosa PilP. Our attempts to heterologously express mature full-length PilP (i.e. the protein minus its signal sequence – residues 18–174) in Escherichia coli indicated that the protein was highly sensitive to proteases (data not shown). Sequence-based secondary structure predictions suggested that the N-terminus of the protein lacks regular secondary structure, and predictions of intrinsic disorder (Jones and Ward, 2003) suggested that residues between positions 30–50 and 70–90 are disordered (Fig. S1). N-terminal sequencing of the smallest stable proteolytic fragment revealed that the first 71 amino acids were missing. Therefore, structural studies of this stable fragment (PilPΔ71) were pursued. Given its small size (12.5 kDa), and the difficulty in growing diffraction-quality crystals, the structure was determined using NMR spectroscopic techniques. A series of 3D triple resonance experiments (see Experimental procedures) were used to assign the resonance frequencies to backbone and side-chain 1H, 15N and 13C in a uniformly labelled sample, and 3D NOESY-HSQC data were used to obtain 1417 experimental nuclear Overhauser effect (NOE) distance restraints. A summary of the experimental restraints and structural statistics is provided in Table 1.
Table 1. Experimental restraints and structural statistics for the 20 structure ensemble of PilPΔ71.
RMS deviations from TALOS dihedral angle restraints
RMS deviations from idealized geometry
Bond angles (o)
Bond lengths (Å)
RMSD (relative to average structure for residues 106–167)
0.39 ± 0.07
All heavy atoms (Å)
0.81 ± 0.08
Ramachandran plot statistics
There are no distance violations > 0.5 Å.
Most favoured (%)
Additionally allowed (%)
Generously allowed (%)
Much of the N-terminus of PilPΔ71 (residues 72–90) is disordered, consistent with predictions, based on both minimal chemical shift differences from random coil (Figs S2 and S3) and the small number of mid- to long-range tertiary NOE contacts (Fig. S4). However, by these measures a few residues in this region including Lys78, Lys84, Pro85 and Asp86 make stable interactions, which likely tether these residues to the structured portion of PilP. Long range NOEs to Lys84, Pro85 and Asp86 were observed (Fig. S4), and Lys78, Lys84 and Pro85 were predicted to have reduced mobility by the Random Coil Index (RCI) (Berjanskii and Wishart, 2005), which calculates the mobility of a residue based on the difference between experimentally assigned and random coil chemical shift values (Fig. S3).
To experimentally examine the backbone dynamics of this tethered region, 1H-15N heteronuclear NOE data were collected. These data indicated that residues before Arg76, residue Lys81 and residues after Lys171, as well as some loops between the β-strands, are significantly more mobile than the ordered β-rich region of PilPΔ71 on the ps-ns timescale (Fig. S5). Residues Gly79–Asn80 and Ile84–Phe93 have heteronuclear NOE values of the same magnitude as the β-domain of PilPΔ71 demonstrating that these regions have restricted mobility, as suggested by the chemical shift-derived flexibility predictions and the observed interproton NOEs.
The ordered region (residues 93–169) contains a short 310 helix followed by seven anti-parallel β-strands arranged into two β-sheets (Fig. 1). Sheets 1 and 2 are comprised of β-strands 1–3 and 4–7 respectively. Each sheet has one side that contributes to the core of the modified β-sandwich (inside face), and another side wholly exposed to solvent (outside face). A similar β-sandwich fold was observed previously in the published NMR structure of a comparably sized PilPNm C-terminal fragment (residues 69–181; PilPΔ68Nm) (Golovanov et al., 2006). Superimposition of the two proteins gave a root mean square deviation of 1.35 Å over 60 Cα atoms in the β-domain (Fig. 2A). Although this level of structural homology is not surprising given the 31% sequence identity (55% similarity) in these domains, some differences were observed. For example, PilPΔ68Nm has 14 more residues in the disordered C-terminal region than PilPΔ71 and while no specific function has been ascribed to this region, it may play an important role in forming inter-protein contacts in the N. meningitidis T4P complex.
A second notable difference between the PilP and PilPNm structures is the presence of a van der Waals interaction in PilPΔ71 between the conserved residues Pro85 (in the N-terminal region; corresponding to Pro78 of PilPNm) and Trp161 (in the second β-sheet; corresponding to Trp154 of PilPNm) (Fig. 2A). In the lowest energy model of PilPΔ71 the distance between the Hδ of Pro85 and the Hε1 of Trp161 is 2.2 Å. This interaction is absent in the PilPΔ68Nm structure, as the distance between the same protons is 7.1 Å in the lowest energy model (Fig. 2B). We were able to assign nine and seven NOEs between Trp161 and the side-chain protons of Pro85 and Lys84 respectively. Although this interaction may also occur in PilPΔ68Nm, it was not observed in the structure because the side-chain protons of the analogous residue, Pro78, were not assigned (BMRB Accession No. 7209). These interactions provide support for the tethering of these regions as observed in the RCI and heteronuclear NOE analyses, further verifying that the interaction between the highly conserved Pro and Trp exists in solution.
Another significant difference between the P. aeruginosa and N. meningitidis proteins is their divergent surface properties. The predicted isoelectric points (pI) of mature PilP and PilPNm are 8.0 and 4.8, respectively, while the pIs of the fragments whose structures have been solved are 8.3 and 4.7 for PilPΔ71 and PilPΔ68Nm respectively. Examination of electrostatic surface potential for both proteins illustrates these significant disparities (Fig. S6). There are a number of surface residues in PilP that add a basic charge and/or remove an acidic charge relative to the N. meningitidis protein. These residues in PilPNm are Glu112, Glu114, Gln128, Glu134, Ser140, Glu148 and Ala159, which are substituted with Lys119, Ala121, Arg135, Val141, Lys147, Pro155 and Arg166 in PilP respectively. Five of these seven residues (Lys119, Ala121, Val141, Pro155 and Arg166) are located in the region of the β-sandwich domain that is partially covered by the N-terminal disordered region, while the other two (Arg135 and Lys147) are located on the outer face of sheet 2 (Fig. S6A), resulting in two species-specific surfaces.
Structural and functional homologues of PilP
As tertiary structures are not available for all of the proteins that may be functionally similar to PilP, we examined proteins that share similar secondary structure characteristics with particular emphasis on those involved in other secretion systems. Using bioinformatics tools (see Experimental procedures), we found that PilP is similar to the GspC family of integral inner membrane proteins of the T2SS. GspC family proteins contain an N-terminal transmembrane helix anchoring the protein in the inner membrane; a region termed the homology region (HR) consisting of a Pro-rich region with no regular predicted secondary structure, followed by a β-strand region with six to seven predicted β-strands; and finally, a protein interaction domain such as a coiled coil or a PDZ motif (Fig. 3A). A sequence alignment of the HR of various members of the GspC family with PilP (Fig. 3B) indicates that the predicted secondary structure elements of the HR domains align very well with the known secondary structure elements of PilP and PilPNm, despite their low sequence identity (range between 3% and 29%, with an average of 11%).
Indeed this analysis was confirmed with the recent release of the co-ordinates for a co-crystal structure of GspC/GspD (PDBID: 3OSS, Fig. S7). The structures of PilPΔ71 and GspC can be superposed with a RMSD of 2.36 Å over 130 Cα atoms, and share the same modified β-sandwich motif. The comparisons of the structure have been discussed elsewhere (Korotkov et al., 2011), but do support our analysis that the HR region of GspC type proteins and PilP share structural homology.
Characterization of a transposon mutant in pilP
Mutagenesis of the pilMNOPQ operon in several different organisms (Martin et al., 1995; Carbonnelle et al., 2005; 2006; Nudleman et al., 2006; Rumszauer et al., 2006) confirmed that the products of these genes were essential for T4P function. However, only PilQ has a well-defined role. To address the role of PilP in T4P assembly and function, we characterized a strain mPAO1 pilP mutant from the University of Washington collection (Jacobs et al., 2003) with a transposon insertion after nucleotide 254 (Fig. 4). Using a series of assays that measure pilus-specific phage sensitivity (a correlate of retractable pilus expression), twitching motility and extracellular PilA levels, the phenotype of the pilP mutant was found to resemble that of a non-piliated strain (pilA) (Table 2 and Fig. 5). Western blot analysis indicated that the protein expression of PilQ was not decreased in the pilP mutant, and sucrose density gradient fractionation indicated that PilQ was trafficked to the outer membrane as in the wild-type strain suggesting that there were no polar effects on PilQ expression (data not shown).
Table 2. Phenotypic characterization of mutant strains and complementation constructs.
−, no significant amount of protein detected; +, less than mPAO1; ++, equivalent to mPAO1.
0.08 ± 0.01
1.0 ± 0.2
1.1 ± 0.2
1.06 ± 0.03
0.81 ± 0.01
0.74 ± 0.06
++ (PilP) + (PilPΔ71)
0.07 ± 0.01
0.1 ± 0.02
1.00 ± 0.01
0.67 ± 0.1
0.08 ± 0.01
Complementation of the pilP mutant with full-length pilP restored T4P function to wild-type levels, as assessed by restoration of both phage sensitivity and wild-type levels of twitching motility (Table 2 and Fig. 5), but resulted in recovery of less than 20% wild-type levels of extracellular PilA as measured by more than three independent sheared surface protein preparations (Fig. 5C). This is not due to a reduction of PilA expression as intracellular levels of PilA were equivalent in all of the strains examined (Fig. 5C). We have previously noted a lack of correlation between motility and recoverable surface pili in ectopically complemented mutants that may arise due to changes in the stoichiometry of assembly components (Ayers et al., 2009; Giltner et al., 2010).
Absence of PilP does not cause mislocalization of T4P assembly proteins
Previous studies from our lab showed that PilP colocalized to the inner membrane with PilM/N/O and that deletion of PilP had no significant effect on the post-translational stability of the other components of the T4P inner membrane subcomplex (Ayers et al., 2009). To test whether loss of PilP caused mislocalization of any of the components in the inner or outer membrane subcomplexes, sucrose gradient fractionation and Western blot analyses were performed. No change in localization of PilM/N/O/Q or PilF compared with mPAO1 was observed for the pilP mutant (data not shown). Expression of pilP in trans in all strains examined (mPAO1 and pilP::Tn5) also had no effect on the localization of any these proteins (data not shown). These data suggest that PilP is not required for the correct targeting of the PilM/N/O/Q or PilF proteins.
Species-specific differences between PilP from P. aeruginosa and N. meningitidis
Given the structural similarity between PilP and PilPNm, we tested whether PilPNm could complement the P. aeruginosa pilP mutant. Two constructs with internal histidine tags were produced, one containing the native N. meningitidis signal sequence (pilPNm6Hi) and the second a chimeric construct with the signal sequence from P. aeruginosa PilP (pilPPaNm3Hi). The histidine tag was inserted between amino acids 74 and 75 (Balasingham et al., 2007) and expression was tested in both the pilP mutant and mPAO1. PilPNm6Hi was unable to restore wild-type T4P function as determined by our functional assays (Table 3), and expression was reduced compared with a similarly tagged version of P. aeruginosa PilP (PilP6Hi) (Fig. 6). Although the chimeric construct, pilPPaNm3Hi, was expressed well, the complemented strain had the same phenotype as pilPNm6Hi (Fig. 6) indicating that N. meningitidis PilP cannot complement the P. aeruginosa pilP mutant.
Table 3. Complementation of P. aeruginosa strains with pilPNm.
−, no significant amount of protein detected; +, less than mPAO1; ++, equivalent to mPAO1.
1.1 ± 0.2
1.2 ± 0.4
0.99 ± 0.2
0.95 ± 0.2
0.1 ± 0.02
0.1 ± 0.03
0.1 ± 0.01
0.65 ± 0.1
The failure of PilPNm6Hi to complement the pilP mutant is not a consequence of the presence of an internal tag as complementation with pilP6Hi restored twitching to ∼65% relative to wild-type pilP and had no effect on twitching in mPAO1 (Table 3), and untagged versions of PilPNm in the same background as the tagged versions were similarly unable to complement, although expression of these proteins could not be confirmed. Together these results suggest that an internal His tag can reduce the function of the T4P apparatus, but that the tag is not responsible for lack of complementation by pilPNm.
C-terminal domain of PilP has dominant-negative properties
To test whether the proteolytically stable C-terminal domain used for structure determination was capable of restoring T4P function in the P. aeruginosa pilP mutant, a periplasmic complementation construct with an in-frame deletion of residues 19–71, but retaining the native signal sequence, was generated (pilPssΔ71). Complementation with pilPssΔ71 was unable to restore T4P function (Table 2). When pilPssΔ71 was expressed in mPAO1, a partial dominant-negative effect on twitching motility was observed, such that the twitching was reduced to 74 ± 6% relative to mPAO1 (Table 2) or to 67 ± 12% relative to mPAO1 with the empty vector (+pUCP20Gm). The inhibition is not due to reduced levels of endogenous, full-length PilP, as Western blot analysis of whole-cell lysates shows that PilP expression levels are unaltered in the presence of PilPssΔ71 (Fig. 5B).
To determine whether PilPssΔ71 inhibited T4P function by titrating protein partners in the inner and/or outer membrane, strains expressing PilPssΔ71 were subjected to sucrose gradient membrane fractionation. Membrane fractionation showed that, similar to full-length mature PilP, PilPssΔ71 was localized to the inner and mixed membrane fractions (Fig. 7) despite the absence of most of its N-terminus, including the putative lipidation site (Table 2). The observation that PilPssΔ71 localized to inner membrane fractions suggests either that the C-terminal fragment is sufficiently hydrophobic to partition with the membrane, or that it is capable of interacting with other integral membrane proteins (i.e. the Lol lipoprotein processing complex).
Absence of a putative lipidation site is not essential for PilP localization or function but is important for protein stability
As an indirect method to probe the importance of lipidation, the putative lipidation site (Cys18) of PilP was modified by site-directed mutagenesis to Ala (PilPC18A). The PilPC18A protein restored twitching motility levels to ∼60% compared with wild-type pilP in the pilP mutant (Table 2). These data suggest that, while not essential, the putative lipidation site of PilP is important for optimal T4P function. As observed previously, when the mutant strain was complemented with pilPC18A, extracellular PilA levels were not restored to mPAO1 levels, but were similar to the levels recovered upon complementation with native pilP (Table 2 and Fig. 5), and analysis of the Western blots showed that PilPC18A was processed correctly and of the expected molecular weight.
To determine whether the reduced twitching motility observed upon complementation with pilPC18A was related to decreased stability or mislocalization, whole-cell lysates expressing PilPC18A were analysed by Western blot. In the mutant background, PilPC18A levels were reduced compared with both endogenous PilP in mPAO1 and wild-type PilP expressed in trans from the same vector (Table 2 and Fig. 5B). This result suggests that lipidation – and by extension, membrane anchoring of the full-length protein – are important for its stability. Analysis of membrane and soluble fractions indicated that although a significant proportion of PilPC18A continued to fractionate with membranes, there was a larger proportion in the soluble fraction compared with both endogenous PilP and PilP expressed in trans (data not shown). Sucrose density fractionation of strains expressing PilPC18A indicated that it remained associated with the inner membrane fraction (Fig. 7). Together, the reduced expression but correct localization of the protein suggests that PilPC18A associates with the inner membrane through contacts with other inner membrane components. In the absence of lipidation, such interactions are likely less efficient, making PilPC18A more susceptible to proteolytic degradation.
Co-purification of PilN/PilO/PilP
The inner membrane localization of PilP in the absence of its putative lipidation site suggested that the protein is capable of interacting with one or more protein components of the T4P inner membrane subcomplex, likely PilN and PilO (Ayers et al., 2009). We could co-purify PilPΔ18-6His with PilNΔ44/PilOΔ51 (18.5 kDa, 17.1 kDa and 17.4 kDa respectively) as all three proteins co-eluted from a Ni-agarose column (Fig. 8A) even though only PilP contained a 6-His tag. To evaluate the stability of the putative complex, the 75 mM imidazole fraction was concentrated and analysed by size exclusion chromatography. A single peak containing all three proteins was observed (Fig. 8B). After 2 weeks at 4°C, no significant changes were observed either in the Kav for the peak on the same gel filtration column or on Western blots, indicating that the complex was stable over time (data not shown). Careful examination of the Western blots revealed that the N-terminus of PilPΔ18-6His was not susceptible to proteolytic degradation in the presence of PilN/PilO, in contrast to when PilP was expressed in the absence of its partners, pointing to the presence of protective PilN/PilO interactions with the disordered regions in the N-terminus of PilP (data not shown).
Having established that PilN, PilO and PilP were present in a single peak upon gel filtration, we examined their stoichiometry in the complex. Using a standard curve generated by separating five proteins ranging in mass from 13.7 to 440 kDa under the same experimental conditions, the molecular mass of the complex was estimated to be ∼112 kDa, closely matching the expected mass of a 2:2:2 complex of PilNΔ44:PilOΔ51:PilPΔ18-6His of 106 kDa. Analysis of a PilNΔ44-6His/PilOΔ51/PilPΔ25 construct where PilN was His-tagged behaved comparably upon Ni-agarose purification and gel filtration (Fig. S8). Size determination by gel filtration and analytical ultracentrifugation analyses of the PilNΔ44-6His/PilOΔ51/PilPΔ25 purified complex indicated that this species existed as a heterotrimer with a mass estimation of 60 kDa and 50 kDa respectively (Fig. S8 and Supplemental data). Together these data indicate that PilN/PilO/PilP likely exists as a stable heterotrimer, or a dimer of heterotrimers in vitro.
PilPΔ18 does not co-purify with homodimers of PilOΔ51
From the co-purification data, it was not clear whether PilP interacted directly with PilN, PilO or both. PilO was previously shown to form stable homodimers as well as PilN/PilO heterodimers in solution (Sampaleanu et al., 2009), although complementation studies suggested that PilN/PilO heterodimers were the functionally relevant state (Ayers et al., 2009). To test whether PilP could interact with PilO homodimers, PilPΔ18-6His was coexpressed with PilOΔ51 and the lysate loaded on a Ni-agarose column. Only PilPΔ18-6His was recovered in the high imidazole fractions, suggesting that it was not able to interact stably with PilO alone (Fig. S9). Co-purification studies using a 6-His-tagged fragment PilOΔ43 and untagged PilPΔ18 gave the same result – PilP and PilO could not be co-purified from a Ni-agarose column (data not shown). We were unable to explicitly test the interaction between PilN and PilP, as PilN is insoluble when expressed in the absence of PilO (Sampaleanu et al., 2009). Together, these data suggest that PilN is required for the PilP interaction and that PilN/PilO/PilP is most likely the relevant physiological association for these proteins.
N-terminal residues of PilP are important for its interaction with PilN/PilO
As PilPssΔ71 exhibited a dominant-negative effect on T4P function in our in vivo twitching assays, we wondered whether this fragment could bind to the PilN/PilO heterodimer. Neither coexpression (data not shown) nor mixing the pellets of cells expressing PilN/PilO and PilP before lysis resulted in co-purification of PilNΔ44/PilOΔ51/PilPΔ71-6His (Fig. S10), indicating that the C-terminal domain of PilP is unable to form stable interactions with PilN/PilO. These data, in addition to the stabilizing effects of PilN/PilO on the N-terminus of PilP, provide indirect evidence that the N-terminal disordered region of PilP is important for the stable interactions between these three proteins.
Our detailed characterization has revealed a number of previously unknown facets of PilP function. We have demonstrated PilP participates in high-affinity protein–protein interactions that can localize it to the inner membrane even in the absence of the putative lipidation signal. PilP co-purifies with the inner membrane components PilN and PilO in a stable ternary complex that remains together over extended periods, suggesting that this subcomplex is likely to be stable in vivo as well. The stoichiometry of the three proteins in vivo remains to be determined. In the T2SS, the PilN and PilO homologues (GspL and GspM respectively) have been reported to form homodimers, heterodimers and heterodimers of homodimers (Sandkvist et al., 1999; Py et al., 2001; Robert et al., 2002; Abendroth et al., 2004). In vitro we found evidence that the PilN/PilO/PilP proteins exist in either a 1:1:1 or a 2:2:2 complex. The stoichiometry observed varied depending upon which protein contained the 6-His tag. As His-tags have been shown to influence the oligomeric state of complexes through steric or electrostatic effects (Cheung et al., 2010), the location of the tag could therefore potentially influence complex formation and be the source of the discrepancies observed. However, we cannot rule out that the PilP construct used in the PilNΔ44-6His/PilOΔ51/PilPΔ25 studies was shorter at the N-terminus by seven amino acids compared with the His-tagged version (PilPΔ18-6His), and that these residues may also influence multimerization and/or formation of a higher-order complex. In the absence of PilN/PilO, PilP appears to be a monomer, while in isolation PilN/PilO forms a heterodimer (Sampaleanu et al., 2009). We propose that the minimal unit of this inner membrane complex is a single copy of each protein that may form higher-order oligomers upon formation of the functional complex.
Subsequent analysis of the interaction between PilO and PilP suggested that they do not form a stable interaction in the absence of PilN. PilN's insolubility makes it difficult to determine whether PilP interacts with both PilN and PilO in the heterodimer, or with PilN alone. The lack of PilO–PilP interaction in the absence of PilN is consistent with our earlier observations in a pilO mutant strain, where the levels of PilN and PilP were decreased, and twitching motility and protein expression levels were restored only when pilM/N/O/P were coexpressed from a single plasmid (Ayers et al., 2009). Therefore, it is likely that stoichiometry and possibly spatiotemporal effects on expression of PilN/PilO/PilP are important for the formation and function of the inner membrane complex.
Interestingly, PilPΔ71 did not co-fractionate with PilN and PilO, providing indirect evidence that the N-terminal region of PilP is essential for the PilN/PilO/PilP interaction – a logical conclusion given the anchoring of each of these three proteins in the membrane by their N-termini. This finding was supported by the inability of our PilPΔ71 fragment to complement the pilP mutant in vivo. Interestingly, a dominant-negative phenotype was observed when this fragment was expressed in mPAO1, suggesting that the C-terminal region is competent for protein–protein interactions that are different from those observed in vitro between larger fragments of PilP and PilN/PilO. The obvious candidate for this interaction is the outer membrane protein PilQ, as interactions between PilP and PilQ have been verified in other species (Drake et al., 1997; Balasingham et al., 2007) and between the PilP and PilQ homologues in the T2SS [GspC and GspD (Bleves et al., 1999; Possot et al., 1999; Lee et al., 2000; Robert et al., 2005b)]. The recently released co-crystal structure of GspC and GspD from T2SS in E. coli (PDB 3OSS) confirms the predicted structural homology between PilP and GspC-type proteins (Fig. S7), and further supports the hypothesis that PilP and PilQ interact.
Both the N- and C-terminal regions of PilP were implicated in the PilP–PilQ interaction in N. meningitidis (Balasingham et al., 2007). The observation that PilQ monomers localize primarily with the inner membrane upon sucrose gradient fractionation (Koo et al., 2008) may help to explain this observation. We propose that PilP may interact with PilQ monomers in the inner membrane via the N-terminal region on PilP, and possibly even with the PilN/PilO/PilP complex prior to their transit and insertion into the outer membrane, and then again in the outer membrane with the fully assembled PilQ multimers via the folded β-region of PilP (Fig. 9).
It has been reported that the region of GspC from Erwinia chrysanthemi (OutC) involved in the interaction with the outer membrane secretin (GspD/OutD) was limited to residues 139–158 in the HR domain of GspC, which the authors termed the secretin interacting peptide (SIP) (Login et al., 2010). Using the structural data presented here, as well as the alignment of PilP and the HR domains from GspCs in Fig. 3, we show that this region corresponds to the most distal two C-terminal β-strands in PilP. However, in the GspC–GspD crystal structure, it is the two N-terminal β-strands of GspC involved in the interaction. It is reasonable to predict, in the absence of other data, that PilP and GspC-type proteins may be able to form a protein interaction interface on either free edge of the modified β-sandwich (i.e. with either strand 1 or strand 7/6), and that different systems use different interaction interfaces while achieving the same goal of linking the inner and outer membrane complexes.
The inability of PilPNm to complement our P. aeruginosa pilP mutant highlights the fact that structural similarity is not enough to confer function upon a protein, even though – based on bioinformatic and functional data – a similar arrangement of the T4P complex likely exists in both organisms. We envision a situation where the tertiary structures of the protein members of the T4P and T2SS systems (and indeed, many other secretion systems) (Fig. 9) have been conserved due to their importance for function, but the protein surfaces have been free to diverge in relation to species-specific and environmental pressures. The observation that proteins or domains from closely related species cannot be exchanged is consistent with this theory (He et al., 1991; de Groot et al., 2001; Gerard-Vincent et al., 2002).
Both experimental and predictive evidence indicate that large portions of the N-terminal region of PilP are disordered. There are a number of reports describing the key roles of intrinsically disordered proteins in protein complex formation (Fong et al., 2009; Tompa et al., 2009). The advantages of having disordered regions participate in protein–protein interactions include the formation of protein interaction interfaces with large surface areas, rapid association times and the flexibility to undergo conformational rearrangements to facilitate complexation (Mittag et al., 2010; Chouard, 2011). PilP could take advantage of such properties while forming interactions with other T4P proteins such as PilN and PilO. Binding PilN/PilO with an extended PilP N-terminus would contribute a larger surface area for the interaction than if PilP were to interact in a fully folded state, and a highly flexible N-terminus may be essential for accommodating the very stable and well-folded PilN/PilO heterodimer. The pilus is also a highly dynamic organelle and these dynamics may necessitate the quick association and disassociation of the transenvelope complex as part of the regulation of T4P function.
Based on the similarity of PilP to the HR region of GspC, and the evidence that PilP forms a stable ternary complex with PilN/PilO in the T4P system we believe PilN/PilO/PilP form a complex that is analogous to the inner membrane complexes identified in T2SS (Robert et al., 2005a; Lybarger et al., 2009). Additional domains present in GspC could modulate the affinity of the interaction between GspC and GspL/M, as both the transmembrane helix and protein interaction domain (PDZ or coiled-coil) have been shown to be important in homo- and heterointeractions (Bouley et al., 2001; Gerard-Vincent et al., 2002; Korotkov et al., 2006; Login and Shevchik, 2006). As PilP lacks these additional domains, it is unlikely that the functions of the two proteins are identical. The presence of common elements (or, if you like, repeating themes) between the T2SS and T4P systems (as noted elsewhere: Russel, 1998; Peabody et al., 2003; Ayers et al., 2010) will provide important clues for general strategies regarding the assembly of multiprotein transenvelope secretory complexes in bacteria.
In conclusion the data provide a picture for the role of PilP in the T4P apparatus, namely that it is the most likely candidate to bridge the inner and outer membrane subcomplexes. This arrangement would allow for the formation of a continuous channel across the periplasm through which the pilus could extend and retract, thereby ensuring that the pore formed by PilQ is aligned with the inner membrane assembly platform, including the cytoplasmic motor ATPases, for efficient pilus extension and retraction. Finally, we demonstrated important species-specific differences present in PilP from P. aeruginosa and N. meningitidis, the functional implications of which are not yet fully understood.
The bacterial strains and vectors used in this study are listed in Table 4. Growth media included Lennox L broth base, with or without agar (BioShop) and Pseudomonas Isolation Agar (PIA; Difco). Antibiotics were used where indicated at the following concentrations: 50 µg ml−1 kanamycin, 35 µg ml−1 chloramphenicol, 100 µg ml−1 ampicillin, 15 µg ml−1 gentamicin in E. coli and 30 µg ml−1 gentamicin in P. aeruginosa. Plasmids were introduced into chemically competent E. coli strains by heat shock. Transformation into P. aeruginosa was achieved by electroporation of cells as described previously (Koo et al., 2008).
The oligonucleotides used in this study are summarized in Table S1. P. aeruginosa mPAO1 genomic DNA was purified using the Bio-Rad InstaGene matrix, and primers were designed to amplify two constructs of pilP, with and without the first N-terminal 18 or 71 amino acid residues (pP5′ D18 NcoI and pP5′ D71 NcoI). These constructs are denoted PilP, PilPΔ18-6His and PilPΔ71-6His respectively. To facilitate subsequent cloning, the 3′ primer for all constructs contained an XhoI site (pP3′ XhoI). The amplified product was initially cloned into pSTBLUE-1 (Novagen), and then subcloned into pET28a (PilPΔ18-6His and PilPΔ71-6His) vectors using the corresponding restriction sites. These vectors allow for the expression of C-terminally 6-His-tagged protein from a T7 promoter upon induction with 1-thiogalactopyranoside (IPTG).
For complementation of the P. aeruginosa pilP::Tn5 strain (Table 4), pilP was amplified from the pSTBLUE-1-pilP vector and cloned into a modified pUC19 vector (pUCP20Gm) using the BamHI and SphI restriction sites in the multiple cloning site (pP5′ BamHI, pP3′ SphI). A similar strategy was used to amplify pilP from N. meningitidis genomic DNA (pilPNm) which was subsequently cloned into pUCP20Gm for expression in P. aeruginosa (NmP5′ BamHI-3, NmP3′ SphI-2). A version of pilPNm with an internal 6-His tag (pilPNm6Hi with an in-frame 6-His insertion between base pairs 222 and 223 in the pilPNm sequence) was ordered from GenScript (USA), and subcloned into the pUCP20Gm vector. A chimeric version of pilP, consisting of a fusion between the P. aeruginosa signal sequence and the N. meningitidis mature sequence, was generated using overlap extension PCR. Using primers, NmP17-23 and NmP3′ SphI-2, the mature sequence of pilPNm (bp 49–543) was amplified, and with primers PilP sig.seq-1 and 5′pUCP20Gm the signal sequence of P. aeruginosa pilP (bp 1–54) and 270 bp of the pUCP20Gm vector were amplified using 24 rounds of thermo cycling. The products of these reactions were then gel purified (Fermentas) and used as the template in a second reaction including primers pP5′ BamH1 and NmP3′ SphI-2. The amplified fragment was gel purified, digested with BamHI and SphI, and subsequently cloned into appropriately digested pUCP20Gm vector. An internal in-frame 3His tag (in the same location as describe above for wild-type pilPNm6Hi) was generated using divergent phosphorylated primers (NmPilP_3His fwd and NmP_3His rev) to make pilPPaNm3Hi. The amplified linear fragment was gel purified, and then blunt end ligated to re-circularize the plasmid with an internal 3His tag.
A cysteine to alanine site-specific mutant at the conserved lipidation site in pilP (Cys18) was made using the QuikChange site-directed mutagenesis protocol (Strategene) with primers pilPC18A fwd and pilPC18A rev. A complementation construct that encoded residues 72–181 of pilP (i.e. PilPΔ71) fused to the native signal sequence with Cys18 mutated to Ala (PilPssΔ71) was made using a set of divergent 5′ phosphorylated primers (pilPssΔ71 fwd and pilPssΔ71 rev). One primer annealed to bp 54–41 (5′–3′) on the anticodon strand with a 5′ overhang to create a Cys18 to Ala substitution, while the other annealed to bp 217–231 (3′–5′) on the coding strand. The PCR-amplified 5.3 kb fragment was gel purified using a Qiagen spin column gel extraction kit. The template DNA was subsequently digested with DpnI, and the fragment treated with 1 unit of T4 DNA ligase overnight at room temperature to self-ligate the blunt ends and re-circularize the vector. All constructs were verified by DNA sequencing (AGCT Corp., Toronto, Ontario).
To produce the PilN/PilO/PilP coexpression system, the generation of the pET-28a-pilNΔ44-pilOΔ51 coexpression vector was achieved by amplifying the entire region encoding pilNΔ44 and pilOΔ51 using pET-Duet-pilNΔ44-pilOΔ51 as a template (Sampaleanu et al., 2009) with primers containing a 5′ NcoI site and a 3′ SacI site. This fragment was cloned into an appropriately digested pET28a vector giving an untagged version of both PilN and PilO. The construct was verified by sequencing (AGCT Corp., Toronto, Ontario). The pET-Duet-pilOΔ51 construct was generated as a cloning intermediate in the production of the pET-Duet-pilNΔ44-pilOΔ51 using protocols described in Sampaleanu et al. (2009).
Expression and purification of PilPΔ71
The pET28a-pilPΔ71-6His expression vector was transformed into E. coli BL21 Codon Plus Cells (Stratagene). A 10 ml overnight culture was used to inoculate 1 l of LB broth that was grown to mid-log phase before induction with 1.0 mM IPTG. Cells were grown overnight at 25°C and harvested by centrifugation (6300 g, 4°C, 15 min). The cell pellet was resuspended in 20 ml of lysis buffer (50 mM Tris, 150 mM NaCl, pH 7.5) before being lysed by sonication (pulses of 30 s followed by 60 s rest period, for a total processing time of 1 min per 1 g of cells). Lysates were clarified by centrifugation (11 000 g, 4°C, 20 min). The clarified lysate was bound to 5 ml of Ni-nitrilotriacetic acid (Ni-agarose) resin pre-equilibrated with binding buffer (20 mM Tris pH 7.5, 150 mM NaCl, 20 mM imidazole) and washed with 25 ml of binding buffer. Protein was eluted step-wise from the column using 10 ml each of binding buffer with 75 mM, 150 mM and 300 mM imidazole. The 150 mM and 300 mM elution fractions were pooled and dialysed overnight in 20 mM Tris pH 7.5, 150 mM NaCl. Dialysed samples were analysed by SDS-PAGE and estimated to be > 90% pure. These samples were subsequently concentrated to 20 mg ml−1 and loaded on to an S75 gel filtration column equilibrated in the same buffer. Fractions containing PilPΔ71 were pooled and used for structural studies.
NMR structure determination
NMR studies on 1H,15N,13C PilPΔ71 were carried out at a protein concentration of 2 mM in 20 mM Tris pH 7.5, 150 mM NaCl. To produce uniformly labelled protein, cells were grown up in minimal media supplemented with 1 g of 15N-NH4Cl and 2 g of 13C-glucose per litre, and the protein was expressed and purified as described above. The NMR spectra were collected at 25°C on either a Varian Inova 500 MHz spectrometer equipped with a pulse field gradient unit and triple resonance probe, or a Varian Inova 600 MHz spectrometer equipped with a pulse field gradient unit and triple resonance cryoprobe. Backbone resonances were assigned using HNCACB and CBCA(CO)NNH triple resonance experiments and side-chain resonances were assigned using CCC-TOCSY, HCC-TOCSY and CT-HSQC experiments (Kanelis et al., 2001). A CN NOESY-HSQC was recorded with a mixing time of 150 ms (Pascal et al., 1994) to obtain NOE distance restraints for structural purposes. Data were processed with NMRPipe (Delaglio et al., 1995) and analysed with NMRView (Johnson, 2004; Kirby et al., 2004). Structures were calculated using cyana (Guntert et al., 2004), and the models were visualized using PyMOL (DeLano, 2002) and molmol (Koradi et al., 1996). 1H,15N heteronuclear NOE experiments were performed as described in Kay et al. (1989), and peak intensities were fit using FuDA (Hansen, 2008).
Structural and sequence analysis tools
The accessible surface area was calculated using the VADAR web server (Willard et al., 2003). Conserved residues were identified using a clustalw alignment of multiple sequences (Larkin et al., 2007) with identities ranging from 26% to 72% and an average identity of 31.4%. Residues > 90% conserved were considered highly conserved, residues that were < 20% accessible to solvent were considered buried, and residues with side-chains > 50% exposed were considered to be important for the surface properties of the protein. Protein properties were predicted using ProtParam from Expasy (Gasteiger et al., 2005).
Extracellular proteins were harvested as described previously (Voisin et al., 2007) and the gels were analysed as described in Ayers et al. (2009). Briefly plate-grown cells were harvested and after addition to PBS cells were vortexed to release surface-associated proteins. The proteins were precipitated with MgCl2 and analysed by SDS-PAGE.
Twitching motility assays were performed as described in Gallant et al. (2005) with 1% agar LB plates. The area of twitching observed was measured using ImageJ (NIH), and the average area calculated from at least 15 replicates. To control for plate variation, wild-type mPAO1 was assayed in replicate with every mutant strain tested, and the values reported are normalized to the mPAO1 values for each experiment.
Bacteriophage sensitivity assays
Sensitivity to pilus-specific bacteriophage PO4 (titre, approximately 108 PFU per ml) was tested qualitatively using a protocol adapted from Whitchurch and Mattick (Whitechurch and Mattick, 1994) as described in Koo et al. (2008).
Whole-cell lysate preparation and analysis
Lysates of whole cells were made by harvesting 3 ml of culture grown overnight at 37°C in LB broth with the appropriate antibiotic and resuspending the resulting pellet to a concentration of 100 µg cells per 100 µl of 1× SDS sample buffer [63 mM Tris HCl, 2% m/v sodium dodecylsulphate (SDS), 10% v/v glycerol, 0.005% m/v bromophenol blue]. Lysis was facilitated by heating the samples for 10 min at 95°C and vortexing. The samples were then loaded onto an SDS-PAGE gel. When probing for PilP, PilN, PilO, PilA and PilPNm 10 µl of sample was loaded into a 16% polyacrylamide SDS-PAGE gel, while 12% polyacrylamide SDS-PAGE gels were used for PilF, PilM and PilQ. Proteins were transferred onto polyvinylidene difluoride membranes (PVDF, Pall and GE Healthcare), and P. aeruginosa proteins were probed using purified rabbit polyclonal antibodies as described previously (Ayers et al., 2009). The secondary anti-rabbit antibody conjugated to alkaline phosphatase was used as per manufacturers directions (Bio-Rad). N. meningitidis PilP6His was probed using a monoclonal anti-hexa-histadine antibody pre-conjugated to alkaline phosphatase (Sigma). All blots were developed using 5-bromo-4-chloro-3′-indolylphosphate (BCIP) p-toluidine salt and nitroblue tetrazolium chloride (Pierce). The relative expression levels were determined by densitometry with ImageJ (NIH), all intensities were normalized to a cross-reactive band to control for differences in loading.
Crude membranes were harvested upon lysing cells (grown to an OD600 1.0) by passing twice through a French pressure cell at 18 000 psi. The lysate was clarified at 3000 g and the supernatant then subjected to a high-speed spin at 95 000 g for 90 min. The supernatant contained all the soluble proteins, and pellet contained the cell membranes and associated proteins. The membrane pellet was resuspended in 1× SDS-PAGE sample buffer and subjected to analysis by SDS-PAGE and Western blotting.
Separation of the inner and outer membrane fractions was achieved using the sucrose gradient fractionation method described previously (Koo et al., 2008), with the following modifications: (i) bacterial strains were grown to an OD600 of 0.7 before harvesting and the pellets were resuspended in 11 ml of lysis buffer, and (ii) all sucrose solutions were prepared fresh in 20 mM Tris pH 8.0. The membrane fractions (IM-inner membrane, MM-mixed membrane and OM-outer membrane) were mixed 1:1 with 2× SDS-PAGE sample buffer and 10 µl were loaded onto an SDS-PAGE gel. The individual proteins were probed with specific antibodies.
The NADH oxidase activity was assayed (Osborn et al., 1972) and the amount of 2-keto-3-deoxyoctonate (KDO) present in each sample determined (Lee and Tsai, 1999) to monitor the efficiency of the fractionation process in separating the inner and outer membrane components respectively.
Co-purification of PilNΔ44/PilOΔ51/PilPΔ18-6His
To co-purify a complex of the PilN/PilO/PilP proteins pET28a-pilPΔ18-6His and pET-Duet-pilNΔ44-pilOΔ51 expression vectors were transformed separately into E. coli BL21 DE3 cells (Novagen) and used to inoculate 5 ml of LB containing 50 µg ml−1 Kanamycin. Each overnight culture was used to inoculate 1 l of LB-kanamycin and the cells were grown at 37°C until OD600 of 0.6–0.7, when protein expression was induced by adding IPTG to a final concentration of 1 mM. The cells were grown for 4 h after induction at 37°C and were harvested by centrifugation (7700 g, 30 min, 4°C) and stored at −20°C until required. The frozen cells were thawed, resuspended and then mixed together in 40 ml (or 2 × 20 ml) of Tris buffer (20 mM Tris HCl pH 7.5, 600 mM NaCl) with one Complete EDTA-free protease inhibitor cocktail tablet (Roche) and the cells lysed by three rounds of homogenization at 15–20 kpsi. The cellular debris was removed by centrifugation (39 000 g, 30 min, 4°C). The cell lysate was mixed with 2 ml of Ni-agarose (Qiagen) at 4°C for 2 h. The resin was subsequently packed into a column and washed with 2 × 20 ml of Tris buffer containing successively 10 and 30 mM imidazole. Bound protein was eluted from the column in three 10 ml batches with Tris buffer and 75, 150 and 300 mM imidazole respectively. The 75 mM imidazole fraction was subsequently concentrated to 2 ml (Amicon unit with 10 kDa cut-off, Millipore), to a final protein concentration of 10–15 mg ml−1. The concentrated protein sample was applied to a Superdex S-200 column (GE Health) and eluted using 20 mM Tris HCl pH 7.5 and 150 mM NaCl. The protein concentration of the peak fractions was determined using the Bradford assay (Pierce). Analysis of the protein present in the peak fraction was performed by SDS-PAGE analysis.
Coexpression and co-purification of PilOΔ51 and PilPΔ18
The pET-DUET-pilOΔ51 and pET28a-pilPΔ18 vectors were co-transformed into BL21(DE3) codon Plus cells using a similar protocol as described previously. Protein expression was performed as already described using LB broth containing both Ampicillin and Kanamycin. Lysis, purification and analysis were carried out as previously described above.
Co-purification of PilNΔ44/PilOΔ51 and PilPΔ71-6His
The expression, lysis, co-purification and analysis were preformed as described above for the PilNΔ44/PilOΔ51/PilPΔ18-6His co-fractionation.
PDB accession codes
The atomic co-ordinates for the PilPΔ71 structure have been deposited in the Protein Data Bank, http://www.rcsb.org (PDB ID code 2LC4). Chemical shift assignments have been deposited in the BioMagResBank database (Accession No. 17598).
We thank Dr Scott Gray-Owen for the generous gift of N. meningitidis genomic DNA. This work was supported by Grant MOP 93585 from the Canadian Institutes of Health Research (CIHR) to P.L.H. and L.L.B. and by funding from the US Cystic Fibrosis Foundation to J.D.F. M.A. and S.T., and J.K. are recipients of graduate scholarships from the Canadian Cystic Fibrosis Foundation and the CIHR respectively. L.M.S. was funded, in part, by a fellowship from the CIHR Strategic Training Program in Membrane Proteins Associated with Disease. P.L.H. and L.L.B. are recipients of a Canada Research Chair and CIHR New Investigator award respectively.