Pilotin–secretin recognition in the type II secretion system of Klebsiella oxytoca

Authors

  • Tommaso Tosi,

    1. Institut de Biologie Structurale, Bacterial Pathogenesis Group, Université Grenoble I, 41 rue Jules Horowitz, 38027 Grenoble, France.
    2. Commissariat à l'Energie Atomique, 41 rue Jules Horowitz, 38027 Grenoble, France.
    3. Centre National de la Recherche Scientifique, 41 rue Jules Horowitz, 38027 Grenoble, France.
    Search for more papers by this author
    • These authors contributed equally to the work reported.

  • Nicholas N. Nickerson,

    1. Institut Pasteur, Unité de Génétique moléculaire, 25 rue du Docteur Roux, 75015 Paris, France.
    2. Centre National de la Recherche Scientifique URA2172, 25 rue du Docteur Roux, 75015 Paris, France.
    Search for more papers by this author
    • These authors contributed equally to the work reported.

  • Luca Mollica,

    1. Commissariat à l'Energie Atomique, 41 rue Jules Horowitz, 38027 Grenoble, France.
    2. Centre National de la Recherche Scientifique, 41 rue Jules Horowitz, 38027 Grenoble, France.
    3. Institut de Biologie Structurale, Protein Dynamics and Flexibility Group, Université Grenoble I, 41 rue Jules Horowitz, 38027 Grenoble, France.
    Search for more papers by this author
  • Malene Ringkjøbing Jensen,

    1. Commissariat à l'Energie Atomique, 41 rue Jules Horowitz, 38027 Grenoble, France.
    2. Centre National de la Recherche Scientifique, 41 rue Jules Horowitz, 38027 Grenoble, France.
    3. Institut de Biologie Structurale, Protein Dynamics and Flexibility Group, Université Grenoble I, 41 rue Jules Horowitz, 38027 Grenoble, France.
    Search for more papers by this author
  • Martin Blackledge,

    1. Commissariat à l'Energie Atomique, 41 rue Jules Horowitz, 38027 Grenoble, France.
    2. Centre National de la Recherche Scientifique, 41 rue Jules Horowitz, 38027 Grenoble, France.
    3. Institut de Biologie Structurale, Protein Dynamics and Flexibility Group, Université Grenoble I, 41 rue Jules Horowitz, 38027 Grenoble, France.
    Search for more papers by this author
  • Bruno Baron,

    1. Institut Pasteur, Biophysics of Macromolecules and their Interactions Platform, Proteopole and Structural Biology and Chemistry Department, rue du Dr. Roux, 75015 Paris, France.
    2. CNRS URA2185, rue du Dr. Roux, 75015 Paris, France.
    Search for more papers by this author
  • Patrick England,

    1. Institut Pasteur, Biophysics of Macromolecules and their Interactions Platform, Proteopole and Structural Biology and Chemistry Department, rue du Dr. Roux, 75015 Paris, France.
    2. CNRS URA2185, rue du Dr. Roux, 75015 Paris, France.
    Search for more papers by this author
  • Anthony P. Pugsley,

    1. Institut Pasteur, Unité de Génétique moléculaire, 25 rue du Docteur Roux, 75015 Paris, France.
    2. Centre National de la Recherche Scientifique URA2172, 25 rue du Docteur Roux, 75015 Paris, France.
    Search for more papers by this author
  • Andréa Dessen

    Corresponding author
    1. Institut de Biologie Structurale, Bacterial Pathogenesis Group, Université Grenoble I, 41 rue Jules Horowitz, 38027 Grenoble, France.
    2. Commissariat à l'Energie Atomique, 41 rue Jules Horowitz, 38027 Grenoble, France.
    3. Centre National de la Recherche Scientifique, 41 rue Jules Horowitz, 38027 Grenoble, France.
      E-mail andrea.dessen@ibs.fr; Tel. (+33) 4 38 78 95 90; Fax (+33) 4 38 78 54 94.
    Search for more papers by this author

E-mail andrea.dessen@ibs.fr; Tel. (+33) 4 38 78 95 90; Fax (+33) 4 38 78 54 94.

Summary

A crucial aspect of the functionality of bacterial type II secretion systems is the targeting and assembly of the outer membrane secretin. In the Klebsiella oxytoca type II secretion system, the lipoprotein PulS, a pilotin, targets secretin PulD monomers through the periplasm to the outer membrane. We present the crystal structure of PulS, an all-helical bundle that is structurally distinct from proteins with similar functions. Replacement of valine at position 42 in a charged groove of PulS abolished complex formation between a non-lipidated variant of PulS and a peptide corresponding to the unfolded region of PulD to which PulS binds (the S-domain), in vitro, as well as PulS function in vivo. Substitutions of other residues in the groove also diminished the interaction with the S-domain in vitro but exerted less marked effects in vivo. We propose that the interaction between PulS and the S-domain is maintained through a structural adaptation of the two proteins that could be influenced by cis factors such as the fatty acyl groups on PulS, as well as periplasmic trans-acting factors, which represents a possible paradigm for chaperone–target protein interactions.

Introduction

The type II and type III secretion systems (T2SS and T3SS) of Gram-negative bacteria secrete toxins and hydrolytic enzymes into the surrounding milieu or translocate them into other cells. The sole common element in these systems is the ring-like outer membrane component formed by proteins belonging to the secretin superfamily. Secretins associate into 12–14 subunit homo-oligomers to create a central pore that is large enough to allow the passage of folded proteins (Izoréet al., 2011; Korotkov et al., 2011). Structurally, secretins can be divided into two main domains: a C-terminal, membrane-embedded ‘secretin homology region’ that forms the channel (Chami et al., 2005), and an N-terminal periplasmic region that confers substrate specificity (Korotkov et al., 2009). Visualization of secretins from T2SS, T3SS, as well as type IV piliation systems (T4PS) by electron microscopy revealed that they share similar overall architectures (Collins et al., 2004; Chami et al., 2005; Reichow et al., 2010; Jain et al., 2011; Schraidt and Marlovits, 2011).

Many secretins depend on small lipoproteins, collectively referred to as secretin chaperones, for their correct targeting to and/or assembly in the outer membrane (Bayan et al., 2006). The term pilotin was coined for a subclass of these chaperones that target the secretin monomer to the outer membrane. The archetypical pilotin is PulS, a 12 kDa lipoprotein (d'Enfert and Pugsley, 1989) that targets monomers of the Klebsiella oxytoca secretin PulD to the outer membrane (Collin et al., 2011). PulS interacts with PulD through a 70 residue extension of the C-domain, the intrinsically unfolded S-domain (Daefler et al., 1997; Nickerson et al., 2011) and with the lipoprotein chaperone LolA (Collin et al., 2011) to transit the periplasm to the outer membrane, where PulD oligomerizes, apparently without assistance from any other factor (Guilvout et al., 2006; Collin et al., 2007; Guilvout et al., 2011). In the absence of PulS, PulD is prone to proteolytic trimming and degradation, and self-oligomerizes in the inner membrane, where it causes membrane perturbation, dissipation of the transmembrane electrochemical potential, and overexpression of the phage shock response protein PspA (Guilvout et al., 2006).

PulS shares significant sequence homology with the Erwinia chrysanthemi T2SS pilotin OutS, which targets the OutD secretin to the outer membrane and protects it from proteolysis (Shevchik and Condemine, 1998). Lipoprotein pilotins related to the PulS/OutS family are absent from most other characterized examples of the T2SS, and it is not clear how many of these secretins reach the outer membrane. Aeromonas and Vibrio species require GspA and GspB for efficient secretin assembly in the outer membrane. GspA and GspB are inner membrane proteins that do not share any sequence homology with PulS, and are involved in the reorganization of the peptidoglycan to allow secretin complex assembly (Strozen et al., 2011). T3SS and T4PS secretin chaperones, some of which have pilotin activities and most of which seem to be lipoproteins, differ completely in sequence and predicted secondary structure from PulS. For example, a central, hydrophobic tunnel-like cavity in secretin chaperone MxiM, which reportedly pilots the T3SS secretin MxiD to the outer membrane in Shigella flexneri (Lario et al., 2005), is filled by a lipid molecule that might be displaced upon binding to the C-terminal region of MxiD. Like PulS, MxiM remains attached to the outer membrane via its lipid anchor (d'Enfert and Pugsley, 1989; Okon et al., 2008) but, unlike PulD, MxiD oligomerization also requires a second secretin chaperone, MxiJ (Schuch and Maurelli, 2001). Like GspA and GspB, but unlike PulS, certain T4PS secretin chaperones promote secretin assembly in the outer membrane. For example, the Neisseria meningitidis secretin chaperone PilW, which is a lipoprotein, is required for T4PS PilQ multimerization, but does not function as a pilotin (Trindade et al., 2008; Szeto et al., 2011). The related lipoprotein PilF from the T4PS of Pseudomonas aeruginosa (Kim et al., 2006) was described as a pilotin, but its absence does not completely abolish the outer membrane localization of PilQ (Koo et al., 2008) and it might be involved, like PilW, in PilQ multimerization. Thus, even though secretin chaperones share several key features (small size, high pI and N-terminal lipidation), they play different roles in secretin targeting and assembly.

A more in-depth understanding of the molecular mechanisms underlying pilotin–secretin interactions is essential to understand secretin targeting and assembly in the outer membrane. In this work, we solved the high-resolution crystal structure of PulS, studied the role of a charged groove on the surface of PulS in S-domain binding and used nuclear magnetic resonance (NMR) to characterize further the interaction between the S-domain and PulS. These results confirm recent proposals that PulD interacts via its 30 C-terminal amino acids and demonstrate that this domain shows residual dynamics even in its bound form. The results reveal a unique structure for PulS and a novel mechanism of interaction with its secretin.

Results

PulS displays an all-helical fold

A PulS construct starting at residue Q19 (after the fatty acylated cysteine; PulSQ19) failed to produce crystals suitable for structural analysis. In contrast, a truncated variant of PulS, obtained by limited proteolysis with thermolysin and starting at residue V30 (PulSV30, Fig. 1A; see Experimental procedures) yielded crystals diffracting X-rays to 1.24 Å. The structure was solved by performing a SAD experiment at the European Synchrotron Radiation Facility (ESRF) in Grenoble employing a BaCl2-soaked crystal (see Table S1 for structure solution and refinement data).

Figure 1.

Crystal structure of PulSV30. A. Topology of PulSV30. The residues not present in the crystal structure are coloured in red. Cysteine residues 53 and 107, involved in the intramolecular disulfide bridge, are underlined in the sequence. The residues included in the PulS groove are marked with an asterisk. B. PulSV30 structure front view (left) and 180° rotation (right). The intramolecular disulfide bridge is coloured in yellow. Residues S41 (in double conformation), V42, G45, K74, R75 and L96, included in PulS groove, are coloured in purple.

The PulSV30 monomer folds into a C-shaped, slightly elongated bundle of four helices connected by short loops (Fig. 1). Helix α2 runs almost parallel to the longer N-terminal α1 helix, which in turn is flanked on two sides by perpendicular helices, α3 and α4. The disulfide bridge between C53 and C107, which is essential for PulS function (Pugsley et al., 2001), connects the first loop to α4 (Fig. 1B). A net of intramolecular van der Waals interactions between the side-chains of W77 and Q35, F117 and V42, F110 and L96, L85 and I69, S89 and R47 stabilizes the entire structure (Fig. S1). The hydrodynamic modelling of PulS from the crystal structure using Hydropro (Garcia de la Torre et al., 2000) and US-Somo (Brookes et al., 2010) gave calculated sedimentation coefficients of 1.4S and 1.5S respectively, in excellent agreement with the analytical ultracentrifugation experimental value (1.4 ± 0.1 S) (Nickerson et al., 2011). The dimensions and shape of PulS in the crystal are therefore consistent with that of the protein in solution. Because this structure is quite different from those of other secretin chaperones, we set out to investigate the PulD-recognition mechanism by site-directed mutagenesis, surface plasmon resonance (SPR) and NMR.

A charged groove of PulS is involved in PulD interaction

The analysis of the surface electrostatic potential of PulS (using Pymol's APBS plugin) showed the presence of a groove located on its concave side (Fig. 2A). This 27 × 7 Å groove is characterized by a positively charged patch that is closely flanked by the side-chain of S41 and includes the side-chains of K74 and R75, and that is in close proximity to a hydrophobic region defined by V42, G45 and L96 (Figs 1B and 2A). To examine the involvement of the PulS groove in the binding of the PulD S-domain, we created S41A, V42D, G45S, K74I, R75G and L96Q substitutions in MalE-PulSQ19, and tested their interaction to isolated PulD S-domain by size exclusion chromatography (Fig. 2B). In control experiments, MalE-PulS eluted as a single peak at 1.4 ml, while the stable complex between the S-domain and wild-type MalE-PulS eluted at 1.3 ml. MalE-PulS-S41A behaved like the wild-type complex, and the K74I variant was degraded upon purification; these variants were not studied further. Substitution of residues V42, G45, R75 and L96 abolished or severely reduced complex formation, as indicated by the elution of the MalE-PulS variants and the S-domain in separate peaks or, in the case of the G45 variant, eluted as an intermediate peak at 1.35 ml (Fig. 2B). Thus, residues V42, G45, R75 and L96 all influence the interaction with the S-domain in vitro.

Figure 2.

In vitro functional characterization of the PulS/S-domain complex. A. Side view (top) and front view (below) of the surface electrostatic potential of PulSV30. Negatively charged, positively charged and hydrophobic areas of the structure are coloured in red, blue and white respectively. The potential PulD S-domain binding groove is marked with an arrow. The residues located within the groove are indicated in bold. B. Size exclusion chromatography elution profile of PulS variants in the presence of PulD S-domain. The column was calibrated using Conalbumin (75 kDa, elution volume 1.27 ml), Ovalbumin (43 kDa, elution volume 1.38 ml), Carbonic Anhydrase (29 kDa, elution volume 1.55 ml) and Ribonuclease A (13.7 kDa, elution volume 1.76 ml). MalE-PulS elutes at 1.4 ml, consistent with a species of 50 kDa, while PulD S-domain elutes at 1.7 ml (upper panel). The higher apparent molecular weight of the S-domain (around 14 kDa instead of 9 kDa) is due to its lack of secondary structure. The wild-type complex elutes as a single peak at 1.3 ml, consistent with a macromolecule of 60–70 kDa (middle panel); MalE-PulS variants V42D, G45S, R75G and L96Q (purple) do not form a complex and elute as separate species from PulD S-domain (lower panel). Their retention volume is the same as wild-type MalE-PulS (1.4 ml) except for G45S (1.35 ml).

PulS variants have significantly reduced binding affinities for the S-domain

Circular dichroism (CD) analysis confirmed that the PulSV30 variants V42, G45, R75 and L96 (after thermolysin cleavage of MalE-PulSQ19) have similar secondary structures to the wild-type protein, being predominantly α-helical (see Fig. S2 for comparison of wild-type PulS and V42D variant). In SPR experiments, PulSV30 and PulSV30 variants were flowed over a surface onto which the MalE-PulD S-domain had been captured non-covalently. The SPR signal was monitored in real time (Fig. S3), and the concentration dependence of the steady-state response was used to calculate the equilibrium dissociation constants relative to wild-type PulS (Table 1, see Experimental procedures). The affinity of the V42D variant could not be quantified, as very little binding signal was detected above background. The G45 variant had an affinity approximately sevenfold lower than the wild-type protein, consistent with the intermediate elution observed by size exclusion chromatography. The affinities of the R75 and L96 variants were at least 50-fold lower than that of the wild-type protein. These observations confirm and refine the size exclusion chromatography data and indicate the critical role of V42.

Table 1.  Summary of the functionality of the PulS–PulD complex in the presence of PulS variants.
Protein/constructIn vitro interactionIn vivo phenotypes
SECaKd/KdwtbProtectioncPspAdSecretioneTDf
  • a. 

    Complex formation between MalE-PulS and PulD S-domain analysed by size exclusion chromatography (SEC).

  • b. 

    Equilibrium dissociation constants (Kd) measured by surface plasmon resonance were expressed as a ratio to wild type (Kd/Kdwt).

  • c. 

    Protection of PulD momoner from proteolysis (see Experimental procedures).

  • d. 

    Relative percent of PspA compared with empty vector control (100%).

  • e. 

    Secretion of PulA; secretion positive is relative values > 80%, secretion negative is relative values < 20%.

  • f. 

    TD – MalE-PulS transdominance determined from PspA production and PulA secretion assay.

  • NA, not applicable; ND, not determined.

VectorNANA100
PulS+1+48.6++
PulS V42DND68.7
PulS G45S−/+6.7+26++
PulS R75G>50+20.9++
PulS L96Q>50+31.8+

Residue V42 is crucial for PulS functionality

To assess the importance of the substituted amino acids in PulD stability and targeting, Escherichia coli strains carrying the complete K. oxytoca T2SS lacking PulS were transformed with plasmids encoding wild-type PulS or PulS variants V42D, G45S, R75G or L96Q, and were tested for secretion of pullulanase (the amylolytic enzyme secreted by this T2SS), induction of the phage shock response protein PspA (indicative of incorrect PulD targeting to the inner membrane) and PulD trimming and degradation (Guilvout et al., 2006; 2011) (Fig. 3A and B, Table 1). For these experiments, full-length PulS and PulS variants contain the wild-type signal peptide for correct outer membrane targeting and for fatty acylation of Cys18. Only one of the substitutions, V42D, had any measurable effect on these parameters. Strains producing PulS-V42D behaved almost identically to control strains transformed with the empty vector: pullulanase was poorly secreted, PspA levels were high and PulD was trimmed (presence of second band indicated by arrow in Fig. 3A) and was less abundant than in strains with wild-type PulS. Thus, PulS-V42D is non-functional in vivo. All of the other variants (G45S, R75G and L96Q) behaved like wild-type PulS (pullulanase was efficiently secreted, PspA levels were low, PulD levels were normal and the PulD monomer was not trimmed, Table 1).

Figure 3.

Functionality of PulS variants in vivo. Total cell extracts of E. coli PAP105(pCHAP1219) carrying all the Pul factor genes with a pulS deletion (ΔpulS) was complemented with pulS (PulS WT) or pulS variants (PulS – V42D, G45S, R75G and L96Q). An equivalent of 0.05 A600 units of culture for PulD (A) or 0.015 A600 units of culture for PspA (B) were separated by SDS-PAGE and immunoblotted with PulD- or PspA-specific antiserum respectively. PulD samples were treated with phenol for dissociation of PulD multimers (Mu) into monomer (Mo) subunits; arrows indicate presence of trimmed monomer. Relative PspA values are given in Table 1.

These results clearly show that V42 is essential for the PulS–PulD interaction and that its substitution abolished pilotin function. The fact that substitution of the other groove residues (G45, R75, L96) abolished or diminished complex formation in vitro but still permitted in vivo functionality indicates that they might play secondary roles in the interaction with the S-domain or that the effects of these substitutions are attenuated in vivo (see Discussion). To address this issue, in vivo trans-dominance experiments were performed using the MalE-PulSQ19 proteins, lacking the signal peptide and lipoprotein processing site. MalE-PulS is dominant over the wild-type copy of PulS, causing mistargeting of PulD to the inner membrane, blocking pullulanase secretion and inducing the phage shock response (Hardie et al., 1996). As expected, production of wild-type MalE-PulS abolished secretion and caused a massive phage shock response (high production of PspA protein; Fig. 4 and summarized in Table 1). Pullulanase secretion was normal and little PspA protein was detected with the V42 variant, confirming that this substitution completely disrupts PulS–PulD interaction. The G42 and R75 variants of MalE-PulS were both transdominant, indicating that MalE–PulS interacts with full-length PulD in vivo. However, the L96 variant gave a similar response to the V42 variant and empty vector control, i.e. efficient pullulanase secretion and low PspA levels, indicating inability to compete efficiently with wild-type PulS for binding to PulD. Together, these results reveal a range of defects exhibited by the four substitutions created, from severe (V42D) to mild (G45S).

Figure 4.

Transdominance of MalE-PulS variants. Total cell extracts of E. coli strain PAP7232 carrying all the Pul secretion factors on the chromosome, expressing MalE (lane 1, Vector), MalE-PulSQ19 (lane 2, PulS) or MalE-PulSQ19 variants (lane 3–6, MalE-PulSQ19) in trans. An equivalent of 0.015 A600 units of culture were separated by SDS-PAGE and immunoblotted with PspA-specific antiserum. Cultures grown to 1.5 A600 units were used to measure the relative percent of PulA secretion between whole and lysed cells (see Experimental procedures). Values are averaged from two independent experiments.

PulS–PulD complex formation: NMR characterization of PulD in the free and bound forms

A peptide corresponding to the S-domain (residues 590–660 of PulD) was prepared in 15N- and/or 15N/13C labelled form for NMR studies (Fig. 5A, see Fig. S4 for residue assignments). Assignment of the intrinsically disordered PulD S-domain revealed between 10% and 20% of alpha helical propensity in regions (598–601, 612–618, 623–635, 641–643, 654–660) as illustrated by the positive Secondary Structure Propensity (SSP) values (Fig. 5B). The S-domain exhibits classical features of a highly disordered domain, as seen from the narrow 1H peak dispersion of the 15N-HSQC spectrum from 7.5 and 8.5 ppm (Fig. 5A). The {1H}15N NOE values for the majority of the residues are < 0.5, signifying substantial flexibility. Residues 605–607 and 647–660 at the C-terminus show negative NOEs, indicating that these residues are highly mobile (Fig. 5C). Titration experiments were carried out to characterize the interaction of PulD S-domain with PulSV30 (Fig. 5A, D and E). In Fig. 5A, the HSQC spectrum of PulD S-domain shows significant peak shifts upon binding PulS. The titration results show a significant loss of peak intensity from residues 629–660 as the concentration of PulS increases (Fig. 5D, green and blue bars), and complete loss of peak intensity is seen with saturating levels of PulS (Fig. 5D, orange bars). Peak intensity changes as a function of PulS concentration indicate that only region 630–660 is implicated in the interaction. This is in agreement with recent results derived from limited proteolysis (Nickerson et al., 2011). Spin relaxation measurements show very high transverse relaxation rates, rising to 24 s−1 for residue 657 and up to 30 s−1 for unassigned peaks in the interaction site (Fig. 5E, blue bars). The remaining peaks in the interaction site are broadened beyond detection. These values are indicative of the presence of exchange contributions, which strongly imply that the interaction site of PulD S-domain remains dynamic in the complex with PulS.

Figure 5.

NMR Assignments and titration studies of PulD S-domain in complex with PulSV30. A. Superposition of the 1H-15N HSQC spectrum of the isolated PulD S-domain (red) and in complex with PulSV30 (blue). Peaks in the spectrum of the PulD S-domain that shift upon binding to PulSV30 are identified in Supporting information (Fig. S4). B. Secondary Structure Propensity [SSP (Marsh et al., 2006)] calculated from experimental Cα and Cβ chemical shifts of isolated PulD S-domain. Positive and negative values indicate α-helical and β-sheet propensity respectively. C. 1H-15N Heteronuclear Nuclear Overhauser Effect (nOe) measured on isolated PulD S-domain (14.1 Tesla, 298 K). Note that this construct contains a histidine tag at the N-terminal for which nOe values are not shown here. D. Ratios between peak intensities from resonances in the HSQC of free PulD S-domain compared with the HSQC spectra recorded in the presence of increasing amounts of PulSV30. The concentrations and ratios for the different titration points of PulD : PulS are: Red – 0.097:0.029 (ratio 0.3), green – 0.095:0.057 (ratio 0.6), blue – 0.091:0.011 (ratio 1.2), orange – 0.083:0.20 (ratio 2.4). Reduced intensity in the region 590–628 results from simple dilution, while the higher rate of intensity loss observed from residue 629–659 results from binding of PulD S-domain to PulS. E. 15N R2 relaxation rates of isolated PulD S-domain (red) and in complex with PulSV30 (blue) measured at 23.5 Tesla and 298 K. A recent comparison of hydrodynamic properties of folded domains with long flexible tails (Bae et al., 2009) suggests a rotational diffusion correlation time of the complex of PulS–PulD in the range of 12ns (giving R2 values of around 20 s−1 at 23.5 Tesla). The much higher values measured here are therefore indicative of the presence of exchange contributions.

Discussion

Here we report the first structural and functional characterization of a T2SS pilotin. We show that PulS has a novel, all-helical fold, completely different from all other secretin chaperones studied to date. Despite the fact that PulD is a large protein and oligomerizes into a dodecameric outer membrane complex (Nouwen et al., 1999; Chami et al., 2005; Guilvout et al., 2011), the PulS binding site is mainly or completely localized within the last 30 residues of the intrinsically unfolded S-domain [(Nickerson et al., 2011) and this study]. We propose that PulS interacts with the S-domain through a groove located on one side of the structure that is composed of both charged and hydrophobic regions. In vitro and in vivo experiments support this hypothesis (see below) and suggest that binding of PulS to the S-domain depends on a very limited area of the groove. Substitution of valine at position 42 by aspartate disrupts the complex both in vitro and in vivo. These results clearly suggest that the hydrophobic region around V42 plays a key role in secretin recognition. Because V42 is close to F117 (Fig. S1), a residue that provides hydrophobic interactions that potentially stabilize α4, its replacement probably changes the surface properties of a critical area of the groove dramatically, hampering the interaction to an extent that other stabilizing factors (see below) cannot restore functionality.

Despite the fact that substitutions of other residues in the groove (G45, R75 and L96) severely disrupted the interaction of non-lipidated PulS with isolated S-domain in vitro, as revealed by size exclusion chromatography and SPR, they did not exert a major effect on PulS function and, by inference, the interaction between lipidated PulS and full-length PulD monomers in vivo. Furthermore, strains expressing these variants produced less PspA than wild-type PulS in the complementation studies (Fig. 3 and Table 1), further underscoring the differences between the in vitro interaction data with PulSV30 and the functionality of lipidated PulS in vivo. MalE-PulS variants G45 and R75 interacted poorly or very poorly with isolated S-domain in vitro but they were transdominant in vivo, indicating that they could interact with PulD in the periplasm. In these cases, the same proteins behaved differently in vivo and in vitro. The MalE-PulS variant L96Q failed to interact with the S-domain in vitro and with PulD in vivo (absence of transdominance), but lipidated PulS L96Q was functional in vivo. The only substitution that completely abolished the interaction with the S-domain, as measured by SPR, involved the crucial V42, and was the only one to abolish in vivo function.

NMR relaxation experiments suggest that the interaction between PulS and PulD involves adaptation of the unfolded S-domain of PulD (Nickerson et al., 2011) to the PulS groove. The range of phenotypes observed in transdominance and transcomplementation assays with MalE-PulS or PulS variants bearing substitutions that, in vitro, dramatically reduced PulS–PulD S-domain interactions, suggests that defective interactions in this groove can be compensated (or aggravated) by cis factors, such as the lipids on PulS, the interaction between these lipids and LolA (Collin et al., 2011), and the N and C domains of PulD (Chami et al., 2005), or the trans effect of the environment in the periplasm, including non-specific periplasmic chaperones.

The mechanism by which PulS and PulD interact is clearly different from that of the only other well-characterized secretin–chaperone interaction, that between the T3SS secretin MxiD and its specific chaperone, MxiM. The C-terminal region of MxiD that interacts with MxiM is similar in size to the region of the S-domain of PulD that interacts with PulS (Daefler et al., 1997; Lario et al., 2005; Okon et al., 2008; Nickerson et al., 2011) and, like the S-domain, is intrinsically unfolded. However, unlike the S-domain, which appears to acquire only limited secondary structure and undergoes conformational exchange on the micro-millisecond timescale upon binding to PulS, the MxiM-binding peptide of MxiD clearly folds into a well-ordered structure with a hydrophobic helix that caps a lipid binding cavity in its chaperone (Lario et al., 2005; Okon et al., 2008). Free lipids are not involved in the interaction between PulS and the S-domain, the interaction relying mainly on the surface properties of the PulS groove. These structural and mechanistic results argue against the existence of a generalized, universal mechanism for secretin interactions with their cognate chaperones. The general principles established here and previously (Nickerson et al., 2011) for the interaction between a disordered peptide and a binding groove on its cognate chaperone that can be modulated by cis and trans factors is a possible paradigm for other chaperone–protein interactions.

Experimental procedures

Crystallization and structure determination of PulSV30

DNA encoding PulS starting at residue Q19 (PulSQ19) was cloned without the N-terminal signal peptide and N-terminal cysteine residue (amino acids 1 to 18) in-frame with a His6-MalE tag, and was expressed and the protein purified as previously described (Nickerson et al., 2011). Purified protein was proteolysed overnight at 4°C with thermolysin (Sigma) at a protease : protein ratio of 1:1000 (w/w) yielding a PulS variant lacking the first 29 residues (PulSV30); the size and homogeneity of which were confirmed by mass spectrometry and N-terminal sequencing. PulSV30 was further purified by size exclusion chromatography in a Superdex 200 10/60 Prep grade column (GE Healthcare) in a buffer composed of 30 mM HEPES pH 7.5, 150 mM NaCl.

Purified PulSV30 was concentrated to 10 mg ml−1 and crystallized in collaboration with the High Throughput Crystallization (HTX) Lab of the Partnership for Structural Biology in Grenoble. Initial hits were obtained using 2-Methyl-2,4-Pentanediol as a precipitant. Data collection of manually reproduced native crystals was performed at beamline ID14-EH1 at the European Synchrotron Radiation Facility (ESRF; Grenoble). 1080 images were indexed, scaled and merged using XDS (Kabsch, 2010). The native crystals diffracted to 1.24 Å and belonged to space group I222 (Table S1). Structure factor amplitudes were obtained from the observed intensities using TRUNCATE (Cowtan et al., 2011). Phasing was performed using the single anomalous dispersion (SAD) method on diffraction data obtained for PulSV30 crystals grown in the presence of 10 mM BaCl2. An isomorphous dataset of 360 images was collected at a X-ray energy of 9 kEv at beamline BM14 (ESRF) and processed as the native dataset. AUTOSHARP (Vonrhein et al., 2007) was used to calculate experimental phases from two barium sites and for density modification. ARP-WARP (Langer et al., 2008) was used to build an initial model that was completed manually using COOT (Emsley et al., 2010) and refined against the native dataset using REFMAC5 (Murshudov et al., 2011).

NMR studies

For all the NMR experiments, recombinant 15N- and 15N/13C-labelled PulD S-domain was produced as above in M9 minimal bacterial growth media (Miller, 1972) and purified according to Nickerson et al. (2011). Protein concentrations were determined by measuring the UV light absorbance at 280 nm with bovine serum albumin as a standard.

NMR spectra of the free and bound proteins were recorded at 298 K on 600 and 800 MHz (Varian) instruments equipped with inverse triple-resonance cryoprobe and pulsed-field gradients. The 1H,13C,15N assignments of the backbone resonances of the S-domain were obtained from a series of BEST type triple resonance experiments [intra-residue HNCA, HN(CO)CA, HNCO, HN(CA)CO, intra-residue HNCACB and HNCOCACB (Lescop et al., 2007)] performed at pH 6.0. The protein concentration was 0.3 mM. All spectra were processed with NMRPipe (Delaglio et al., 1995) and analysed using Sparky (Goddard and Kneller, 2002). The program MARS was used for automatic assignment of spin systems (Jung and Zweckstetter, 2004).

15N-backbone spin relaxation experiments were performed at 298 K on a 1000 MHz (Bruker) spectrometer (R2) and on a 600 MHz (Varian) spectrometer (heteronuclear nOe) equipped with a cryogenic triple resonance probe. The measurements were performed on a 15N-labelled sample of the S-domain at a concentration of 0.3 mM, while the PulD–PulS complex was at a molar ratio of 1:3 (0.1 mM:0.3 mM). 15N R2, and (1H)-15N heteronuclear NOE experiments were performed using standard pulse sequences (Farrow et al., 1994). R2 spectra were acquired with 9 relaxation delays (14, 28, 57, 72*, 86, 130, 172*, 200 and 230 ms – asterisk refers to repeat measurements). Relaxation rates were extracted from the peak intensity decay curves using a two parameter exponential curve fit.

Titration of 15N labelled PulD with unlabelled PulSV30 was performed in order to identify the interacting region of PulD. PulS and PulD samples used for the titrations were concentrated at 0.25 mM and 0.1 mM, respectively, and titrated in 240 ml volume Shigemi tubes. 15N-labelled S-domain was titrated with unlabelled PulS up to a molar ratio 1:2.

Mutagenesis of pulS

PulS residues S41, V42, G45, K74, R75 and L96 were selected for mutagenesis due to their location within the concave region of the PulS cleft, which suggested that they could play a role in partner recognition. PulS variants S41A, V42D, K74I, R75G, G45S, L96Q were constructed according to the QuikChange site-directed mutagenesis kit protocol (Stratagene) using appropriately designed mutagenic primers. Residues were mutated into variants that would modify their charge characteristics and/or their hydrophobic nature. Mutated plasmids were sequenced to verify the absence of secondary mutations. PulSQ19 variants were produced and purified for analysis as described above for PulSQ19 (Nickerson et al., 2011).

Size exclusion chromatography of PulS/S-domain complexes

DNA encoding the PulD S-domain (residues 590–660) was cloned into a pQE30 vector with an N-terminal His6-tag and used to transform cells of E. coli strain BL21(λDE3) [E. coli B ompT hsdS gal (λDE3) endA] carrying pDIA17 [pACYC184 derivative carrying lacIq under tet promoter control (Boyd et al., 2000)]. Production and purification of the protein was as previously reported (Nickerson et al., 2011). Purified MalE-PulSQ19 or PulS variants were mixed with purified PulD S-domain at a 1:1 molar ratio. After 20 min of incubation at room temperature, the sample was loaded onto a Superdex 75 5/150 GL column (GE Healthcare) in 30 mM Tris pH 8.0, 150 mM NaCl.

SPR

Assays were performed on the Biacore 2000 instrument (GE Healthcare) equilibrated at 25°C in 20 mM Tris-Cl pH 8.0, 150 mM NaCl supplemented with 0.005% Tween 20. The monoclonal α-MalE mAb565 antibody (England et al., 1997) was immobilized on three flow cells of a CM5 sensor chip (GE Healthcare) to a level of 12 000–15 000 resonance units (RU; RU ∼ pg mm−2). Different densities (∼ 1200–2000 RUs) of ligand [MalE-PulD S-domain (Nickerson et al., 2011)] were captured on the surface of two different flow cells and one free flow cell was maintained as an α-MalE blank. PulSV30 or PulSV30 variants were injected over the surface loaded with MalE-PulD S-domain for 4 min at 50 µl min−1 to allow steady-state equilibrium to be reached. Concentration ranges were fine-tuned to allow determination of affinities for all PulS variants. Profiles were double referenced by subtracting the signals from the reference surface and the blank injections of running buffer using the Scrubber 2.0 software (BioLogic Software). Data analysis and affinity calculations were carried out using the BIAevalution software (GE Healthcare). Under these conditions, wild-type PulS had an affinity of 16 nM, lower than the previously determined affinity using a competition experimental set-up (Nickerson et al., 2011).

Circular Dichroism (Far-UV) Spectroscopy

Far-UV circular dichroism spectra were performed on PulSV30 and PulSV30 variants to verify the absence of gross impact of residue substitutions on secondary structure. CD spectra were recorded between 195 and 260 nm on an Aviv 215 spectropolarimeter (Aviv Biomedical), using a cylindrical cell with a 0.01 cm path length and an averaging time of 1 s per step, with protein samples at 0.4–0.6 mg ml−1 in 20 mM Tris-Cl pH 8.0, 150 mM NaCl. Three consecutive scans from each sample were merged to produce an averaged spectrum and corrected using buffer baselines measured under the same conditions. Data were normalized to the molar peptide bond concentration and path length and expressed as Mean Residue Ellipticity ([θ] degree·cm2·dmol−1).

In vivo complementation and transdominance experiments

For complementation studies, E. coli strain PAP105 [Δ(lac-pro) F′ (lacIq1ΔlacZM15 proAB+ Tn10)] was transformed with pCHAP1219 [a derivative of pCHAP231 carrying the full pul operon with a pulS deletion (pulS::Tn5) (d'Enfert and Pugsley, 1989; Possot et al., 2000)], and complemented with either wild-type PulS [pCHAP580 (Possot et al., 2000)] or PulS variants. For transdominance studies, strain PAP7232 carrying all the pul secretion factors on the chromosome [malP::(pulS pulAB pulCO) (Possot et al., 2000)] was transformed with vector pMal-p2 (New England BioLabs) or pMal-p2 expressing MalE-PulS or MalE-PulS variants. For all experiments, strains were grown under aeration at 30°C in LB broth or LB with 10% M63 minimal media, with 0.4% maltose for induction of the secretion machinery and appropriate antibiotics (chloramphenicol, 25 µg ml−1; ampicillin, 100 µg ml−1; tetracycline, 10 µg ml−1). Enzymatic assays measuring pullulanase secretion were as previously described (Michaelis et al., 1985). Pullulanase secretion was expressed as the proportion of the enzyme accessible to its substrate (pullulan) in whole cells compared with that in cells lysed with 0.5% octylpolyoxyethylene. Each assay was performed separately at least twice. For Western blots, exponentially growing cells were harvested and dissolved in SDS loading buffer or treated with phenol to dissociate PulD multimers. Samples were separated by 10% or 12% SDS-PAGE, transferred to nitrocellulose membrane, probed with protein-specific antibodies and detected using enhanced chemiluminescence (GE Healthcare). For quantification, Western blots were developed using the ECL Plus Western Blotting Detection System (GE Healthcare) and chemifluorescence visualized in a Molecular Dynamics Storm 840 imaging system using ImageQuant software (GE Healthcare). Bands were quantified by densitometry and background subtracted using the ImageJ software (Rasband, 1997–2011).

Acknowledgements

The authors wish to thank Ingrid Guilvout and Séverine Collin for their helpful discussions and members of the Molecular Genetics Unit at the Institut Pasteur for their support. We are also grateful to J. Marquez and the HTX Lab team (Partnership for Structural Biology, Grenoble, PSB) for access to and help with high throughput crystallization; Hassan Berlhali (EMBL Grenoble) for help with collection of PulS-Barium data and Carlos Contreras-Martel for his support in solving PulSV30 structure; the European Synchrotron Radiation Facility (ESRF, PSB) for access to beamlines; Bertrand Raynal (Centre of Biophysics of Macromolecules and their Interactions, Institut Pasteur) for hydrodynamic modelling calculations; and Jacques d'Alayer (Protein Microsequencing and Analysis Platform, Institut Pasteur) for N-terminal sequencing. This work was supported by Grant ANR-09-BLAN-0291. N.N.N. was supported by a postdoctoral fellowship from the Canadian Louis Pasteur Foundation during part of this work. The co-ordinates of PulS have been deposited in the Protein Database under code 4A56.

Author contributions

T.T, N.N., L.M., M.R.J. and B.B. did experiments; T.T., N.N., M.B., P.E., A.P.P. and A.D. analysed the data; T.T., N.N., A.P.P. and A.D. wrote the paper.

Conflict of interest

The authors declare that they have no conflict of interest.

Ancillary