Insights into pilus assembly and secretion from the structure and functional characterization of usher PapC

Overview of PapC Core Structure and Its Functional Implications.

The β-barrel of PapC core consists of 24 β-strands, the most β-strands and the largest β-pore for a single-protein β-barrel observed (Fig. 1). The β-barrel is kidney-shaped, with the β1 strand H-bonding to β24, positioning both the N-terminal and the C-terminal domains of PapC in the periplasmic region for participation in pilus assembly and secretion (Fig. 1). The PapC β-barrel is ≈43 Å in height and has a larger pore size on the periplasmic side (≈46 × 28 Å) than on the extracellular side (≈40 × 24 Å). Based on available crystal structures of pilus subunits (PapG, PapE, PapK, and PapA) and chaperone (PapD), the size of the PapC pore is only large enough to allow the passage of individual folded pilus subunit in an upright pilus polymerization direction (Fig. 2), but is insufficient for accommodating the chaperone or any of the chaperone–subunit complexes. Thus, DSE must occur in the periplasmic region at the base of the β-barrel. In addition, the height of the β-barrel (43 Å) is shorter than any of the pilus subunits. Thus, the β-pore is fully occupied by 1 pilus subunit, which is assumed to be secreted out during each polymerization cycle (Fig. 2).

A β-sandwich plug domain, formed by residues 259–335, is laterally situated in the lumen of the β-barrel; it blocks the passage of proteins or other solutes through the pore in the pilus nonproductive state (Fig. 1). The plug domain of PapC is formed by a long sequence insertion between strands β6 and β7, rather than at the N terminus as found in other outer membrane proteins (34). Sealing of the β-pore by the plug domain is presumably beneficial to bacteria as it prevents the entry of toxic substances and leakage of important cellular components in the pilus nonproductive state. Thus, the current PapC core structure only presents a snapshot of the usher structure in the pilus nonproductive state, suggesting that the opening of the channel requires the dislocation of the plug domain, presumably caused by binding of PapD–PapG to the usher.

Head-to-Head Dimers in Crystals Suggest Possible Structure of Solution Dimers.

Previous structural information and biochemical evidence suggest that ushers form and function as dimers both in vitro and in vivo (27, 31, 32). To further confirm the oligomerization state of PapC in solution, we carried out size exclusion chromatography and glutaraldehyde cross-linking (35) (Fig. S2 C and D). These experiments consistently support that PapC forms a dimer in solution. In the crystal structure reported by Remaut et al. (15) both the C2 and F222 crystal forms of PapC130–640 contained a twinned dimer formed around a crystallographic 2-fold axis, but there were no direct interactions found between the 2 protomers (Fig. S3). Cryo-EM studies also suggest a similar dimerization fashion for PapC in lipids (15, 31). By contrast, our crystals of the PapC core do not contain a twinned dimer assembled in parallel via the flat side of the 2 β-barrels. Rather, 4 PapC core protomers are assembled around a 4-fold crystallographic symmetry axis with strands β2 and β3 from one protomer approaching β-strands β20 and β21 of another protomer (Fig. 3A), but there are no close lateral contacts (within 4.5 Å distance) found between the 4 protomers. However, the 2 PapC core protomers in head-to-head orientation that constitute 1 asymmetric unit are connected through a symmetric, antiparallel β-strand interaction via residues 199–205 of β4 (Fig. 3B). This finding is very unusual because, in general, β-barrel proteins avoid edge-to-edge β-sheet docking by having no edges exposed and have continuous β H-bonding all of the way around a cylindrical barrel (36). This packing feature is conserved in all 3 published crystal forms (I422, C2, and F222) of PapC, suggesting that a specific interaction is mediated by docking of exposed β4 strand edges between the 2 protomers. The nonspecific detergent-mediated interaction between the monomers in the twinned dimers and the specific interaction between the head-to-head dimer conserved in 3 different crystal forms suggests that PapC most likely forms a head-to-head dimer in the detergent-solubilized state in solution.

Functional Characterization of Usher PapC Using in Vitro Reconstitution of Fiber Assembly.

Previous in vitro reconstitution of type 1 pilus assembly demonstrated that efficient pilus assembly requires no cellular components other than pilus subunits (FimA and FimH), chaperone (FimC), and usher (FimD), and both pilus assembly and secretion proceed independent of cellular energy (33). To test whether usher PapC functions as a pilus assembly catalyst and further dissect PapC domain functions, we overproduced and purified WT PapC, PapC-ΔNT (residues 1–125 deleted), PapC-ΔCT (residues 650–809 deleted), and PapC-Δplug (residues 259–335 deleted, PapC-Δplug reconstituted in lipid bilayer shows apparent channel-forming activity, suggesting that it is well folded) from E. coli. strain sf120 outer membranes and analyzed their ability to polymerize PapE, the major structural subunit of the P pilus tip, from purified monomeric PapD–PapE complex. We used the PapD–PapE complex as substrate to replace the PapD–PapA complex for 2 reasons: First, although PapA is the major rod subunit of P pilus, we had difficulties obtaining homogeneous monomeric PapD–PapA complex. The full-length PapA (in complex with PapD) has a strong propensity for self-polymerization/aggregation even at low protein concentration (37), whereas the PapD–PapE complex forms stable and homogeneous monomers even at high protein concentration. Second, PapE subunits, similar to PapA and FimA, are self-complemented and are the major tip subunits of P pilus (10–100 copies in the tip) (Fig. S1) (11, 15). Therefore, the PapD–PapE complex is suitable as substrate for detection by negative staining with electron microscope upon polymerization (11).

P Pilus Assembly Proceeds in a Way Similar to Type 1 Pilus in Vitro.

We performed in vitro reconstitution of P pilus assembly experiments in a way similar to that described by Nishiyama et al. (33) for type 1 pilus. WT PapC, preincubated with the PapD–PapG complex in equimolar concentrations, was added to a solution containing the PapD–PapE complex (for 1 h at room temperature), and PapE filaments with lengths of ≈0.1–0.15 μm were generated and could be clearly detected by negative staining electron microscopy, as shown in Fig. 4 A and B. The addition of PapF subunits, which link PapG to PapE in vivo, was not required. As the individual PapE subunit is ≈6 nm in length, the generated PapE filaments consist of ≈20 copies of PapE subunits. Although generated PapE filaments are longer than the pilus tip of natural bacteria, they are close in length to the pilus tips of PapE overexpressed piliated bacteria (11). In contrast, without preincubation of PapD–PapG, PapC does not efficiently catalyze the formation of PapE filaments under the similar experimental condition (Fig. 4C). As a control, in the absence of either usher PapC (Fig. S4A) or substrate PapD–PapE (Fig. S4B), there are clearly no filaments generated. Therefore, analogous to FimC–FimH, PapD–PapG is also critical for the function of PapC as a PapE polymerization catalyst, and the assembly of P pilus proceeds in a way similar to that of type 1 pilus in vitro (33).

Both the Plug Domain and the C-Terminal Domain of PapC Are Required for Pilus Subunit Assembly.

Thanassi and coworkers (27, 32) showed that the N-terminal domain (residues 1–124) of PapC is responsible for recruiting chaperone–subunit complexes, shifting chaperone–subunit complexes to the usher C terminus and/or preparing subunits for donor strand exchange, and both the N-terminal domain and the C-terminal domain (residues 667–809) of PapC are required for pilus biogenesis. To test whether our in vitro assembly experiments are consistent with those observations in bacteria, we tested the polymerization capability for both PapC-ΔNT (residues 1–125 deleted) and PapC-ΔCT (residues 667–809 deleted) mutant proteins. As shown in Fig. 4D, the PapC-ΔCT mutant protein (preincubated with PapD–PapG) completely lost its polymerization capability, and no PapE filaments were generated, suggesting that the C-terminal domain is required for filament assembly both in vitro and in bacteria. However, deletion of the N-terminal domain of PapC (PapC-ΔNT) did not abrogate its capability on PapE polymerization, and significant amount of filaments were generated as shown in Fig. 4E. This inconsistency between pilus assembly in vivo and in vitro may result from the protein concentrations used in our reconstitution experiments (PapC-ΔNT: 0.1 μM; PapD–PapG: 0.1 μM; PapD–PapE: 5 μM), which are significantly higher than the corresponding protein concentrations in the periplasmic region of bacteria. It is likely that under our in vitro conditions, the concentration of chaperone–subunit complexes (both PapD–PapG and PapD–PapE) in proximity to the usher is high enough for PapE polymerization to proceed efficiently without the help of N-terminal domain. Thus, the N-terminal domain of PapC may not be required for pilus subunit polymerization and subsequent secretion.

To further test whether the plug domain participates in pilus assembly, we examined the catalytic capability of PapC-Δplug protein. Addition of preincubated PapC-Δplug/PapD–PapG protein complex into the pool PapD–PapE did not catalyze the formation of PapE filaments at all (Fig. 4F), demonstrating that the plug domain is required for PapE polymerization. These observations together suggest that pilus assembly requires the participation of both the plug domain and the C-terminal domain of PapC. Thus, the plug domain has a dual function: closing the channel in pilus nonproductive state and participating in pilus assembly.

Usher PapC Monomers may Be Sufficient for Fiber Assembly.

Remaut et al. (15) propose that FimD dimers only use 1 pore for fiber formation, but 2 N-termini of the twinned FimD dimer are required for the progressive stepwise growth of pilus fiber. Our observations do not support this model, but support the hypothesis of pilus assembly and secretion in a nonspecific manner upon targeting of chaperone–subunit complexes to the usher (27, 32). The catalytic activity of PapC-ΔNT in pilus assembly demonstrated that the N-terminal domain of the usher may be mainly responsible for increasing the local concentration of chaperone–subunit complexes, but does not participate in subsequent pilus polymerization and secretion. So and Thanassi (32) have shown that usher FimD, when coexpressed with chaperone-adhesin FimC–FimH, is able to rescue a PapC-ΔCT mutant for the assembly of P pili, and that FimD was copurified with PapC when coexpressed in bacteria. Because the N-terminal domain of FimD shows no affinity for P pilus subunits, this observation indicates that pilus assembly and secretion in bacteria does not require 2 N-terminal domains, and the activated usher FimD is competent for P pilus assembly and secretion using the Pap chaperone–subunit complexes provided by PapC-ΔCT. These observations suggest that PapC monomers may be sufficient for pilus assembly and secretion. Nonetheless, it is likely that usher tends to form oligomers (dimers), like trimeric porins, in the outer membrane, but they are not functionally required (38).

As mentioned, PapC most likely forms head-to-head dimers in solution but nevertheless efficiently catalyzes the polymerization of PapE subunits. Because the head-to-head usher dimer is assembled in a reverse orientation, the presence of the other copy of the usher at the extracellular side must impede the growth of the pilus in vitro. Therefore, the growth of the fiber in vitro must first disrupt the head-to-head dimer. We propose that the binding of chaperone–adhesin complex to the usher may cause a conformational change of the usher PapC, which, on the one hand, may result in the straightening of the inward β hairpin, complementing the exposed β4 edges, which replaces the other copy of PapC; on the other hand, the binding may cause the dislocation of the plug domain, priming pilus assembly and secretion.

Protein Preparation, Crystallization, Data Collection, and Structure Determination.

For detailed information regarding the expression and purification of PapC, PapC mutant proteins, PapD–PapG and PapD–PapE complexes, and crystallization, data collection, and structure determination of PapC core, see SI Text.

Filament Assembly Assay and Electron Microscopy.

According to experimental requirements, WT PapC or PapC deletion mutant proteins (PapC-ΔNT, PapC-ΔCT, and PapC-Δplug) were preincubated with an equimolar amount of the PapD–PapG complex for 3–4 h in 20 mM Tris·HCl (pH 8.0), 150 mM NaCl, and 0.02% (wt/vol) DDM at room temperature. Filament assembly was initiated by the addition of PapD–PapE (final concentration: 5 μM) to the preincubated protein mixtures (except controls) (the final protein concentrations were ≈0.1 μM for PapC and PapD–PapG complex). The reactions were performed at room temperature. The buffer contained 20 mM Tris·HCl (pH 8.0), 150 mM NaCl, and 0.02% (wt/vol) n-dodecyl-beta-d-maltoside (DDM) for all of the proteins before mixing. Protein concentrations were determined via the specific absorbance at 280 nm. The following molar extinction coefficients were used: PapC, 156,330 M⁻¹·cm⁻¹; PapC-Δβ hairpin, 157,820 M⁻¹·cm⁻¹; PapC-Δplug, 151,860 M⁻¹·cm⁻¹; PapC-ΔNT, 140,735 M⁻¹·cm⁻¹; PapC-ΔCT, 135,235 M⁻¹·cm⁻¹; PapC-ΔNT-ΔCT, 132,455 M⁻¹·cm⁻¹; PapD–PapE, 40,590 M⁻¹·cm⁻¹; and PapD–PapG,108,555 M⁻¹·cm⁻¹.

All filament assembly reactions lasted for 1 h at room temperature, and samples were removed at the end of the assembly reactions, negatively stained with 1% (wt/vol) uranyl acetate, and imaged with a FEI Tecnai F20 microscope at the University of Texas Southwestern Molecular and Cellular Imaging Center. Micrographs were recorded at an accelerating voltage of 120 kV with a magnification of either SA = 8,000 or SA = 60,000. The pictures were taken with an exposure time of 1.0 s. For length estimates, the lengths of 100 filaments were measured from a 5-fold dilution of the original reaction mixture, and the lengths of the filaments are ≈120 nm.