Unraveling the molecular basis of subunit specificity in P pilus assembly by mass spectrometry

Abstract

P pili are multisubunit fibers essential for the attachment of uropathogenic Escherichia coli to the kidney. These fibers are formed by the noncovalent assembly of six different homologous subunit types in an array that is strictly defined in terms of both the number and order of each subunit type. Assembly occurs through a mechanism termed “donor-strand exchange (DSE)” in which an N-terminal extension (Nte) of one subunit donates a β-strand to an adjacent subunit, completing its Ig fold. Despite structural determination of the different subunits, the mechanism determining specificity of subunit ordering in pilus assembly remained unclear. Here, we have used noncovalent mass spectrometry to monitor DSE between all 30 possible pairs of P pilus subunits and their Ntes. We demonstrate a striking correlation between the natural order of subunits in pili and their ability to undergo DSE in vitro. The results reveal insights into the molecular mechanism by which subunit ordering during the assembly of this complex is achieved.

Gram-negative bacteria assemble a variety of surface organelles to mediate recognition and attachment to host tissues (1, 2). These pili or fimbriae are multisubunit filamentous appendages that typically present an adhesin at their tip that binds to a specific host receptor and is essential for infectivity (3, 4). Pili are formed by at least four known pathways in Gram-negative bacteria, one of which, the chaperone–usher pathway, assembles a large number of adhesive pili essential for pathogenesis (5). In this pathway, pilus subunits are synthesized in the cytoplasm and transported, via the Sec machinery, into the periplasm (2). Here, the same periplasmic chaperone binds each subunit, capping its interactive surfaces and maintaining it in an assembly-competent conformation (Fig. 1A) (6–8). The chaperone–subunit complexes are then taken to a site of assembly consisting of an outer-membrane protein, the “usher,” where the chaperone is released and subunit polymerization occurs (Fig. 1A) (9).

Fig. 1.

The P pilus, DSC, and DSE complexes and conserved Nte-binding sequences. (A) Schematic diagram of P pilus biogenesis. Subunits (colored), bound to a chaperone (PapD, brown) in the periplasm, assemble at the usher (PapC, gray) in the order PapG, PapF, PapE, PapK, PapA, and PapH. (B) Topology diagram of the DSC interaction. In this complex, the chaperone G₁ strand (brown) completes the Ig fold of each subunit (green) by binding into the hydrophobic groove between strands A and F (indicated). The Nte is unstructured. (C) Crystal structure of the DSC interaction between the subunit PapK (shown in space fill) in complex with the chaperone PapD (brown) (PDB ID code 1PDK). The P1–P4 residues and the exposed (unoccupied) P5 pocket are indicated. (D) Sequences of Ntes of different Pap subunits. Full-length Ntes were used for PapA, PapK, PapE, and PapF, whereas the long N terminus of the Nte of PapH [which does not participate in donor-strand exchange (18)] was truncated to aid solubility. The peptides used in this study to mimic each Nte (see Methods) are outlined in black. Residues composing the hydrophobic binding motif (P2–P5) are shown in red. (E) Topology diagram of the DSE interaction. The Nte from one subunit (green) completes the Ig fold of the subunit previously assembled (yellow) by forming a new intermolecular β-strand. (F) Crystal structure of the subunit PapE (shown in space fill) bound to the Nte of PapK (green) (PDB ID code 1N12). The P2–P5 residues that bind to the P2–P5 pockets are indicated.

The structures of the chaperones and pilus subunits (“pilins”) have been solved for several chaperone–usher pilus systems (10). The chaperone consists of two immunoglobulin (Ig)-like domains (11). In the periplasm, each pilus subunit forms an incomplete Ig fold in which the C-terminal β-strand is absent. This leaves a hydrophobic groove into which the chaperone inserts the G strand of its domain 1 (strand G₁), completing the subunit's fold (Fig. 1 B and C) (12–15). In this binding interface, four residues from the donated chaperone β-strand (named P1–P4 residues, see below) insert into complementary hydrophobic pockets (named P1–P4 pockets) formed within the pilin groove, leaving another pocket (P5) in the groove exposed (Fig. 1C). This interaction is known as “donor-strand complementation” (DSC). With the exception of the adhesin, each subunit also possesses an unstructured N-terminal extension (Nte), usually 11 residues in length that is not involved in its Ig-like fold (Fig. 1B). This region contains a conserved binding motif of alternating hydrophobic residues (named P2–P5 residues) (Fig. 1D). In the interaction between two subunits, as in the polymeric pilus itself, the Nte from one subunit occupies the hydrophobic groove of an adjacent subunit, binding to the P2–P5 pockets and completing its Ig fold (Fig. 1 E and F) (8, 14, 15). The reaction whereby the chaperone's strand is exchanged with the Nte of the subunit next in assembly is termed “donor-strand exchange” (DSE).

The P pilus is a prototypic chaperone–usher system, assembled by uropathogenic Escherichia coli and encoded by the pap (pyelonephritis-associated pilus) operon. P pili are highly complex fibers, composed of six different subunit types that are assembled in a specific order (Fig. 1A) (3, 16–18). Subunit ordering in these pili is highly conserved and likely to be important for preserving optimal function and hence the pathogenicity of the bacterium (4, 17). The first subunit to be assembled is a single copy of PapG, an adhesion molecule or “adhesin,” that binds Galα(1–4)Gal moieties on kidney tissue and is necessary for establishing infection (19). This is followed by a single copy of an adaptor subunit, PapF, the incorporation of which precedes assembly of ≈10 copies of PapE, which forms the bulk of the flexible tip fibrillum. Subsequent incorporation of a single copy of another adaptor subunit, PapK, leads to the incorporation of thousands of PapA subunits that form the rigid helical PapA rod. Finally, pilus biogenesis is halted by the incorporation of a single copy of the termination subunit PapH (Fig. 1A) (18).

Despite the importance of controlled subunit ordering for the production of functional pili, how the incorporation of correct subunits is coordinated during pilus biogenesis remains an open question. Differences in affinity between the various chaperone–subunit complexes and the usher could contribute to subunit ordering because the PapD/PapG and PapD/PapF complexes exhibit the highest affinity for the PapC usher (9, 20). On the other hand, the interaction of the Nte with the acceptor groove may also play a role in determining subunit ordering. Indeed, a recent mutagenesis study of the PapF subunit has demonstrated that its Nte sequence is crucial for the interaction of PapF with PapG and its function as an adaptor subunit (21). However, the role of Nte/groove interactions in determining subunit/subunit recognition in the Pap system has never been comprehensively explored. Here, we have used noncovalent mass spectrometry (MS) (22) to perform a systematic study of all 30 possible pairwise interactions between the six Pap subunits with the five Ntes of PapF, PapE, PapK, PapA, and PapH (Fig. 1 A and D). Despite the high structural homology of the Pap subunits (8, 12, 18, 23), and the conserved binding motifs of all Ntes (Fig. 1D), we show dramatic differences in the reactivity between different subunit/Nte pairs that mirror closely the order of subunit assembly found in the assembled pilus. The data provide insights into the origins of subunit ordering in pilus formation, suggesting that specificity is determined, at least in part, by the interaction between the Nte sequence and the acceptor binding groove and that, within this interaction, the sequence of the Nte at and around the P5 residue is of particular importance.

Results

Donor-Strand Exchange Monitored by Noncovalent MS.

To examine the specificity of DSE in Pap assembly, an in vitro assay was established in which all six chaperone/subunit complexes were incubated individually in the presence of peptides corresponding to each of the five Pap Ntes (Fig. 1D). PapG lacks an Nte because it is positioned at the distal end of the pilus and acts only as a donor-strand acceptor (Fig. 1A). For reactions involving PapE or PapA, the only two self-polymerizing subunits, truncated constructs were used to prevent self-polymerization. In PapE, the entire Nte was removed, creating PapE_Ntd1 (8), for PapA, the region N-terminal to the conserved Nte-binding sequence was removed, and the conserved glycine at the P4 site was substituted with Asn, creating PapA_Ntd1G15N (23). These constructs have been shown not to self-polymerize and to be suitable models as donor-strand acceptors (8, 23). A stable PapH construct, referred to as PapH_Ntd1, was used whereby the proline-rich N-terminal region of the Nte, known not to be involved in DSE, was removed (18). Previous studies have established the validity of adding peptides in trans to mimic DSE: Lee et al. (21) showed in vivo that the Nte and Ig-folded domain function separately in the Pap system, whereas Remaut et al. (15) demonstrated that the relative rate of DSE initiated by different Nte peptides in Saf pilins from Salmonella enterica mirrored that of their intact subunit counterparts.

Of the 30 reactions examined, 7 mimic the interactions known to occur in the pilus structure in vivo (referred to hereafter as “cognate” interactions), whereas the remaining 23 reactions involve noncognate pairings. A tenfold molar excess of peptide over chaperone/subunit complex was chosen as the optimal concentration ratio [see supporting information (SI) Text]. Under these conditions, DSE proceeds with pseudo first-order kinetics, allowing the reactivity of different Nte/subunit pairs to be compared. In each case, the progress of the reaction was monitored by noncovalent nanoelectrospray ionization MS (nanoESI-MS) (15, 22).

Mass spectra of the reactions between PapD/PapK (theoretical mass 42,224.8 Da, experimental mass 42,225.9 Da) and the Nte peptide of PapA (A_Nte, cognate) or the Nte peptide of PapK (K_Nte, noncognate) are shown as examples in Fig. 2. Individual components within the sample mixtures were identified by their unique mass (accuracy <0.01%), confirmed by the presence of a minimum of three consecutive charge state peaks (see Methods). The ESI mass spectra of PapD/PapK shortly after addition of each Nte (Fig. 2 A and E) show PapD/PapK to be the predominant species (red peaks). However, with time, changes in the spectra reveal the progress of DSE, with the peaks corresponding to PapD/PapK decreasing in intensity, and new charge state distributions relating to the PapK/A_Nte or PapK/K_Nte complexes (blue peaks) and released PapD (yellow peaks), the products of DSE, appearing and increasing in intensity (Fig. 2 B, C, F, and G). Integration of the peaks from each spectrum showed a marked difference in the progression of DSE for the cognate and noncognate pairs (Fig. 2 D and H). Thus, although the reaction between PapD/PapK and A_Nte is effectively complete within 72 h, a significant amount of the PapD/PapK complex remains in the reaction with K_Nte at this time, with only small peaks relating to the PapK/K_Nte product observed (Fig. 2G). The data demonstrate the effectiveness of nanoESI-MS to monitor DSE in real time, allowing the potentials of different cognate/noncognate pairs to undergo DSE to be compared.

Fig. 2.

ESI-MS of cognate/noncognate DSE reactions. (A–G) NanoESI mass spectra of the DSE reactions between PapD/PapK and A_Nte (A–C) and K_Nte (E–G): 30 min (A and E), 24 h (B and F), and 72 h (C and G) after reaction initiation. The chaperone/subunit complex is shown in red, unbound chaperone in yellow, and subunit/Nte product in blue. (D and H) Relative amount of chaperone/subunit complex (red) and subunit/Nte complex (blue) quantified at different times during DSE by using ESI-MS.

Chaperone/Subunit Complexes React Preferentially With Cognate Ntes.

The ability of each chaperone/subunit complex to undergo DSE with each peptide Nte, monitored by using nanoESI-MS, is shown in Fig. 3. The results reveal a striking span of reactivities, with the most rapid reactions reaching completion within 20 h (PapD/PapE_Ntd1 plus K_Nte or E_Nte) (Fig. 3C), whereas other combinations showed no significant reaction even after 300 h [e.g., PapD/PapG plus H_Nte, A_Nte, K_Nte, or E_Nte (Fig. 3A)]. With the exception of PapD/PapH_Ntd1,which does not undergo DSE with any Nte, consistent with the role of this subunit in capping assembly (Fig. 3F) (18), all other chaperone/subunit complexes were able to undergo DSE.

Fig. 3.

Discrimination in pilus assembly revealed by the apparent rate of DSE of different cognate/noncognate chaperone/subunit Nte pairs. (A–F) The apparent rate of DSE, monitored by the loss of chaperone/subunit complex versus time followed by nanoESI-MS for PapD/PapG (A); PapD/PapF (B); PapD/PapE_Ntd1 (C); PapD/PapK (D); PapD/PapA_Ntd1G15N (E), and PapD/PapH_Ntd1 (F) when challenged with F_Nte (orange), E_Nte (yellow), K_Nte (green), A_Nte (light blue) or H_Nte (dark blue). The stability of each chaperone/subunit complex in the absence of Nte peptide is shown in black. Corresponding apparent pseudo first-order rate constants (k_obs) for each chaperone/subunit Nte pair are shown in (G–K), with cognate interactions indicated (*). For reactions in which <20% of the substrate reacts within 14 days, k_obs were estimated based on the extent of substrate loss at this time (hatched bars). Errors are given in Table S1.

Interestingly, individual chaperone/subunit complexes showed different reactivities with each Nte, with some complexes being highly specific for their cognate Ntes (e.g., PapD/PapG, Fig. 3A), whereas others were more promiscuous (e.g., PapD/PapF, Fig. 3B). To compare the reactivity of each chaperone/subunit complex for each Nte, the progress of DSE (measured by chaperone/subunit loss) was fitted to a single exponential function (see SI Text), yielding an apparent rate constant, k_obs, for each chaperone/subunit Nte pair (Fig. 3 G–K and Table S1). The results reveal a striking correlation of the apparent rate of DSE with the nature of the Nte (cognate or noncognate), with the most rapid reactions uniformly occurring with the cognate partner(s). The data suggest, therefore, that the complementarity between each subunit groove and its cognate Nte plays a significant role in determining subunit assembly.

The apparent rate of DSE for each chaperone/subunit complex with its cognate Nte is compared in Fig. 4. The figure highlights the wide range in k_obs for the different cognate pairs, ranging from 2.1 × 10⁻¹ hr⁻¹ for PapD/PapE_Ntd1 plus K_Nte to 5.5 × 10⁻³ hr⁻¹ for PapD/PapG plus F_Nte. Interestingly, the slowest two reactions occur for PapG and its adaptor PapF, both of which are essential for pilus function (17, 21) and have the highest affinity for the usher protein (9, 20). By contrast, the subunits that self-polymerize in vivo (PapE and PapA) show some of the most rapid rates of DSE.

Fig. 4.

Comparison of the reactivity of each chaperone/subunit complex with its cognate Nte(s). The apparent rate of DSE of each chaperone/subunit complex for all cognate reactions: PapD/PapG plus F_Nte (red), PapD/PapF plus E_Nte (orange), PapD/PapE_Ntd1 plus E_Nte (yellow), PapD/PapE_Ntd1 plus K_Nte (olive), PapD/PapK plus A_Nte (green), PapD/PapA_Ntd1G15N plus A_Nte (light blue) and PapD/PapA_Ntd1G15N plus H_Nte (dark blue), measured as the loss of the initial chaperone/subunit complex monitored by nanoESI-MS.

Subunit Reactivity Is Determined by Complementarity at the P5 Pocket.

The range of reactivities between different chaperone/subunit Nte pairs is remarkable given the structural homology between the subunit folds (mean C_α RMSD = 2.0Å, see SI Text) and the highly conserved Nte binding motif, comprised of the P2 to P5 residues of the different Ntes (Fig. 1D). Despite the close similarity of the different Nte sequences (Fig. 1D), the subunits themselves differ significantly in sequence [overall similarity of ≈25% (see Fig. S1)]. Previous studies of DSE in the Saf pilus system demonstrated a key role for the P5 residue in the Nte and the P5 pocket in the subunit groove in determining the rate of DSE (15). To determine whether this is also the case in the P pili examined here, different chaperone/subunit complexes were challenged with chimeric Nte peptides, comprising the N-terminal half of a cognate Nte (encompassing the P2 and P3 residues) and the C-terminal half of a noncognate Nte (encompassing the P4 and P5 residues) and vice versa, and the reactivity of each was examined by using nanoESI-MS. Because the P4 residue is a conserved glycine in all Ntes, only the P5 site and adjacent residues differ in the C-terminal region (Fig. 1D), and hence these chimeric peptides serve as a good model to determine the role of this region in determining chaperone/subunit/Nte reactivity. Two chaperone/subunit complexes were chosen for analysis, the rapidly reacting PapD/PapE_Ntd1 and the more slowly reacting PapD/PapF (Fig. 3 B and C). PapD/PapE_Ntd1 was incubated with chimeric peptides comprised of the N-terminal half of E_Nte and the C-terminal half of H_Nte (named E/H_Nte), as well as the complementary peptide H/E_Nte. These peptides were carefully chosen to represent a cognate (E_Nte) and a noncognate (H_Nte) Nte that react rapidly and slowly, respectively (Fig. 3 C and I). Similarly, PapD/PapF was incubated with E/K_Nte and K/E_Nte (Fig. 3 B and H). The progress curves of these reactions (Fig. 5) showed that the apparent rate of DSE is determined by the C-terminal half of the Nte. Thus, DSE of PapD/PapE_Ntd1 with E/H_Nte mirrors that of H_Nte, whereas H/E_Nte mirrors E_Nte (Fig. 5A). Similar results were obtained for PapD/PapF, with K/E_Nte mirroring E_Nte, and E/K_Nte mirroring K_Nte (Fig. 5B). The results demonstrate that residues at, or adjacent to, the P5 site play a key role in determining Nte-groove reactivity in P pili, the Nte presumably then zipping into the complementary groove in a mechanism that is likely conserved in all pili belonging to this family (15, 24).

Fig. 5.

Chimeric peptides show that the rate of DSE is determined by the C-terminal half of the Nte. (A) Decrease in PapD/PapE_Ntd1 signal intensity when challenged with E_Nte (yellow), H_Nte (blue), E/H_Nte (black), or H/E_Nte (gray). (Inset) Sequences of E_Nte, H_Nte, E/H_Nte, and H/E_Nte. (B) Decrease in PapD/PapF signal intensity when challenged with E_Nte (yellow), K_Nte (green), E/K_Nte (black), or K/E_Nte (gray). (Inset) Sequences of E_Nte, K_Nte, E/K_Nte, and K/E_Nte. In the sequences, the P2–P5 residues are highlighted in red.

Discussion

P pili possess a complex architecture involving the precise ordering of a well defined number of six different subunit types. Despite detailed structural studies of different chaperone/subunit and subunit/Nte complexes (10), how the correct assembly of this complex organelle, which is essential for the virulence of a number of pathogenic Gram-negative bacteria (5), is determined has remained unclear. In a recent study of the PapF adaptor, the Nte of PapF was shown to be essential to assemble functional pili, although the PapG/PapF/PapE order could be altered to a PapG/PapE order simply by replacing the Nte of PapE with that of PapF in the PapE subunit (21). However, how subunit ordering is achieved in vivo, whether this can be emulated in DSE reactions in vitro and the molecular basis of subunit discrimination have not been explored systematically. Here, using noncovalent nanoESI-MS and experiments that mimic DSE in vitro, we provide evidence of an inherent specificity between different pairs of acceptor subunit grooves and donor subunit Ntes that mirrors subunit order in the intact pilus. Thus, all subunits located adjacently in P pili (cognate pairs) react more rapidly than their associated noncognate pairs (Fig. 3). The results suggest that the inherent compatibility of each Nte for each subunit groove plays an important role in determining the order of subunit assembly in vitro.

A further important feature of DSE to emerge from this study is the exquisite sensitivity of the reaction to sequence changes. Thus the chimeric peptides H/E_Nte and E/H_Nte share 66% and 60% sequence identity with E_Nte and H_Nte, respectively, yet show apparent rates of DSE with PapD/PapE_Ntd1 that differ markedly (Fig. 5). Given that the P4 site is a conserved Gly in all five Ntes studied here, the data demonstrate that residues at, or close to, the P5 site play a significant role in determining subunit/Nte reactivity. Similar results have been observed for Saf, in which single amino acid substitutions at the P5 site result in significant alterations in the rate of DSE (15). It is likely that residues adjacent to the one that inserts into the P5 pocket are also important, potentially due to interactions formed at the surface of the pocket. However, the precise structural basis for the rate differences is unknown, and future kinetic analyses combined with structures of all six subunits will be needed to determine the importance of individual residues in determining subunit ordering. Coevolution of the pilin genes within the pilus operon has thus resulted in a family of pilins in which the complementarity between binding grooves and their cognate Nte sequences are finely tuned so as to enable subtle changes in sequence to influence subunit/subunit interactions.

The ability of chaperone/subunit complexes to undergo DSE in vitro with noncognate Ntes, albeit less efficiently than their cognate counterparts, could provide an evolutionary advantage for subunit assembly in vivo, enabling the development of new pilus sequences and preventing mutations from ablating the ability of a pathogenic bacterium using the chaperone–usher system to export these essential virulence factors. For example, pili grown in E. coli that lack a papK gene are able to assemble with the usual architecture of a tip fibrillum and a rod, but the fibrillae in these bacteria are approximately twice as long as those found in wild-type pili (3). This suggests that the order of assembly must be altered in these pili, with PapE undergoing DSE with A_Nte, rather than the usual cognate reaction with K_Nte (Fig. 1A). The increased length of the fibrillum in the papK⁻ bacteria is consistent with the slower rate of DSE observed here in vitro for PapD/PapE with A_Nte compared with K_Nte (Fig. 3). The data suggest, therefore, that functional pili with different orders and numbers of subunits can be assembled in vivo, although some architectures will be more favored by the inherent complementarity of different subunit/Nte pairings.

Although our results demonstrate the importance of the groove/Nte interactions in determining subunit/subunit interactions in vitro, other factors play a role in determining subunit ordering during pilus assembly in vivo. Firstly, as in any multimolecular interaction, the rate of subunit incorporation during pilus assembly in vivo will depend on the concentration of each reaction partner. Secondly, the usher itself may play a role in determining subunit ordering. Indeed, the usher is known to bind chaperone/subunit complexes with different affinities (9, 20), which could increase the probability of incorporation of some subunits over others.

The observation that pilus formation occurs more rapidly in vivo (minute timescale) (25), compared with the timescale of DSE observed here in vitro, suggests an active role for the usher in subunit polymerization. Indeed, recent experiments have shown that the FimD usher of Type I pili acts as a catalyst of DSE (26). The recently solved crystal structure of the PapC usher, combined with cryoelectron microscopy images of the FimD usher caught in the act of DSE, suggest a structural rationale for the catalytic power of the usher, whereby the two periplasmic N-terminal domains of the usher dimer alternately recruit chaperone–subunit complexes for pilus assembly, most probably bringing them into close proximity with previously assembled subunits in the usher pore and orienting them optimally for DSE (27). However, although such a proximity effect may provide a rationale for the observed catalytic effect of the usher on DSE, the extent to which the usher actively modulates or controls subunit ordering remains elusive.

Methods

Protein Expression and Purification.

Details of the cloning of different pap genes and the expression and purification of chaperone/subunit complexes are described in SI Text.

N-terminal Extension Peptides.

Peptides representing the Ntes of PapH, PapA, PapK, PapE, and PapF, plus hybrid E/K_Nte, K/E_Nte, E/H_Nte, and H/E_Nte peptides, were purchased from CSS Albachem. The peptides were dissolved in 5 mM ammonium acetate (pH 5.5). Two lysine residues were added C-terminally to each Nte sequence (not shown in Fig. 1D) to increase solubility. The C terminus of each peptide was amidated. Control experiments comparing the rates of DSE using the A_Nte or H_Nte peptides with and without the additional Lys–Lys sequence, confirmed that these residues do not affect the kinetics of DSE (data not shown).

Mass Spectrometry Data Acquisition and Processing.

Forty-micromolar concentrations of each chaperone/subunit complex were incubated individually with 400 μM of each Nte peptide in 5 mM ammonium acetate (pH 5.5) at 22°C. The temperature and pH were optimized to minimize complex dissociation during data acquisition and to allow sufficient data to be collected over an appropriate time period. The reactions were sampled from an initial time point of ≈1 min after addition of the peptide Nte to 14 days, and analyzed by using a Q-Tof1 mass spectrometer with a nanoESI source (Waters) (see SI Text for further details).

MS data were processed by using the MassLynx software supplied with the Q-Tof1 (see SI Text). The areas of all peaks (including salt adducts) for a given species were summed and expressed as a fraction of the total area under all peaks in the spectrum. For quantitative analysis, the chaperone/subunit loss data were then normalized between the maximum concentration (i.e., concentration of reactant before reaction initiation) and the baseline signal (for more details see SI Text). Analysis of triplicate reactions showed a standard deviation of ± 2.5%, demonstrating the high reproducibility of the experiment. Data were fitted to a single exponential, assuming that each reaction goes to completion, to give an apparent rate constant of DSE (k_obs). Where k_obs was estimated (see SI Text).