Structural Basis of Type 2 Secretion System Engagement between the Inner and Outer Bacterial Membranes

ABSTRACT Sophisticated nanomachines are used by bacteria for protein secretion. In Gram-negative bacteria, the type 2 secretion system (T2SS) is composed of a pseudopilus assembly platform in the inner membrane and a secretin complex in the outer membrane. The engagement of these two megadalton-sized complexes is required in order to secrete toxins, effectors, and hydrolytic enzymes. Pseudomonas aeruginosa has at least two T2SSs, with the ancestral nanomachine having a secretin complex composed of XcpQ. Until now, no high-resolution structural information was available to distinguish the features of this Pseudomonas-type secretin, which varies greatly in sequence from the well-characterized Klebsiella-type and Vibrio-type secretins. We have purified the ~1-MDa secretin complex and analyzed it by cryo-electron microscopy. Structural comparisons with the Klebsiella-type secretin complex revealed a striking structural homology despite the differences in their sequence characteristics. At 3.6-Å resolution, the secretin complex was found to have 15-fold symmetry throughout the membrane-embedded region and through most of the domains in the periplasm. However, the N1 domain and N0 domain were not well ordered into this 15-fold symmetry. We suggest a model wherein this disordering of the subunit symmetry for the periplasmic N domains provides a means to engage with the 6-fold symmetry in the inner membrane platform, with a metastable engagement that can be disrupted by substrate proteins binding to the region between XcpP, in the assembly platform, and the XcpQ secretin.

inner membrane protein complex that functions as an assembly platform and ATPdependent motor, and from this platform, the pili are driven into a cavernous secretin complex embedded in the outer membrane that spans the periplasm to engage the inner membrane. The T2SS itself encompasses a pilus-like structure, referred to as a pseudopilus since it is not driven outside the cell for locomotion but rather driven in a short motor stroke. The pseudopilus is driven into the T2SS secretin complex by either a piston or a screw-like mechanism, thereby promoting secretion of substrate proteins (toxins, effectors, and hydrolytic enzymes) across the outer membrane. Recent highresolution structures of the Vibrio-type and Klebsiella-type secretins showed them to be of 15-fold symmetry (4). These structural breakthroughs highlighted a perplexing problem, given that the secretins interact intimately with 6-fold symmetric structures in the T2SS inner membrane platform (5)(6)(7).
Pseudomonas aeruginosa is a bacterial pathogen of particular concern to patients with burns, wounds, or cystic fibrosis (8). Species of Pseudomonas have two characteristic T2SSs, defined by distinct secretins, XcpQ and HxcQ (9)(10)(11). Phylogenetic analysis has suggested that while the T2SS containing HxcQ was acquired more recently by lateral gene transfer from betaproteobacteria, the XcpQ secretin defines the ancestral Pseudomonas-type T2SS (10). The T2SS containing XcpQ is responsible for the secretion of at least 19 substrates, most of them hydrolases ranging from proteases to hemolytic lipases that are critical for pathogenesis (11)(12)(13). After crossing the inner membrane via the SecYEG or TAT translocases, substrates of the T2SS enter the secretin chamber via an unknown mechanism (5,6,9,14). In order to address how the outer membrane and inner membrane components engage together and yet can allow for substrate uptake, high-resolution structural data are required.
Previous detailed sequence analysis has demonstrated the sequence-based features that distinguish the Klebsiella-type secretins from the Vibrio-type secretins (15). Cluster Analysis of Sequence data (CLANS) is a sensitive tool to depict sequence relationships (16): in a CLANS plot, each protein is represented as a dot clustering with other proteins of similar sequence. A core set of T2SS secretin sequences was extracted from the InterPro "GspD/PilQ family" (IPR001775) (see Text S1 and Table S1 in the supplemental material), and a CLANS plot of these sequences from diverse bacterial lineages showed the evolutionary distances separating the XcpQ and HxcQ secretins of Pseudomonas from other functionally characterized secretins (Fig. 1A). Great sequence drift separates the XcpQ and HxcQ secretins from each other and from the previously characterized Klebsiella-type and Vibrio-type secretins, while the secretin from the recently discovered T2SS of Acinetobacter baumannii (17) clusters with the XcpQ secretin from species of Pseudomonas and a grouping of marine bacteria (Fig. 1A). Despite their very close sequence-based clustering, the Klebsiella-type and Vibrio-type secretins are structurally distinct. Thus, with the XcpQ secretin being even more diverse in sequence and representing an archetype for the Pseudomonas-type secretins in general, the secretin complex of P. aeruginosa PAO1 was purified for structural analysis by single-particle cryo-electron microscopy.
XcpQ oligomers were detergent solubilized and purified by size exclusion chromatography, and the~1-MDa species was shown to be a pure XcpQ multimer when characterized by SDS-PAGE (Fig. S1). XcpQ samples were flash frozen on vitreous ice and were imaged on a Titan Krios transmission electron microscope equipped with a Gatan K2 Summit direct electron detector (Fig. 1B). After analysis of 18,000ϩ particles, class averages revealed that XcpQ is a pentadecameric secretin complex ( Fig. 1B and  S2) with no other symmetries observed in the two-dimensional (2D) class averages.
In the T2SS secretins, each subunit consists of four N domains followed by the highly conserved secretin domain (Pfam PF00263), followed by an S domain. These features, as defined in the secretin structures of the Klebsiella-type secretins, are present in the structure of XcpQ ( Fig. 2A and B). In the Klebsiella-type secretins and Vibrio-type secretins, the S domain is necessary for interactions with the pilotin that mediates secretin trafficking to the outer membrane and assembly (15,(18)(19)(20). The disposition of the S domain in the XcpQ complex structure is equivalent to that of the other pilotin-dependent secretins (Fig. 2C). Features previously named the "upper chamber," "inner gate," and "cap" were resolved ( Fig. 2A and B). Comparison of the XcpQ structure to that of the Klebsiella-type secretin complex from Escherichia coli K-12 strain DH5␣ (Protein Data Bank [PDB] accession no. 5WQ7) revealed a striking structural homology (Fig. 2C) despite the differences in their sequence characteristics (Fig. 1A).  Table S1 in the supplemental material) shows that proteins from diverse species cluster into six types. The Klebsiella type and Vibrio type (red) are barely resolved at this scale, and a diverse group of gammaproteobacterial sequences (yellow) are clustered between the Klebsiella-type/Vibrio-type group and the other groups. The Pseudomonas-type HxcQ secretin clusters among a group of betaproteobacterial T2SS sequences from the Burkholderiales (orange). Pseudomonas aeruginosa XcpQ clusters among three well-resolved groups: one comprised of various XcpQs from various Pseudomonas species, one comprised of marine bacteria, and one comprised of the recently described T2SS secretins from species of Acinetobacter. Insets show the distinct structures of the Klebsiella-type (PDB accession no. 5WQ7) and Vibrio-type (PDB accession no. 5WQ8) secretins. (B) XcpQ was purified by size exclusion chromatography, the fraction containing the~1-MDa complex was frozen, and individual XcpQ particles were visualized by electron microscopy. The N domains of XcpQ contribute to the chamber that needs to dock onto the XcpP subunits of the inner membrane platform (5,14). At 3.6-Å resolution, individual ␣-helices and ␤-strands were resolved throughout the upper part of the secretin structure and within the N3 domain, and the majority of the protein could be visualized (Fig. 2C). There is a relatively unstructured linker or hinge between the N3 and N2 domains (corresponding to residues 239 to 246 in the mature form of XcpQ), where the resolution of the structure starts to decline (Fig. 2D, S2D, and S3). Less well resolved density was present for the N1 domains, and no density for the N0 domains could be observed (Fig. 2D, S2D, and S3). The failure to identify these N domains raises the prospect that this region of the secretin complex, proximal to XcpP in the inner membrane, is not arranged in a uniform 15-fold symmetry.
The T2SS has been an enigma to structural and molecular biologists. In order to drive protein secretion, a secretin like XcpQ must dock to an integral inner membrane complex with 6-fold symmetry (XcpP) (5-7). These intimate interactions raise a further structural dilemma, in that substrates access the T2SS in the periplasm and must therefore interpose between the seals formed by the N domains of the secretin (5,7). Given the recent findings of Douzi et al. (21), in combination with our data, we suggest a plausible scenario that would resolve this conundrum (Fig. 2E).
The three-dimensional (3D) reconstruction of the XcpQ oligomer was performed under a strict cyclic-15-fold (C15) symmetry. This symmetry restriction was judged to be appropriate by unambiguous 2D class averages of "top-down" projections, clearly and only showing a C15 symmetry ( Fig. 1B and C). As can be seen from the 2D class averages (Fig. S2), the best-resolved density is that of the secretin ring and N3 domain, leaving the proximal N domains poorly resolved. This holds true for our 3D model, which during the Fourier reconstruction favored the high-contrast features of the secretin ring and N3 domains. Our data suggest that the N0-N2 domains may not fully conform to the C15 symmetry, and attempts to resolve this part of the structure were not successful. We propose that a pseudo-6-fold symmetry exists in the proximal region of the N domains (Fig. 2E). This might be acquired by the XcpQ dimer-based structures defined by Douzi et al. (21) via a radially symmetric array wherein three N domains are displaced away from the XcpP structure to create a metastable arrangement. The flexibility with which the features pack in order to transition from 15-fold to pseudo-6-fold symmetry provides insight into the mechanism for substrate entry into the T2SS, with the substrate-XcpP interactions competing for the XcpQ-XcpP interactions in order to gain entry into the XcpQ chamber; success in this competition would see the substrate be secreted via pseudopilus movements.
Detailed descriptions of methods are available in Text S1 in the supplemental material, with the map and model parameters for the structures presented as Fig. S2 and Table S2, respectively. T2SS secretin sequence accession numbers and annotations used for the CLANS analysis are presented in Table S1. Electron microscopy data were collected at the Ramaciotti Centre for Cryo-Electron Microscopy, Monash University. All figures were generated with either PyMOL (22) or UCSF Chimera (23).
Accession number(s). The electron density map has been deposited in the EMDB (EMD-8860). The model of XcpQ has been deposited in the PDB (5WLN).