Report

Three-Dimensional Model of Salmonella’s Needle Complex at Subnanometer Resolution

See allHide authors and affiliations

Science  04 Mar 2011:
Vol. 331, Issue 6021, pp. 1192-1195
DOI: 10.1126/science.1199358

Abstract

Type III secretion systems (T3SSs) are essential virulence factors used by many Gram-negative bacteria to inject proteins that make eukaryotic host cells accessible to invasion. The T3SS core structure, the needle complex (NC), is a ~3.5 megadalton-sized, oligomeric, membrane-embedded complex. Analyzing cryo–electron microscopy images of top views of NCs or NC substructures from Salmonella typhimurium revealed a 24-fold symmetry for the inner rings and a 15-fold symmetry for the outer rings, giving an overall C3 symmetry. Local refinement and averaging showed the organization of the central core and allowed us to reconstruct a subnanometer composite structure of the NC, which together with confident docking of atomic structures reveal insights into its overall organization and structural requirements during assembly.

Type III secretion systems (T3SSs) are widespread among many pathogenic Gram-negative bacteria such as Salmonella, Yersinia, Shigella, enteropathogenic Escherichia coli (EPEC), and Pseudomonas. Bacteria deploy the T3SS to transfer bacterial proteins (effectors) into eukaryotic cells. These effectors modulate various cellular functions, which make the host accessible to bacterial invasion (1, 2). The central core element of the T3SS is the needle complex (NC) required to make contact with the host cell (3). It is a large, cylindrically shaped macromolecular complex that is embedded in the inner and outer membrane of bacteria, spans the periplasmic space, and extends into the extracellular environment with a needle filament. Three-dimensional (3D) reconstructions from cryo (4) and negatively stained (5) electron microscopy (EM) images of NCs isolated from Salmonella or Shigella, respectively, provided the first glimpse of the overall architecture of the core element of the T3SS. In both organisms, NCs are organized as a series of ringlike structures built from individual subunits, suggesting a close structural and evolutionary relationship between different T3SSs and the flagellar system (68). Assembly of the ~3.5 MD NC occurs in a stepwise, coordinated manner, during which multiple copies of the inner and outer membrane proteins organize as rings laterally in their respective membranes to form the base of the NC (Fig. 1A) (9, 10).

Fig. 1

Analysis of symmetry and structural heterogeneity of wild-type NC of Salmonella typhimurium. (A) Organization of individual substructures within the NC. (B) (Left and middle) Representative class average of OR and IR substructures after selective NC disassembly and staining with methylamine vanadate. The OR showed 15 repeating subunits (9904 total particles analyzed; 1241 particles/class average) and the IR a concentric ring organization with 24 repeating subunits (1906 total particles analyzed; 928 particles/class average). (Right) Class-average image [n = 2172 (total), n = 224 (class average)] of top views (cryo-EM images) of intact NC after changing the preferred orientation. For the IR, a 24-fold symmetry (widest diameter) and for the OR, a 15-fold symmetry (small diameter) could be observed. Scale bars, 10 nm. (See figs. S1 and S2.) (C) (Left) Class-average image of cryo-EM images of aligned, nontilted NCs. Letters (a to d) indicate the position of substructures (a, OR; b, neck; c, IR1; d, IR2). The NC is positioned in a square box with dimensions of 256 pixels, and specific pixel levels are indicated with numbers. (Right) Variance analysis of the nontilted NCs shown in the left panel. Arrow indicates high-variance region below IR2. (D) (Left and middle) Density below IR2 is only seen in some class averages (arrow) and is absent in others. (Right) Class-average image shows a NC with a slightly bent neck (arrow), indicating a flexible connection between the upper and lower neck region.

Recently, atomic structures of individual domains of major base proteins from various species [PrgH periplasmic domain from Salmonella (amino acid 170 to 392), the PrgK homolog EscJ from EPEC (amino acid 21 to 190), and the InvG homolog EcsC from EPEC (amino acid 21 to 174)] became available, and it was speculated that a common structural motif is required for ring formation (11). Moreover, the topology of InvG, PrgH, and PrgK within the NC could finally be directly visualized by EM (12), allowing their rough placement within the complex.

Much less is known about how many copies of the three base proteins are required to build a NC. Strong evidence for 24 subunits comes from the atomic structure of the PrgK homolog EscJ that crystallized as a superhelix with 24 monomers covering one full turn (13) and from the image analysis of projection images obtained from disassembled mutant NCs (12) showing that PrgK builds the smaller of two concentric rings with 24 modulations per turn. The larger ring, which is composed of PrgH in Salmonella, also contains 24 subunits, establishing a 1:1 relative stoichiometry (PrgH:PrgK) (12). Similarly, 24 modulations have been observed in top views from Shigella NCs (5, 14). Crystal contacts from the InvG homolog of the outer membrane protein EscC from EPEC did not provide any information about subunit number. Nevertheless, for other family members of EscC/InvG (secretin family), 12 to 14 monomers have been discussed to form a ring structure, suggesting that the secretin family has the potential to adopt multiple symmetries (5, 1520). In this study, we determined the absolute stoichiometry of PrgH:PrgK:InvG and solved the structure of the NC to subnanometer resolution by cryo-EM. Docking of all three proteins revealed the handedness of the NC and provided insights into its organization at near-atomic resolution.

Although, in principle, analyzing top views of ring-forming complexes is a suitable method to determine their symmetry, the analysis of the T3SS NC is complicated by having two ring structures [inner rings (IR) (PrgH/PrgK) and outer rings (OR) (InvG)] stacked on top of each other (Fig. 1A), resulting in overlapping protein density in EM projections. To unambiguously determine symmetries of the NC inner and outer rings, we therefore aimed to obtain and analyze these structures separately but from previously fully assembled NCs. Using an extended high-pH treatment allowed us to selectively disassemble purified wild-type NCs from Salmonella into IR and OR structures (Fig. 1B and fig. S1A). We negatively stained such IR and OR structures using methylamine vanadate (pH 8) and analyzed EM images of top views using multivariate statistical analysis (eigen image analysis, fig. S1B). During the analyses, no symmetry operations were performed. The IR showed a concentric ring organization with 24 subunits at the two largest rings. This is essentially the same organization as previously described for the less stable mutant complex (PrgH∆4), which lacks four amino acids at the C terminus of PrgH (12). The OR clearly showed 15 circularly arranged subunits, suggesting that, together with the 24 subunits present in the inner membrane ring formed by PrgH/PrgK, the entire NC of Salmonella displays an overall C3 symmetry. To investigate whether other symmetries can be found in intact NCs, we analyzed top views of fully assembled NCs by cryo-EM. For this, we changed the typically preferred side-view orientation of the NC particles to a top-view orientation by protonation of a polyhistidine tag fused to the N terminus of PrgH. The analysis showed exclusively class averages of NCs with an overall C3 symmetry, composed of a larger ring of 24 and a smaller ring of 15 subunits (Fig. 1B and fig. S2A).

For reconstituting the 3D structure of the NC, we subsequently collected >37,000 particles from vitrified samples. Previous studies described variations in NC diameter and in the structure of the cytoplasmic localized IR2 (4, 5), which is composed of the N-terminal domain of PrgH and the C-terminal domain of PrgK (12). Such variations, which might reflect different conformational states, structural flexibilities, or additionally bound proteins, would need to be considered for a 3D reconstruction because they would influence the overall achievable resolution. We therefore studied the structural heterogeneity of individual parts of the NC by single-particle analysis: From the entire particle data set (tilted and nontilted), we used multivariate statistical analysis to generated reference-free class averages and determined their tilt angle by projection matching. To decrease the complexity of analysis and interpretation, only members of nontilted class averages were subsequently analyzed for their variance (Fig. 1, C and D; fig. S4; and movie S1). We found that the most rigid part of the NC is composed of the IR1, the periplasmic localized domain of the IR, and the connecting part of the neck (between levels 80 and 135) (Fig. 1C). The most flexible part of the core NC was at IR2 and the connection between the outer ring and the neck (levels 136 to 200). The greatest variance was observed below IR2, where unidentified additional density was seen in some class averages. The flexibility of IR2 has also been interpreted as an “open” and “closed” conformation (5) and was speculated to be required for the reprogramming of the NC in Salmonella (4). Due to the strong variance, we excluded the areas of IR2 and below from subsequent alignments (fig. S3) but kept the data during the final 3D reconstitution.

We reconstructed the structure of the NC (in C1) from ~37,000 cryo-EM images: For the structure determination, two major substructures (IR1/lower neck and OR/upper neck) were first treated separately to recover structural details at higher resolution, and the resulting 3D volumes were then combined into a composite structure. During this procedure, no symmetry operations were applied (image data processing details provided in supporting online material). Subsequently, a threefold symmetrization allowed us to obtain a structure of the Salmonella typhimurium NC to ~10 Å (Fig. 2). Clearly, newly discernible features are visible at all levels, in particular details in the individual subunits of the 24-fold IR1 ring and the 15-fold OR/neck. The top-view surface representation shows the organization of the 15- and the 24-fold rings in which five of the OR/neck subunits cover eight of the IR1 subunits (Fig. 2C). The structure also revealed details about the interior elements; in particular, cutting away a vertical segment showed the organization of the socket into individual subunits that directly connect to the inner rod (Fig. 2A). At the cytoplasmic side, the cup, which is believed to be the entry point for substrates, is also separated into individual subunits that can be viewed from the bottom (Fig. 2C). This organization raises the possibility that, upon intimate host cell contact, the delivery of effectors could be triggered because of conformational changes in the needle filament (2123) that are directly transmitted via the inner rod and socket to the cytoplasmically located cup. Subsequent local symmetrization of IR1 (24-fold), lower neck (15-fold), and OR/upper neck (15-fold) from the 3D structure reconstructed in C1 allowed us to reconstruct substructures at subnanometer resolutions (Fig. 2, B and D, and fig. S5).

Fig. 2

Three-dimensional surface views of Salmonella typhimurium’s NC. (A and C) Surface views of NC 3D reconstruction with C3 symmetry applied and filtered to a resolution of ~10 Å. Side, cut-away (A), top and bottom (C) view show the exterior and interior subunit organization (IR2 removed in bottom view for better visualization of IR1 and cup). The most probable configuration of cup and socket is displayed with respective regions highlighted in blue and yellow (see image data processing details in supporting online material). (B) Applying symmetries to substructures of different components of the NC, that is, a 24-fold for the IR1 and a 15-fold for the neck and OR, led to higher resolutions. (D) Tilted view (45°) of IR1 after symmetrization (C24) revealed structural details to subnanometer resolution.

In agreement with the recently determined topology of PrgH, PrgK, and InvG (12), we were able to unequivocally dock all available atomic structures of the base proteins into the cryo-electron density map (atomic models for PrgK and InvG were generated from their respective homologous structures, using EscJ and EscC as templates) (fig. S6) and to determine not only the arrangement of individual monomers within their respective rings but also their organization between rings (Fig. 3). (For details, see figs. S8 to S11 and movies S2 to S4.) In the resulting model, all three base proteins build up individual rings and are independently docked as rigid bodies guided by their two-domain structure and their secondary structural elements: 15 monomers of the N-terminal domain of InvG (orange) are arranged such that the loop containing lysine-38 is located in the vicinity of IR1. This is in agreement with a chemical cross-link found between lysine-38 of InvG and lysine-367 of PrgH (12, 24) (fig. S11). Within the IR1, 24 monomers of PrgH (gray) and PrgK (yellow) were independently docked into the larger and the smaller of the two concentric rings, respectively. The docked PrgH also conforms to the accessibility of an internal his-tag (between amino acid 267 and 268; ** in Fig. 3A) introduced for nano-gold labeling (12). A six–amino acid fragment (201 to 206; * in Fig. 3A; helix 1 in fig. S7) close to the N-terminal part of the periplasmic domain extends out of the electron density map, suggesting that this domain is relaxed when crystallized as an individual protein outside the complex. This could occur as a result of the absence of the connecting transmembrane and cytoplasmic domain of PrgH (Fig. 3B). PrgK is positioned such that its lipidated N-terminal domain is able to interact with the outer leaflet of the inner membrane (fig. S11). The subsequent domain, however, reaches vertically further up into the complex. From there, PrgK continues with ~60 amino acids, for which no atomic structure is available as of yet, and folds back from its periplasmic location through the membrane toward the cytoplasm (12). This second transmembrane segment is visible in the EM map (Fig. 3B and fig. S11). Furthermore, in PrgK amino acids 135 to 146, which are not present in its homolog EscJ, are modeled as a beta-hairpin (fig. S6) that extends toward the socket. An electron density could be visualized in the corresponding region at elevated density levels (* in Fig. 3B). Taken together, we were able to generate a molecular model of the major part of the NC base (Fig. 4A and movie S5) and established the absolute stoichiometry of PrgH:PrgK:InvG = 24:24:15.

Fig. 3

Docking of atomic structures into the 3D cryo-EM map of the NC provides near-atomic models of the IR1 and the neck region. (A) InvG (orange), PrgH (gray), and PrgK (yellow) were independently docked as rigid bodies guided by their domain structure and secondary elements. Tubelike densities, which correspond to α helices, are clearly visible. (For a 3D representation, see also movies S2 to S4.) Positions of amino acids within PrgH are indicated: 267 and 268 (**), and 201 to 206 (*). (B) A slice through the NC (side view) with docked atomic structures shows the organization of PrgH and PrgK within IR1 and demonstrates that the cup/socket region is not formed by these proteins. The asterisk marks a density that connects the modeled PrgK loop with the cup/socket structure. The arrow marks the position of two tubelike densities that most probably correspond to the transmembrane helices (TM) of PrgH and PrgK. (C) Magnified top view of neighboring monomers in their respective NC positions. (Left) Organization of two InvG monomers. (Right) Organization of two PrgH and two PrgK monomers. (D) 15 InvG (left) and 24 PrgH/PrgK (right) monomers were docked into the lower neck ring and IR1 structure, respectively. PrgH and PrgK are organized as concentric rings with different diameters.

Fig. 4

Atomic model of the NC base and validation of PrgH organization. (A) Side and top view of atomic structures of InvG (orange), PrgH (gray), and PrgK (yellow) organized as rings with 15/24/24 subunits, respectively (for a 3D model, see movie S5). (B) Point mutations at the predicted interface between PrgH monomers prevent effector protein (SptP) secretion. E252 and Y239 are located at the interaction surface (side chains displayed as spheres). Single-point mutations to alanine showed wild-type level secretion of SptP (sup., supernatant), whereas a double mutant (Y239A/E252A) did not secrete SptP. Cellular SptP levels (cells) in all strains were unchanged. In the PrgH (Y239A/E252A) strain, PrgH protein levels were decreased, indicating that the assembly of NC is compromised, leading to a nonsecreting phenotype. w.t., wild type; ∆prgH mutant, negative control; +prgH, ∆prgH mutant complemented with prgH-containing plasmid; Y239A and E252, single-point mutated, and Y239A/E252A double-mutated prgH-containing plasmid complementing the ∆prgH mutant.

To validate the docking, we tested whether mutations at the PrgH-PrgH interaction surface have an influence on the function of the NC. Our model predicts that amino acid Y239 in one PrgH monomer is located in the vicinity of amino acid E252 of a second PrgH monomer when assembled as a ring structure (Fig. 4B). Single-point mutations to alanine were not sufficient to change the secretion behavior for the effector protein SptP, indicating that the interaction surface is not sufficiently disturbed and that the individual mutations did not affect the correct folding. The point mutations are located at opposing sites of one monomer, and it thus seems likely that they act independently and that proper folding of PrgH would also not be affected in a double mutant. Yet the level of PrgH is decreased in Y239A/E252A, and no SptP secretion could be detected. Taken together, this suggests that the double mutant does not properly assemble a ring structure and that free PrgH is degraded rapidly, indicating that shortly after expression, wild-type PrgH forms a ring complex and is therefore protected from further degradation.

We showed that the three proteins PrgH, PrgK, and InvG, which dictate the overall shape of the NC base, organize as individual rings of 24 (PrgH/K) and 15 (InvG) subunits and that, as a consequence, the highly oligomeric S. typhimurium NC in fact assembles into a complex with an overall low symmetry (C3). This implies that a local symmetry mismatch exists, in which five of the OR/neck subunits are positioned over eight IR1 subunits. To accommodate this symmetry mismatch, the binding interface must have some structural tolerance that could be important for an efficient assembly when the IR and OR dock onto each other. Structural tolerance could be provided by a weakly ordered interaction surface, which overall could also grant the flexibility needed for the reprogramming phase. Indeed, such a flexibility is supported by the crystallographic analysis of the C-terminal domain of PrgH, in which the last 30 amino acids (363 to 392) could not be resolved (11).

For the cytoplasmically localized ring (IR2), the extensive structural deviations might reflect different conformational or even compositional states and could suggest that IR2 in fact plays an additional functional role during secretion, for example, as a binding platform for a putative C ring (10, 25).

This study provides an experimentally validated near-atomic model of the majority of the NC organelle, and it shows how its central core is organized and engages with the inner rings. Because of the high degree of conservation of T3SSs among pathogens, this structure could be the basis for the understanding of the design, assembly, and biology of this important translocation machinery and may also pave the way for the development of therapeutic drugs against microbial infections.

Supporting Online Material

www.sciencemag.org/cgi/content/full/331/6021/1192/DC1

Materials and Methods

Figs. S1 to S17

References

Movies S1 to S5

References and Notes

  1. We would like to thank G. Resch for collaborating on the development of a semi-automated image acquisition software, the Institute of Molecular Pathology/Institute of Molecular Biotechnology/Gregor Mendel Institute (IMP/IMBA/GMI) EM and bioinformatics facility for technical support, W. Weber for generating and analyzing the mutational analysis, and all members of the Marlovits laboratory for critically reading the manuscript. We furthermore would like to thank J. E. Galán (Yale University) and members of his laboratory for continuous discussion, generous support, and materials. This work was supported by a Ph.D. fellowship from Boehringer Ingelheim Fonds (to O.S.) and a research grant from the Zentrum für Innovation und Technologie (ZIT), Center for Molecular and Cellular Nanostructure Vienna (CMCN) (to T.C.M.). Accession numbers are EMD-1871, EMD-1874, EMD-1875, and PDB ID code 2Y9K and 2Y9J.
View Abstract

Navigate This Article