Structure of a type III secretion needle at 7-Å resolution provides insights into its assembly and signaling mechanisms

Cryo-EM Image Reconstruction of Wild-Type Needles.

We sought to obtain a high-resolution structure of the needle using cryo-EM. However, the needle diameter is only 7 nm and its subunit just 9 kDa, making image contrast of frozen-hydrated specimens in vitreous ice extremely poor and therefore accurate image alignment for high-resolution analysis difficult. To surmount this, we made methodological improvements (16, 17) to obtain higher image contrast (Fig. 1A). We used an electron cryomicroscope, JEM-3200FSC (JEOL), equipped with a field-emission electron gun, a liquid helium-cooled specimen stage, an Ω-type in-column energy filter, and a 4,096 × 4,096 CCD camera, to collect cryo-EM images of purified needles, made artificially long by MxiH overexpression and then sheared off the bacterial surface and into smaller (approximately 300 nm) fragments. We operated the electron microscope at an accelerating voltage of 200 kV, and collected images at specimen temperatures of around 50 K. We collected cryo-EM images of wild-type needle from approximately 300 CCD frames, carried out helical image reconstruction using approximately 100,000 needle segments (Table 1), and obtained the structure at 7.7-Å resolution (based on the Fourier shell correlation = 0.5 criterion; Fig. S2). The 3D-density map is shown in stereoview in Fig. 1B. The image processing statistics are given in Table 1. The average of power spectra of all the image segments used for 3D-image reconstruction shows sharp layer lines (Fig. S3), indicating the high homogeneity of the needle structure.

The density map clearly resolved α-helices as rod-like densities (Fig. 1B). The refined helical parameters are 4.30 Å for the axial rise and 64.1° for the azimuthal rotation along the 1-start helix (Table 2) confirming previous data (4, 14). However, none of the available atomic models of MxiH and its homologs (Fig. S1) (12, 13, 18, 19) can be fitted into our density map.

Initial Remodeling of the Needle Protein.

Because the 3D map allowed the identification of correspondences with secondary structural elements of the MxiH crystal structure, we tried to fit the MxiH “head” (residues 26–51), including the short Pro-Ser-Asn-Pro (PSNP; residues 41–44) loop into the map. The conformation of this portion is nearly the same in all the NMR and X-ray structures of monomeric proteins available for this protein family (12, 13, 18, 19) (Fig. S1). Indeed, the head motif fitted well into the map. The head portion has two ends. One extends toward the central channel, forming a long rod-shaped density labeled H1 (Fig. 1C). The other forms the external surface of the needle structure and includes two rod-shaped densities labeled H2 and H3 and a “protrusion” between them (Fig. 1C). Given their size and shape the H1, H2, and H3 densities were modeled as α-helices in Fig. 1D. H1 is approximately 60-Å long, corresponding to an α-helix of approximately 40 residues, which is in good agreement with the full-length N-terminal helix extended from the MxiH crystal structure (12). This fact led us to assign H1 to the N-terminal half and H2 and H3 to the C-terminal half of MxiH.

Experimental Validation of the Model.

To confirm this assignment, we examined the effect of termini alterations. In the 3D map, H3 of one subunit interacts with the head of the subunit immediately below (Fig. 1B, Upper, pink arrowhead ) whereas its H1 lies free (Figs. 1B, Lower, and 2A), suggesting that the C terminus plays a more critical role than the N terminus for needle formation. It has been shown that three to five residue deletions at the MxiH C terminus cause absolute polymerization defects, implying that this region is directly involved in the intermolecular interactions for polymerization (5). The five C-terminal residues of Salmonella MxiH homolog PrgI are essential for T3SS function whereas its six N-terminal residues are dispensable (13). Removal of any number of MxiH N-terminal residues beyond three generated a null phenotype (Fig. S4). Therefore, we generated shorter MxiH truncations and extensions (duplications of the first and last three residues; Fig. 3). We tested for T3SS assembly by examining overnight secretion of early-acting virulence proteins into the culture medium and for T3SS functionality using an artificial inducer of secretion, the small amphipathic dye Congo red (CR) (5). CΔ1 was still partially functional for needle assembly and T3SS activation, but CΔ2, CΔ3, and C + 3 were not (Fig. 3 A and B and Fig. S5). In contrast, NΔ2 and NΔ3 still allowed needle formation, but only NΔ2 permitted substantial T3SS activation (Fig. 3 A and B) although it led to unstable polymers. N + 3 allowed better than normal MxiH secretion (Fig. S5) and polymerization (Fig. 3A) although it displayed reduced sensitivity to CR (Fig. 3B). In the N + 3 needle cryo-EM image reconstruction, the extension was invisible and hence probably disordered. However, because the MxiH N terminus tolerates more substantial modifications than its C terminus, H1 extending into the central channel must be N terminal and H2 and H3 C terminal.

Refinement of the Model.

Given previous sequence analysis suggesting the presence of a β-structure for residues L54–S65 (5, 13), the protrusion between H2 and H3 can be interpreted as a β-hairpin. We therefore built a preliminary atomic model of MxiH within the needle using the A-form of the MxiH crystal structure for residues G21-Y50 as the head. Then, the N-terminal α-helix was extended from the head to S2, and the β-hairpin (Q51–Q64) and C-terminal α-helix (S65–R83) were built based on the map. Because there are heptad repeats in both the N- and C-terminal regions, we placed the three α-helices with their hydrophobic residues facing to each other as should occur in an α-helical coiled-coil structure and optimized the model as described in SI Materials and Methods.

Structures of Mutant Needles Deregulated for Secretion.

To advance our understanding of how the needle acts to transduce the signal of host-cell contact, we examined the structures of needles purified from strains overexpressing either of two key needle protein mutants, D73A and Q51A (Fig. 4), which we have previously characterized as having altered-tip-complex composition and deregulated secretion (5, 9). These mutants were selected because they show the highest needle stability and order amongst all those previously analyzed (14). MxiH(D73A) lacks the tip complex, secretes early-virulence proteins constitutively, and is uninducible by CR. MxiH(Q51A) displays minor alterations in composition of the tip complex, is constitutively active for secretion, and is still inducible by CR.

We carried out cryo-EM helical-image reconstruction of these two mutant needles and obtained the density maps at 7.7- and 8.3-Å resolution, respectively (based on the Fourier shell correlation = 0.5 criterion; Fig. S2). As shown in Table 2 the refined helical parameters are identical to those of wild-type, confirming previous data (4, 14). Comparison of the wild type, D73A, and Q51A mutant needles (Fig. 5) revealed that the structures are also nearly identical and that no significant structural change is observed. Given that they are independent reconstructions, their similarity demonstrates the reproducibility of the structural analysis performed. The results suggest that any structural changes in the needle protein responsible for the functional alteration are either too small to detect at the present resolution or not retained in needles isolated from their base and tip.

Refolding of Needle Proteins During Polymerization.

The structures of the wild-type and two mutant needles presented here reveal a previously undescribed fold of the needle protein that is distinct from those previously seen in X-ray crystal structures and NMR solution structures. Using FTIR of the Salmonella typhimurium needle protein (PrgI*), soluble or polymerized into needles, as well as solid state nuclear magnetic resonance (ssNMR) of the latter, Poyraz et al. provided evidence for a change in secondary structure from α-helix to β-structure within the C-terminal region upon polymerization (13). The present structure identifies the precise region involved as the β-hairpin made of residues Q51–Q64, which overlaps well with the prediction of sequence analysis (5). Relatively small proteins without an extensive hydrophobic core can fold into different conformations depending on chemical and physical environments as well as upon assembly into larger structures. Therefore, it is not completely unexpected that MxiH, which is only 85 residues long, refolds upon assembly into the needle structure. The result shown here illustrates why caution should be exercised in interpreting the crystal structure of small monomeric proteins that normally function in assembled complexes.

Needle Stability.

The model confirms that the needle structure corresponds to only the innermost tube of the flagellum (4, 17, 20, 21). What makes a tubular filament as thin as the needle so stable? The innermost tubes of the flagellar-axial structures are consolidated by interactions between coiled-coils of the D0 domain of their subunits but H2 and H3 are not so closely associated between neighboring needle subunits. However, the β-hairpin lies at the interface of subunits along all major three helical lines (6, -5, and -11 start; Fig. 4), perhaps working like a “linchpin” to stabilize the structure. To test this, we altered, shortened, or removed the β-hairpin by mutagenesis (Table S1). These mutants displayed poor MxiH expression or intrabacterial stability, very reduced secretion, and lack of needle assembly (Table S2). Interestingly, the mutated region corresponds to a portion of the Pseudomonas and Yersinia needle proteins known to interact tightly with their intracellular chaperone PscG/YscG in an extended conformation, suggesting that a similar complex may also exist in Shigella even though the PscG/YscG analog(s) remains unidentified (22, 23). We then altered residues to alanine where the β-hairpin interacts in neighboring molecules, along the 6-start helix where the head joins up with H1 and the PSNP loop (Table S1). These changes also led to reduced MxiH expression and/or secretion (Table S2), preventing examination of whether these residues are also important for assembly. However, a previously generated mutant, Y57A, is informative (5). Y57A is expressed and secreted but leads to a reduced number of needles that are shorter than wild type. In our model, Y57 is located at the tip of the β-hairpin and plays an important role in intersubunit interactions along the 6-start helix, stabilizing the association of adjacent protofilaments, which are near-axial arrays of subunits along the 11-start helical line (Fig. 4). Therefore, the model explains the high conservation of this residue.

Needle Assembly.

In addition, the model allows us to propose a mechanism for needle polymerization. The polymerization of each external portion of the bacterial flagellum requires an “assembly cap.” The filament cap is a flat pentameric structure with “legs” that rotates above the helically growing structures to aid selection of the insertion site for the next flagellin subunit. The cap also has a central cavity open to the outside to promote the refolding and binding of the approximately 50-kDa subunits to the appropriate location (24). Although initially proposed to exist (2), no assembly cap has been found for the T3SS needle. However, given its small size, the needle protein may not need a polymerization-promoting complex (11). In view of the β-hairpin structure found, we now propose that the assembly process is as follows. Each needle subunit is exported possibly with its N terminus first, where its secretion signal is located (25). The terminus of each subunit would then bind to one of the five possible insertion sites and integrate into the lining of the channel. Then, the folding of the β-hairpin would stabilize the subunit position within the needle sufficiently to allow its C terminus to fold on the outside of the structure. When the right needle length is reached and a ruler molecule is secreted, the secretion specificity is switched to tip complex components (26). The addition of the first tip-complex protein (IpaD) is then sufficient to halt needle elongation (13).

Needle Length Control.

The structure also provides insight into needle-length control. Needle subunits, like axial-flagellar proteins, are exported by T3SS through the central channel of the growing structure and assemble at its distal end (27, 28). The lengths of the needle and the flagellar hook are both controlled by conserved ruler proteins (29, 30). In the absence of their rulers, extra-long needles and hooks are produced. Two models have been proposed to explain how the rulers work (30–33). In one model, polymer elongation is controlled by several ruler molecules passing stochastically through the growing structure, with subunits and rulers traveling alternatively in the channel (30, 31, 33). In the other model, one ruler molecule positioned inside the channel is stretched gradually during polymer assembly, where the ruler coexists with polymer proteins translocated through the channel (32). The ruler is thought to adopt an α-helical conformation inside the channel based on a linear correlation between ruler size and polymer length, with the polymer extending by 1.9 Å per ruler residue (29, 32), because the axial translation of C^α atoms between adjacent residues is approximately 1.5 Å in an α-helix and approximately 3.7 Å in an extended chain. The present structure shows that the channel diameter is approximately 13 Å (Fig. 2). Because the diameter of an α-helix is approximately 10 Å, it is impossible for a ruler and a polymer protein to coexist in the channel, ruling out the second model.

Transduction of the Host Cell Contact Signal.

Does our density map help us understand how the needle transduces the host cell contact signal to the export apparatus? The four residues (P44 + Q51, Y57, and D73) that lead to an inability to transmit the activation signal when altered to alanine (5) lie in positions where they may directly (Y57 and Q51) or indirectly (P44 and D73) affect the orientation or conformation of the β-hairpin. The location of the MxiH β-hairpin is topologically analogous to that of the β-hairpin in domain D1 of flagellin in the flagellar filament (34). In flagellin, the β-hairpin is crucial to the polymorphic supercoiling of the filament, which involves a 0.8 Å change in the axial repeat along the 11-start protofilament, in response to a change in the direction of motor rotation (34). The comparison of the two protofilament conformations of the flagellar filament called L and R type has shown that the conformational change of flagellin involved in the polymorphic supercoiling of the filament is the switch in the orientation of the outer core domains relative to the invariant structure of the tubular assembly of the inner core domains (35). However, the axial repeat is unaltered in needle mutants leading to deregulated secretion (14), as confirmed by our cryo-EM image analysis here (Table 2) and the density maps of D73A and Q51A mutant needles did not reveal significant conformational changes compared to that of wild type (Fig. 5). Nevertheless, the present model of the needle suggests that H2, H3, and the β-hairpin on the outer surface can be displaced relative to the tightly packed H1 helices in the inner core (Fig. 1B), perhaps by a conformational change in the PSNP loop. Such a displacement could be caused by a change in the conformation of the tip complex upon host-cell contact. The tight packing of the H1 helices could provide a solid framework for the cooperative propagation of the host contact signal down to the base of the needle and eventually to the export apparatus to activate effector secretion into host cells without changing the repeat and helical symmetry of the needle structure, just as speculated previously (11). In addition, it is quite possible that such changes would not be preserved when needles are isolated from their base and tip during purification. To unveil the signal transduction mechanism, the necessary next step is the structural examination of wild-type and mutant needles that have retained their tip and base by cryo-EM image analysis.