The role of the FliD C-terminal domain in pentamer formation and interaction with FliT

Flagellar biogenesis is controlled by a negative feedback loop. When FliD was secreted at the late step of flagellar assembly, the FliD-FliT complex disassembled and free FliT bound to the FlhDC complex, a master regulator of flagellar biogenesis, subsequently inhibiting the overall expression of flagellar proteins. In this study, we analyzed the role of the FliD C-terminal domain in pentamer formation and interaction with FliT. Our study showed that the FliD L443R mutant exists as a monomer in solution, indicating that the Leu443 residue of FliD, which contributes to its interaction with FliT, plays a crucial role in the pentameric oligomerization of FliD. Consistently, the increased levels of free FliT proteins caused by FliD L443R mutation had negative effects on the gene expression of flagellar synthesis and reduced the expression of flagellar proteins. The lengths of flagella in each cell were significantly reduced in L443R mutant strain, suggesting that normal flagellar biogenesis was impeded. These results suggest that the C-terminal domain of FliD plays a crucial role in the pentameric oligmerization of FliD and the binding of FliT to the C-terminal domain of FliD is critical to inhibit the premature assembly of the FliD pentamer in the cytosol.

(HAP2) are exported as partially unfolded monomers 11,12 . To transport unfolded monomers of HAPs though the HBB, flagella-specific chaperones such as FlgN and FliT are required 13 . FlgN functions as an export chaperone for FlgK and FlgL, while FliT binds to the flagellar filament capping protein FliD 14 . In Salmonella Typhimurium, FlgN and FliT form homodimers, and prevent oligomerization of HAPs by direct interaction using their C-terminus or N-terminus, respectively 14 , indicating that FlgN and FliT are substrate-specific flagellar chaperones that prevent oligomerization of the HAPs by binding to their helical domains before export from the cell 13 .
During the final stage of flagellar assembly, one might expect that the expression of flagellar proteins is repressed to prevent the cell from using energy to synthesize unnecessary proteins. The overall production of flagellar proteins is reduced by sensing flagellar morphology via FliD-FliT complex. Formation of the FliD-FliT complex was known to play diverse roles; first, by inhibiting the premature assembly of the FliD pentamer in the cytosol 13 ; second, by facilitating the secretion of FliD from cells by binding to the FlhA cytoplasmic domain only in complex with FliT 15 ; third, by increasing flagellar biosynthesis by enhancing FlhD 4 C 2 activity 16 . FlhD 4 C 2 activity is negatively regulated by FliT, thus, FliD was proposed to act as an anti-anti-regulator, in a manner similar to the σ 28 -FlgM regulatory circuit originally by Yamamoto and Kutsukake 17 . Indeed, FliD acts as an anti-anti-FlhD 4 C 2 factor because the FliD and FlhD 4 C 2 binding sites on FliT overlap 18 . FliT also interacts with the FlhC in FlhD 4 C 2 complex and increases the presentation of the FlhC recognition region to ClpX, resulting in FlhC proteolysis 17,19 .
Recently, solution structure of the FliD-FliT complex, containing a smaller fragment of FliD compared to our crystal structure, was reported 16 . However, it remains unclear why the binding of FliT to the C-terminal domain of FliD is critical to inhibit the premature assembly of the FliD pentamer in the cytosol. To further analyze the detailed interactions at the FliD-FliT interface and understand the negative feedback loop in molecular details, we performed structural and functional analysis of the FliD-FliT interaction by multi-experiment analysis.

Results and Discussion
Interaction between StFliD and StFliT. Previously, it was shown that the C-terminal region of FliD from Salmonella Typhimurium (StFliD) is responsible for binding to FliT from S. Typhimurium (StFliT) 16 . To further analyze the interaction between StFliD and StFliT, various constructs truncating the N-or C-terminal regions of StFliD (StFliD 1-300 , StFliD 45-467 , StFliD 339-467 , StFliD 401-467 , and StFliD 428-467 ) were designed and their interactions with StFliT 1-94 were examined by purifying individual complexes ( Fig. 1a and b and Supplementary  Fig. S1). Previously, it was reported that the C-terminal segment of StFliT (residues 95-122) does not contribute to its binding to StFliD, but negatively regulates StFliT activity 4 , thus StFliT 1-94 was used. A stable binary complex of full-length FliD and C-terminus-truncated FliT from S. Typhimurium (StFliD and StFliT, respectively) could be reconstituted by coexpressing StFliD 1-467 and StFliT  in Escherichia coli cells. The binary complex of StFliD 1-467 and StFliT 1-94 was co-eluted from an affinity column and co-migrated on size exclusion columns ( Fig. 1a and b). All constructs except StFliD 1-300 in combination with StFliT 1-94 were purified and found to be stable ( Supplementary Fig. S2A). To analyze the oligomeric states of the StFliD-StFliT complex in solution, we measured the molecular weights (61 kDa and 18.7 kDa) of StFliD 1-467 -StFliT 1-94 (theoretical molecular weight of 1:1 complex: 61 kDa) and StFliD 401-467 -StFliT 1-94 (theoretical molecular weight of 1:1 complex: 18.7 kDa) complexes, respectively, using size-exclusion chromatography with multi-angle light scattering (SEC-MALS). The molecular masses measured for both complexes are consistent with a 1:1 complex of StFliD and StFliT in solution (Fig. 1b), which are the same results with the previous reports published in other groups 4, 18 . Crystal structure of the StFliD-StFliT complex and its comparison with the solution structure. Recently, the solution structure of StFliT fused with the C-terminal helix of StFliD was solved 16 . To solve the structure of FliT in complex with longer C-terminal helix of FliD, we determined the 2.3-Å resolution crystal structure of StFliD 401-467 bound to StFliT 1-94 (Fig. 1c). Nearly the entire sequences of both StFliD 401-467 and StFliT 1-94 were found to be ordered, with the exception of the N-terminal flexible loop (11 residues) in StFliD 401-467 ( Supplementary Fig. S2B). StFliD 401-467 forms a long coil over the entire length (~80 Å) and bound to the concave surface of the StFliT 1-94 monomer ( Fig. 1c and Supplementary Fig. S2). Overall crystal structure of StFliD-StFliT complex was similar to the solution structure 4,16 (Fig. 1c). However, the N-terminal region of the C-terminal helix of StFliD was slightly different (Fig. 1c). When both StFliD-StFliT structures were superimposed, the amino terminal parts of StFliD did not overlap well (Fig. 1c). We speculated that the bending of the StFliD C-terminal helix in the solution structure might be induced by the fusion of StFliT with the C-terminal helix of StFliD. The interface between StFliD and StFliT buries several hydrophilic residues (Arg58, Glu75, and Arg85) of StFliT against those of StFliD (Gln439, Thr453, Ser454, Met464, and Lys466) (Fig. 1d). Between these regions, two hydrophobic cores are buried; one consists of five residues (Trp30, Leu33, Leu34, Leu78, and Leu82) of StFliT and three residues (Leu443, Met446, and Leu450) of StFliD (Fig. 1d). The other hydrophobic core consists of seven residues (Phe7, Trp11, Ile14, Tyr40, Ile44, Ile68, and Leu72) of StFliT and three residues (Tyr456, Leu457, and Met464) of StFliD (Fig. 1d).
Crucial role of Leu443 residue in the StFliD-StFliT complex. To evaluate the contributions of the interfacial residues of the StFliD-StFliT complex, the interfacial residues (Gln439 and Leu443) of StFliD were substituted with Arg and their interactions with StFliT were examined using GST pull-down and isothermal titration calorimetry (ITC) ( Fig. 2a and b). When the StFliD Gln439, which forms a hydrogen bond with StFliT Arg85, was mutated, the interaction between StFliD and StFliT remained (Fig. 2a). However, when the StFliD Leu443, which is part of the hydrophobic core of the StFliD-StFliT complex, was mutated, the binding affinity between StFliD and StFliT was significantly reduced ( Fig. 2a and b), suggesting that the Leu443 residue of StFliD plays a crucial role in StFliD-StFliT complex formation. Full-length StFliD wild-type (WT) strongly bound to StFliT 1-94 with a dissociation constant of K d = 235.1 ± 0.8 nM (Fig. 2b), whereas the L443R mutation reduced the binding affinity (K d = 3.01 ± 0.1 μM), suggesting the importance of Leu443 in the direct interaction between StFliD and StFliT in cells.

SAXS structures of various StFliD-StFliT complexes.
We further investigated the molecular details of full-length StFliD and StFliT proteins in an SEC-MALS experiment. The StFliD 1-467 -StFliT 1-94 complex showed a concentration-dependent quaternary structure in solution. At a concentration of 100 μM, the molecular mass of the StFliD 1-467 -StFliT 1-94 complex was 126 kDa (theoretical molecular weight of 1:1 complex: 61 kDa), consistent with a 2:2 complex, while the molecular mass was 64 kDa at a concentration of 10 μM, corresponding to a 1:1 complex (Fig. 2c). We hypothesized that the StFliD 1-467 -StFliT 1-94 heterodimer dimerizes via the N-terminal domains of StFliD, as the C-terminal part of StFliD are engaged in StFliT binding. Consistently, StFliD 1-300 (100 μM) was a dimer (64 kDa) in solution (theoretical molecular weight of dimer: 65 kDa) (Fig. 2c). To determine the overall conformation of the StFliD 401-467 -StFliT 1-94 complex in solution, small-angle X-ray scattering (SAXS) data were collected ( Fig. 2d and Supplementary Fig. S3-5). The crystal structure of the StFliD 401-467 -StFliT 1-94 complex was successfully incorporated into the molecular envelope ( Fig. 2d and Supplementary Fig. S6A), indicating that the C-terminal region of StFliD is crucial for its pentamer assembly. Interestingly, our study showed that the StFliD L443R mutant exists as a monomer in solution at both high and low concentrations (100 μM and 10 μM, respectively) (Fig. 2c). This indicates that the Leu443 residue of StFliD, which contributes to its interaction with StFliT, plays a crucial role in the pentameric oligomerization of StFliD. These finding suggest that the C-terminal helix of StFliD is critical for the pentameric assembly of StFliD and the binding of StFliT to the C-terminal helix of StFliD can efficiently prevent the pentamer assembly of StFliD. Additionally, the solution structures of the StFliD 45-467 -StFliT 1-94 heterotetramer and StFliD 1-300 dimer revealed that the N-terminal domains can dimerize in a head-to-head fashion and both complexes showed a 2-fold symmetrical and elongated molecular shape ( Supplementary Fig. S6). In this orientation, the directions of both C-terminal parts of StFliD are opposite, and thus this may not be the orientation of the functional StFliD pentamer at the tip of flagella. These data suggest that the heterotetrameric form of the StFliD-StFliT might exist in the cytosol. Physiological effect of the StFliD L443R mutation in Salmonella motility. We further investigated whether the L443R mutation affects its own activity in facilitating the polymerization of Salmonella flagellin. It was previously reported that the Salmonella ΔfliD mutant strain was not motile, but became motile after extracellular addition of purified StFliD protein and subsequent incubation in culture medium 20 . In order to evaluate whether the StFliD L443R mutant can induce the polymerization of flagellin, two purified proteins (StFliD WT or L443R mutant) were added to the ΔfliD mutant culture, respectively. Indeed, transmission electron microscopy (TEM) supported that flagellar filaments were synthesized in ΔfliD mutant culture in the presence of either the StFliD WT or StFliD L443R mutant ( Supplementary Fig. S7). These observations suggest that the StFliD L443R mutant lost its ability to bind to StFliT, but could still synthesize filament at the distal flagellar tip. Thus, we further used the StFliD L443R mutant as a representative mutant to study the interaction between StFliD and StFliT. To investigate the physiological effect of the StFliD-StFliT complex in Salmonella motility, we evaluated the behavior of the ΔfliD mutant strain in the presence or absence of the pACYC184 vector expressing either the StFliD WT or L443R mutant (pPint-FliD or pPint-FliD (L443R), respectively) under its intrinsic promoter (Fig. 3). Bacterial motility was completely lost in the ΔfliD mutant strain, and the introduction of a plasmid expressing StFliD rescued motility to a similar level as in WT (ΔfliD + pPint-FliD in Fig. 3a). However, expression of the StFliD L443R mutant resulted in decreased motility as compared to StFliD + (ΔfliD + pPint-FliD (L443R) in Fig. 3a). Because the L443R mutation in StFliD significantly lowered its binding to StFliT, unbound StFliT would be free to inhibit FlhD 4 C 2 -dependent flagellar gene expression.
The role of StFliD Leu443 residue in the negative feedback loop. Previously, it was shown that FliD regulates the ability of FliT to inhibit FlhD 4 C 2 from activating flagellar gene expression 17 . Thus, StFliD was proposed to act as an anti-regulator of StFliT, suggesting that the secretion of StFliD is essential for the disassembly of the FliD-FliT complex, so that free FliT can bind to the FlhD 4 C 2 activator 18 . We further studied the negative feedback loop using our StFliD L443R mutant. In order to check whether the StFliD L443R mutation affects the expression of genes involved in flagellar synthesis, qRT-PCR and western blot analysis was carried out using the ΔfliD mutant strain transformed with the pACYC184 vector expressing either StFliD WT or L443R mutant under its intrinsic promoter. Representative genes belonging to three classes of the transcriptional hierarchy were selected and monitored ( Fig. 3c and Supplementary Fig. S8). We hypothesized that the increased levels of free StFliT proteins caused by StFliD deletion or StFliD L443R mutation had negative effects on the gene expression of flagellar synthesis by binding and inactivating the FlhD 4 C 2 activator and may reduce the expression of flagellar proteins. Indeed, the overall expression of flagellar genes was decreased in the ΔfliD mutant strain (ΔfliD + pACYC184) and restored by the expression of StFliD WT (ΔfliD + pPint-FliD) (Fig. 3c). The StFliD L443R mutant (ΔfliD + pPint-FliD (L443R)), however, could not restore gene expression in contrast to the ΔfliD mutant strain expressing exogenous WT StFliD (Fig. 3c). Additionally, western blot analysis revealed decreased levels of flagellar protein in the ΔfliD and L443R mutant strains, respectively ( Supplementary Fig. S9).
qRT-PCR analysis was also carried out using the ΔfliD mutant strain transformed with the pUHE21-2lacI q vector overexpressing either StFliD WT or StFliT WT (Fig. 3b and e). The overall expression of flagellar genes was restored in the ΔfliD mutant strain overexpressing StFliD (ΔfliD + pPlac-FliD), but significantly decreased by the overexpression of StFliT (WT + pPlac-FliT) (Fig. 3b and e). These results also suggested that large amounts of StFliT negatively affected the gene expression of flagellar synthesis by binding and inactivating the FlhD 4 C 2 activator. The model in which the secretion of StFliD has the ability to inactivate the FlhD 4 C 2 activator predicts that the fusion of bulky GFP protein to the N-terminus of StFliD might fail to be exported and might not inactivate the FlhD 4 C 2 activator. Indeed, the GFP-StFliD fusion increased the amount of GFP-StFliD protein found in the cytosol compared to that of StFliD without GFP (Fig. 3d). Cells expressing the GFP-StFliD protein were defective in motility because of the absence of StFliD cap at the tip of flagellar (Fig. 3b) and the decreased expression of flagellar genes in the ΔfliD mutant strain (ΔfliD) was restored by the expression of GFP-StFliD (ΔfliD + pPlac-EGFP-FliD-FLAG) (Fig. 3e). These results support the previously shown model that the secretion of StFliD is essential for the disassembly of the StFliD-StFliT complex, so that free StFliT can bind to the FlhD 4 C 2 activator 18 .

Perturbation of the StFliD-StFliT interaction by L443R mutation affects the lengths of flagellar filaments.
We next investigated whether perturbation of the StFliD-StFliT interaction affects the flagellar phenotype. As a representative phenotype of flagellar biogenesis, the lengths of flagellar filaments per cell were monitored ( Fig. 4a and b). Interestingly, the lengths of flagella in each cell were significantly reduced in StFliD deletion or L443R mutant strains (ΔfliD + pACYC184 or ΔfliD + pPint-FliD (L443R), respectively) compared to in the WT Salmonella strain (WT + pACYC184) (Fig. 4a), suggesting that normal flagellar biogenesis was impeded because of reduced gene expression of flagellar proteins. The synthesis of flagella was nearly abolished when StFliD was not exported by fusing the bulky GFP to the N-terminus of StFliD (second panel in Fig. 4b). We next hypothesized that the lengths of flagellar were increased and decreased by the overexpression of StFliD and StFliT (ΔfliD + pPlac-FliD and WT + pPlac-FliT, respectively). Indeed, the lengths of flagellar slightly increased when StFliD was overexpressed (from 6.35 nm to 9.25 nm) (first panel in Fig. 4b), whereas the synthesis of flagellar was significantly abolished when StFliT was overexpressed (fourth panel in Fig. 4b). These results support the model that the interaction between StFliD and StFliT is crucial for the overall regulation of flagellar gene expression.

Concluding remarks.
The bacterial flagellum is a model system for how very large molecular machines can be assembled in order. According to the morphological development of bacterial flagellum, the expression of flagellar genes is regulated in coordination with assembly. The bacterial flagellum is composed of thousands of protein subunits; thus, the synthesis of unnecessary flagellar proteins may be a burden to bacteria and the negative feedback loop by StFliD secretion may be essential for the life of bacteria 17,18,21 . Nevertheless, the feedback loop by the StFliD secretion may not lead to complete repression of flagellar genes. Each cell contains more than one flagellum; thus, the morphological development of each flagellum may differ at different times and basal expression of flagellar genes may be necessary. Consistent with this hypothesis, perturbation of the StFliD-StFliT interaction by expressing the StFliD L443R mutant did not lead to an all-or-nothing response, but basal expression of flagellar genes was detected (Fig. 3c). The feedback loop may control the levels of flagellar protein to adjust the expression of flagellar proteins. Although additional studies are needed to evaluate the molecular mechanism of gene regulation in flagellar biogenesis, the results and model presented here provide a foundation for further analysis.  and restrained refinement was performed using the REFMAC5 program 26 . Several rounds of model building, simulated annealing, positional refinement, and individual B-factor refinement were performed Supplementary  Table S1 lists the refinement statistics. The crystallographic asymmetric unit of the StFliD 401-467 -StFliT 1-94 complex contains one heterodimer. The refined model includes 193 water molecules, and 100% of the residues are in the most allowed region of the Ramachandran plot. No electron density was observed for residues 401-411 in StFliD. Atomic coordinates and structure factors for the StFliD 401-467 -StFliT 1-94 complex have been deposited in the Protein Data Bank (PDB ID code 5GNA).

Size exclusion chromatography with multi-angle light scattering (SEC-MALS). SEC-MALS
experiments were performed using an FPLC system (GE Healthcare) connected to a Wyatt MiniDAWN TREOS MALS instrument and a Wyatt Optilab rEX differential refractometer (Santa Barbara, CA, USA). A Superdex 200 10/300 GL (GE Healthcare) gel-filtration column pre-equilibrated with buffer A was normalized using ovalbumin protein. Proteins were injected (0.5-5 mg) at a flow rate of 0.4 mL/min. Data were analyzed using the Zimm model for fitting static light-scattering data and graphed using EASI graph with a UV peak in ASTRA V software (Wyatt). Sample-to-detector distances of 4.00 m and 1.00 m for SAXS were used. The magnitude of the scattering vector, q = (4□/L) sin l, was 0.1 nm −1 < q < 6.50 nm −1 , where 2l is the scattering angle and λ is the wavelength of the X-ray beam source. The scattering angle was calibrated with polystyrene-b-polyethylene-b-polybutadiene-b-polystyrene block copolymer standard. We used a quartz capillary with an outside diameter of 1.5 mm and wall thickness of 0.01 mm as solution sample cells. All scattering measurements were carried out at 4 °C by using a FP50-HL refrigerated circulator (JULABO, Allentown, PA, USA). SAXS data were collected in six successive frames of 0.1 min each to monitor radiation damage. Measurements of protein solutions were carried out over a wide concentration range (1.0-10.0 mg/mL). Each 2D SAXS pattern was radial-averaged from the beam center and normalized to the transmitted X-ray beam intensity, which was monitored using a scintillation counter placed behind the sample. The scattering of specific buffer solutions was used as the experimental background. The R g,G (radius of gyration) values were estimated from the scattering data using Guinier analysis 27 . The molecular mass was calculated from a BSA standard protein. The pair distance distribution p(r) function was obtained using the indirect Fourier transform method in the program GNOM 28 .

Construction of 3D structural models.
To reconstruct the molecular shapes, the ab initio shape determination program DAMMIF 29 was used. For each model reconstruction, 10 independent models were selected and the averaged was aligned using the program DAMAVER 30 . The SAXS curves were calculated from the atomic models using the program CRYSOL 31 . To compare the overall shapes and dimensions, ribbon diagrams of the atomic crystal models were superimposed onto the reconstructed dummy atom models using the program SUPCOMB 32 .
Bacterial strains, plasmids, and culture conditions. Salmonella was genetically manipulated using the phage λ Red recombination system 33 and phage P22-mediated transduction 34 with Salmonella enterica serovar Typhimurium SL1344 as the parent strain. All bacterial strains and plasmids used in this study are listed in Supplementary Table S4. Bacteria were grown aerobically at 310 K in LB medium supplemented with antibiotics. Antibiotics were used at the following concentrations: 50 μg/mL ampicillin, 25 μg/mL chloramphenicol, and 50 μg/mL kanamycin.
Construction of bacterial strains. The phage λ-derived Red recombination system was used to delete genes in frame or to fuse genes/proteins with peptide tags 33 . To construct SR7031 (ΔfliD), the Km R cassette from pKD13 was amplified using primers fliD-del-F and fliD-del-R, and the resultant PCR products were introduced into the SL1344 strain containing the plasmid pKD46. Recombinant bacteria containing the Km R cassette in place of target genes were determined using kanamycin resistance and diagnostic PCR. The Km R cassette was further removed using the plasmid pCP20 21 as described previously. FlhC, FliA, FliT, or FliC with the FLAG peptide at the C-terminus was also constructed using the phage λ-mediated recombination system 33,35 . The primers used to construct the bacterial strains are listed in Supplementary Table S5.

Plasmid construction.
To construct pPint-FliD producing FliD under its putative intrinsic promoter, DNA containing the Salmonella fliD gene was amplified by PCR using the primers fliD-comple-F and fliD-comple-R, and the PCR products were introduced between the BamHI and SphI sites of pACYC184. Plasmids expressing FliD with L443R substitution were constructed using a QuickChange II site-directed mutagenesis kit (Stratagene, La Jolla, CA, USA) with the primers fliD-L443R-F and fliD-L443R-R for FliD (L443R). pPlac-FliD or pPlac-FliT plasmids producing FliD or FliT under the lac promoter, respectively, were constructed. DNA containing the Salmonella fliD or fliT genes was amplified by PCR using the primers fliD-over-F and fliD-over-R or fliT-over-F and fliT-over-R, and the PCR products were introduced between the BamHI and SalI sites of pUHE21-2lacI q . pPlac-EGFP-FliD-FLAG or pPlac-FliD-FLAG plasmids producing EGFP-FliD-FLAG or FliD-FLAG under the lac promoter, respectively, were constructed. DNA containing the Salmonella fliD gene was amplified by PCR using the primers pET28a-EGFP-FliD-F and pET28a-EGFP-FliD-R, and the PCR products were introduced between the BamHI and HindIII sites of pET28a-EGFP. Afterwards, DNA containing the EGFP-fliD gene was amplified by PCR using the primers EGFP-fliD-FLAG-F and EGFP-fliD-FLAG-R, and the PCR products were introduced between SalI and HindIII sites of pUHE21-2lacI q (pPlac-EGFP-FliD-FLAG). DNA containing Salmonella fliD gene was also amplified by PCR using fliD-FLAG-F and fliD-FLAG-R, and the PCR products were introduced between SalI and HindIII sites of pUHE21-2lacI q (pPlac-FliD-FLAG). The primers used for plasmid construction are listed in Supplementary Table S5.
Flagella growth on living cells. Wild-type and ΔfliD mutant strains were cultured in LB medium at 310 K with shaking (220 rpm) overnight. Bacterial cultures were inoculated into LB medium again at a 1/100 dilution supplemented with purified FliD or FliD (L443R) proteins (approximately 0.2 mg/mL) and incubated at 310 K for 3 h to the mid-log phase.
Transmission electron microscopy. Salmonella cells were prepared in growth medium that had not been centrifuged, as centrifugation may disrupt fine bacterial structure such as pili, flagella, and capsules. A droplet of the bacterial suspension was placed onto a 50-nm-thick Formvar film that had been transferred onto a copper index TEM grid (Canemco & Marivac, Quebec, Canada). After waiting for 15 min, the remaining solution was wicked away using a piece of filter paper. The samples were then rinsed with 2 mL of 2 mM HEPES buffer, pH 6.8, followed by rinsing with 1 mL of MilliQ water. The samples were then negatively stained with phosphotungstic acid and observed using a LIBRA 120 Energy-Filtering Transmission Electron Microscope (Zeiss, Jena, Germany) at a magnification of ×8000. Bar, 1 μM.
Motility assay and measurement of filament length. One microliter of an overnight culture was spotted onto the middle of a swim plate (LB medium, 0.3% agar supplemented with antibiotics at the following concentrations: 25 μg/mL chloramphenicol or 50 μg/mL ampicillin, or 0.1 mM IPTG). Plates were allowed to dry for 1 h at room temperature. All plates were incubated at 310 K for 6 h. The flagella images in electron micrographs were analyzed with an image processor (IBAS, Zeise) equipped with a Videoplan program (Zeiss).

Analysis of EGFP-FliD-FLAG protein in Salmonella. Bacterial cells containing pPlac-EGFP-FliD-FLAG
plasmid (a pUHE21-2lacI q derivative expressing EGFP-FliD-FLAG under the lac promoter) exponentially grown in LB medium with IPTG (0.1 mM) was centrifuged (16000 g, 1 min) and resuspended in PBS. The cells were washed twice with PBS buffer and observed by epifluorescence microscopy (DE/Axio Imager A1 microscope, Carl Zeiss, Oberkochen, Germany) with a GFP filter (470/40 nm excitation, 495 nm dichroic, 525/50 nm emission). To compare FLAG-tagged StFliD protein levels in ΔfliD mutant strain with or without pPlac-EGFP-FliD-FLAG or pPlac-FliD-FLAG (a pUHE21-2lacI q derivative expressing EGFP-FliD-FLAG or FliD-FLAG, respectively) under the lac promoter, protein samples were isolated from each culture grown in LB medium at mid-log phase in the presence of IPTG (0.1 mM) and subjected to western blot analysis. Flagellins were removed by votexing prior to the western blot analysis.

SDS-PAGE and western blotting analysis.
Whole-cell fractions (extracted by sonication or pellets) were dissolved in Laemmli's sample buffer and boiled for 5 min. The protein samples were loaded onto a 12% SDS-polyacrylamide gel. Proteins were separated on the gel depending on the molecular weights and transferred to a polyvinylidene difluoride membrane. The membrane was blocked with 5% nonfat dry milk in 1× Tris-buffered saline-Tween 20 buffer and probed with anti-FLAG antibody (3:2000 dilution, F1804, Sigma, St. Louis, MO, USA) or anti-DnaK antibody (1:10,000 dilution, ADI-SPA-880, Enzo Life Science, Farmingdale, NY, USA) as primary antibodies. Anti-mouse IgG conjugated to peroxidase (3:5000 dilution, sc-2005, Santa Cruz Biotechnology, Santa Cruz, CA, USA) was used as the secondary antibody in all western blotting experiments. The chemiluminescent signals were developed with a West-Zol plus western blot detection system (Intron Biotechnology, Daejeon, Korea).
RNA isolation and quantitative real-time RT-PCR. Salmonella strains were grown in LB medium and total RNA was isolated at mid-log phage using an RNeasy mini kit (Qiagen, Hilden, Germany). After DNase treatment of the isolated total RNA, cDNA was synthesized by RNA to the cDNA EcoDryPremix (Random Hexamers) (Clontech, Mountain View, CA, USA). cDNA was mixed with 2 × iQ SYBR Green Supermix (Bio-Rad, Hercules, CA, USA) and real-time PCR was performed using the CFX 3.1 (Bio-Rad). The mRNA expression level of the target gene was normalized relative to the gyrB (DNA gyrase subunit B) expression level. All primer sets for qRT-PCR used in this study are listed in Supplementary Table S6. Data availability. Atomic coordinates and structure factors for the StFliD401-467-StFliT1-94 complex have been deposited in the Protein Data Bank (PDB ID code 5GNA).