The flexible linker of the secreted FliK ruler is required for export switching of the flagellar protein export apparatus

The hook length of the flagellum is controlled to about 55 nm in Salmonella. The flagellar type III protein export apparatus secretes FliK to determine hook length during hook assembly and changes its substrate specificity from the hook protein to the filament protein when the hook length has reached about 55 nm. Salmonella FliK consists of an N-terminal domain (FliKN, residues 1–207), a C-terminal domain (FliKC, residues 268–405) and a flexible linker (FliKL, residues 208–267) connecting these two domains. FliKN is a ruler to measure hook length. FliKC binds to a transmembrane export gate protein FlhB to undergo the export switching. FliKL not only acts as part of the ruler but also contributes to this switching event, but it remains unknown how. Here we report that FliKL is required for efficient interaction of FliKC with FlhB. Deletions in FliKL not only shortened hook length according to the size of deletions but also caused a loose length control. Deletion of residues 206–265 significantly reduced the binding affinity of FliKC for FlhB, thereby producing much longer hooks. We propose that an appropriate length of FliKL is required for efficient interaction of FliKC with FlhB.

FliK L contains ten proline residues (Fig. 1a) 39 and hence is intrinsically disordered 41 . Hook lengths of the Salmonella fliK(∆238-269) (32 residues deletion) and fliK(∆248-269) (22 residues deletion) mutants are 40.2 ± 6.1 nm [mean ± standard deviation (SD)] and 51.0 ± 8.8 nm, where their average lengths are shorter than that of the wild-type strain (52.7 ± 4.5 nm) 42 . The length of the hook produced by the fliK(∆161-216) mutant (56 residues deletion) is 48.7 ± 22.3 nm, where the average is also shorter than that of the wild-type strain by 4 nm. However, the SD value is larger than the wild-type one, indicating a much looser length control of the hook structure 42 . Furthermore, deletions of residues 161-223 and 161-244 cause polyhooks without the filament attached whereas a deletion of residues 208-269 results in the polyhooks with the filament attached (polyhook-filament phenotype) 42 . These observations raise the possibility that FliK L not only acts as part of the ruler but also contributes to substrate specificity switching of the flagellar protein export apparatus. To clarify this hypothesis, we constructed a series of mutant variants of FliK with in-frame deletions within FliK L . We show that a proper length of FliK L between FliK N and FliK C is required for efficient interaction of FliK C with FlhB C .

Results
Effect of deletions of five amino-acid residues within residues 206-235 on hook length control.
It has been shown that the N-terminal portion of FliK L is responsible for proper measurement of hook length 42 . To clarify the role of residues 206-235 of FliK L in the hook length control, we constructed a series of mutant variants of FliK with sequential 5-amino-acid deletions within a region of residues 206-235, namely FliK(∆206-210), FliK(∆211-215), FliK(∆216-220), (∆221-225), (∆226-230) and FliK(∆231-235) ( Table 1). These six fliK deletion variants fully restored motility of the ∆fliK mutant in 0.35% soft agar plates when they were expressed from the pTrc99A-based plasmid (Fig. 1b). Consistently, the levels of FlgE and FliC secreted by these deletion mutants were detected at the wild-type levels (Fig. 1c, 2nd and 3rd rows). These fliK deletions did not affect either protein stability or protein secretion into the culture media (Fig. 1c, 1st row). Therefore, we conclude that these in-frame deletions do not affect FliK function at all.
The length of the most extended polypeptide chain is 0.37 nm per residue. If FliK L adopts a fully extended conformation to act as part of the ruler, we predicted that these 5-amino-acid deletions within residues 206-235 of FliK L would reduce the hook length by 1.9 nm. Therefore, we measured the hook length of these fliK deletion mutants. The average hook length of the fliK(∆206-210), fliK(∆211-215), fliK(∆216-220), fliK(∆221-225), fliK(∆226-230) and fliK(∆231-235) mutants were 48.5 ± 4.5 nm (n = 113), 49.3 ± 5.2 nm (n = 168), 49.0 ± 5.6 nm (n = 130), 48.6 ± 4.2 nm (n = 118), 49.2 ± 4.2 nm (n = 232) and 49.6 ± 5.4 nm (n = 126), respectively, which were shorter than the length of the wild-type hook (53.3 ± 6.5 nm, n = 154) (Fig. S1). Over-expression of FliK slightly shortens the hook length. In contrast, when the expression level of FliK is reduced, the cell produces polyhooks, sometimes with the filament attached 27,48 . Polyhooks are frequently observed when FlgE is overproduced in wild-type cells 27,43 . Thus, the balance between the secretion levels of FlgE and FliK seems to be critical for the proper termination of the hook assembly. Since 5-amino-acid deletions within residues 206-235 shorten the hook length by 4 nm, which is shorter than the predicted value, we assume that these deletions may affect not only hook length measurements but also the secretion process of FliK by the type III protein export apparatus and/or the export switching process of the protein export apparatus induced by the interaction of FliK C with FlhB C .
To further understand the ruler function of residues 206-235 in FliK L , we constructed larger deletion variants, FliK(∆206-215), FliK(∆216-225), FliK(∆226-235), FliK(∆206-220), FliK(∆221-235) and FliK(∆206-235) and analyzed their motility in 0.35% soft agar plates (Fig. S2a). Motility of the fliK(∆206-215), fliK(∆226-235), fliK(∆206-220), fliK(∆221-235) and fliK(∆206-235) cells was almost the same as wild-type motility whereas that of the fliK(∆216-225) mutant was slightly less than the wild-type level (Fig. S2a). The cellular and secretion levels of these deletion variants were essentially the same as the wild-type levels (Fig. S2b). Because these deletion mutants were expressed from the pTrc99AFF4 vector, it is also possible that their over-expression results in motility comparable to the wild-type level. To verify this possibility, the wild-type fliK allele on the chromosomal DNA was replaced by the fliK(∆206-235) allele. Motility of the fliK(∆206-235) mutant was slightly less than that of wild-type cells (Fig. S2c). To test whether the deletion of residues 206-235 affect hook length, we isolated hook-basal bodies from the fliK(∆206-235) mutant and measured the hook length (Figs. 2 and S3). A major peak of the hook length distribution was shifted to a shorter value than that of the wild-type (Fig. 2). While the majority of wild-type hook length was distributed within a range from 50 nm to 60 nm, the hook length distribution of the fliK(∆206-235) mutant showed a major peak population between 41 nm and 50 nm. However, longer hooks were also observed albeit shorter than those of polyhooks produced by the ∆fliK mutant [362.8 ± 200.9 nm (N = 146)]. As a result, the average hook length of the fliK(∆206-235) mutant was 54.5 ± 15.5 nm (n = 112), compared to 53.8 ± 5.6 nm (n = 130) for the wild-type. The SD value of the fliK(∆206-235) mutant was larger than the wild-type one, indicating that the deletion of residues 206-235 in FliK L cause a looser hook length control (2020) 10:838 | https://doi.org/10.1038/s41598-020-57782-5 www.nature.com/scientificreports www.nature.com/scientificreports/ than the wild-type. Therefore, we suggest that a deletion of residues 206-235 not only shortens the hook length according to the size of deletion but also affects the interaction of FliK C with FlhB C during hook assembly.
Effect of much larger deletions within FliK L on hook length control and substrate specificity switching. To clarify the role of FliK L in the export switching process of the flagellar type III protein export apparatus, the wild-type fliK allele on the chromosome was replaced by the fliK(∆206-245), fliK(∆206-255) or www.nature.com/scientificreports www.nature.com/scientificreports/ fliK(∆206-265) allele. Motility of these mutants were worse than wild-type motility (Fig. 3a) although the cellular and secreted amounts of FlgE and FliC were detected almost at the wild-type levels (Fig. 3b, 2nd and 3rd rows). Because neither cellular nor extracellular FliK level was affected by these larger deletions within FliK L (Fig. 3b, 1st row), this suggests that these deletions reduce FliK function.
To investigate whether these larger deletions directly affect the export switching function of FliK C , we introduces deletions residues 206-245, 206-255 or 206-265 into the fliK(∆2-99) allele. Motility of the cells over-expressing FliK(∆2-99) with these three deletions was reduced significantly (Fig. 3d), and especially the deletion of residues 206-265 reduced the cellular and extracellular levels of FliC by about 3-fold, thereby reducing motility considerably (Fig. 3e). This suggests that these three deletions reduce the export switching function of FliK C .

Effect of the 100 residues suppressor insertion on the length of hook produced by the
The amino acid sequence of FliK(∆206-265SP) is longer by 40 amino-acids than that of wild-type FliK, from which the average hook length is predicted to be about 70 nm, thereby reducing motility. To verify this hypothesis, we isolated hook-basal bodies from the ∆fliK mutant carrying pMMK1015 [FliK(∆206-265)] or pMMK1015SP (2020) 10:838 | https://doi.org/10.1038/s41598-020-57782-5 www.nature.com/scientificreports www.nature.com/scientificreports/ [FliK(∆206-265SP)] and measured their hook length (Figs. 6c and S5). The average hook length of the ∆fliK mutant carrying pMMK1015 was 225.4 ± 167.7 nm (n = 127), which were longer than that of the MMK1015 strain (107.1 ± 65.0 nm). Since FliK(∆206-265) was expressed from the pTrc99AFF4 vector, we assume that such a length difference may be a consequence of the multicopy effect of FliK(∆206-265SP). The hook length distribution of the ∆fliK mutant carrying pMMK1015SP showed a major peak population between 61 nm and 100 nm, but much longer hooks and polyhooks were observed as well. As a result, the average hook length of the ∆fliK mutant carrying pMMK1015SP was 204.0 ± 183.9 nm (n = 518), which was shorter than that of the ∆fliK mutant carrying pMMK1015. This suggests that the suppressor insertion mutation increases the probability of the interaction between FliK C with a deletion of residue 206-265 and FlhB C , thereby increasing the export switching probability of the flagellar type III protein export apparatus.

Discussion
The bacterial injectisome directly transports virulence effector proteins into the cytosol of host cells for bacterial infection. The injectisome consists of basal body rings and a tubular structure called the needle and looks similar to the flagellar hook-basal body 50 . The injectisome uses a secreted molecular ruler, SctP (originally referred to as YscP and InvJ in the Yeshinia and Salmonella injectisomes, respectively) to determine the needle length in a way similar to FliK 51,52 . The core domain of FliK C is conserved among FliK/SctP family 53 and possesses a fold similar to the C-terminal domain of SctP of the injectisome of Pseudomonas aeruginosa 54 . It has been shown that residues 301-350 of FliK C are directly involved in substrate specificity switching of the flagellar type III protein export apparatus 47 . Recent photo-crosslinking experiments have demonstrated that the conserved core domain of FliK C directly binds to FlhB C 45 . Similar protein-protein interactions have been observed in the injectisome 55 . These suggest that length control and substrate specificity switching mechanisms are conserved in both flagellar and injectisome systems. However, it remains unknown how the length measurement process by the secreted ruler is linked to the substrate specificity switching process of the type III protein export apparatus.
It has been reported that FliK L forms part of the ruler to determine hook length, but a deletion of residues 208-269 results in polyhooks with the filament attached 42 , having led to a hypothesis that residues 208-235 may contribute to efficient substrate specificity switching of the flagellar type III protein export apparatus. To verify this hypothesis, we introduced systematic deletions into FliK L and found that a deletion of residues 206-235 not only shortened hook length according to the size of deletion but also caused a loose hook length control (Fig. 2). The hook length control became much worse in the fliK(∆206-255) and fliK(∆206-265) mutants (Fig. 3c). The deletion of residues 206-265 considerably reduced the export switching function of FliK C (Fig 3d,e). An insertion of 100 amino-acids between Glu-270 and Trp-271 residues in the core domain of FliK C considerably improved the switching function of FliK(∆2-99 + ∆206-265), thereby shortening the hook length considerably and increasing the probability of filament formation significantly (Figs. 4 and 5). Consistently, this inserted sequence allowed FliK(∆2-99 + ∆206-265 + I304pBPA) to form a photo-crosslinked product with FlhB C in a way similar to FliK(I304pBPA) (Fig. 4e). Therefore, we suggest that the inserted sequence of the suppressor mutant significantly increases the binding affinity of the core domain of FliK C for FlhB C . Although FliK(∆2-99) is not secreted via the flagellar type III protein export apparatus during hook assembly, it retains the ability to catalyze the substrate specificity switching of the flagellar type III protein export apparatus to a considerable degree 49 . Therefore, we suggest that FliK L is also required for efficient interaction between the core domain of FliK C and FlhB C . When the length of FliK L was shortened by deletions, the export switching activity of FliK C was reduced depending on the size of deletions (Fig. 3). Furthermore, when the linker length became longer by 40 amino-acids compared to the wild-type length, the switching function of FliK C became worse (Fig. 6). Therefore, we propose that a proper length of FliK L between FliK N and FliK C may be important for FliK C to bind to FlhB C to switch the substrate specificity of the flagellar type III protein export apparatus. Assuming that FliK N suppresses the switching activity of FliK C when FliK N gets close to FliK C via deletion of residues in FliK L , it is also possible that FliK L may push FliK N away from FliK C to allow these two domains to fully exert their own functions. www.nature.com/scientificreports www.nature.com/scientificreports/ The core domain of FliK C consists of four β-strands, β1, β2, β3 and β4 and two α-helices, α1 and α2. Three parallel β1, β3 and β4 strands and one anti-parallel β2 strand form a hydrophobic core with the α1 and α2 helices (Fig. 4d) 46 . Highly conserved Val-302 and Ile-304 residues in the β2 strand, which are critical for the switching function of FliK 47 , form hydrophobic interaction networks in FliK C 46 . Photo-crosslinking experiments have shown that Val-302 and Ile-304 are in very close proximity to FlhB C , suggesting that these two residues are exposed on the molecular surface of FliK C upon binding to FlhB C 45 . Since FliK(∆2-99 + ∆206-265SP + I304pBPA) formed a photo-crosslinked product with FlhB C whereas neither FliK(∆2-99 + I304pBPA) nor FliK(∆2-99 + ∆206-265 + I304pBPA) did (Fig. 4e), this suggests that the inserted sequence of the suppressor mutant induces a conformational change of the N-terminal portion of the core domain of FliK C to allow Ile-304 to bind to FlhB C . Therefore, we propose that FliK L may be required for efficient conformational rearrangements of FliK C to interact with FlhB C . However, it is also possible that deletion of residues 206-265 makes FliK N very close to FliK C to suppress the interaction of FliK C with FlhB C and that the inserted sequence of the suppressor mutant may push FliK N away from FliK C , allowing FliK to bind to FlhB C . www.nature.com/scientificreports www.nature.com/scientificreports/ FliK C associates with and dissociates from FliK N in solution 41 . When FliK L adopts a fully extended conformation, the N-terminal portion of the core domain of FliK C becomes disordered 41 . Because the length of the most extended polypeptide is 0.37 nm per residue, the stretch of FliK sequence inside the channel of the hook-basal body must be longer than 250 residues to measure the hook length of about 55 nm together with the rod length of 35 nm. Therefore, we propose that FliK L may adopt a fully extended conformation when the hook length reaches about 55 nm, allowing Val-302 and Ile-304 in the hydrophobic core domain of FliK C to be in very close proximity to FlhB C to catalyze substrate specificity switching of the flagellar type III protein export apparatus.  www.nature.com/scientificreports www.nature.com/scientificreports/ and fliK(∆206-265) alleles, using the λ Red homologous recombination system 56 . L-broth contained 10 g of Bacto-Tryptone, 5 g of yeast extract and 5 g of NaCl per liter. Soft agar plates contained 10 g of Bacto Tryptone, 5 g of NaCl and 0.35% Bacto-Agar per liter. Ampicillin was added at a final concentration of 100 μg/ml if necessary. DNA manipulations. DNA manipulations were carried out as described 57 . A series of mutant variants of FliK with deletions within FliK L were generated by inverse PCR using pKM002 48 or pNM201 44 as a template. The fliK(∆206-265SP) allele were generated by overlap PCR method. All of the fliK deletions were confirmed by DNA sequencing. DNA sequencing reactions were carried out using BigDye v3.1 (Applied Biosystems) and then the reaction mixtures were analyzed by a 3130 Genetic Analyzer (Applied Biosystems).
Motility assays in soft agar. Fresh colonies were inoculated onto 0.35% soft tryptone agar plates and incubated at 30 °C. At least seven independent measurements were carried out. Secretion assays. Secretion assays were performed as described previously 58 . Salmonella cells were grown in 5 ml L-broth containing 100 μg/ml ampicillin at 30 °C with shaking until the cell density had reached an OD 600 of ca. 1.2-1.6. 1.5 ml of each culture was transferred into a 1.5 ml Eppendorf tube. After centrifugation (15,000 g, 5 min, 4 °C), cell pellets and culture supernatants were collected, separately. The cells were resuspended in OD 600 × 250 μl of SDS-loading buffer (62.5 mM Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 0.001% bromophenol blue) containing 1 μl of 2-mercaptoethanol and heated at 95 °C for 3 min. Trichloroacetic acid was added to each culture supernatant at a final concentration of 10%. After leaving on ice for 1 h, proteins in the culture supernatants were precipitated by centrifugation at 20,000 g for 20 min. Pellets were suspended in OD 600 x 25 μl of a Tris-SDS loading buffer (one volume of 1 M Tris, nine volumes of 1 × SDS loading buffer) containing 1 μl of 2-mercaptoethanol and heated at 95 °C for 3 min. After sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), immunoblotting with polyclonal anti-FlgE, anti-FliC or anti-FliK antibody was carried out as described previously 16 . Detection was done with an ECL prime western blotting detection reagent (GE Healthcare). Chemiluminescence signals were captured by a Luminoimage analyzer LAS-3000 (GE Healthcare). The regions of interest were cropped from original immunoblots shown in the Supplemental information using a software, Photoshop CS6, and then the contrast and brightness were adjusted. At least three independent experiments were performed.
Electron microscopy. Osmotically shocked Salmonella cells were prepared described previously 45 . After centrifugation (18,500 g, 30 min), the cell pellets were resuspended in 200 μl of H 2 O. Samples were applied to carbon-coated copper grids and were negatively stained with 1.0% (W/V) phosphotungstic acid, pH 7.0. Micrographs were recorded at a magnification of × 5,000 with a JEM-1200EXII transmission electron microscope (JEOL, Tokyo, Japan) operating at 80 kV.
Hook-basal bodies and polyhook-basal bodies were isolated as described before 27 . Salmonella cells were grown in 5 l L-broth containing ampicillin at 30 °C with shaking until the cell density had reached an OD 600 of ca. 1.0. The cells were harvested by centrifugation (10,000 g, 10 min, 4 °C) and suspended in 20 ml of ice-cold 0.1 M Tris-HCl pH 8.0, 0.5 M sucrose, followed by addition of EDTA and lysozyme at the final concentrations of 10 mM and 0.1 mg/ ml, respectively. The cell suspensions were stirred for 30 min at 4 °C and then were solubilized on ice for 1 hour by adding Triton X-100 and MgSO 4 at final concentrations of 1% and 10 mM, respectively. The cell lysates were adjusted to pH 10.5 with 5 M NaOH and then centrifuged (10,000 g, 20 min, 4 °C) to remove cell debris. After ultracentrifugation (45,000 g, 60 min, 4 °C), pellets were resuspended in 10 mM Tris-HCl, pH 8.0, 5 mM EDTA, 1% Triton X-100, and the solution was loaded a 20-50% (w/w) sucrose density gradient in 10 mM Tris-HCl, pH 8.0, 5 mM EDTA, 1% Triton X-100. After ultracentrifugation (49,100 g, 13 h, 4 °C), intact flagella were collected and ultracentrifuged (60,000 g, 60 min, 4 °C). Pellets were suspended in 50 mM glycine, pH 2.5, 0.1% Triton X100, and were incubated at room temperature for 30 min to depolymerize flagellar filaments. After ultracentrifugation (60,000 g, 60 min, 4 °C), pellets were resuspended in 50 μl of 10 mM Tris-HCl, pH 8.0, 5 mM EDTA, 0.1% Triton X100. Samples were negatively stained with 2%(w/v) uranyl acetate. Samples were applied to carbon-coated copper grids and were negatively stained with 2%(w/v) uranyl acetate. Electron micrographs were recorded with a JEM-1011 transmission electron microscope (JEOL, Tokyo, Japan) operated at 100 kV and equipped with a F415 CCD camera (TVIPS, Gauting, Germany). Hook length was measured by ImageJ version 1.48 (National Institutes of Health). 59 and a pETDuet-based plasmid encoding both FliK with an amber mutation and FlhB C were exponentially grown at 30 °C in L-broth containing 1 mM pBPA. Then, 100 μM IPTG and 0.02% arabinose were added and the incubation was continued until the culture density had reached an OD 600 of ca. 1.4-1.5. Photo-crosslinking was carried out as described previously 60 . The cell pellets were harvested by centrifugation, suspended in SDS-loading buffer, and heated at 95 °C for 3 min. After SDS-PAGE, immunoblotting with polyclonal anti-FliK antibody was carried out. Detection was done with an ECL prime western blotting detection reagent. Chemiluminescence signals were captured by a Luminoimage analyzer LAS-3000. The regions of interest were cropped from original immunoblots shown in the Supplemental information using a software, Photoshop CS6, and then the contrast and brightness were adjusted. At least three independent experiments were performed.

Data availability
All data generated or analyzed during this study are included in this published article and its Supplementary Information files.