The flagellar export apparatus consists of six integral membrane proteins (FlhA, FlhB, FliO, FliP, FliQ and FliR) and three soluble proteins (FliH, FliI and FliJ), and is thought to be located at the base of the flagellum1,2. Export of bacterial flagellar proteins has characteristics in common with the type III secretion system (T3SS) of virulence factors of pathogenic bacteria, and these two systems actually consist of many homologous component proteins5.

The soluble export component FliI is an ATPase3. Because mutants lacking the ATPase activity cannot export flagellar proteins, FliI was thought to convert chemical energy released by ATP hydrolysis into mechanical work necessary for the export3. FliI shows extensive similarity to the α/β subunits of proton-driven F0F1-ATPase for its entire molecular structure6, although sequence similarity is limited to their respective ATPase domains7,8. Unlike F1-ATPase, however, which forms the α3β3 hexameric ring, FliI self-assembles into a homo-hexamer9,10. When FliI oligomerization is suppressed by a small deletion in its amino-terminal region, flagellar protein export does not occur efficiently, suggesting that FliI hexamerizes on docking to the export gate made of the six integral membrane components, for which the cytoplasmic domains of FlhA and FlhB are thought to form the docking platform11. FliI binds to FlgN (chaperone) and to a FlgN–FlgK (HAP1) complex12, suggesting that FliI has a critical role in substrate recognition as well. As Salmonella InvC—a virulence T3SS homologue of FliI—binds to chaperone–effector complexes and induces chaperone release from, and unfolding of, the effector to be secreted in an ATPase-dependent manner13, FliI has been thought to act in a similar manner.

FliH binds to the extreme N-terminal region of FliI to form the FliH2–FliI complex in the cytoplasm, and suppresses FliI oligomerization and ATPase activity4,8. However, as FliI cannot efficiently dock to the export gate when FliH is missing14, FliH is postulated to provide a link between ATP hydrolysis by FliI and flagellar protein export, although it is not known how this is done. We therefore investigated this by characterizing a Salmonella fliH fliI double null mutant and by isolating gain-of-function mutants from this mutant.

A fliH null mutant was weakly motile whereas a fliI null mutant was non-motile14,15. Notably, when both FliH and FliI were missing (Fig. 1b, lane 4), the cells formed swarms on soft agar plates after prolonged incubation (Fig. 1a). Consistently, intact flagella as well as immature flagellar structures were occasionally observed by electron microscopy (data not shown). We then examined the effect of FlhB deletion, which makes wild-type cells non-motile (Fig. 1a). The motility of the double null mutant was totally abolished by a flhB::Tn10 mutation (Fig. 1c), indicating that some flagella are produced at a low probability even in the absence of FliH and FliI, but in a manner that still requires FlhB.

Figure 1: Characterization of a fliH fliI double null mutant.
figure 1

a, Swarming motility of SJW1103 strain (wild type, WT), and MKM11 (ΔfliH), MKM30 (ΔfliI), MMHI001 (ΔfliH fliI, indicated as ΔfliHI) and MKM50 (ΔflhB) mutant strains on soft agar plates. b, Immunoblotting, using anti-FliH and anti-FliI antibodies, of whole cell proteins. Weak bands just below those of FliI are due to nonspecific reactions. A slight downward shift of the FliI band in lane 1 is caused by an invisible band of a large amount of FliC (51.5 kDa) just above it. c, Effect of a flhB::Tn10 mutation on motility of the ΔfliH fliI mutant.

FliH and FliI interact not only with FlhA and FlhB but also with flagellar chaperones such as FliJ and export substrates, leading to the proposal that the FliH–FliI complex is responsible for delivery of the substrates to the export gate4,11,12,14. To test this, gain-of-function mutants were isolated from the fliH fliI double null mutant by streaking an overnight culture out on soft agar plates, incubating at 30 °C for 2 days and looking for motility haloes emerging from the streak. In total 22 pseudo-revertants were obtained. When we analysed the secretion of FliC (flagellin) into the culture supernatant by Coomassie brilliant blue staining, FliC was seen on SDS–polyacrylamide gel electrophoresis gels from the MMHI0117 strain, but not from the others (data not shown). Therefore, we decided to carry out further characterization of MMHI0117. The motility of MMHI0117 was significantly better than that of the parent mutant although not as good as that of wild type (Fig. 2a). Most cells of this strain had a couple of long flagellar filaments, whereas wild-type cells produce five or more (data not shown). Most flagellar proteins were detected in the culture supernatant of this mutant (Fig. 2b, lane 6). The amounts of FlgD (hook cap protein) and FliK (hook-length control protein) secreted by this pseudo-revertant were even larger than those of the wild type, whereas the secretion levels of FlgG (distal rod protein) and FlgE (hook protein) were threefold and tenfold lower than wild-type levels, respectively. Consistently, flagellar basal bodies, often with the hook and filament attached, were observed by electron microscopy (Fig. 2c). The secretion levels of these proteins were restored to wild-type levels by introduction of a plasmid encoding FliH and FliI into the mutant (Supplementary Fig. 1). These results indicate that the export of FlgE and FlgG depends on the FliH–FliI complex more strongly than the export of FlgD and FliK. As all of these proteins belong to the same rod-type and hook-type export class16, the FliH–FliI complex not only guides these proteins to the export gate but also controls their export order and levels. The cellular levels of FliC, FlgK and FlgL (HAP3) in the MMHI0117 strain were much lower than those in the wild type because expression of these genes occurs only after hook completion17, which is evidently retarded in this strain (Fig. 2b).

Figure 2: Isolation of pseudo-revertants from the Δ fliH fliI mutant.
figure 2

a, Motility of wild type, ΔfliH fliI mutant (ΔfliHI) and its pseudo-revertant (ΔfliH fliI, flhB(P28T), indicated as ΔfliHI, flhB*). b, Secretion analysis of FlgG, FlgE, FlgD, FliK, FliC, FlgK and FlgL by immunoblotting. ‘Cell’ and ‘Sup’ indicate whole cell proteins and culture supernatant fractions, respectively. Relative secretion levels are normalized to the wild-type level of each protein. Black bar, wild type; grey bar, pseudo-revertant. c, Electron micrographs of the flagella and basal bodies isolated from ΔfliH fliI, flhB(P28T). Scale bar, 200 nm. d, Position of a suppressor mutation in FlhB. CM, cytoplasmic membrane; cyto, cytoplasm; peri, periplasmic space.

P22-mediated genetic mapping18 showed that the gain-of-function mutation lies in the flhBAE operon (data not shown). DNA sequencing identified a missense mutation, P28T, near the N terminus of FlhB, close to the interface with the first membrane span (Fig. 2d). The position of this P28T mutation was very close to the positions of the FliH-bypass flhB mutations isolated previously14. Therefore, we tested whether the FliH-bypass flhB and flhA mutations can improve motility of the fliH fliI double mutant. These mutations, as represented here by a flhA(V404M) mutation, considerably enhanced both motility and flagellar protein export (Supplementary Fig. 2), although their suppression abilities were much weaker than that of the flhB(P28T) mutation.

Salmonella T3SS virulence factors are secreted by the flagellar export apparatus in the absence of their chaperones19. As InvC and SsaN are the T3SS homologues of FliI in Salmonella20,21, there is a possibility that InvC and/or SsaN may complement the fliH fliI double null mutants for flagellar protein export, although SsaN is not expressed under our experimental condition21. However, this was ruled out by our observations that neither motility nor protein export of the pseudo-revertant was abolished by InvC or SsaN deletions (Supplementary Fig. 3). Therefore, we conclude that the gain-of-function mutations in FlhA and FlhB increase the probability of entry of flagellar proteins into the export gate, thereby increasing export efficiency. An increased gate-opening probability for higher efficiency of protein entry could be deleterious to the cells owing to leakage of small solutes. However, the growth of the mutants was normal (Supplementary Fig. 4a), suggesting that the export gate is not always open. It is likely that the interaction of export substrates with the gate, with or without the FliH–FliI complex, induces opening of the pore.

Unlike the fliH null and the fliH fliI double null mutants, the fliI null mutant was non-motile (Fig. 1a), indicating that FliH inhibits flagellar protein export in the absence of FliI. To test whether the flhB(P28T) mutation would also suppress the fliI null mutation, we analysed the motility of the pseudo-revertant transformed with a plasmid encoding FliH (Supplementary Fig. 5). The motility gained by the flhB(P28T) mutation was significantly suppressed, suggesting that the docking of free FliH to the FlhA–FlhB platform interferes with the entry of export substrates into the export gate, even in the presence of the flhB(P28T) mutation.

As it has been shown that Yersinia enterocolitica type III secretion is inhibited considerably by treatment with the protonophore carbonyl cyanide m-chlorophenylhydrazone (CCCP)22, and that flagellar formation is also dependent on proton motive force (PMF)23, we investigated whether flagellar protein export is still dependent on PMF in the gain-of-function mutant as well as wild-type cells. The rate of cell growth decreased when CCCP concentration increased, and 25 μM CCCP immediately caused growth arrest (Supplementary Fig. 4b). As expected, the flagellar motor rotation, which is driven by the PMF1, was abolished by 25 μM CCCP treatment, indicating that the PMF is significantly collapsed (data not shown). The intracellular levels of FlgD were maintained even at 25 μM (Fig. 3); however, the levels of FlgD secretion by both wild-type and mutant cells decreased significantly above 10 μM CCCP and diminished at 25 μM (Fig. 3). In agreement with previous data22, the intracellular ATP level did not change on CCCP treatment within 1 h (Supplementary Fig. 4c). These results indicate that PMF is absolutely essential for FlgD export regardless of the presence or absence of FliH and FliI.

Figure 3: Effect of CCCP on FlgD secretion.
figure 3

a, Immunoblotting, using the anti-FlgD antibody, of whole cell proteins and culture supernatant fractions prepared from wild-type (WT) and a gain-of-function mutant (ΔfliH fliI, flhB*) grown at 30 °C in the presence of 0, 5, 10 and 25 μM CCCP. b, The decay in the secretion levels shown by plotting band densities of the two blots, which are normalized for the cellular FlgD levels. These data are the average of three independent experiments. The experimental errors are within 10%.

The MotA–MotB complex acts as a proton conductive pathway to couple proton influx with flagellar motor rotation1. Mutations in MotA or MotB do not affect flagellar formation while the motor is paralized1. Consistently, deletions of these two proteins do not interfere with flagellar protein export even in the gain-of-function mutant (data not shown).

Both ATP and PMF seem to be required for pre-protein translocation in Escherichia coli, where ATP is essential for the initial step of protein translocation and PMF increases the rate of translocation24. SecA is the ATPase that uses the energy of ATP binding and hydrolysis to drive pre-protein translocation into and across the cytoplasmic membrane24. Each catalytic cycle of SecA permits the stepwise translocation of pre-proteins across the membrane25,26. In contrast, we have shown that ATP hydrolysis by FliI is not absolutely required for flagellar protein export. As the diameter of the central channel of the flagellum—which is the export path for flagellar proteins—is only 2 nm27, proteins to be exported must be largely unfolded for entry into and translocation through the channel. Our observations—in particular, levels of FlgD and FliK secretion by a gain-of-function mutant that exceed wild-type levels (Fig. 2c)—suggest that the successive process of unfolding and translocation of export substrate is driven by PMF. The role of the FliH–FliI complex for efficient export would probably be to increase the initial docking efficiency of the N-terminal segments of export substrates into the export gate through specific interactions between the FliHx–FliI6 complex and the FlhA–FlhB platform. As the binding of the FliHx–FliI6 complex without ATPase activity inhibits the export process4, the energy of ATP hydrolysis seems to be used to facilitate the release and disassembly of the FliHx–FliI6 complex from the export gate and the protein about to be exported, in order for the subsequent PMF-driven, successive process of unfolding of the protein and translocation of the polypeptide chain through the gate by the export gate complex to proceed without retardation, as well as for efficient use of the FliH2–FliI complexes for the next cycle of initial docking (Fig. 4). Because N-terminal segments of export substrates containing export signals are all natively unfolded in the monomeric form of the proteins28, the energy of ATP hydrolysis would not be required for the process of initial docking and entry to the gate.

Figure 4: Model for flagellar protein export.
figure 4

Top panel: The flagellar protein export apparatus in the basal body. Middle panel: in wild type, the FliH2–FliI complex delivers export substrates to the export gate along with FliJ (left). Once the N-terminal segment of a substrate is docked by formation of the FliHx–FliI6 complex (middle), ATP hydrolysis induces dissociation of the FliHx–FliI6 complex and successive unfolding and translocation of the substrates is driven by the PMF (right). Bottom panel: protein translocation is inhibited by FliH in the fliI mutant (left). The fliH mutant retains the export function to some extent (middle left). The fliH fliI double mutant also retains the export function, albeit at a low level (middle right). Gain-of-function mutations increase the efficiency of substrate entry (right). Pi, inorganic phosphate.

Considering the many similarities between the flagellar export apparatus and F0F1-ATPase, such as almost identical structures of FliI and the α/β subunits6, the PMF-driven functions of the export gate and the F0 portion, and sequence/functional similarities between FliH and the δ/b subunits29, these two remotely related systems may be similar to each other for their entire structural architectures.

Methods Summary

Salmonella strains and plasmids used in this study are listed in Supplementary Table 1. L-broth (LB) and soft tryptone agar plates were prepared as described30. Ampicillin and tetracycline were added as needed at a final concentration of 100 μg ml-1 and 15 μg ml-1, respectively. Motility assays were carried out on soft agar plates as described before4. Export assays were done as described previously30.

The hook–basal bodies were purified and negatively stained with 3% phosphotungstic acid (pH 4.5) on carbon-coated copper grids. Micrographs were recorded at a magnification of ×25,000 with a JEM-1011 transmission electron microscope (JEOL) operated at 100 kV.

The cellular levels of ATP were measured using the ATP bioluminescence assay kit CLS II as described previously22.

Online Methods

Transductional crosses and DNA manipulations

P22-mediated transductional crosses were carried out with p22HTint18. DNA manipulations were carried out as described before4.

Motility assay

Fresh colonies were inoculated on soft tryptone agar plates and incubated at 30 °C.

Export assay

Whole cells and culture supernatant fractions were prepared from Salmonella cells grown at 30 °C in LB, as described previously30.

To test the effect of CCCP on flagellar protein export, the cells were grown with shaking in 5 ml of LB at 30 °C until the cell density had reached an optical density at 600 nm (OD600) of approximately 0.6–0.7. After washing twice with LB, the cells were re-suspended in 5 ml LB with or without CCCP and incubated at 30 °C for 1 h. Cultures were centrifuged to obtain the cell pellets and culture supernatants. Cell pellets were re-suspended in the SDS-loading buffer, and normalized to a cell density to give a constant amount of cells. Proteins in the culture supernatants were precipitated by 10% trichloroacetic acid, suspended in the Tris/SDS loading buffer and heated at 95 °C for 5 min.

Immunoblotting with polyclonal anti-FlgG, anti-FlgD, anti-FlgE, anti-FlgK, anti-FlgL, anti-FliK, anti-FliC, anti-FliH and anti-FliI antibodies was carried out as described before30. Detection was performed with an ECL plus immunoblotting detection kit (GE Healthcare).

Preparation of the hook–basal bodies

The hook–basal bodies were prepared as described previously31 with minor modifications. Samples were negatively stained with 3% phosphotungstic acid (pH 4.5) on carbon-coated copper grids. Micrographs were recorded at a magnification of ×25,000 with a JEM-1011 transmission electron microscope (JEOL) operated at 100 kV.

Construction of a ΔinvC::tetRA strain

An invC null strain (ΔinvC::tetRA) was constructed using the λ Red homologous recombination system, as described before32,33,34.

Measurement of intracellular ATP levels

The cellular levels of ATP were measured using the ATP bioluminescence assay kit CLS II (Roche) as described previously22. Cultures, 1.5 ml each, were centrifuged, and the pellets were re-suspended in 100 mM Tris-HCl pH 7.75, 4 mM EDTA, with adjustment of the OD600 of the cell suspensions to 1.0. The cell suspensions, 300 μl each, were boiled for 2 min at 100 °C. Samples were centrifuged, and 100 μl of each supernatant was transferred to a microtitre plate that was kept on ice until measurement. A 100 μl solution of luciferase reagent was added and then bioluminescence was detected by a LAS-3000 luminescent image analyser (Fujifilm).