In Vitro Reconstitution of Functional Type III Protein Export and Insights into Flagellar Assembly

ABSTRACT The type III secretion system (T3SS) forms the functional core of injectisomes, protein transporters that allow bacteria to deliver virulence factors into their hosts for infection, and flagella, which are critical for many pathogens to reach the site of infection. In spite of intensive genetic and biochemical studies, the T3SS protein export mechanism remains unclear due to the difficulty of accurate measurement of protein export in vivo. Here, we developed an in vitro flagellar T3S protein transport assay system using an inverted cytoplasmic membrane vesicle (IMV) for accurate and controlled measurements of flagellar protein export. We show that the flagellar T3SS in the IMV fully retains export activity. The flagellar hook was constructed inside the lumen of the IMV by adding purified component proteins externally to the IMV solution. We reproduced the hook length control and export specificity switch in the IMV consistent with that seen in the native cell. Previous in vivo analyses showed that flagellar protein export is driven by proton motive force (PMF) and facilitated by ATP hydrolysis by FliI, a T3SS-specific ATPase. Our in vitro assay recapitulated these previous in vivo observations but furthermore clearly demonstrated that even ATP hydrolysis by FliI alone can drive flagellar protein export. Moreover, this assay showed that addition of the FliH2/FliI complex to the assay solution at a concentration similar to that in the cell dramatically enhanced protein export, confirming that the FliH2/FliI complex in the cytoplasm is important for effective protein transport.

T ype III secretion systems (T3SSs) are critical to the function of bacterial injectisomes and flagella. The injectisome is a bacterial protein transporter used to deliver virulence proteins into host cells for infection. The flagellum is a motile organelle critical for infection by many pathogens and is constructed by the T3SS built into the base of the flagellum.
The bacterial flagellum consists of a rotary motor spanning the cell envelope and a filamentous axial structure extending out from the cell surface. The axial structure is a tubular, helical protein assembly composed of more than 20,000 protein subunits of about 10 different proteins. These axial proteins and their assembly scaffold proteins, such as FlgJ, FlgD, and FliD, are translocated via the flagellar T3SS, the flagellar protein export apparatus, across the cytoplasmic membrane into the central channel of the growing flagellum. The flagellar protein export apparatus consists of the transmembrane gate complex formed by FlhA, FlhB, FliP, FliQ, and FliR and the cytoplasmic ATPase complex composed of FliH, FliI, and FliJ (1-3) (Fig. 1A). It was thought that FliO is also one of the transmembrane gate proteins, but a recent study revealed that it only facilitates the assembly of the export gate complex (4). Except for FliO, these proteins share sequence homologies with those of injectisome T3SSs in pathogenic bacteria (5). The gate complex is located in the central pore of the basal body MS-ring, and the ATPase complex is bound just below the gate complex in the cytoplasm through the interactions with the gate and the basal body C-ring (6). FliI is a Walker-type ATPase (7) and is homologous to the ␣ and ␤ subunits of F 1 -ATPase in structure (8). FliI and FliJ together form the FliI 6 /FliJ complex that is similar to the F 1 -␣ 3 ␤ 3 ␥ complex, suggesting that the FliI 6 /FliJ complex hydrolyzes ATP in a way similar to F 1 -ATPase (9). FliH forms a homodimer similar to the peripheral stalk of V-ATPase (10) and binds to the N-terminal domain of FliI in a way similar to F-and V-type ATPases (11,12). While FliH is indispensable for efficient formation of the FliI 6 /FliJ complex at the flagellar base (13), FliH also regulates the ATPase activity of FliI negatively by forming the FliH 2 /FliI complex and suppresses the hexamerization of FliI in solution (11,12). The FliH 2 /FliI complex binds to late export substrates in complex with their cognate chaperones (14)(15)(16) and ensures the interaction between the chaperone-substrate complex and FlhA (17), but the details are unclear. Flagellar protein transport is primarily driven by proton motive force (PMF) (18,19). FlhA has an ion channel activity, and the FlhA-FliJ interaction enables effective utilization of PMF for protein export (20). Since infrequent ATP hydrolysis by FliI ATPase with the E211D substitution is sufficient for processive protein export for flagellar formation, the energy of ATP hydrolysis by FliI is thought to be required primarily for gate activation (18,19,21,22). A study on the virulence T3SS ATPase InvC has suggested that the energy of ATP hydrolysis is used to unfold substrate proteins for export (23). However, it has been shown that the ATPase activity of InvC is not essential for protein unfolding (24). Therefore, it is still controversial how these two types of energy are used in the protein export mechanism.
Despite many genetic and biochemical studies on the T3SSs, the molecular mechanism of protein transport is still obscure due to difficulties in accurate measurements of protein transport under precise control of measurement conditions in vivo. Therefore, an in vitro transport assay system with easy control of measurement conditions is needed for further in-depth mechanistic understanding of protein transport. The inverted membrane vesicle (IMV)-based assay has been used for studying protein translocation across the cell membrane (25), such as the Sec machinery (26) and the twin arginine translocation (TAT) machinery (27), but has never been applied to supramolecular complexes such as the T3SS. We have developed an IMV-based flagellar protein transport assay system that enables accurate measurements of protein transport under well-controlled conditions. The export apparatus in the IMV preserves the protein transport activity. The formation of the flagellar hook about 55 nm in length was reproduced in the IMV, and the export apparatus in IMV retains the protein transport function at a similar level as that in the cell. Surprisingly, even ATP hydrolysis by FliI alone was able to drive flagellar protein export without PMF. We discuss the molecular mechanism of protein transport on the basis of this novel protein transport assay.

RESULTS
Preparation of inverted membrane vesicles. The IMVs were prepared from a Salmonella ΔfliT ΔflgD ΔflhB mutant strain expressing FlhB(N269A) and FlhD/FlhC. We used a fliT null and FlhD/FlhC-overexpressed mutant because deletion of the fliT gene and overexpression of FlhD/FlhC increase the number of the flagellar basal bodies per cell (28). We also introduced a flgD null mutation allele because we used purified FlgD to evaluate the protein transport activity of IMV. FlgD has most frequently been used for our in vivo export analysis of the flagellar T3SS because of the sensitivity and quality of our polyclonal anti-FlgD antibody for measuring the amount of FlgD. FlgD is a scaffolding protein required for the flagellar hook formation. FlgD forms a cap complex at the distal end of the growing hook, and the hook protein FlgE is incorporated into the growing hook just beneath the cap complex (29,30). Therefore, the hook is not formed without FlgD. FlgD and FlgE are rod-hook-type proteins and need no specific chaperones for export. The flagellar export apparatus switches export specificity from rod-hook-type proteins to filament-type proteins upon completion of hook assembly (2); hence, we introduced the FlhB(N269A) mutation to lock the export apparatus in the rod-hook-type protein export mode (31). FlhB undergoes autocleavage between N269 and P270, and the autocleavage is a critical event for the export specificity switch.
In Vitro Reconstitution of Type III Secretion ® Therefore, the FlhB(N269A) mutation completely inhibits the cleavage and locks the export apparatus in the rod-hook-type protein export mode.
The cells were converted to spheroplasts by treatment with EDTA and lysozyme to remove the outer membrane and the peptidoglycan layer, respectively, and were then disrupted by a high-pressure homogenizer to produce inside-out vesicles. The vesicles were purified by sucrose density gradient centrifugation. The vesicles contained FlhA and FlhB (membrane components); FliH, FliI, and FliJ (soluble components); FliF (MSring component); and FliG, FliM, and FliN (C-ring components) as judged by immunoblotting (see Fig. S1A in the supplemental material). To verify the membrane orientation, the vesicles were incubated with proteinase K, followed by immunoblotting with polyclonal anti-FlhA, anti-FliI, and anti-FlgG antibodies. FlhA and FliI were degraded (Fig. 1B, top and middle panels), indicating that they are present outside the membrane vesicles. In contrast, FlgG (a rod protein) was resistant to proteolysis (Fig. 1B, bottom panel), indicating that it is inside the vesicles. FlhA and FliI are located on the cytoplasmic side of the export apparatus, whereas FlgG is located in the periplasm. Thus, we concluded that the orientation of the membrane vesicles is inside-out. The formation of the IMV was further confirmed by electron cryotomography (ECT). We observed the flagellar basal body embedded in the membrane with the C-ring exposed on the outside of the vesicle (Fig. 1C). A disk-like density identified as the cytoplasmic domain of FlhA in the previous ECT analysis of Salmonella and Campylobacter jejuni (6,32) was observed within the C-ring (Fig. 1C), suggesting that the basal body in the IMV retains the membrane components of the export apparatus. No clear spherical density corresponding to the ATPase complex (33) was observed below the disk-like density. However, we often found some density probably corresponding to a partially disrupted ATPase complex during IMV preparation.
IMV preserves the protein transport activity. The protein transport activity of the export apparatus in the IMV was examined by measuring the amount of FlgD transported into the vesicles. After incubation of the IMVs with purified FlgD in a buffer solution containing 20 mM Tris-HCl, pH 7.5, and 125 mM KCl, followed by proteinase K treatment to digest the external proteins, the transported proteins were monitored by immunoblotting with polyclonal anti-FlgD antibody. FliJ was added in the reaction mixture in all experiments except for the experiments below, because it is essential to efficiently couple PMF with protein export (21). FlgD was transported into the IMVs in the presence of ATP, Mg 2ϩ , and FliJ in the external solution ( Fig. 2A). No FlgD transport was observed for the IMVs prepared from the flhB null mutant cells (IMV ΔflhB ), which do not have protein export activity in vivo (18) (Fig. S2A). In agreement with previous in vivo experiments (18,19), protein transport was suppressed by a proton-specific ionophore, carbonyl cyanide m-chlorophenylhydrazone (CCCP) ( Fig. 2A and B). These results indicate that the export apparatus in the IMVs preserves the export activity.
PMF across the cytoplasmic membrane is maintained by proton pumping by the reverse reaction of endogenous F o F 1 -ATP synthase in our assay system, although the initial PMF was produced by using IMVs filled with a solution containing 20 mM morpholinoethanesulfonic acid (MES)-NaOH, pH 6.0, and 300 mM NaCl. Mg 2ϩ and ATP were both indispensable for FlgD transport in the IMV assay, and Mg 2ϩ could not be replaced by Ca 2ϩ (Fig. S2B). No FlgD was transported into the IMV when ADP or AMPPNP was used instead of ATP (Fig. S2B), and the amount of transported FlgD depended on the ATP concentration (Fig. S2C). This Mg 2ϩ -ATP dependence was also detected for the IMVs prepared from the fliH-fliI null mutant cells (IMV ΔfliHI/ϩΔpH/pNa ) (Fig. 2C) or in the IMVs filled with the same solution as the external buffer (IMV fliHIϩ/ϪΔpH/pNa ) to deplete the initial pH and ion concentration gradients (Fig. 2D). Moreover, the FlgD transport was suppressed by the addition of CCCP for IMV ΔfliHI/ϩΔpH/pNa and IMV fliHIϩ/ϪΔpH/pNa ( Fig. 2C and D). These results suggest that endogenous ATPases, such as F o F 1 -ATP synthase, generate PMF using the energy of ATP hydrolysis. Addition of F o F 1 -ATP synthase inhibitors strongly suppressed FlgD transport, supporting the contribution of F o F 1 -ATP synthase to the PMF generation (Fig. 2E). To confirm the contribution of F o F 1 -ATP synthase, we altered the F o F 1 -ATP synthase locus (subunits ␣, ␤, ␥, ␦, , A, B, C, and I) of Salmonella strain STH002 (ΔflhB ΔflgD ΔfliT ΔfliHI) to tetracycline resistance genes (tetRA) and prepared the IMV (IMV ΔFoF1-ΔfliHI ). Deletion of F o F 1 -ATP synthase considerably decreased FlgD transport (see Fig. 5A), suggesting that the protein transport by the IMV is mainly driven by PMF generated by the reverse reaction of F o F 1 -ATP synthase.
The FliH 2 /FliI complex in solution is essential for efficient protein transport. While the FliI 6 ring complex associates with the basal body through the interactions of FliH with FlhA and FliN, FliI shows turnovers between the basal body and the cytoplasmic pool in an ATP-independent manner (34), raising the possibility that the FliH 2 /FliI complex present in the cytoplasm is needed for the protein export. To examine this hypothesis, we added the FliH 2 /FliI complex with FliJ to the external solution and found that the FlgD transport into the IMVs was dramatically enhanced (Fig. 3A, upper panel). The transport enhancement was also observed for the IMVs prepared from the fliH-fliI null mutant cells (IMV ΔfliHI ) (Fig. 3A, lower panel). These results show that the FliH 2 /FliI complex externally added in solution can assemble onto the export gate complex to form a functional export apparatus. Individual addition of either FliJ or the FliH 2 /FliI complex promoted FlgD transport only slightly (Fig. 3A), which is consistent with previous observations that fliH-fliI double null mutant and fliJ null In Vitro Reconstitution of Type III Secretion ® mutant strains both show a weakly motile phenotype (18,19). In contrast, addition of FliI alone did not affect FlgD transport at all (Fig. 3A). These findings indicate that the FliH 2 /FliI complex and FliJ cooperatively act on the protein export mechanism. It is possible that FliH/FliI could fall off during IMV preparation because the FliI density is lost in cryotomograms of lysed cells in the previous studies (33). Then, we preincubated IMVs with the FliH 2 /FliI complex and conducted the transport assay. Preincubation alone showed only a slight enhancement of the FlgD transport (Fig. S3). The effect of the FliH 2 /FliI complex in solution containing purified FliJ was further examined by changing the concentration of the FliH 2 /FliI complex. The transport assay was performed in solution containing FliI and the FliH 2 /FliI complex at various molar ratios with the total amount of FliI kept constant. The concentration of FliJ was adjusted to 1/6 of the total FliI concentration because a small amount of FliJ facilitates the ATPase complex formation but an excess amount of FliJ inhibits FliI hexamer formation (9). The level of FlgD transport was increased considerably with an increase in the amount of the FliH 2 /FliI complex (Fig. 3B). Therefore, we conclude that the FliH 2 /FliI complex in solution is essential for efficient protein export.
FlgD facilitates FlgE export. We next examined the transport of another substrate, FlgE, to investigate hook formation inside the vesicle. We added purified FlgE with or without FlgD in the external solution and performed the in vitro protein transport assay. In the presence of FlgD, the transported FlgE monomers assembled into the hook structure on the basal body, resembling those seen in unperturbed cells ( Fig. 4A and Fig. S4), indicating that hundreds of FlgE molecules were processively transported into the IMV via a single export apparatus to form a bona fide hook. Interestingly, the amount of transported FlgE was increased by about two times in the presence of FlgD, indicating that FlgD facilitates FlgE export into the IMVs in addition to chaperoning assembly (Fig. 4B).
In vitro hook length control. The length of the wild-type hook is regulated at about 55 (Ϯ6) nm by FliK, a molecular ruler that measures the hook length. After reaching its mature length, FliK interacts with FlhB to switch the export specificity from rod-hook- type to filament-type proteins (2). We tried to reproduce the hook length control and the export specificity switch in the IMV prepared from a Salmonella ΔfliT ΔflgD ΔflhB mutant strain expressing wild-type FlhB and FlhD/FlhC by adding purified FliK at various concentrations into the reaction mixture. After a 1-h reaction, we purified the hook-basal bodies from the IMV and measured the hook length by negative-staining electron microscopy (EM). Polyhooks with various lengths were produced without FliK, but with the increase of FliK concentration, the average length became shorter and the length distribution became narrower (Fig. S5). In the presence of 4 M FliK, the average length was successfully controlled at about 55 (Ϯ22) nm, suggesting that no soluble component other than FliK is necessary for the hook length control and substrate specificity switch. The standard deviation (SD) of the hook length was larger than that of the wild-type hook but was similar to that for the cells overproducing FlgE (35). The FlgE concentration used for this assay (4 M) corresponds to about 4,000 molecules/ cell, and this is presumably much higher than the cellular level. The large SD of the hook length may be due to this high concentration of FlgE.
ATP hydrolysis by FliI can drive the protein export without PMF. As described above, the protein transport by the IMV is mainly driven by PMF generated by the reverse reaction of F o F 1 -ATP synthase. However, FlgD transport was still retained to some extent without F o F 1 -ATP synthase, even in the absence of the pH and ion concentration gradients, when the FliH 2 /FliI complex, FliJ, and Mg 2ϩ -ATP were present in the external solution (Fig. 5A), suggesting that protein export can be driven solely by In the presence of FlgD at a sufficiently high concentration, most of the FlgE transported into the IMV was detected in the precipitate, indicating that the transported FlgE formed the hook structure.
In Vitro Reconstitution of Type III Secretion ® ATP hydrolysis by the FliI ATPase. To confirm protein transport without PMF, we added CCCP in the reaction mixture containing FliJ, the FliH 2 /FliI complex, and Mg 2ϩ -ATP. The FlgD export activity of the IMV fliHIϩ was reduced in the presence of CCCP but was still significantly retained ( Fig. 5B and C). To verify protein transport without PMF, we performed a transport assay using IMV ΔFoF1-ΔfliHI without pH and ion concentration gradients (Fig. 5A, lower panel, and D). The FlgD export activity of IMV ΔFoF1-ΔfliHI was not  affected at all by CCCP up to 40 M in the presence of FliJ, the FliH 2 /FliI complex, and Mg 2ϩ -ATP (Fig. 5D). We further tested hook formation without PMF. IMV ΔFoF1-ΔfliHI mixed with FlgD, FlgE, FliJ, and the FliH 2 /FliI complex was incubated with or without Mg 2ϩ -ATP for 1 h. Then, basal bodies were purified from the IMV and observed by negative-staining EM. Polyhooks were formed in the presence of 5 mM ATP and even with 10 M CCCP but not without ATP (Fig. S6). These results indicate that processive transport of FlgE occurs even in the absence of PMF when FliJ, the FliH 2 /FliI complex, and Mg 2ϩ -ATP are present. To examine if ATP hydrolysis by FliI is responsible for the protein export without PMF, we added the FliH 2 /FliI(E211Q) complex to the reaction mixture instead of the FliH 2 /FliI complex. FliI(E211Q) can bind Mg 2ϩ -ATP to form a hexameric ring and associate with the export gate but has no ATP hydrolyzing activity (36). Unlike with wild-type FliI, FlgD was not transported into the IMVs (Fig. 5E). Therefore, we conclude that flagellar protein export can be driven by ATP hydrolysis by FliI in the absence of bulk PMF.

DISCUSSION
The flagellar export apparatus transports 20,000 to 30,000 protein subunits of 14 different proteins. Therefore, the coordination of protein export is important to construct the flagellum. The substrate specificity of the export apparatus is switched from rod-hook-type proteins to filament-type proteins after completion of the hook, and the substrate specificity switch is coupled to the gene expression of each type of protein (2). The export of filament-type proteins is coordinated by the flagellar type III export chaperones. These chaperones bind to their specific cognate substrates (37)(38)(39)(40). The binding affinity of the chaperone/substrate complexes for FlhA, an export gate protein, is thought to regulate the secretion order of the filament-type proteins (41,42). Rod-hook-type substrates bind to a well-conserved hydrophobic pocket of the C-terminal cytoplasmic domain of FlhB (43). Recent genetic analyses have suggested that FliH, FliI, FlhA, and FlhB coordinate hook-type protein export with hook assembly to regulate the hook length at 55 nm in Salmonella (44). However, the coordination of rod-hook-type protein export was not well known, and no such export chaperones were found for rod-hook-type proteins. We observed that export of FlgE is facilitated by FlgD. FlgD and FlgE are both rod-hook-type export substrates and are supposed to compete for their transport. Thus, FlgE transport should be decreased by an addition of FlgD as seen in vivo (45). However, to the contrary, FlgE transport was increased by about two times in the presence of an equal amount of FlgD. This result suggests the presence of a cooperative mechanism between rod-hook-type proteins for efficient and ordered protein export. The enhancement of FlgE export by FlgD is reasonable because the FlgD cap formation is needed prior to the hook formation (29).
The FlgD transport was 20 times enhanced in the presence of 1.5 M FliH 2 /FliI in the reaction mixture. The number of FliI molecules in a cell was estimated to be 1,500 (46). If we assume that the bacterial cell is a rod-shaped cylinder with a length of 1.5 m and a diameter of 1 m capped by hemispherical ends with a diameter of 1 m, the bacterial cell volume is 1.7 fl, and the concentration of FliI in a cell is estimated to be 1.5 M. This value is comparable to our experimental condition, implying that a certain amount of the FliH 2 /FliI complex in the cytoplasm is needed for efficient protein export. Then, what is the role of the FliH 2 /FliI complex in solution? In situ electron cryotomography revealed that the FliI hexamer complex is associated with the basal-body C-ring through FliH (6,10,34), but the FliI hexamer complex has never been observed in the purified basal body in vitro, or lysed cells for electron cryotomography (33), indicating that an interaction between the FliH/FliI complex and the basal body is not so strong. Moreover, turnover of FliI between the basal body and the cytoplasmic pool has been demonstrated by fluorescence recovery after photobleaching (FRAP) experiments (34). Thus, a certain amount of the FliH 2 /FliI complex in the cytoplasm may be required for stable association and turnover of the FliI hexamer to the basal body.
Previous in vivo studies have revealed that flagellar type III protein export is driven by proton motive force (PMF) and facilitated by ATP hydrolysis by FliI ATPase (18,19).
In agreement with the in vivo experiments, protein transport of IMV ΔfliHI was observed and inhibited by CCCP. Because the diameter of the central channel of the rod and hook for the substrate passage is only 1.3 nm (47), the substrate proteins must be unfolded during their translocation through the channel. We used purified FlgD and FlgE for a transport assay, and IMV ΔfliHI transported these substrate without any help from other proteins except for FliJ ( Fig. 2C and 3A). Size exclusion chromatography, which is the final purification step, showed a symmetrical elution peak around the expected molecular weight for each protein. The purified FlgD has been crystallized, and the structure will be published elsewhere. These results indicate that the purified FlgD and FlgE are both well folded in solution. Therefore, our results suggest that the transmembrane gate complex itself has the unfolding activity of substrate proteins. On the other hand, our IMV assay indicates that ATP hydrolysis by FliI can drive the protein export without bulk PMF. Protein transport without PMF was not detected in the previous in vivo experiment. In our in vitro experiments, the concentration of the substrate was higher than in the in vivo experiments and was kept constant. Moreover, the concentrations of the FliH 2 /FliI complex and FliJ were also kept constant. These conditions may enable us to detect the protein transport without PMF. It is still unclear how the cytoplasmic ATPase complex unfolds the export substrate and opens the transmembrane gate. Since the entire structure of the ATPase complex is very similar to that of F o F 1 -ATP synthase (8)(9)(10), and ATP hydrolysis is closely linked to efficient proton translocation (48), one possibility is that local PMF is generated by the ATPase complex, and the transmembrane gate complex exports the substrates using this local PMF. In vivo local pH measurement using pHluorin(M153R) revealed that the ATPase activity of FliI reduced local pH near the export apparatus (48). Moreover, ΔpH is not required for protein export in the presence of FliH and FliI (21). These previous results support this idea, but further investigation is needed to understand the role of ATP hydrolysis by FliI. The Sec protein translocation system requires both PMF and ATP (26), while the TAT system needs only PMF (27). Our study revealed that, unlike these two systems, both PMF and ATP hydrolysis can drive the flagellar type III protein export, although they are both needed for the maximum secretion activity. This energetic redundancy may contribute to the robustness of the flagellar protein export apparatus that is essential for bacterial motility.
Development of an in vitro transport assay system with easy control of measurement conditions is important for mechanistic understanding of protein transport. The IMVbased assay has been used for studying various transporters and greatly contributed to the elucidation of the molecular mechanism of substrate transport across the cell membrane (25)(26)(27). However, the IMV-based technique has been used only for the transporter consisting of a few to a dozen protein subunits. We applied this technique to the flagellar export machinery, which is composed of more than 250 protein subunits of 12 different types of proteins (FliF, FliG, FliH, FliI, FliJ, FliM, FliN, FliP, FliQ, FliR, FlhA, and FlhB), including the housing of the export apparatus. The export apparatus translocated purified substrate proteins from the external solution into the lumen of the IMV. In addition, the hook length control and the export specificity switching were reproduced in the IMV. These results suggest that the export apparatus in the IMV retains its native function as in the cell. The IMV-based in vitro assay described here may also be applied to other large membrane complex systems that are difficult to purify or isolate from the cell membrane in the fully functional form. The in vitro system of course cannot fully reproduce the cellular condition, including molecular crowding and complex interactions between various molecules, and therefore the measurement results with the IMV-based system do not directly reflect the cellular events. However, even with such limitations, this IMV-based method will be a useful tool for studying the functional mechanisms and structures of such large complex systems.

MATERIALS AND METHODS
Bacterial strains and plasmids. Bacterial strains and plasmids are listed in Table S1  Preparation of the cells. The cells used for IMV preparation were constructed from a Salmonella mutant strain with deletion of flhB, flgD, and fliT. The cells were transformed with a pBAD33-based plasmid harboring flhB(N269A) and flhDC, into the mutant cell. To increase the number of flagellar basal bodies, we deleted fliT, the negative regulator of the flagellar class 2 gene, and expressed flhDC, the master regulator of the flagellar genes. The flgD gene was also deleted, because we selected FlgD as a standard substrate for the transport assay. To prevent the substrate specificity switching from rod-hooktype to filament-type proteins, flhB was deleted and the flhB N269A mutant gene was introduced. The IMVs used for the hook length control and export switch experiment were prepared from the cells harboring wild-type flhB instead of flhB N269A.
Preparation of inverted membrane vesicles. Overnight cell culture was inoculated into 1 liter of fresh LB broth in a 5-liter flask with 1/100 dilution and cultured at 30°C for 1 h. L-Arabinose was added to the final concentration of 0.02% (wt/vol), and the culture was continued at 18°C for 12 to 16 h (until optical density at 600 nm [OD 600 ] reached around 1.5). The cells were collected, suspended into 75 ml of sucrose solution (10 mM Tris-HCl, pH 8.0, 0.75 M sucrose) and stirred with 22.5 mg of lysozyme on ice. Then, 150 ml of 1.5 mM EDTA was poured onto the suspension on ice, and stirring was continued for 1 h at 4°C. The cells were collected at 5,000 ϫ g and suspended into 25 ml solution A (20 mM MES-NaOH, pH 6.0, 300 mM NaCl) or solution B (20 mM Tris-HCl, pH 7.5, 125 mM KCl) with 1 tablet of protease inhibitor Complete EDTA-free (Roche). The suspension was passed through a high-pressure cell homogenizer (Stansted) at 90 MPa to produce inverted-membrane vesicles. After removal of debris by centrifugation at 20,000 ϫ g for 10 min, IMVs were precipitated by ultracentrifugation at 100,000 ϫ g for 1 h. IMVs were suspended into 1 ml of solution A or B and purified by sucrose density gradient centrifugation (60% [wt/wt] 5 ml, 50% [wt/wt] 9 ml, 45% [wt/wt] 9 ml, 40% [wt/wt] 6 ml stepwise gradient in a Beckman ultraclear tube) at 60,000 ϫ g (SW32 Ti rotor; Beckman) for 16 h. A brown layer, which is the fraction containing IMVs, was recovered and precipitated by ultracentrifugation at 100,000 ϫ g for 1 h. The pellet was suspended into 900 l of solution A or B, divided into 300-l aliquots, frozen by liquid nitrogen, and stored at Ϫ80°C until use.
Protein purification. Cells expressing FlgD, FlgE, or the FliH 2 /FliI complex were suspended in solution C (50 mM Tris-HCl, pH 8.0, 500 mM NaCl) and disrupted by sonication. After removal of the cell debris by centrifugation, the supernatant was loaded onto a HisTrap HP (GE Healthcare) or nickelnitrilotriacetic acid (Ni-NTA) agarose column (Qiagen), and proteins were eluted with imidazole solution. To remove the His tag, FlgD and FlgE solutions were incubated with thrombin (GE Healthcare) at room temperature for 3 h and then passed through the HiTrap benzamidine FF column (GE Healthcare) to remove thrombin. The FliH 2 /FliI complex was not treated with thrombin; therefore, the His tag of the complex was retained. Finally, proteins were purified using a Superdex 200 column (GE Healthcare) equilibrated with solution B. Cells expressing FliK were suspended in solution D (50 mM Tris-HCl, pH 8.0, 150 mM NaCl) and disrupted by sonication. After removal of the cell debris by centrifugation, supernatant was loaded onto a HisTrap HP (GE Healthcare) column, and proteins were eluted with imidazole solution. To remove the His tag, FliK solution was incubated with thrombin (GE Healthcare) at room temperature for 2 h and then dialyzed with solution A. The FliK solution was passed through the HisTrap HP (GE Healthcare) column to remove His-tag-retaining FliK and then passed through a HiTrap benzamidine FF column (GE Healthcare) to remove thrombin. Finally, proteins were purified using a Superdex 200 column (GE Healthcare) equilibrated with solution B. FliI and FliJ were purified as previously reported in the crystallographic reports (8,9). Each protein and protein complex eluted as a symmetrical peak corresponding to the expected molecular weight for the products. The purity of the purified proteins was examined by SDS-PAGE.
Transport assay. Three hundred microliters of the frozen stock solution of IMVs was thawed at room temperature and homogenized 10 times through an 0.8-m polycarbonate membrane with the Avanti Mini-Extruder (Avanti Polar Lipids). The IMV solution was loaded on the Sephadex G50 fine column (GE Healthcare) equilibrated with solution B, and the IMV solution was eluted with 1.5 ml of solution B. The transport assay solution was prepared by mixing 100 l of the IMV solution with 375 l of external buffer containing Tris-HCl, pH 7.5, KCl, MgCl 2 , dithiothreitol (DTT), and purified substrate proteins, and protein export was initiated by adding 25 l of 0.1 M ATP solution. The ATP solution was prepared by dissolving ATP (dipotassium salt) in Tris-HCl solution (final concentration of 20 mM) followed by neutralization with KOH. The final reaction mixture contained 20 mM Tris-HCl, pH 7.5, 115 mM or 133 mM KCl, 5 mM MgCl 2 , 1 mM DTT, and 5 mM ATP. The concentration of each chemical component in the external buffer was adjusted to match these values. After incubation at 37°C for 1 h, proteinase K was added at a final concentration of 10 g/ml to stop the transport reaction, and the reaction mixture was further incubated for 30 min to degrade nontransported substrate molecules. The assay mixture was ultracentrifuged at 100,000 ϫ g for 30 min, and the precipitant containing IMV was washed by 1 ml of solution B. After removal of the washing buffer, the precipitant was incubated with 1 ml of solution B containing 10 g/ml proteinase K for 10 min at room temperature, followed by washing again with 1 ml of solution B, and then IMVs were dissolved with 45 l of 1% (vol/vol) Triton X-100. The transported proteins were precipitated by trichloroacetate and were detected by immunoblotting analyses with polyclonal anti-FlgD or anti-FlgE antibody.
Purification of the hook-basal body from IMV. The hook-basal body complexes were purified according to the method shown by Aizawa et al. with minor modification (49). After the transport reaction, a part of the IMV solution was treated with 0.1% (vol/vol) Triton X-100 and ultracentrifuged (150,000 ϫ g, 30 min). The precipitates containing the hook-basal body were suspended in TET solution In Vitro Reconstitution of Type III Secretion ® (10 mM Tris-HCl, pH 8.0, 1 mM EDTA, 0.1% [vol/vol] Triton X-100), and the alkali solution (10% [wt/vol] sucrose, 0.1% [vol/vol] Triton X-100, 0.1 M KCl; pH was adjusted to 11 by KOH) was added to the suspension to completely dissolve the membrane. The suspension was layered on an equal volume of the 35% (wt/vol) sucrose solution prepared by dissolving sucrose in the TET solution in an ultracentrifuge tube, incubated on ice for 30 min, and ultracentrifuged at 38,000 rpm (Beckman TLA100.3 rotor) for 30 min to precipitate the hook-basal body. The precipitates were suspended in the TET solution and then observed by electron microscopy.
Electron cryotomography. R0.6/1.0 Quantifoil grids (Quantifoil Micro Tools, Jena, Germany) were glow discharged and pretreated with a solution of 10-nm colloidal gold particles (MP Biomedicals, USA). A 2.6-l solution of purified IMV was applied to the grid. The grids were blotted briefly with filter paper and then were rapidly plunged into liquid ethane using Vitrobot Mark II (FEI) for freezing. Electron microscopy images were collected at the liquid nitrogen temperature using a Titan Krios electron microscope (FEI) equipped with a field emission gun and a Falcon II direct electron detector (FEI). The microscope was operated at 300 kV and a nominal magnification of ϫ37,000 with a calibrated pixel size of 4.46 Å after 2 ϫ 2 binning. Images of single-axis tilt series were collected covering an angular range from Ϫ70°to ϩ70°with a nonlinear Saxton tilt scheme at 4-to 7-m underfocus using the Xplore three-dimensional (3D) software package (FEI) and a cumulative dose of~120 e Ϫ /Å 2 . The IMOD package (50) was used to align tilted projection images and to generate the final 3D density map from the aligned image stack. The final 3D density map was obtained by the simultaneous iterative reconstruction technique (SIRT).
Negative-staining electron microscopy. Sample solutions were applied to carbon-coated copper grids and negatively stained with 2.0% (wt/vol) phosphotungstic acid or 2.0% (wt/vol) uranyl acetate. The samples for the hook length measurement were stained at 4°C to make the hooks straight. Images were observed with a JEM-1011 transmission electron microscope (JEOL, Tokyo, Japan) operating at 100 kV using a TVIPS TemCam-F114 charge-coupled device (CCD) camera or a TemCam-F415 CCD camera.