The transmembrane export complex of the fT3SS is located inside the MS-ring formed by the transmembrane protein FliF (Figure 1; Johnson et al., 2021; Kawamoto et al., 2021; Takekawa et al., 2021; Tan et al., 2021). It consists of nine copies of FlhA, a single copy of FlhB, five copies of FliP, four copies of FliQ, and a single copy of FliR (Abrusci et al., 2013; Kuhlen et al., 2018, 2020; Johnson et al., 2019).

FliP and FliR assemble into the FliP₅-FliR₁ complex with the help of the FliO scaffolding protein and form the polypeptide channel for the translocation of export substrates across the cytoplasmic membrane (Figure 3, left panel) (Fabiani et al., 2017; Fukumura et al., 2017). Four FliQ subunits bind to the outside of the FliP₅-FliR₁ complex to form the FliP₅-FliQ₄-FliR₁ complex (Figure 3, middle panel). A flexible loop formed by the highly conserved Met-209, Met-210, and Met-211 residues of FliP (the M-loop) on the cytoplasmic side of the polypeptide channel and a plug loop composed of residues 106–122 of FliR (the R-plug) seem to prevent the leakage of small molecules during high-speed protein translocation (Figure 3, right panel) (Ward et al., 2018; Hüsing et al., 2021). The FliP₅-FliQ₄-FliR₁ complex has a helical arrangement of subunits similar to the rod (Figure 3), so FliE, which is the first export substrate transported by the fT3SS (Minamino and Macnab, 1999; Minamino et al., 2000), can directly assemble at the distal end of the FliP₅-FliQ₄-FliR₁ complex to form the most proximal part of the rod. Interactions between FliE and FliF not only firmly connect the rod with the MS ring but also open the exit gate of the polypeptide channel through conformational changes of FliP and FliR (Hendriksen et al., 2021).

Salmonella FlhB consists of an N-terminal transmembrane domain (FlhB_TM) with four transmembrane helices (TMHs) (residues 1–211) and a large C-terminal cytoplasmic domain (FlhB_C) (residues 212–383) (Minamino et al., 1994; Kinoshita et al., 2021). FlhB_TM associates with the FliP₅-FliQ₄-FliR₁ complex to form the FliP₅-FliQ₄-FliR₁-FlhB₁ complex (Figure 3, middle panel), and the cytoplasmic loop connecting TMH-2 and TMH-3 (FlhB_Loop) wraps around the entrance gate of the FliP₅-FliQ₄-FliR₁ complex through interactions of the loop with each FliQ subunit (Figure 3, right panel). It is thus plausible that FlhB may coordinate gate opening for substrate entry into the polypeptide channel. Recent genetic analysis has suggested that the N-terminal cytoplasmic tail of FlhB and FlhB_C are involved, along with the cytoplasmic ATPase complex, in the gating function of FlhB (Kinoshita et al., 2021).

Salmonella FlhA is divided into two distinct regions: an N-terminal transmembrane region (FlhA_TM) with eight TMHs (residues 1–327) and a large C-terminal cytoplasmic region (residues 328–692) (Figure 4A; Minamino et al., 1994; Kinoshita et al., 2021). The crystal structure of the C-terminal cytoplasmic region is composed of a compactly folded domain (FlhA_C, residues 362–692) and a flexible linker (FlhA_L, residues 328–361) connecting FlhA_C with FlhA_TM (Figure 4A; Saijo-Hamano et al., 2010). FlhA assembles into a homo-nonamer through intermolecular interactions between FlhA_C subunits, and the interactions of FlhA_L with its neighboring FlhA_C subunit stabilize the FlhA_C-ring structure (Figure 4B; Terahara et al., 2018; Kuhlen et al., 2021). FlhA_TM associates not only with the FliP₅-FliQ₄-FliR₁ complex but also with the MS-ring (Kihara et al., 2001).

If either MS-ring or the FliP₅FliQ₄FliR₁ complex is missing in Salmonella cells, FlhA cannot efficiently form the oligomer at the flagellar base as monitored with FlhA labeled with yellow fluorescent protein (YFP), suggesting that FlhA assembles into the export complex along with other export-gate proteins during MS-ring formation (Morimoto et al., 2014). The highly conserved Arg-94, Lys-203, Asp-208, and Asp-249 residues of FlhA_TM are critical for H⁺-coupled protein export (Hara et al., 2011; Erhardt et al., 2017). Over-expression of FlhA in Escherichia coli decreases the intracellular pH. Furthermore, over-expression of FlhA increases intracellular Na⁺ concentration in the presence of 100 mM NaCl. These observations suggest that FlhA forms a pathway for the transit of both H⁺ and Na⁺ across the cytoplasmic membrane. The flhA(D208A) mutation facilitates the H⁺-channel activity of FlhA, suggesting that Asp-208 of FlhA may coordinate H⁺ flow though the FlhA channel with protein export. However, this mutation does not affect the Na⁺-channel activity of FlhA at all, suggesting that the Na⁺ pathway is different from the H⁺ pathway (Minamino et al., 2016b).

FlhA_C and FlhB_C project into the cytoplasmic cavity of the basal body C-ring and form a docking platform for the cytoplasmic ATPase complex, flagellar export chaperones, and export substrates (Minamino and Macnab, 2000c; Minamino et al., 2003, 2010, 2012a; Bange et al., 2010). The FlhA_C-FlhB_C docking platform determines the order of substrate export to facilitate efficient flagellar assembly and also regulates gate opening of the FlhA ion channel and the FliP₅-FliQ₄-FliR polypeptide channel (Minamino and Macnab, 2000a; Kinoshita et al., 2013; Inoue et al., 2019; Minamino et al., 2020a,2021b).

A highly conserved hydrophobic dimple including Phe-459, Asp-456, and Thr-490 of FlhA is critical for substrate recognition by the fT3SS during flagellar assembly (Figure 4A; Xing et al., 2018). The C-terminal part of FlhA_L binds to its neighboring FlhA_C subunit to stabilize the open conformation of FlhA_C in the nonameric ring, allowing flagellar export chaperones associated with their cognate substrates to bind to the conserved hydrophobic dimple with a nanomolar affinity (Inoue et al., 2021).

The F₁ ATPase is composed of three copies of the α subunit, three copies of the β subunit, a single copy of the γ subunit and a single copy of the ε subunit (Figure 1). The α and β subunits form a hetero-hexameric α₃β₃ ring, and the γ subunit binds within the central pore of the α₃β₃ ring (Abrahams et al., 1994). The ε subunit binds to the γ subunit to control the ATP hydrolysis activity of the F₁ ATPase in an ATP-dependent manner (Kato-Yamada et al., 2000). The α₃β₃γ₁ subcomplex is the minimum unit that can function as an ATP-driven rotary motor to couple ATP hydrolysis with the rotation of the γ subunit within the α₃β₃ ring. ATP binds to three catalytic sites in the α₃β₃ ring, each of which is located at an interface between the α and β subunits. Three catalytic β subunits in the α₃β₃ ring undergo highly cooperative and sequential conformational changes in their C-terminal domains during ATP hydrolysis. These conformational changes drive the γ subunit to rotate within the α₃β₃ ring (Watanabe and Noji, 2013). The FliI₆-FliJ₁ subcomplex of the fT3SS, which looks similar to the α₃β₃γ₁ subcomplex, can act as the ATPase at the base of the flagellum (Figure 1; Ibuki et al., 2011).

FliI is the flagellum-specific ATPase. It has highly conserved Walker A and B motifs (Vogler et al., 1991; Fan and Macnab, 1996). Salmonella FliI consists of three domains: N-terminal (residues 2–97, FliI_N), ATPase (residues 109–380, FliI_CAT) and C-terminal (residues 381–456, FliI_C) (Figure 5A; Imada et al., 2007). Residues 98–105, most of which are invisible in the electron density map, form a flexible hinge connecting FliI_N and FliI_CAT, and this flexible hinge loop undergoes conformational changes during ATP binding and hydrolysis (Minamino et al., 2001).

The structures of FliI and its fT3SS homolog SctN are remarkably similar to the α and β subunits of the F₁ ATPase (Zarivach et al., 2007). However, in contrast to the F₁ ATPase, FliI and SctN form homo-hexamers in an ATP-dependent manner (Figure 5B, C), and both hexamers themselves can hydrolyze ATP at the interface between FliI/SctN subunits (Claret et al., 2003; Kazetani et al., 2009). Thus, the ATPase ring complex of the T3SS has six catalytic sites. The FliI₆ and SctN₆ ring structures have been identified at the base of the flagellum and injectisome, respectively, by electron cryotomography and sub-tomogram averaging (Chen et al., 2011; Kawamoto et al., 2013).

Intermolecular interactions between FliI_N domains are required for FliI ring formation (Figure 5C; Okabe et al., 2009). The core structure of FliI_N can be superimposed onto the N-terminal domains of the α and β subunits of the F₁ ATPase within α₃β₃ hetero-hexamer. In the FliI₆-ring model, which was generated by fitting the crystal structure of FliI into the structures of the α and β subunits, FliI_N shows steric hindrance at the subunit interfaces, suggesting that a conformational change in FliI_N is required for FliI ring formation. Deletion of residues 2–7 of FliI_N suppresses FliI hexamerization and decreases the ATPase activity of FliI (Minamino et al., 2006), suggesting that the extreme N-terminal region of FliI regulates FliI oligomerization. Recently, it has been reported that Arg-33, Asn-73, and Arg-76 are also responsible for well-regulated FliI ring formation (Figure 5C; Kinoshita et al., 2021).

Amino acid residues in the F₁ ATPase that are known to be involved in ATP hydrolysis are highly conserved in the FliI/SctN family. FliI_CAT contains the highly conserved P-loop (residues 182–188), the catalytic glutamate (Glu-211), and the arginine finger (Arg-374) (Figure 5A; Walker, 2013). ADP binds to the P-loop of FliI, as it does in the F₁ ATPase. The carboxyl group of Glu-190 in the β subunit of the thermophilic Bacillus F₁ ATPase, which corresponds to Glu-211 of FliI, polarizes a water molecule for the nucleophilic attack on the γ-phosphate of ATP, and the G190Q substitution results in a complete loss of ATPase activity (Shimabukuro et al., 2003). The fliI(E211Q) mutation completely abolishes ATPase activity but does not affect the binding of ATP to the P-loop, and FliI with the E211Q substitution can form the hexamer ring in the presence of Mg²⁺-ATP. Thus, Glu-211 of FliI_CAT acts as the catalytic glutamate.

Arg-373 in the α subunit of the F₁ ATPase, which corresponds to Arg-374 of FliI, functions as the arginine finger that protrudes into the nucleotide-binding site of the adjacent β-subunit. The side chain of this arginine residue forms a positively charged binding pocket for the negative charge of the γ-phosphate of ATP (Rees et al., 2012). The fliI(R374A) mutation inhibits FliI ring formation significantly and decreases ATPase activity. This effect indicates that Arg-374 of FliI stabilizes the binding of ATP to the P-loop in a way similar to Arg-373 of the α subunit. These observations suggest that FliI and the F₁ ATPase share a similar catalytic pathway for ATP hydrolysis.

The binding of ADP to the P-loop induces a conformational change in FliI_C relative to FliI_CAT, suggesting that the FliI hexamer may undergo conformational changes in its C-terminal domains that are coupled with the catalytic reaction cycle in the same way as in the F₁ ATPase. This idea is supported by the asymmetric cryoEM structure of the SctN₆-SctO₁ ring complex with a non-hydrolyzable ATP analog (Majewski et al., 2019).

FliJ and its vT3SS homolog SctO adopt an antiparallel coiled-coil structure that is similar to the two-stranded α-helical coiled-coil part of the γ subunit of the F₁ ATPase (Figure 5D; Ibuki et al., 2011). FliJ binds to the C-terminal region of the first α-helix of FliI_C (residues 382–406 of Salmonella FliI), which corresponds to the region of the β subunit that is responsible for interaction with the γ subunit. This interaction facilitates FliI ring formation and increases the ATPase activity of FliI. FliJ penetrates the central cavity of the FliI₆ ring like the γ subunit in the F₁ ATPase (Figure 5C). These observations have been confirmed by the cryoEM structure of the SctN₆-SctO₁ ring complex. FliJ has been shown to exert a rotor-like function in both rotary F₁ and V₁ ATPases (Kishikawa et al., 2013; Baba et al., 2016). Thus, the FliI₆-FliJ₁ ring complex may function as an ATP-driven rotary motor that couples ATP hydrolysis with the rotation of FliJ within the FliI hexamer.

The b and δ subunits of the F_OF₁ ATP synthase form the peripheral stalk that connects the α₃β₃γ₁ε₁ ring complex to the membrane-embedded F_O unit (Figure 1). The extreme N-terminal region of the b subunit binds to F_O, whereas the δ subunit interacts with the extreme N-terminal region of the α subunit of F₁ (Walker and Dickson, 2006). The N-terminal and C-terminal regions of FliH and its vT3SS homolog SctL are homologous to the b and δ subunits of the ATP synthase (Pallen et al., 2006). This is confirmed by the crystal structure of an N-terminally truncated variant of Salmonella FliH consisting of residues 99–235 in complex with FliI (Figure 6; Imada et al., 2016).

Salmonella FliH consists of 235 amino-acid residues and forms a homo-dimer through residues 101–140, which form a coiled-coil structure (Minamino and Macnab, 2000b; González-Pedrajo et al., 2002). The FliH dimer binds to each FliI_N domain of the FliI₆ ring (Figure 6, left panel) and also to the FliN protein in the C-ring (González-Pedrajo et al., 2006; McMurry et al., 2006; Paul et al., 2006). The interactions of FliH with FliN and FliI_N are required for efficient and robust association of the FliI₆-FliJ ring complex with the flagellar basal body (Figure 1; Minamino et al., 2009). The N-terminal domain of FliH (residues 1–140, FliH_N) adopts a quite elongated α-helical coiled coil structure similar to that of the b subunit of the ATP synthase, and the extreme N-terminal region of FliH is involved in the interaction with FliN (Hara et al., 2012). Both C-terminal domains (residues 141–235, FliH_C) in the FliH dimer are involved in the interaction with FliI (Minamino et al., 2002). These two FliH_C domains have completely different conformations; one binds to the extreme N-terminal α-helix of FliI consisting of residues 2–21, and the other binds to a positively charged cluster consisting of Arg-26, Arg-27, Arg-30, Arg-33, Arg-76, and Arg-93 of FliI_N (Figure 6, middle and right panels). Because FliI cannot localize to the flagellar base in the absence of FliH, FliH seems to act as a peripheral stalk to firmly anchor the FliI₆-FliJ₁ ring complex to the C-ring.

The PMF consists of the electric potential difference (Δψ) and the proton concentration difference (ΔpH) across the cytoplasmic membrane. When the cytoplasmic ATPase ring complex works properly for flagellar assembly, the transmembrane export gate complex uses the Δψ component to drive H⁺-coupled protein export under a variety of environmental conditions (Paul et al., 2008; Minamino et al., 2011, 2021b). However, when the ATPase ring complex becomes non-functional under certain physiological conditions, the export gate complex prefers to use the SMF over a wide range of external pH, indicating that the transmembrane export complex is intrinsically a dual-fuel export engine that can use either H⁺ or Na⁺ as the coupling ion (Minamino et al., 2016b,2021a). This in turn suggests that the cytoplasmic ATPase ring complex switches the export gate complex from the dual-fuel engine mode to a highly efficient Δψ-driven one.

FliJ binds to FlhA_L with high affinity to activate the H⁺ channel of FlhA_TM and to unlock the entrance gate of the polypeptide channel. As a result, the export gate complex becomes an active H⁺/protein antiporter that couples inward-directed H⁺ flow through the FlhA ion channel with outward-directed protein translocation across the polypeptide channel (Minamino et al., 2011). An inactive export gate complex can also be activated by an increase in Δψ above a certain threshold through an interaction between FliJ and FlhA_L, suggesting that Δψ is required for efficient and stable interaction between FliJ and FlhA_L (Minamino et al., 2021b).

A helix-loop-helix formed by Gln-38, Leu-42, Tyr-45, Tyr-49, Phe-72, Leu-76, Ala-79, and His-83 of FliJ, which are highly conserved residues in FliJ homologs, extends out of the FliI₆-ring (Figure 5D; Ibuki et al., 2013). This is confirmed by the cryoEM structure of the SctN₆-SctO₁ ring. Tyr-45, Tyr-49, and Phe-72 of FliJ are also conserved between FliJ and the γ subunit of the F₁ ATPase. Among these conserved, surface-exposed residues of FliJ, Phe-72 and Leu-76 are critical for the interaction with FlhA_L. Residues in the γ subunit corresponding to Phe-72 and Leu-76 are involved in the interaction with an α-helix of the δ subunit of the bovine mitochondrial F₁ ATPase, which is a homolog of the conserved ε subunit in the F₁ ATPase family (PDB ID: 1E79) (Gibbons et al., 2000). The flhA(E351A/W354A/D356A) triple mutation significantly reduces the binding affinity of FlhA_L for FliJ (Inoue et al., 2021). Because the residues from Val-349 to Val-357 of FlhA form an α-helix, FliJ may bind to this α-helix in FlhA_L as is seen in the γ-δ interaction in the bovine mitochondrial F₁ ATPase.

ATP hydrolysis by the FliI ATPase and rapid protein translocation by the export complex are both linked to efficient H⁺ translocation through the FlhA ion channel (Morimoto et al., 2016). Recently, it has been reported that ATP hydrolysis by the FliI ATPase also unlocks the entrance gate of the polypeptide channel formed by FliP, FliQ, and FliR for efficient entry of export substrates into the channel (Kinoshita et al., 2021). Furthermore, the Salmonella ΔfliHIJ flhB(P28T) flhA(T490M) mutant has been isolated as a revertant of the ΔfliHIJ mutant that has increased motility (Minamino et al., 2021a). The protein-export activity of the transmembrane export complex in cells with both the flhA(T490M) and flhB(P28T) mutations is almost at the wild-type level under a variety of experimental conditions even in the absence of the FliH₁₂-FliI₆-FliJ₁ ring complex (Minamino et al., 2021b). This finding suggests that the export complex normally requires the FliH₁₂-FliI₆-FliJ₁ complex to serve as a H⁺-coupled protein transporter. Because FliJ requires FliH and FliI to bind efficiently to FlhA_L (Minamino et al., 2011), this observation raises the question of how this ATPase ring complex activates the export complex.

The conserved Glu-211 residue of FliI catalyzes ATP hydrolysis. The E211D substitution decreases FliI ATPase activity by about 100-fold (Minamino et al., 2014). Salmonella wild-type cells produce an average of 4.4 ± 1.6 flagellar filaments per cell. In contrast, more than 90% of Salmonella fliI(E211D) cells have an average of 2.3 ± 1.5 flagellar filaments, and the average length of those filament is only half that of the wild type. Because the fT3SS transports 20,000–30,000 flagellin molecules per flagellum to form a 10–15 μm long helical filament (Minamino and Namba, 2004), the rate of ATP hydrolysis by the FliI ATPase cannot determine the rate of filament assembly, as ATP consumption by the fT3SS during flagellar assembly must be relatively small. Because the export complex still transports flagellar axial proteins even with the infrequent ATP hydrolysis provided by FliI(E211D), ATP hydrolysis appears to be required only for activation of the H⁺-driven export engine.

The six FliI_C domains undergo cooperative and sequential conformational changes triggered by ATP hydrolysis. Deletion of residues 401–410 in the first α-helix of FliI_C, which is responsible for the interaction with FliJ, significantly decreases the protein transport activity of the fT3SS although the ATPase activity is still at about 40% of the wild-type level. Because this deletion does not inhibit the interaction between FliI and FliJ, it may affect conformational changes in the FliI_C domains that rotate FliJ within the FliI hexamer. Rotation of FliJ may induce conformational changes in the FlhA_TM domain through an interaction between FliJ and FlhA_L, thereby activating the FlhA ion channel and unlocking the entrance gate of the polypeptide channel (Figure 7).

The elementary step size of γ rotation within the α₃β₃ ring is 120°, which is composed of 80° and 40° sub-steps driven by ATP binding–ADP release and ATP hydrolysis–Pi release, respectively (Watanabe and Noji, 2013). The cryoEM structure of the SctN₆-SctO₁ ring complex has suggested a possible rotational mechanism for catalysis. In this model, the SctO stalk rotates in the SctN₆ ring through an interaction between each SctN_C domain and SctO. Because the SctN₆-SctO₁ ring complex has six catalytic sites, the elementary step size of SctO rotation within the SctN₆ ring is probably 60° (Majewski et al., 2019). The FliI(E211Q) substitution in Salmonella, which completely eliminates ATPase activity but not ATP binding to the P-loop of FliI_CAT, results in only 17% of cells having one or two flagellar filaments about 25% the length of those of the wild type (Minamino et al., 2014). Thus, ATP binding to the P-loop is sufficient to activate the H⁺-driven export engine of the fT3SS to some degree. So, the 60° rotation of FliJ may be divided into two sub-steps, and ATP binding may induce the first sub-step, which may be sufficient to activate the H⁺-driven export engine weakly.

The FliI monomer interacts with the FliH dimer to form a hetero-trimer in the cytoplasm (Minamino and Macnab, 2000b; Auvray et al., 2002). High-resolution imaging of fluorescently labeled FliI in vivo has revealed that FliH₂-FliI₁ complexes are associated with the basal body through interactions of FliH with FlhA and FliN. FliI-YFP shows a rapid exchange between the basal body and a freely diffusing cytoplasmic pool. The FliI(K188I) substitution, which inhibits ATP binding to the P-loop in FliI_CAT, does not affect the exchange rate of FliI-YFP, suggesting that ATP hydrolysis does not drive the association-dissociation cycle (Bai et al., 2014). FliH also suppresses the ATPase activity of the FliH₂-FliI₁ complex (Minamino and Macnab, 2000b). Deletion of flhA decreases the number of FliI-YFP molecules associated with the basal body but does not affect the exchange rate. The highly conserved Trp-7 and Trp-10 residues of FliH_N are directly involved in the interactions of FliH with FliN and FlhA_TM (Hara et al., 2012; Notti et al., 2015). Because the interaction between FliH and FliN is required for efficient localization of the FliH₂-FliI₁ complex to the flagellar base, the FliH-FliN interaction must be highly dynamic to achieve rapid and efficient flagellar protein export by the fT3SS. Flagellar chaperones in complex with their cognate substrates both bind to FliI_C, suggesting that FliI_C is also involved in substrate recognition (Thomas et al., 2004; Imada et al., 2010; Minamino et al., 2012b). Pull-down assays have demonstrated that chaperone-associated export substrates bind to FlhA_C and FlhB_C even in the absence of FliHI (Evans et al., 2013; Kinoshita et al., 2013; Inoue et al., 2019). However, they require the FliH₂-FliI₁ complex to efficiently interact with FlhA_C and FlhB_C in vivo (Minamino et al., 2016b; Inoue et al., 2018; Kinoshita et al., 2021).

In vitro protein transport assays using inverted membrane vesicles have shown that addition of the purified FliH₂-FliI₁ complex considerably increases the transport of flagellar axial protein into the lumen of the membrane vesicles (Terashima et al., 2018, 2020). Thus, the FliH₂-FliI₁ complex acts as a dynamic carrier to deliver chaperone-associated export substrates from the cytoplasm to the flagellar base and to facilitate their docking to FlhA_C and FlhB_C, thereby allowing the activated export complex to unfold and transport export substrates into the central channel of the flagellum.

Salmonella cells lacking the FliH and FliI proteins display a very weak motile phenotype. This defect is considerably alleviated by either an increase in the expression level of export substrates and chaperones or an increase in the PMF (Erhardt et al., 2014). Expression of Vibrio alginolyticus FlhA, which has 73.2% similarity and 52.9% identity in amino acid sequence with Salmonella FlhA, restores motility in the Salmonella ΔflhA mutant but does not increase motility in the Salmonella ΔfliHI flhB(P28T)ΔflhA mutant (Minamino et al., 2016a). Thus, Vibrio FlhA requires FliH and FliI to perform protein export in the Salmonella fT3SS. Deletion of flgM, which encodes the negative regulator of the flagellar regulon, increases the expression levels of FliJ, export substrates and flagellar export chaperones and allows Vibrio FlhA to perform protein transport even in the absence of FliH and FliI. These results suggest that FlhA needs FliH and FliI to buffer protein export against internal perturbations (Minamino et al., 2016a).

The fT3SS utilizes the secreted molecular ruler protein FliK to stop growth of the hook at about 55 nm (Minamino et al., 1999; Erhardt et al., 2011; Kinoshita et al., 2017). The ΔfliHI flhB(P28T) bypass mutant cannot properly control the length of the hook, although it secrets the hook capping protein FlgD and the FliK ruler into the culture media almost at the wild-type level (Inoue et al., 2018). However, secretion level of the hook protein FlgE is about 10-fold lower than the wild-type level. The flhA(F459A) mutation, which targets a residue within FlhA_C (Figure 4), significantly increases the secretion of FlgE, so the secreted FliK ruler can measure the length of the hook more precisely. Neither the secretion levels nor control of hook length is affected by the FlhB(P28T) and FlhA(F459A) substitutions when FliH and FliI are present. Because FlgD, FlgE, and FliK bind to FliH and FliI as well as FlhA_C and FlhB_C (Minamino and Macnab, 2000c), the FliH₂-FliI complex may coordinate targeting of FlgD, FlgE, and FliK to the FlhA_C-FlhB_C docking platform to make control of the hook length more robust.

FlgN, FliS, and FliT act as export chaperones for FlgK/FlgL, FliC, and FliD, respectively (Fraser et al., 1999; Auvray et al., 2001). The chaperone-substrate complexes bind FlhA_C with nanomolar affinity (Figure 4; Kinoshita et al., 2013). This strong interaction of the chaperone with FlhA_C facilitates protein unfolding and transport by the H⁺-driven export complex (Furukawa et al., 2016; Minamino et al., 2021c). In wild-type cells, more than 90% of flagellin molecules transported by the fT3SS assemble into the filament. The ΔfliHI flhB(P28T) flhA(F459A) cannot efficiently produce the hook-filament junction and filament cap structures at the hook tip, and hence more than 90% of the flagellin molecules are secreted as monomer into the culture supernatant. Because FlgN and FliT bind to the FliH₂-FliI₁ complex whereas FliS does not (Thomas et al., 2004; Minamino et al., 2012b; Sajó et al., 2014), the FliH₂-FliI₁ complex may contribute to hierarchical targeting of the flagellar chaperones to FlhA_C, thereby allowing the junction and filament cap structures to be efficiently formed at the hook tip prior to filament formation. Thus, the FliH₂-FliI₁ complex works with the FlhA_C-FlhB_C docking platform to ensure the correct order of protein export.

The information thus far summarized allows us to propose a model for the energetics of protein export by the fT3SS (Figure 7). The transmembrane export complex remains inactive until the cytoplasmic ATPase ring complex localizes to the flagellar base through an interaction between FliH_N and FlN (Step 1). ATP hydrolysis by the FliI ATPase induces the rotation of FliJ within the FliI₆-ring at the FlhA_C-FlhB_C docking platform. The interactions of FliJ and FlhA_L induce conformational changes in the export complex that activate the FlhA ion channel and unlock the entrance gate of the polypeptide channel (Step 2). Then, cytoplasmic FliH₂-FliI₁ complexes escort export substrates and chaperone-substrate complexes from the cytoplasm to the FlhA_C-FlhB_C docking platform through the interactions of FliH with FliN and FlhA (Step 3). The binding of the export substrate to the docking platform induces opening of the gate to the polypeptide channel, and the activated export complex acts as an H⁺/protein antiporter that couples proton flow through the FlhA ion channel with the translocation of export substrates into the polypeptide channel (Step 4). The association-dissociation cycle of the FliH₂-FliI complex with the docking platform allows the transport of flagellar axial proteins in a highly controlled manner.