The flagellar substrate specificity switch protein FlhB assembles onto the extra-membrane export gate to regulate type three secretion

Export of proteins through type three secretion systems (T3SS) is critical for motility and virulence of many major bacterial pathogens. Proteins are transported through an export gate complex consisting of three proteins (FliPQR in flagellar systems, SctRST in virulence systems) that were initially annotated as membrane proteins, but which we have recently shown assemble into an extra-membranous helical assembly. A fourth putative membrane protein (FlhB/SctU) is essential to the export process, and also functions to “switch” secretion substrate specificity once the growing hook/needle structures reach their determined length. Here we present the structure of an export gate containing the switch protein from a Vibrio polar flagellar system at 3.2 Å resolution by cryo-electron microscopy. The structure reveals that the FlhB/SctU further extends the helical export gate assembly with its four putative transmembrane helices adopting an out-of-membrane location, wrapped around the other export gate components at the base of the structure. The unusual topology of the switch protein helices creates a loop that wraps around the bottom of the closed export gate complex. Structure-informed mutagenesis suggests that this loop is critical in gating secretion and we propose that a series of conformational changes in the type 3 secretion system trigger opening of the export gate through the interactions between FlhB/SctU and FliPQR/SctRST.

switch protein from a Vibrio polar flagellar system at 3.2 Å resolution by cryo-electron microscopy. The structure reveals 23 that the FlhB/SctU further extends the helical export gate assembly with its four putative transmembrane helices 24 adopting an out-of-membrane location, wrapped around the other export gate components at the base of the 25 structure. The unusual topology of the switch protein helices creates a loop that wraps around the bottom of the closed 26 export gate complex. Structure-informed mutagenesis suggests that this loop is critical in gating secretion and we 27 propose that a series of conformational changes in the type 3 secretion system trigger opening of the export gate 28 through the interactions between FlhB/SctU and FliPQR/SctRST. Type three secretion is a mechanism of bacterial protein secretion across both inner and outer bacterial membranes. It 4 is found in the virulence-associated injectisome (vT3SS), a molecular syringe, and the bacterial flagellum (fT3SS), a 5 motility organelle (1). Both families contribute in significant ways to bacterial pathogenesis. vT3SS facilitate secretion 6 not only across the bacterial envelope but also insert translocon proteins at the tip of the needle into the eukaryotic 7 host plasma membrane, allowing direct injection of virulence factors in the host cytoplasm. The fT3SS is responsible for 8 construction of the flagellar filament in both Gram-negative and Gram-positive bacteria, and hence imparts 9 pathogenicity (2) for example via the ability to swim towards favourable environments or sense environmental 10 conditions (3). 11 T3SS are multi-megadalton protein complexes that are capable of bridging from the bacterial cytoplasm to the 12 extracellular space. At the core of the secretion system is the highly conserved export apparatus (EA) (4, 5), which is 13 made up of five predicted transmembrane (TM) proteins (SctR, SctS, SctT, SctU and SctV in the vT3SS; FliP, FliQ, FliR,14 FlhB and FlhA in the fT3SS). FlhA/SctV has been shown to form a nonameric ring (6), consisting of a large cytoplasmic 15 domain situated below a hydrophobic domain predicted to contain 72 helices. This structure was proposed to surround 16 an "export gate" through which substrates would enter the secretion pathway. This export gate is constructed from the 17 other 4 EA proteins and was predicted to lie in the inner membrane. However, our recently determined structures of 18 the S. enterica serovar Typhimurium FliP5Q4R1 and the Shigella flexneri SctR5S4T1 complexes (7,8) demonstrated that the 19 export gate is actually embedded within the proteinaceous core of the T3SS basal body, placing it above the predicted 20 location of the inner membrane. Furthermore, the helical structure of the export gate makes it likely that it is 21 responsible for nucleating the helical filaments that assemble above it (9). Interestingly, the export gate complex has 22 also recently been proposed to facilitate inward transport across the inner membrane associated with nanotubes (10, 23 11). The final component of the EA, FlhB/SctU, has long been known to be essential for all T3SS-mediated protein 24 secretion. In addition, FlhB/SctU has a major role in switching the specificity of secretion substrates, mediating the 25 transition from the early components necessary to build the flagellar hook in fT3SS and injectisome needle in vT3SS, to 26 the later subunits required for flagellar filament or injectisome translocon assembly. The FlhB/SctU family of proteins all 27 contain an N-terminal hydrophobic sequence that is predicted to form 4 TM helices (FlhBTM) and a smaller cytoplasmic 1 C-terminal domain (FlhBC). Crystal structures of the FlhB/SctU cytoplasmic domain from a range of species and systems 2 (12-14) demonstrated a compact fold with an unusual autocatalytic cleavage site in a conserved NPTH sequence. 3 Cleavage between the Asn and Pro residues, splitting FlhBC into FlhBCN and FlhBCC, is required for the switching event to 4 occur and a variety of mechanisms have been proposed to explain the need for this unusual mechanism (15). 5 Little was known about the predicted TM portion of FlhB/SctU. Co-evolution analysis and molecular modelling led to 6 suggestions that it forms a 4-helix bundle in the membrane (16), while crosslinks (17) and partial co-purification of FlhB 7 with FliPQR were consistent with FlhB/SctU interacting with the export gate via a conserved site on the FliP5Q4R1 8 complex (7). However, given the inaccuracy of the TM predictions for the other export gate components revealed by the 9 PQR structure, we sought to determine the molecular basis of the interaction of FlhB with FliPQR. Here we present the 10 structure of the TM region of FlhB bound to the FliPQR complex, in addition to two novel structures of the FliPQR 11 homologues from Vibrio mimicus and Pseudomonas savastanoi. The structure reveals a unique topology that presents a 12 loop that wraps around the base of the closed export gate. Mutagenesis studies confirm the crucial role played by the 13 FlhB loop in the export process and suggest potential mechanisms of regulation of opening of the assembly. 14

Conservation of the FliPQR structure 16
Our previously determined structures of S. Typhimurium FliPQR (7) and the vT3SS homologue SctRST from S. flexneri 17 (8) demonstrated that the stoichiometry of the core structure (FliP5Q4R1) is conserved between virulence and flagellar 18 systems. However, classification of the SctRST data revealed variable occupancy of the SctS component (up to a 19 maximum of 4 copies), consistent with our earlier native mass spectrometry measurements (7). Furthermore, our native 20 mass spectrometry had also demonstrated that a small proportion of the P. savastanoi FliPQR complex contained 5 21 copies of FliQ, with the predicted 5 th FliQ binding site beginning to encroach on the predicted FlhB interaction site. In 22 order to further analyse the structural conservation and stoichiometry of the export apparatus core FliPQR we chose 23 the homologous complexes from the fT3SS of two other bacterial species for structural studies: the V. mimicus polar 24 flagellum FliPQR complex, which has a longer FliP sequence including an N-terminal domain conserved in the 25 Vibrionales order ( Supplementary Fig. 1), and the P. savastanoi FliPQR complex, that is a mixture of FliP5Q5R1 and 26 FliP5Q4R1 complexes by native mass spectrometry (7). We determined the structures of both complexes using cryo-EM 27 and single particle analysis to 4.1 Å and 3.5 Å respectively (Fig. 1a, Table 1, Supplementary Fig. 2 and Supplementary Fig.  1 3). Both structures are highly similar to S. Typhimurium FliPQR (7) (RMSD=1.6 Å over all chains) and S. flexneri SctRST (V. 2 mimicus FliPQR and SctRST RMSD=1.9 Å and P. savastanoi FliPQR and SctRST RMSD=2.3 Å) (7,8).  12 Consistent with our previous native mass spectrometry data, the structure of P. savastanoi revealed an additional FliQ 13 subunit in the complex. In the S. Typhimurium and V. mimicus FliPQR structures there are four FliP-FliQ units, each the 14 structural equivalent of a FliR subunit (7), but the fifth FliP is missing a FliQ. In the P. savastanoi structure, FliQ5 binds 15 the remaining FliP subunit in the same way as in the other FliP-FliQ units. This FliQ5 subunit is located close to the site 16 on FliPQR we previously identified as important for interaction with FlhB/SctU (7,17). Mapping of more a extensive in 1 vivo photocrosslinking analysis based on covariance (16) between SctU and SctR, SctS, and SctT supports a binding site 2 for SctU that involves large parts of helix 2 of SctS and helix 4 of SctT (Fig. 1b-c and Supplementary Fig. 4). Mapping of 3 the residues on the structure of FliPQR and a model of FliPQR containing a fifth FliQ subunit reveal a large binding site in 4 the complex containing four FliQ subunits and a more compact binding site when a fifth FliQ subunit is modelled (Fig.  5 1d). In this way FliQ5/SctS5 might be compatible with FlhB/SctU binding, depending on the unknown structure of the 6 FlhB/SctU transmembrane domain (FlhB/SctUTM), but addition of a sixth Fli/SctQ using the same helical parameters, 7 would block this site. 8

Architecture of the FliPQR-FlhB export gate complex 2
We have observed four FliQ subunits in the S. Typhimurium and V. mimicus FliPQR and the S. flexneri SctRST structures 3 but as we have previously observed FliQ to be sensitive to dissociation by detergent in the purification process (7,8), it 4 was possible that the 5 th FliQ is a genuine component of the complex but was lost in the purification of less stable 5 homologues. As the stoichiometry of FliQ has potentially large implications for the placement of FlhB in the system (Fig.  6 1d), we endeavoured to produce the more physiologically relevant FliPQR-FlhB complex. 7 After extensive screening of detergents, constructs with different placement of affinity tags and sequences from a 8 variety of species for co-expression and co-purification of FlhB with FliPQR, we were able to prepare a monodisperse 9 sample of the complex from V. mimicus (Fig. 2a). We analysed this sample by cryo-EM and determined the 3.2 Å 10 structure of the complex (Fig. 2b, Table 1  The soluble, globular, cytoplasmic domain (FlhBC) is not visible, likely due to flexibility in the linker between the two 18 domains. We tested whether this disorder of FlhBC relative to FliPQR-FlhBTM is due to the presence of the detergent 19 micelle in our sample by imaging the sample in the amphipol A8-35, perhaps a better mimic of the proteinaceous 20 environment relevant to the assembled T3SS (7), but we did not observe any additional density resulting from ordering 21 of FlhBC (data not shown). 22 Intriguingly, the two helical hairpins of FlhB are joined by a loop (FlhBL) that literally loops around the (closed) entrance 23 of the FliPQR gate (Fig. 2c). Consistent with our previous prediction (7) and crosslinking analysis (Fig. 1), FlhB contacts 24 the site across FliP5 and FliR, but it additionally contacts cross-linkable residues in the FliP4 subunit (Fig. 2d). As 25 previously predicted (7), hydrophobic cavities between FliP and FliQ, in addition to lateral cavities between the FlhB 26 hairpins and the FlhB/FliQ interface, are observed to contain densities consistent with lipid or detergent molecules 1 although these could not be modelled unambiguously in the current volume ( Supplementary Fig. 6). The density corresponding to FlhB was of sufficient quality to build an atomic model of the structure using only 9 sequence information ( Fig. 3a and Supplementary Fig. 5). The topology of FlhBTM is unusual; the helices 1 and 4 10 neighbour each other in the middle of the structure, while helix 2 and 3 flank the central pair on either side (Fig. 3b,c). 11 In order to further validate the topology of FlhB we compared our model to contacts derived from evolutionary co-12 variation ( Fig. 3d and Supplementary Fig. 7). This shows strong contacts between helices 1 and 2, 3 and 4 and 1 and 4 13 but an absence of contacts between helices 2 and 3, which is inconsistent with a helical bundle, but consistent with our 14 more extended and topologically unusual structure. 15 Despite observing up to five FliQ subunits in FliPQR structures, there are only four FliQ molecules in this structure. In 1 fact, the hairpin composed of FlhB helices 1 and 2 is bound to the site occupied by FliQ5 in our P. savastanoi FliPQR 2 structure, packing on FliP5, whilst helices 3 and 4 pack on FliR. Thus the presence of FliQ5 would block binding of FlhB 3 (Fig. 3e), suggesting that the fifth FliQ binds to the complex in a non-native fashion due to the absence of FlhB in the 4 overexpression system. This superposition of FlhB and FliQ5 also reveals that the hairpin of helices 1 and 2 of FlhB 5 adopts a very similar structure to FliQ despite the fact they are topologically distinct, with helix 2 of FlhB being 6 structurally equivalent to helix 1 of FliQ and vice versa, i.e. the directionality of the hairpin is reversed along the long 7 axis. Modelling FliQ5 and FliQ6 using the same helical parameters by which FliQ1 to FliQ4 are related reveals that FlhB 8 continues the spiral of FliQ subunits and even helices 3 and 4 follow the same parameters (Fig. 3e) despite not 9 interacting with a FliP subunit. Given the very different topologies of the two 10 proteins, the level to which FlhB helices 1 and 2 and FliQ superpose is surprising. An 11 evolutionary relationship between FlhB and FliQ is unlikely due to the topology 12 differences, suggesting that the similarity of the structures is a result of convergent 13 evolution and the need to form this helical assembly.

An extended loop between helices is essential for secretion 26
The unusual topology of FlhBTM means that a long loop, FlhBL, between helix 2 and 3 (residues 110-139) connects the 27 two hairpins of the structure. Most unexpectedly, this loop, c which contains the most highly conserved residues within 28 FlhB ( Supplementary Fig. 8), is seen to wrap around the base of the PQR complex, contacting each of the FliQ subunits 29 in turn and inserting conserved hydrophobic residues into the cavities between the FliQs (Fig. 3f). The loop structure 30 also reveals how a single FliQ residue can co-evolve with multiple FlhBL residues. Although FlhBL in isolation doesn't 31 further constrict the base of the already closed PQR complex, the aperture does become significantly smaller when 1 taking into account the poorly resolved termini of FlhBTM ( Supplementary Fig. 9). Therefore FlhBL could contribute to 2 export gate closure via trapping of the FlhB N-terminus and the linker connecting FlhBTM to FlhBC, in the direct line of 3 the export pathway or by pinning the FliQ subunits closed. A mutation in the FlhB N-terminus had been reported (19) to 4 act as a ∆fliHI bypass mutant (the ATPase and its regulator), presumed to be involved in controlling the opening of the 5 export channel. In the FliPQR-FlhB complex the equivalent residue (FlhBP28 in S. Typhimurium, FlhBA28 in V. mimicus) 6 locates very close to the pore entrance ( Supplementary Fig. 9). In the S. Typhimurium SctRSTU complex the 7 corresponding residue strongly photocrosslinks to SctS ( Supplementary Fig. 4), supporting the notion that SctU 8 mediates gating of the export apparatus core complex. 9 10 We decided to further probe the function of FlhBL using mutagenesis in the motile E. coli strain W ( Fig. 4a and  11 Supplementary Fig. 10). Given that opening of the FliPQR-FlhB aperture would require a conformational change in FlhBL, 12 we hypothesised two mechanisms for FlhBL such conformational changes. FlhBL could either move away from the 13 entrance to the gate through a hinging motion like a lid, or it could extend into a structure with less secondary structure 14 in order to stay in contact with the binding sites on the opening FliQ subunits, reminiscent of a sphincter. 15 Mutations in either the conserved hydrophobic residues of FlhBL that insert between the FliQ subunits (Fig. 3f) or the 16 highly conserved loop of FliQ severely reduced motility (Fig. 4a ) without affecting binding of FlhB to FliPQR 17 ( Supplementary Fig. 11). Although substitution with the bulky, hydrophobic amino acid tryptophan and removal of bulky 18 sidechains only reduced motility, introduction of charged residues completely abolished motility, suggesting that 19 secretion can proceed at lower efficiency when the FliQ-FlhBL interaction is only reduced rather than completely 20 disrupted as in the aspartate mutations. 21 We performed an extensive in vivo photocrosslinking analysis to validate the interactions and functional relevance of 22 the corresponding SctUL in the vT3SS-1 of S. Typhimurium. While no crosslinks to SctS could be identified with the 23 artificial crosslinking amino acid p-benzoyl-phenylalanine (pBpa) introduced into SctUL itself ( Supplementary Fig. 12), 24 numerous crosslinks were identified with pBpa in the lower part of SctS that faces SctUL (Fig. 4b). Using 2-dimensional 25 blue native/SDS PAGE, we could show that the crosslink observed with SctSQ41X occurred not only in the SctRSTU 26 assembly intermediate but also in the assembled needle complex (Fig. 4c), adding further support to the idea that the 27 structure of the isolated complex represents the structure of the complex in the full assembly. The observed crosslinks 1 were independent of functional secretion of the vT3SS, indicating that assembly of needle adapter, the inner rod, and 2 needle filament does not lead to a conformational change of this part of the SctRSTU complex (Fig. 4b). Strikingly, 3 introduction of pBpa at several positions of SctS led to a strong defect in secretion but not SctS-SctUL interaction, 4 highlighting the relevance of this site for secretion function of T3SS (Fig. 4d), while pBpa substitutions within SctUL 5 were more functionally neutral. In total, we found a large number of residues at the FliQ/SctS-FlhBL/SctUL interface that 6 are required for type three secretion (Fig. 4e). 7 Strong cross-linking between SctS and SctU even in the assembled needle complex and loss of function in more 8 disruptive mutations in which the interaction between FliQ and FlhBL is altered through the introduction of charged 9 residues suggest that this interaction is important for activity and FlhBL is not simply one of the closure points of the 10 complex in assembly intermediates. If this interaction is maintained in the open state of the export gate, a more 11 extended conformation of FlhBL would be required. Consistent with this idea, deletions of six or more residues in FlhBL 12 led to loss of motility (Fig. 4a). 13 We further investigated the function of FlhBL through more targeted mutations. FlhBL is a largely extended polypeptide 14 with little secondary structure, but a short stretch at its C-terminus is helical. Interestingly, a mutation of a glycine in this

11
Typhimurium SctS pBpa mutants. e, Structure of FliPQR-FlhB highlighting mutation sites that impaired motility or secretion in red and mutation sites 12 that had no or only a small effect in yellow.

14
In this study we show that FlhBTM is part of the export gate complex together with FliPQR. Two pairs of helices of FlhB 15 bind to FliPR through a structure mimicking the shape of FliQ, despite topological reversal, an example of molecular 16 convergent evolution. The unusual topology of FlhB places helices 2 and 3 apart from each other allowing them to 17 mount a loop, FlhBL, onto the cytoplasm-facing surface of the export gate. Although the way in which FlhBL wraps 18 around the closed pore suggests a role in maintaining the closed state, our structures of FliPQR/SctRST in the absence of 19 FlhB/SctU are also closed (7,8), as is the complex in the context of the assembled T3SS (21), suggesting instead that 1 FlhB may be involved in opening of the gate rather than locking it closed. We further propose that in the process of 2 opening FlhBL forms a more extended structure, implying that it undergoes cycles of extension and contraction as the 3 export gate opens and closes. 4 The location and the topology of FlhBTM place the N-terminus, FlhBCN and FlhBL in close proximity just underneath the 5 aperture of the gate. Although the resolution of the map is poor in the region of the cytoplasmic face of the complex, it 6 is possible to trace the approximate position of the FlhB N-terminus and FlhBCN (Supplementary Fig. 9). The close 7 association of the N-terminus and FlhBCN is consistent with the strong contacts derived from evolutionary co-variation 8 between the N-terminus and the N-terminal part of the cytoplasmic domain (FlhBCN) (Fig. 3c) and a genetic interaction 9 in S. Typhimurium between E11 and E230, in the N-terminus of FlhB and FlhBCN respectively (22). Furthermore,the 10 direction in which FlhBCN leaves the export gate implies that, in the context of the assembled T3SS nanomachine, FlhBC 11 could be located anywhere between the FliPQR-FlhB gate and the nonameric ring of the FlhA cytoplasmic domain below 12 (Fig. 5a). It is conceivable that the conformational changes required for opening of the export gate are propagated via 13 pulling forces imparted on helix 4 of FlhBTM, which is linked to the other helices of FlhBTM via a conserved network of 14 buried charged residues (Fig. 5b) including D208 that has been demonstrated to play a role in motility (23). The mechanism of suppressor mutations in the N-terminal residues of FlhB, such as the P28T mutation that rescues 23 motility of a strain deleted for the ATPase complex (∆fliHI) (19, 28), has long been mysterious. Our structure, 24 demonstrating the clustering of the N-terminus, FlhBCN and FlhBL at the cytoplasmic entrance of the gate, suggests that 25 they may rescue function by altering the dynamics of the closure point in order to facilitate opening of the export gate. 26 This notion is further supported by the interaction we observe of SctUF24pBpa (equivalent to P28T) with SctS. This direct 27 functional link between the ATPase and the gating mechanism, in conjunction with a host of other mutational data in 1 FlhA and FlhB, suggests that cycles of ATP hydrolysis may induce conformational changes in the export gate, presumably 2 via the FliJ stalk of the ATPase complex interacting with the FlhA ring that is positioned between the ATPase and the 3 export gate. 4 Finally, FlhB/SctU is known to play a key role in substrate switching, an event which requires autocatalytic cleavage of 5 the NPTH sequence in FlhBC/SctUC (14, 15, 26, 29). Although we do not observe the residues involved in switching in our 6 export gate structure, the fact that we are able to cross-link SctU in 7 fully assembled basal bodies, using the same residues as in the 8 purified complex, suggests that the gating mechanism discussed here 9 is likely applicable regardless of substrate. Clearly future studies will 10 need to focus on observing gating and switching events. 11 In summary, our structure of FlhB as part of the export gate complex 12 extends our understanding of the regulation of the T3SS export 13 apparatus and suggests possible mechanisms of export gate opening. 14  (100 μg/mL) or kanamycin (30 μg/mL or 60 μg/mL for large scale expression in TB medium). S. Typhimurium strains were 16 cultured with low aeration at 37°C in Luria Bertani (LB) broth supplemented with 0.3 M NaCl to induce expression of 17 genes of SPI-1. As required, bacterial cultures were supplemented with tetracycline (12.5 µg/ml), streptomycin (50 18 µg/ml), chloramphenicol (10 µg/ml), ampicillin (100 µg/ml) or kanamycin (25 µg/ml). Low-copy plasmid-based 19 expression of SctRSTU FLAG was induced by the addition of 500 µM rhamnose to the culture medium. 20 21

Generation of chromosomal deletion mutants 22
Electrocompetent E. coli W expressing λ Red recombinase from plasmid pKD46 were transformed with DNA fragments 23 containing a chloramphenicol resistance cassette surrounded by sequences homologous to the gene of interest as 24 described in Supplementary Table 3. Colonies were selected on LB agar containing chloramphenicol (20 μg/mL) and 25 transformed again with pCP20 and grown on LB agar containing ampicillin (100 μg/mL) at 30 °C. Finally, clones were 1 grown in LB media at 37 °C. Deletion mutations were confirmed by PCR. All Salmonella strains were derived from 2 Salmonella enterica serovar Typhimurium strain SL1344 (Hoiseth and Stocker, 1981) and created by allelic exchange as 3 previously described (Kaniga et al., 1994). 4 Purification of export gate complexes 5 FliOPQR or FliOPQR-FlhB were expressed in Escherichia coli BL21 (DE3) as a single operon from a pT12 vector 6 (Supplementary Table 3) as described previously (7). Briefly, cells were grown at 37 °C in TB media containing rhamnose 7 monohydrate (0.1%), harvested by spinning at 4,000 g, resuspended in TBS (100 mM Tris, 150 mM NaCl, 1 mM EDTA, pH 8 8) and lysed in an EmulsiFlex C5 homogenizer (Avestin). Membranes were prepared from the cleared lysate by 9 ultracentrifugation at 40,000 rpm in a 45 Ti rotor (Beckmann) for 3 hours. Membranes were solubilized in 1% (w/v) 10 LMNG and proteins were purified using a StrepTrap column (GE Healthcare). The resin was washed in TBS containing 11 0.01% (w/v) LMNG and proteins were eluted in the wash buffer supplemented with 10 mM desthiobiotin. Intact 12 complexes were separated from aggregate by size-exclusion chromatography in TBS containing 0.01% (w/v) LMNG 13 (S200 10/300 increase or Superose 6 increase, GE Healthcare). 14 For preparation of FliPQR-FlhB solubilised by the amphipol A8-35, the protein was purified in DDM using 1% (w/v) for 15 extraction from the membrane and 0.02% (w/v) subsequently. Eluate from the StrepTrap column was mixed with 16 amphipol at a ratio of amphipol to protein of 10:1. After incubating for one hour, the sample was dialysed into TBS using 17 a 10,000 MWCO Slide-A-Lyzer device (ThermoFisher Scientific) overnight followed by size-exclusion chromatography on 18 a Superose 6 increase column using TBS as the running buffer. 19 Sample preparation for cryo-EM 20 3 μl of purified complex at 1 to 3.6 mg/ml were applied to glow-discharged holey carbon-coated grids (Quantifoil 300 21 mesh, Au R1.2/1.3). Grids were blotted for 3 s at 100% humidity at 22 °C and frozen in liquid ethane using a Vitrobot 22 Mark IV (FEI). For samples solubilised in detergent, blotting was preceded by a wait time of 5 to 10 seconds. V. mimicus 23 FliPQR was supplemented with 0 mM, 0.05 mM, 0.5 mM or 3 mM fluorinated Fos-Choline prior to grid preparation. 24 EM data acquisition and model building 1 All data contributing to the final models were collected on a Titan Krios (FEI) operating at 300 kV . All movies were 2 recorded on a K2 Summit detector (Gatan) in counting mode at a sampling of 0.822Å/pixel, 2.4 e-Å -2 /frame, 8 s 3 exposure, total dose 48 e-/ Å -2 ,20 fractions written. Motion correction and dose weighting were performed using 4 MotionCor implemented in Relion 3.0 (32) (V. mimicus FliPQR-FlhB and P. savastanoi FliPQR) or using Simple-unblur (33) 5 (V. mimicus FliPQR). CTFs were calculated using CTFFIND4 (34). Particles were picked in Simple and processed in Relion 6 2.0 (35) and 3.0 (32) as described in Supplementary Fig. 2, Supplementary Fig. 3 and Supplementary Fig. 5. 7 Atomic models of FliPQR and FlhB were built using Coot (36) and refined in Phenix (37). 8