Bacterial flagellar capping proteins adopt diverse oligomeric states

Flagella are crucial for bacterial motility and pathogenesis. The flagellar capping protein (FliD) regulates filament assembly by chaperoning and sorting flagellin (FliC) proteins after they traverse the hollow filament and exit the growing flagellum tip. In the absence of FliD, flagella are not formed, resulting in impaired motility and infectivity. Here, we report the 2.2 Å resolution X-ray crystal structure of FliD from Pseudomonas aeruginosa, the first high-resolution structure of any FliD protein from any bacterium. Using this evidence in combination with a multitude of biophysical and functional analyses, we find that Pseudomonas FliD exhibits unexpected structural similarity to other flagellar proteins at the domain level, adopts a unique hexameric oligomeric state, and depends on flexible determinants for oligomerization. Considering that the flagellin filaments on which FliD oligomers are affixed vary in protofilament number between bacteria, our results suggest that FliD oligomer stoichiometries vary across bacteria to complement their filament assemblies. DOI: http://dx.doi.org/10.7554/eLife.18857.001


Introduction
Pathogenic bacteria cause a multitude of deadly human diseases. Many of these microbes possess flagella, molecular machines responsible for cell motility, adherence to host cells, and pathogenicity (Duan et al., 2013;Haiko and Westerlund-Wikstrom, 2013). Flagella are helix-shaped hollow attachments formed predominantly by thousands of copies of the protein flagellin (also called FliC), anchored in the bacterial membrane by a hook (or joint) that is attached to the basal body and that is composed of rotary motor proteins (Arora et al., 1998). A proton motive force typically drives the propeller motion of flagella (Berg, 2003), resulting in swimming motility. A FliD (also called HAP2) oligomer forms the cap protein complex that is located at the tip of the flagellar filament (Yonekura et al., 2000). This complex controls the distal growth of the filament by regulating the assembly of FliC molecules, which are transported through the hollow filament from the cytoplasm to the tip of the flagellum.
The dynamic movement of FliD in this assembly was modeled based on low (~26 Å )-resolution cryo-electron microscopic (EM) structures of the Salmonella serovar Typhimurium flagellum-cap complex Yonekura et al., 2000Yonekura et al., , 2003, which adopts the shape of a fivelegged stool with flexible leg domains that regulate the assembly of new FliC molecules onto the tip of the growing flagellum . It has been suggested that the plate of the stool is formed by core regions of the FliD molecule, and that disordered/flexible regions form the five leg structures  that are known to interact with the FliC filament. FliD exhibits low sequence similarity to the flagellar hook proteins and to FliC. Nevertheless, it shares the disordered terminal regions of these flagellar proteins, a common structural characteristic that is thought to control the polymerization of flagellar proteins and to play an important role in interaction with the FliC filament . These regions are the most conserved in FliD sequences across bacteria. Flagellum-mediated motility is crucial for the virulence and pathogenicity of numerous bacteria, including Campylobacter jejuni (Black et al., 1988), Salmonella (Allen-Vercoe and Woodward, 1999; Marchetti et al., 2004), Escherichia coli (La Ragione et al., 2000), Vibrio cholera (Krukonis and DiRita, 2003), and Pseudomonas aeruginosa (Arora et al., 2005), as well as the major causative agent of gastric cancer Helicobacter pylori (Kim et al., 1999). To date, however, no high-resolution structure of any FliD protein exists. To better define the roles of FliD in bacterial motility and pathogenesis, we determined the first X-ray crystal structure of FliD at 2.2 Å resolution, and assessed the structural contributions of its flexible regions using a multitude of complementary biophysical and functional analyses. eLife digest Many bacteria, including several that cause diseases in people, have long whip-like appendages called flagella that extend well beyond their cell walls. Flagella can rotate and propel the bacteria through liquids, such as water or blood, and they are constructed primarily from thousands of copies of a single protein called flagellin. When flagella are built, the flagellin proteins are placed in their proper positions by another protein called FliD, several copies of which form a cap on the end of flagella. Without FliD, bacteria cannot properly assemble flagella and, thus, can no longer swim; this also hinders their ability to cause disease.
Determining the three-dimensional structure of a protein, down to the level of its individual atoms, can provide unique insights into how the protein operates. However, no one had resolved the structure of a FliD protein from any bacterium to this level of detail before. Now, Postel et al. report the high-resolution structure of a large fragment of FliD from the bacterium Pseudomonas aeruginosa. The structure reveals that parts of this FliD protein are shaped like parts of other proteins from which flagella are constructed, including the flagellin protein that FliD places into position. Some parts of the FliD protein are also very flexible and these parts of the protein are responsible for holding numerous FliD proteins together as a cap. Finally, Postel et al. saw that six copies of FliD bind to one another to form a protein complex on the end of flagella. This last finding was particularly unexpected since it was thought that all FliD proteins formed fivemembered cap complexes, an assumption that was based largely on studies of FliD from another bacterium called Salmonella.
The current structure covers about half of the FliD protein, and so the next challenge is to determine the structure of the full-length protein. An improved understanding of the structure of FliD may, in future, help researchers to design drugs that stop bacteria from building flagella and, therefore, from swimming and causing disease.

Results
Crystal structure of the FliD protein from P. aeruginosa PAO1 To facilitate crystallization of FliD from the P. aeruginosa PAO1 strain, we deleted the predicted coiled-coil domains on both the N-and C-termini of full length FliD, which has 474 residues (FliD 1-474 ), to generate the truncated FliD 78-405 (Figure 1a, Figure 1-source data 1). We expressed FliD 78-405 in E. coli with an N-terminal His 6 -tag and purified it to homogeneity by Ni 2+ -NTA, size exclusion and anion exchange chromatography. We improved initially weakly diffracting crystals of FliD 78-405 by random matrix microseed screening (Bergfors, 2003), yielding crystals that diffracted to 2.2 Å resolution. In the absence of any homologous protein that could be used as a model for molecular replacement, we crystallized a seleno-methionine derivative of FliD 78-405 that included four leucine-to-methionine mutations (FliD 78-405 /L 4 -M 4 ). This crystal provided phase information sufficient to build an initial model, which we used subsequently for molecular replacement with the native FliD 78-405 dataset (Figure 1-source data 1). We modeled residues 80-273 into clear electron density, including all side chains, but observed density of increasingly poor quality in the C-terminus beyond residue 273 (Figure 1-figure supplement 2a). Thus, we were able to model with confidence only a single a helix in this region, corresponding to residues 274-308, with incomplete side chain structures. To determine whether the remaining region of the protein actually existed in the crystals and not just in the protein preparation used for crystallization, we analyzed crystals using liquid chromatography-mass spectrometry (LC-MS) and SDS-PAGE. Both analyses indicated that the crystals consisted of an approximate 50:50 mixture of the FliD 78-405 protein used for crystallization and a further proteolyzed version with a molecular weight of about 27 kDa. The N-terminal His 6 -tag is still detectable by Western blot (Figure 1-figure supplement 2b). Thus, the proteolyzed form corresponds approximately to residues 78-319 of FliD. The 86 residues absent from the C-terminus in a population of FliD proteins are clearly not required for crystal packing, suggesting that they are highly flexible even in a crystalline environment.

FliD is structurally similar on the domain level to FliC and FlgK
Our crystal structure of FliD 78-405 reveals that it consists of two discreet regions with distinct conformational properties, corresponding to a stable head region and a flexible and/or disordered leg region ( Figure 1b). The head region is itself comprised of two separate, but entwined, protein domains. Residues 80-101 form two b strands that belong to the first domain (Head 1), the second domain (Head 2) is formed in its entirety by the contiguous residues 104-230, and residues 231-273 then complete the first domain. The second domain is, thus, a loop insertion of the first domain. We searched for structural homologs of these domains in the Protein Data Bank and found that the first domain of the head region (Head 1) exhibits high structural similarity (RMSD=2.5 Å ), despite low sequence identity (14%), to the FlgK/HAP1/hook filament capping protein of Burkholderia pseudomallei (PDB code 4UT1; Figure 1c). Likewise, the second head region domain (Head 2) exhibits high structural similarity (RMSD=2.7 Å ), despite low sequence identity (17%), to the FliC/flagellin protein of P. aeruginosa (PDB code 4NX9; Figure 1d). In contrast to the head region, the leg region of FliD is highly flexible, as indicated by the paucity of electron density corresponding to residues 274-405 ( Figure 1-figure supplement 2a). Despite this, we were able to model the initial a helical structural element, corresponding to residues 274-308, of this region. This helix extends from the axis of the head region at a nearly perpendicular angle, resulting in an L-shaped monomer subunit structure ( Figure 1b).

FliD from P. aeruginosa PAO1 forms a hexamer
In the crystal, FliD 78-405 monomer subunits are arranged in hexamers, resulting in a shape akin to a six-pointed star when viewed from the top of the FliD oligomer (Figure 2a), which corresponds to the distal end of the growing flagellum. This star shape has a minimum inner diameter of 48 Å and a maximal outer diameter of 136 Å . When viewed from the side (Figure 2b), the FliD hexamer appears as a six-legged stool, the legs of which extend 55 Å below the bottom of the head region. Additional crystallographic symmetry results in the stacking of hexamers in alternating head-to-head and leg-to-leg orientations ( Figure 2c). The leg-to-leg stacking forms dodecamers, resulting from the helix-helix interaction of the residues 274-302 of stacked molecules and the interaction of residues 303-308 of one FliD 78-405 molecule with Head 1 domain of a stacked molecule, burying a surface area of 1362 Å 2 . The formation of dodecamers may be unique to the FliD 78-405 fragment, as this strand could potentially be replaced by additional N-terminal residues in the full-length FliD 1-474 protein. All of the morphologies observed for Pseudomonas FliD are highly reminiscent of the pentamer/decamer oligomeric organization of Salmonella FliD as determined by low-resolution cryo-EM analysis   (Figure 2d). Indeed, despite the difference in the stoichiometries of the Pseudomonas and Salmonella FliD oligomers, the gross measurements are nearly identical for these two proteins of similar molecular weight. In our crystal structure, Pseudomonas FliD measures 136 Å in diameter with a head region that is 30 Å deep and a leg region that is 55 Å long; whereas, in the cryo-EM structure, Salmonella FliD measures 145 Å in diameter with a head region that is 30 Å deep and a leg region that is 55 Å long .
The stoichiometry of Pseudomonas FliD differs from that of Salmonella FliD and our Pseudomonas FliD crystals belong to the P6 space group, which could possibly force a non-physiological oligomeric organization of subunits. Thus, we confirmed that the hexameric assembly of FliD 78-405 occurs not only in the crystalline environment but also in solution using negative stain EM class averaging (Tang et al., 2007), resulting in an image that clearly exhibits a hexameric assembly ( Figure 3a). These data are consistent with the oligomerization state that we detected in the crystal structure, in which a view looking down the a-helices of the leg reveals a six-membered ring organization. We also verified that both FliD 78-405 and full-length FliD 1-474 form oligomers by both analytical ultracentrifugation (AUC) and crosslinking analyses. We found that the FliD 78-405 fragment that we crystallized oligomerizes up to a dodecameric state (Figure 3b,c), similar to the crystallographic assembly ( Figure 2c). Additionally, we collected small-angle X-ray scattering (SAXS) data of FliD 78-405 , for which the calculated radial distribution function ( Figure 3d) is characteristic of an oligomeric assembly forming a hollow sphere (Svergun and Koch, 2003). These data produce a molecular envelope that superimposes well with our dodecameric FliD 78-405 X-ray crystal structure ( Figure 3d). Fulllength FliD 1-474 , by contrast, forms up to hexamers in solution (Figure 3e,f), which are likely to represent the physiologically relevant oligomerization state of this protein on the tip of the flagellum. The predominant tetrameric species in solution identified by AUC may be a stable intermediate on the path to hexamer formation ( Figure 3e).
As the hexameric assembly of full-length Pseudomonas FliD 1-474 proved to be unstable in solution in the absence of the flagellar filament, we sought to stabilize it using our newfound understanding of its structure. We used our crystal structure of FliD 78-405 , as input to Disulfide by Design 2.0 (Craig and Dombkowski, 2013), to identify cysteine mutations that would lead to a stable, disulfide-bridged hexameric FliD 1-474 . We found that when two residues within neighboring head domain subunits, I167 and D253, were each mutated to a cysteine residue ( Figure 4a) a stable, hexameric full-length FliD 1-474(I167C/D253C) resulted under non-reducing conditions, as shown by SDS-PAGE To show that the hexameric form of Pseudomonas FliD is functional in vivo, we complemented the fliD transposon strain PW2975 (DfliD) with wildtype fliD 1-474 and hexamer-stabilized fliD 1-474 (I167C/D253C) , resulting in Pseudomonas PAO1 strains DfliD/fliD 1-474 and DfliD/fliD 1-474(I167C/D253C) , respectively. We found that swimming motility that was lost in the DfliD strain was restored in both DfliD/fliD 1-474 and DfliD/fliD 1-474(I167C/D253C) complementation strains, similar to our observations in the wildtype Pseudomonas PAO1 strain ( Figure 4d). Using antibodies that we generated by phage display to Pseudomonas PAO1 FliD, we confirmed the expression of full-length FliD proteins by Western blot analysis from preparations of flagella isolated from live bacteria from both DfliD/fliD 1-  Together, these data indicate that FliD that is covalently locked in its hexameric assembly can form functional flagella that allow Pseudomonas bacteria to swim like Pseudomonas with wildtype FliD. Thus, the hexamer oligomeric state of Pseudomonas FliD is functional in vivo.
We also tested whether Salmonella FliD, which is known to form pentamers when capping the flagellar filament, could function as a capping protein for Pseudomonas flagella. In contrast to the in vivo functional hexameric forms of P. aeruginosa FliD -474 , complementation of the PAO1 PW2975 transposon strain with fliD from Salmonella typhimurium (DfliD/fliD StyFliDe ) did not restore swimming  (Figure 4-figure supplement 3). As this clone was codon-optimized for expression in Escherichia coli, we also confirmed that a wildtype PAO1 full-length FliD 1-474 encoded by a similarly codon-optimized gene did restore swimming motility in the DfliD strain (DfliD/fliD PAOfliDe ; Figure 4figure supplement 3). Although there exist many possible reasons other than oligomeric state that could explain the inability of Salmonella FliD to functionally complement Pseudomonas, these data suggest that Pseudomonas flagella may prefer FliD proteins that adopt hexameric rather than pentameric states.

Regions outside of the head domains drive FliD oligomerization
Although the FliD 78-405 crystal structure exhibits intermolecular contacts between the head regions of FliD 78-405 subunits comprising the hexamer, each of these interfaces is small, with a buried surface area of only 665 Å 2 , and contains few intermolecular contacts (Figure 5a) relative to typical proteinprotein interactions (Jones and Thornton, 1996). To determine whether these interactions were sufficient to drive oligomerization of FliD, we expressed and purified the head region only, FliD 78-273 . By AUC (Figure 5b), chemical crosslinking ( Figure 5c) and SAXS analysis (Figure 5d), we observed that when FliD lacks the leg region and the N-and C-terminal coiled-coil domains, it is present predominantly in the form of monomers (and dimers to a lesser extent) in solution, but fails to form higher-order oligomers as do the longer versions of FliD that we analyzed. Because different buffer conditions, including changes in pH, have been shown to affect the polymerization states of flagellar filaments (Shibata et al., 2005) and capping proteins , we performed additional AUC experiments and found that FliD 78-273 is entirely monomeric at pH 8.0 and becomes approximately one-third dimeric at pH 5.0; we observed no higher-order oligomers of FliD 78-273 regardless of buffer conditions (Figure 5b). We also assessed, by chemical cross-linking, the oligomerization states of FliD variants lacking only the N-terminal coiled-coil domain (FliD 78-474 ; Figure 5e) or the C-terminal coiled-coil domain (FliD 1-405 ; Figure 5f). We found them to be mainly monomeric with a minority of species appearing to dimerize, although the latter exhibit weak higher-order oligomerization potential. Kratky plots and radial distribution functions calculated from SAXS data of the variants lacking either the N-or C-terminal coiled-coil domain, FliD 78-474 or FliD 1-405 respectively, reveal that these proteins adopt extended shapes with flexible regions that are clearly represented in the resulting molecular envelopes (Figure 6a,b). These data indicate that the driving force for hexamerization of Pseudomonas FliD resides in molecular determinants outside of the head region and, at a minimum, involves residues in the N-terminal and C-terminal coiled-coil domains. FliD 78-405 is lacking the C-terminal and N-terminal coiled-coil domain but still assembles into dodecamers, as shown in the crystal structure, cross-linking experiments and AUC, which is likely caused by strand replacement in the head region domain 1 and helix-helix (residues 274-308) interaction of stacked molecules ( Figure 2c).

The N-and C-terminal regions of FliD are highly flexible
A large extent of FliD sequence currently remains inaccessible to high-resolution structural analysis, including Pseudomonas FliD residues 1-79 and 309-474. Consequently, we performed hydrogen/ deuterium (H/D) exchange-mass spectrometry (HDX-MS) experiments with FliD 78-405 to define its solvent accessible regions and to evaluate its dynamic behavior. We subjected FliD 78-405 to H/D exchange for 10 s to 2 hr and observed that the head region (residues 80-273) of FliD 78-405 adopts a largely stable exchange-protected fold with greater stability observed for domain 2 relative to that Figure 6. Small angle X-ray scattering (SAXS) data of FliD 1-405 , FliD 78-474 and FliD 1-474 . Log-scale intensity SAXS profiles, Kratky Plot (I Â q 2 versus q), radial distribution function calculated by GNOM and SAXS envelopes calculated by DAMMIF are shown for: (a) FliD 1-405 at 10.4 mg/mL (blue), 5.2 mg/ mL (red) and 2.6 mg/mL (grey); (b) FliD 78-474 at 9.9 mg/mL (blue), 4.95 mg/mL (red) and 2.5 mg/mL (grey); and (c) FliD 1-474 at 11 mg/mL (blue), 5.5 mg/ mL (red), 2.7 mg/mL (grey), 1.38 mg/mL (green) and 0.69 mg/mL (yellow). DOI: 10.7554/eLife.18857.014 The following figure supplement is available for figure 6: of domain 1. The leg region, particularly residues C-terminal to the a helix observed in the crystal structure, is more disordered or less stable (Figure 7a). Residues linking head domain 1 to the leg helix display cooperative unfolding behavior as indicated by EX1 kinetics (Weis et al., 2006) that result in double isotopic envelopes (Figure 7b). The FliD 78-405 protein used in these experiments was folded properly as shown by circular dichroism, as were all other FliD protein fragments that we produced (Figure 7-figure supplement 1). When mapped to our crystal structure of FliD 78-405 , the degree of H/D exchange over time on the peptide level corresponds to the degree of conformational stability on the residue level (Figure 7c). When we used a difference plot to compare the extent of H/D exchange of full length FliD 1-474 with that of FliD 78-405 , we observed that residues 165-225 exhibit relatively greater stability in FliD 78-405 (Figure 8a). Within this stretch of residues in head domain 2 are those residues, 165-171, which form the interface between the two head region domains in the hexameric complex (Figure 8b). We observed an additional region of relative stabilization for residues 298-324, part of which, residues 300-308, correspond to the b strand replacement in the opposing head region domain 1 that drives dodecamer formation of FliD 78-405 (Figure 8a). Kratky plots calculated from SAXS data of full-length FliD 1-474 at pH 11, which is monomeric under these conditions ( Figure 6-figure supplement 1), confirms the overall flexible nature of this protein. Accordingly, heterogeneous molecular envelopes calculated from SAXS data collected at different concentrations of FliD 1-474 also exhibit significant conformational flexibility ( Figure 6c).

Multiple regions of FliD interact with one another
To gain further insight to the dynamics and interactions of the different regions of FliD, we performed HDX-MS experiments with the remaining Pseudomonas FliD fragments that we had generated (Figure 1a, Figure 1-figure supplement 1). HDX-MS heat maps show the overall high degree of flexibility of FliD regions outside the head domain (Figure 8-figure supplement 1), and difference plots of FliD truncations in comparison to the full length FliD 1-474 indicate that the head region, and more specifically head region domain 1, is stabilized by FliD regions outside of the head region ( Figure 8c). Indeed, the N-terminal coiled coil is responsible for this stabilization of head region domain 1, in addition to stabilization of residues between the leg region a helix and the C-terminal coiled coil (Figure 8c). Finally, the C-terminal coiled coil predominantly stabilizes residues in the N-terminal coiled coil (Figure 8c). The stabilization of one FliD region by another is most likely to be direct and intramolecular, rather than allosteric and intermolecular, as the full-length protein FliD 1-474 adopts predominantly monomeric species under the conditions that we used for HDX-MS analysis ( Figure 8-figure supplement 2). This leads to a transient structural model of the entire FliD monomer subunit (Figure 8d) in which the N-terminal coiled coil bridges the head and foot regions and is itself pinned to these structural elements by the C-terminal coiled coil. Notably, none of the flexible regions of FliD, save the C-terminal 20 residues, are consistently predicted by sequence analysis to be intrinsically disordered (Figure 8-figure supplement 3); rather, they are inherently capable of adopting a limited number of conformations required to be in an 'on-state' (actively chaperoning and sorting a FliC protein) or an 'off-state' (engaging only structured and previously positioned FliC proteins).

Discussion
At the gross structural level, our studies show that the oligomeric states differ between FliD protein assemblies in diverse bacteria. Salmonella has long served as the model for bacterial flagellum structure and function. Since Salmonella FliD performs its native flagellar capping function as a pentamer Maki et al., 1998;Vonderviszt et al., 1998;Yonekura et al., 2000), it could reasonably be assumed that all FliD proteins form pentamers at the distal ends of all bacterial flagella. We found that Pseudomonas FliD instead forms hexamers both in crystals and in solution. Additionally, we showed that Pseudomonas FliD constrained to its hexameric state by inter-subunit disulfide bonds is functional in vivo, resulting in the formation of flagella and swimming motility. Conversely, Salmonella FliD that assembles as pentamers does not allow flagellar formation and swimming motility in Pseudomonas bacteria.
Much like cryo-EM studies of flagellar filaments from diverse bacteria unequivocally showed that the number of protofilaments ranges, at least, from 11 in Salmonella  to seven in Campylobacter (Galkin et al., 2008), our structural analyses show that the proteins that cap these filaments also vary in their oligomeric states. The end of the Salmonella filament exhibits a non-planar surface with five indentations (Yonekura et al., 2000) into which the five legs of the corresponding FliD pentamer have been modeled . Although no structure of the Pseudomonas filament has been determined even at low resolution, our hexameric FliD structure suggests that it may incorporate an even greater number of protofilaments than does the Salmonella filament. This would allow the formation of an additional molecular cavity on the distal end of FliD that accommodates its unique capping protein hexamer. Indeed, Pseudomonas flagellar filaments have been found to be macroscopically different from those of Salmonella (Shibata et al., 2005).
At 2.2 Å resolution, our crystal structure reveals several previously unknown structural features of FliD that are likely to be critical to its function. First, each one of the six FliD head regions that appear as discreet lobes when visualized by EM analysis is actually composed of the first and second domains of neighboring monomer subunits, as opposed to both domains belonging to the same subunit. This is likely the case for FliD oligomers from diverse bacteria, including Salmonella FliD for which cryo-EM analyses showed five head region lobes Yonekura et al., 2000Yonekura et al., , 2003. Second, despite vanishingly low sequence similarity, each domain within the head region exhibits high structural similarity to the corresponding domains of other flagellar proteins, including the hook-filament junction protein/FlgK and flagellin/FliC. Thus, at least one protein domain from the proteins that occupy the distal ends of the hook (FlgK) and the cap (FliD) adopts a common structural fold, which may be required for and unique to their function at these related positions within the flagellum. This domain conservation among flagellar proteins confirms a previous study proposing a structural relationship of FliD to other flagellar proteins despite their low sequence similarities . This structural conservation suggests that a structurally similar cap complex may form at both hook and on the distal end. It also suggests that FliD may play a functional role earlier in the flagellar assembly process, prior to its known FliC sorting function. Notably, in mutants of Salmonella that lack filaments, FlgK (HAP1), FliD (HAP2) and FlgL (HAP3) form a layered structure at the distal end of the hook-basal body (Homma and Iino, 1985). In addition, the shared domain structure of portions of FliC and FliD suggests that the chaperoning activity of FliD could derive, at least in part, from its presentation of a like-structured template against which FliC subunits may fold prior to being positioned into the growing filament.
Owing to the conformational flexibility of FliD necessitated by its FliC sorting function, we still lack high-resolution structural information for certain regions of this protein. However, using a number of biophysical methods, we found that molecular determinants outside of the conformationally stable head regions control FliD oligomerization. Previous analysis of a trypsinized fragment of Salmonella FliD that lacks the N-terminal 42 residues and C-terminal 51 residues, which is similar to the Pseudomonas FliD 78-405 protein that we crystallized, showed that this fragment can form pentamers but not decamers in solution and dissociates into smaller oligomers at low temperatures . This suggests that determinants outside of the head regions of FliD proteins of diverse bacteria control their oligomeric states. Accordingly, we showed that FliD, when constrained to its hexameric state by disulfide bonds formed between neighboring head domain subunits, is fully functional in vivo. Thus,conformational flexibility in or relative repositioning of the head regions are not functional requirements of FliD but flexibility in regions outside of the head regions undoubtedly is. Considering that inhibiting protein-protein interfaces with small molecules remains a major technical challenge (Arkin et al., 2014), the dependence of FliD oligomeric assembly on its flexible regions enhances the prospects of developing small molecule inhibitors of FliD oligomerization, and consequently of flagellar function, as a novel class of antibiotic agents. coiled-coil domains were codon-optimized for expression in E. coli and cloned into pGEX5x2 in frame with an N-terminal GST-tag followed by a TEV protease site. To obtain a stable hexameric full-length FliD mutant (FliD 1 -474(I167C/D253C) ), two residues (I167 and D253) located at the interface between neighboring head domains were mutated to cysteine residues in wildtype FliD 1-474 .

Recombinant protein expression and purification
All FliD constructs from the P. aeruginosa PAO1 strain were expressed in LB medium for 4 hr at 37˚C in E. coli BL21(DE3)pLysS cells after induction with 1 mM IPTG at an OD 600nm of 0.6. Selenomethionine (SeMet)-labeled FliD 78 -405 /L 4 -M 4 was produced using metabolic inhibition of methionine biosynthesis (Van Duyne et al., 1993) and growth in M9 medium containing 60 mg/L SeMet as the sole source of methionine for 6 hr after induction with 1 mM IPTG. Cells were harvested (5000 g for 15 min) and lysed in PBS including 5 mM b-mercaptoethanol by sonication. For crystallization of Histagged FliD 78 -405 and His-tagged FliD 78 -405 /L 4 -M 4, the soluble fraction was purified using HisPur NiNTA Resin (Thermo Scientific). The protein was further purified by size exclusion chromatography (Superdex 200 10/300 GL, GE Healthcare) in PBS followed by anion exchange chromatography (MonoQ 5/50 GL, GE Healthcare). For crystallization the protein was dialyzed into 30 mM Tris pH 8.0, 80 mM sodium chloride and concentrated to approximately 13 mg/mL. FliD-GST-fusion constructs were purified using a Glutathione Sepharose (BioVision) column. Following 16 hr digestion with TEV protease, the GST tag was removed by Glutathione Sepharose and TEV was removed by NiNTA (Thermo Scientific) chromatography. Cleaved FliD constructs were further purified using size exclusion (Superdex 200 10/300 GL, GE Healthcare) in PBS followed by anion exchange chromatography (MonoQ 5/50 GL, GE Healthcare) using 20 mM CHES pH 9.0 and a linear salt gradient from 0 to 1 M NaCl over 12 min.
X-ray diffraction data processing, structure determination and refinement X-ray diffraction data for the SeMet-labeled FliD 78 -405 /L 4 -M 4 crystal were collected using a Dectris 6M PILATUS detector on the 12-2 beamline at the Stanford Synchrotron Radiation Lightsource, SSRL, processed using XDS (Kabsch, 2010b), scaled in AIMLESS (Evans and Murshudov, 2013;Winn et al., 2011), and phases obtained using the SSRL multi-wavelength anomalous dispersion (MAD) script by A. Gonzalez with SHELX options based on a script by Qingping Xu, including the programs SHELX (Schneider and Sheldrick, 2002), SOLVE (Terwilliger and Berendzen, 1999) and RESOLV (Terwilliger, 2000). The initial FliD 78 -405 /L 4 -M 4 model was improved manually by rebuilding the peptide chain backbone in Coot (Emsley and Cowtan, 2004) and refining using Phenix (Adams et al., 2010). Diffraction data for native, wildtype FliD 78 -405 were collected using a MARmosaic 300 CCD detector on the 23ID-B beamline at the Advanced Photon Source, Argonne National Laboratory, APS, and processed using XDS (Kabsch, 2010b) and XSCALE (Kabsch, 2010a). The partially built and refined SeMet-FliD 78 -405 /L 4 -M 4 model was used as a molecular replacement model for phasing the native FliD 78-405 data using Phaser (McCoy et al., 2007). The initial native FliD 78-405 model was build using Autobuild and improved by manual model rebuilding in Coot (Emsley and Cowtan, 2004) and by iterative rounds of refinement using Phenix (Adams et al., 2010).

Mass spectrometry
FliD 78-405 crystals were crosslinked using 2% formaldehyde, harvested and washed in mother liquor, dissolved in water and the crosslinking reversed by heating the samples to 95˚C for 20 min. The samples were analyzed by liquid chromatography (LC)-electrospray ionization (ESI)-mass spectrometry (MS) using a gradient of mobile phase A (0.1% formic acid in water) and mobile phase B (0.1% formic acid in acetonitrile) increasing from 0% B to 90% B in 20 min. The Accela LC System was attached to a LXQ linear ion trap mass spectrometer (Thermo Scientific). Raw MS data were analyzed using Xcalibus Qual Browser (Thermo Scientific) and deconvoluted using BioWorks (Thermo Scientific, Waltham, MA).

Circular dichroism
10 mM FliD protein in 10 mM sodium phosphate pH 7.0 was used to record a spectrum ranging from 190 nm to 260 nm at 15˚C. CD melting curves were analyzed at 222 nm or 205 nm by increasing the temperature by 1˚C per minute starting at 15˚C using a JASCO J810 CD instrument according to the manufacturer's instructions.

Electron microscopy
An aliquot of a FliD 78-405 protein sample was negatively stained with 2% (weight/volume) uranyl acetate and imaged using a Tecnai F20 (FEI) electron microscope operating at 120 keV. Approximately 3500 particles were selected from 70 micrographs and used to generate class averages in EMAN2 (Tang et al., 2007). Six classes were generated, and Figure 3a shows the single largest class.

Hydrogen/deuterium exchange-mass spectrometry
The coverage maps for FliD 1-474 and FliD 78-405 were obtained from undeuterated controls as follows: 3.5 mL of~40 mM FliD in 30 mM TrisHCl, 150 mM NaCl pH 8.0 was diluted with 31.5 mL of the same buffer at room temperature followed by the addition of 100 mL of ice cold quench (100 mM Phosphate buffer, 1.5 M Guanidine-HCl, pH 2.4). The quenched samples were injected into a Waters HDX nanoAcquity UPLC (Waters, Milford, MA) with in-line pepsin digestion (Waters Enzymate BEH pepsin column). Peptic fragments were trapped on an Acquity UPLC BEH C18 peptide trap and separated on an Acquity UPLC BEH C18 column. A 7 min, 5% to 35% acetonitrile (0.1% formic acid) gradient was used to elute peptides directly into a Waters Synapt G2 mass spectrometer (Waters, Milford, MA). MS E data were acquired with a 20 to 30 V ramp trap CE for high energy acquisition of product ions as well as continuous lock mass (Leu-Enk) for mass accuracy correction. Peptides were identified using the ProteinLynx Global Server 2.5.1 (PLGS) from Waters. Further filtering of 0.3 fragments per residues was applied in DynamX.
For each construct, the HD exchange reactions were performed as follows: 3.5 mL of~40 mM FliD in 30 mM TrisHCl, 150 mM NaCl pH 8.0 was incubated in 31.5 mL of 30 mM TrisDCl, 99.99% D 2 O, pD 8.0, 150 mM NaCl. All reactions were performed at 25˚C. Prior to injection, deuteration reactions were quenched at various times (10 s, 1 min, 10 min, 1 hr and 2 hr) with 100 mL of 100 mM Phosphate buffer, 1.5 Guanidine-HCl, pH 2.4. Back exchange correction was performed against fully deuterated controls acquired by incubating 3.5 mL of 40 mM FliD 1-474 in 31.5 mL 30 mM TrisDCl, 99.99% D 2 O, pD 8.0, 150 mM NaCl containing 6 M deuterated Guanidine DCl for 2 hr at 25˚C prior to quenching (without guanidine HCl). All deuteration time points and controls were acquired in triplicates.
The deuterium uptake by the identified peptides through increasing deuteration time and for the fully deuterated control was determined using Water's DynamX 2.0 software. The normalized percentage of deuterium uptake (%D) at an incubation time t for a given peptide was calculated as follows: With m t the centroid mass at incubation time t; m 0 the centroid mass of the undeuterated control, and m f the centroid mass of the fully deuterated control. Heat maps and percent deuteration difference plots (D%D) were generated using the percent deuteration calculated. Confidence intervals for the D%D plots were determined using the method outlined by Houde et al. (2011), adjusted to percent deuteration using the fully deuterated controls. Confidence intervals (98%) were plotted on the D%D plots as horizontal dashed lines. EX1 type cooperative unfolding was analyzed using HX-Express2 (Guttman et al., 2013).

Determination of peptide coverage of FliD 1-474(I167C/D253C) under reducing and non-reducing conditions
Coverage maps of FliD 1-474(I167C/D253C) in the presence and absence of reducing agent were obtained similarly as above except for the following: 3 mL of 66 mM FliD 1-474(I167C/D253C) were incubated for 2 hr with 15 mL of 8 M Guanidine-HCl and 2 mL of 1 M TCEP (reducing conditions) or 2 mL H 2 O (non-reducing conditions). Subsequently, 180 mL of quench buffer (100 mM potassium buffer, pH 2.4) was added and the mixture immediately injected into the Waters HDX nanoAcquity UPLC. The remainder of the workflow, MS method, peptide identification and coverage map determination was unchanged. In addition, Biopharmalynx 1.3.5 (Waters) was used to search for and to identify disulfide-bridged peptides. A filter of 15% b/y ions identified was applied. The search was performed both in the context of the expected C167-C253 disulfide bridge and forthe C167-C167 and C253-C253 disulfide bridges as negative controls.