Structural insight into the rotational switching mechanism of the bacterial flagellar motor.

The bacterial flagellar motor can rotate either clockwise (CW) or counterclockwise (CCW). Three flagellar proteins, FliG, FliM, and FliN, are required for rapid switching between the CW and CCW directions. Switching is achieved by a conformational change in FliG induced by the binding of a chemotaxis signaling protein, phospho-CheY, to FliM and FliN. FliG consists of three domains, FliG(N), FliG(M), and FliG(C), and forms a ring on the cytoplasmic face of the MS ring of the flagellar basal body. Crystal structures have been reported for the FliG(MC) domains of Thermotoga maritima, which consist of the FliG(M) and FliG(C) domains and a helix E that connects these two domains, and full-length FliG of Aquifex aeolicus. However, the basis for the switching mechanism is based only on previously obtained genetic data and is hence rather indirect. We characterized a CW-biased mutant (fliG(ΔPAA)) of Salmonella enterica by direct observation of rotation of a single motor at high temporal and spatial resolution. We also determined the crystal structure of the FliG(MC) domains of an equivalent deletion mutant variant of T. maritima (fliG(ΔPEV)). The FliG(ΔPAA) motor produced torque at wild-type levels under a wide range of external load conditions. The wild-type motors rotated exclusively in the CCW direction under our experimental conditions, whereas the mutant motors rotated only in the CW direction. This result suggests that wild-type FliG is more stable in the CCW state than in the CW state, whereas FliG(ΔPAA) is more stable in the CW state than in the CCW state. The structure of the TM-FliG(MC)(ΔPEV) revealed that extremely CW-biased rotation was caused by a conformational change in helix E. Although the arrangement of FliG(C) relative to FliG(M) in a single molecule was different among the three crystals, a conserved FliG(M)-FliG(C) unit was observed in all three of them. We suggest that the conserved FliG(M)-FliG(C) unit is the basic functional element in the rotor ring and that the PAA deletion induces a conformational change in a hinge-loop between FliG(M) and helix E to achieve the CW state of the FliG ring. We also propose a novel model for the arrangement of FliG subunits within the motor. The model is in agreement with the previous mutational and cross-linking experiments and explains the cooperative switching mechanism of the flagellar motor.


Introduction
Bacteria such as Escherichia coli and Salmonella enterica swim by rotating multiple flagella, which arise randomly over the cell surface. Each flagellum is a huge protein complex made up of about 30 different proteins and can be divided into three distinct parts: the basal body, the hook, and the filament. The basal body is embedded in the cell envelope and acts as a reversible motor powered by a proton motive force across the cytoplasmic membrane. The hook and the filament extend outwards in the cell exterior. The filament is a helical propeller that propels the cell body. The hook connects the basal body with the filament and functions as a universal joint to transmit torque produced by the motor to the filament. The flagellar motor can exist in either a counterclockwise (CCW) or clockwise (CW) rotational state. CCW rotation causes the cell to swim smoothly in what is termed a run, whereas brief CW rotation of one or more flagella causes a tumble. The direction of motor rotation is controlled by environmental signals that are processed by a sensory signal transduction pathway to generate chemotaxis behavior [1][2][3].
Five flagellar proteins, MotA, MotB, FliG, FliM, and FliN, are involved in torque generation. Two integral membrane proteins, MotA and MotB, form the stator, which converts an inwardly directed flux of H + ions through a proton-conducting channel into the mechanical work required for motor rotation. The FliG, FliM, and FliN proteins form the C ring on the cytoplasmic side of the MS ring, which is assembled from 26 subunits of a single protein, FliF, and this complex acts as the rotor of the flagellar motor [1][2][3]. An electrostatic interaction between the cytoplasmic loop of MotA and FliG is thought to be involved in torque generation [4,5] and in stator assembly around the rotor [6]. The protonation-deprotonation cycle of a highly conserved aspartic acid residue in MotB is coupled to the movement of the MotA cytoplasmic loop to generate torque [7][8][9].
Because FliG, FliM, and FliN are also responsible for switching the direction of motor rotation, their assembly is called the switch complex [10]. Binding of a chemotactic signaling protein CheYphosphate (CheY-P) to FliM and FliN is presumed to induce conformational changes in FliG that result in a conformational rearrangement of the rotor-stator interface, allowing the motor to spin in the CW direction [11,12]. The switching probability is also affected by motor torque, suggesting that the switch complex senses the stator-rotor interaction as well as the concentration of CheY-P [13,14]. Recently, turnover of FliM and heterogeneity in the number of FliM subunits within functioning motors have been reported [15,16]. The turnover rate is increased by the presence of CheY-P, implying that turnover of FliM may be directly involved in the switching process [15].
FliG forms a ring on the cytoplasmic face of the MS ring with 26-fold rotational symmetry [17,18]. FliG consists of three domains, FliG N , FliG M , and FliG C . FliG N is responsible for association with the cytoplasmic face of the MS ring [17,19], and FliG M and FliG C are required for an interaction with FliM [20]. The FliG M domains of adjacent subunits are fairly close to each other in the FliG ring [21]. The crystal structure of FliG MC of Thermotaoga martima (Tm-FliG MC ) shows that FliG M and FliG C are connected by an extended a-helical linker (helix E) [22]. The linker contains two well-conserved Gly residues and hence might be flexible [22]. This finding is supported by genetic analyses of FliG and a computer-generated prediction of its secondary structure [23,24]. Critical charged residues, which are responsible for an interaction with MotA [4][5][6], are clustered together along a prominent ridge on FliG C [25]. It has been shown that the elementary process of torque generation by the stator-rotor interaction is symmetric in CCW and CW rotation [26], although the torque-speed curves are distinct between them [27].
A recent report on the full-length FliG structure of Aquifex aeolicus has shown two distinct conformational differences between the full-length FliG and FliG MC structures [28]. The helix E linker is held in a closed conformation by packing tightly against an ahelix (helix n), which connects FliG N to FliG M in a way similar as helix E connects FliG M and FliG C in the full-length FliG structure. Helix E is dissociated from FliG M in the Tm-FliG MC structure, resulting in its being in an open conformation. The conformation of FliG C is also different in these two structures. Combined with the previous genetic data, it has been proposed that the closed conformation represents FliG during CCW rotation and that switching to CW rotation may be accompanied by the dissociation of helix E from FliG M to form an open conformation.
The S. enterica FliG(DPAA) mutant protein has three-amino-acid deletion at positions 169 to 171. Motors containing this protein are extremely CW biased [29]. The mutant motors remain in CW rotation even in the presence of a cheY deletion, indicating that the motor is locked in the CW state [29]. Therefore, it is likely that binding of CheY-P to FliM may introduce a conformational change in FliG similar to the one introduced by the in-frame PAA deletion. To elucidate the switching mechanism, we crystallized a fragment of a T. maritima FliG mutant variant, FliG MC (DPEV), which contains a deletion equivalent to S. enterica FliG MC (DPAA), and determined its structure at 2.3 Å resolution. Based on the structural difference among full-length A. aeolicus FliG, wild-type Tm-FliG MC , and its deletion variant, we suggest that a reorientation of helix E relative to FliG M is important for switching and propose a new model for the arrangement of FliG subunits in the motor.

Characterization of S. enterica fliG(DPAA) Mutant
The motors of the fliG(DPAA) mutant rotated only CW ( Figure  S1A), whereas wild-type motors rotated exclusively CCW under our experimental conditions. The motors of the deletion mutant produced normal torque under a wide range of external-load conditions, indicating that the deletion does not affect the torque generation step ( Figure S1B). Introduction of a cheA-Z deletion, which causes wild-type motors to spin exclusively CCW [30], into the fliG(DPAA) mutant did not change the CW-locked behavior. These results are in good agreement with a previous report [29].
Switching between the CW and CCW states is highly cooperative [31][32][33][34]. The switching mechanism can be explained by a conformational spread model, in which a switching event is mediated by conformational changes in a ring of subunits that spread from subunit to subunit via nearest-neighbor interactions [34,35]. Therefore we investigated rotation of a single motor composed of wild-type and mutant FliG subunits at different ratios. FliG(DPAA) inhibited expansion of wild-type colonies in semi-solid agar ( Figure 1A), even when its expression level was ca. 5-fold lower than the level of wild-type FliG expressed from the chromosome ( Figure 1B). Bead assays revealed that the decrease in colony expansion results from an increase in both switching frequency and prolonged pausing ( Figure 1C). In addition, a low level expression of FliG(DPAA) partially increased the colony expansion of the DcheA-Z smooth-swimming mutant, presumably because switching now occurred ( Figure 1D, upper and middle panels). These results suggest that even a small fraction of FliG(DPAA) in a motor can affect the CW-CCW switching.
The CW-CCW transition, which is very fast in wild-type motors, became significantly longer in mixed motors (Figure 1), suggesting that, as proposed previously [24], the motor can exist in multiple states. A much higher expression of FliG(DPAA) completely inhibited wild-type motility ( Figure 1D) and did not increase the colony size of the DcheA-Z mutant in semi-solid agar plates because of the extreme CW-biased rotation of its flagella ( Figure 1C and D, lower panel), in agreement with data showing

Author Summary
The bacterial flagellum is a rotating organelle that governs cell motility. At the base of each flagellum is a motor powered by the electrochemical potential difference of specific ions across the cytoplasmic membrane. In response to environmental stimuli, rotation of the motor switches between counterclockwise and clockwise, with a corresponding effect on the swimming direction of the cell. Switching is triggered by the binding of the signaling protein phospho-CheY to FliM and FliN, and achieved by conformational changes in the rotor protein FliG. The actual switching mechanism, however, remains unclear. In this study, we characterized a fliG mutant of Salmonella that shows an extreme clockwise-biased rotation, and determined the structure of a fragment of FliG (FliG MC ) of the equivalent mutant variant of Thermotoga maritima. FliG MC is composed of two domains and covers the regions essential for torque generation and FliM binding. We showed that the mutant structure has a conformational change in the helix connecting the two domains, leading to a domain orientation distinct from that of the wild-type FliG. On the basis of this structure, we propose a new model for the arrangement of FliG subunits in the rotor that is consistent with the previous mutational studies and explains how cooperative switching occurs in the motor.
that a higher expression level of wild-type FliG is required for complementation of the fliG(DPAA) mutant ( Figure S2). Therefore, we conclude that wild-type FliG is more stable in the CCW state than in the CW state, whereas FliG(DPAA) is more stable in the CW state than in the CCW state.

Limited Proteolysis of FliG and FliG(DPAA)
To identify structural differences between the CW and CCW states of FliG, we carried out limited trypsin proteolysis of the wild-type and mutant FliG proteins and analyzed the products by matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry and N-terminal amino-acid sequencing ( Figure 2). Both the wild-type and mutant FliG proteins were cleaved between helix E and FliG C , producing the T1 and T2a fragments. This indicates that there is a flexible region between them. The T1 fragment derived from FliG(DPAA) was less stable than the T1 fragment from wild-type FliG, suggesting that the deletion causes a conformational change in FliG M and helix E. In contrast, the T2a fragment was more stable in FliG(DPAA) than in the wild-type. The T2a fragment derived from the wild-type FliG protein was detected by MALDI-TOF but not on SDS-PAGE gels, indicating that the wild-type T2a fragment is rapidly converted into the T2 fragment. These results suggest that the deletion also influences the conformation in the region between helix E and FliG C .

Structural Comparison of Tm-FliG MC and Tm-FliG MC (DPEV)
We tried crystallizing both wild-type FliG and FliG(DPAA) from S. enterica but did not succeed in obtaining crystals. It has been  reported that the crystal structure of a fragment (residues 104-335) of T. martima FliG (Tm-FliG MC ) consists of FliG M , FliG C , and helix E connecting the two domains ( [22]; PDB ID, 1lkv). FliG C can be further divided into two sub-domains (FliG CN and FliG CC ). Therefore, we introduced the deletion (DPEV), equivalent to DPAA, into Tm-FliG MC (Tm-FliG MC (DPEV)) and determined its structure at 2.3 Å resolution by X-ray crystallography ( Figure 3).
FliG M , FliG CN , and FliG CC are composed of five (n, A-D), three (F-H), and six (I-N) helices, respectively ( Figure 3). Since the residues between G186 and V195 are invisible in the crystal, there are two possible ways to connect FliG M with FliG CN : one is to connect FliG M with its adjacent FliG CN (G186 to V195 in Figure 3A upper panel and Figure S3A), and the other is with a distant FliG CN (G186 to V195' in Figure 3A upper panel and Figure S3A). The Ca distance between G186 and V195, and G186 and V195' is 16.9 Å and 27.9 Å , respectively. Therefore, to connect with the distant FliG CN , the invisible chain would have a fully extended conformation. We thus conclude that the connection with the adjacent FliG CN is more plausible.
Compared with the structure of wild-type Tm-FliG MC , FliG(DPEV) showed a significant conformational change in the hinge between helix E and FliG M , leading to a very different orientation of helix E relative to FliG M ( Figure 3A and B, and Figure 4A and C). As a result, some of the residues in FliG M are exposed to solvent in the Tm-FliG MC (DPEV) structure. This result is in good agreement with the data obtained by limited proteolysis (Figure 2). Thus, the conformational difference in the FliG M -helix E hinge between the wild-type and mutant structures may represent the conformational switch between the CW and CCW states of the motor.
The C-terminal half of helix E is disordered and protrudes into the solvent channel in the Tm-FliG MC (DPEV) crystal ( Figure  S3A). In contrast, helix E in the wild-type crystal is stabilized by forming an anti-parallel four-helix bundle structure with the E helices of three adjacent subunits related by crystallographic symmetry ( Figure S3B) [22]. Therefore, the orientation of FliG C relative to FliG M is different between the wild-type and the deletion variants ( Figure 3A and B upper panel). Because the disordered region of helix E is far from the PEV deletion, we conclude that helix E has a highly flexible nature, which may be responsible for the switching mechanism, as suggested before [23,24].
Tm-FliG MC (DPEV) also showed a conformational difference in the H-I loop, resulting in a rigid body movement of FliG CC relative to FliG CN ( Figure 3A and B middle and lower panels, and Figure 4A). This movement is consistent with the limited proteolysis data because, in the Tm-FliG MC (DPEV) structure, FliG CC almost covers D199, which is the residue corresponding to R198 in S. enterica FliG. It is, however, unclear how the deletion affects the conformation of the H-I loop, because neither direct contact between FliG CC and helix E nor significant structural difference in FliG CN is observed.

Comparison of the Structure of Tm-FliG MC (DPEV) with A. aeolicus FliG
The crystal structure of full-length A. aeolicus FliG (Aa-FliG) showed that the conformation of helix E and the orientation of FliG CN relative to FliG CC are quite distinct from those of wild-type Tm-FliG MC [28]. We compared the Aa-FliG structure with the Tm-FliG MC (DPEV) structure and found that the conformation of helix E and the relative conformation of FliG CC to FliG CN are also different in those two structures ( Figure 3A and C, and Figure 4B and C). The conformational differences are greater than those between Tm-FliG MC and Tm-FliG MC (DPEV). The conformation of helix E in Aa-FliG seems to be stabilized by interactions of helix E with FliG M and helix n in the crystal ( Figure S3C). As mentioned earlier, the conformation of helix E and the orientation of FliG CC to FliG CN are also different between the wild-type and mutant Tm-FliG MC structures. Therefore, these conformational differences among the three structures strongly suggest that both helix E and the linker connecting FliG CN to FliG CC are highly flexible.

Interaction between FliG M and FliG CN
The interaction between FliG M and FliG CN , which share the armadillo repeat motif [36] that is often responsible for proteinprotein interaction, is very tight in the Tm-FliG MC (DPEV) crystal, in agreement with a previous report [28]. FliG M and FliG CN can be identified as a single domain, although it is unclear whether the two domains belong to the same molecule or not because the residues between Gly-186 and Val-195 are invisible in the crystal (Figures 3A and S3A). The interaction surface between FliG M and FliG CN is formed by the C-terminal portion of aB, aC, and aD of FliG M , and aF, aG, and the N-terminal portion of aH of FliG CN , respectively ( Figure 5A and B). The interface is highly hydrophobic. Ala-143, Ala-144, Leu-147, Leu-156, Leu-159, Ile-162, and Ala163 of FliG M , and Ile-204, Met-205, Leu-208, Ile-216, Leu-220, Leu-227, and Ile-231 of FliG CN are mainly involved in the tight domain interaction. Leu-159 is located at the center of the hydrophobic interface ( Figure 5C). Around the hydrophobic core, hydrophilic interactions between Arg-167 and Glu-230, and Gln-155 and Thr-212, also contribute to the domain interaction ( Figure 5C). These interactions are also conserved in the wild-type Tm-FliG MC and Aa-FliG crystals, in which FliG M interacts with FliG CN of an adjacent molecule related by crystallographic symmetry (Figures 3 and S3B). The FliG M -FliG CN unit in the wild-type Tm-FliG MC structure can be superimposed onto that in Tm-FliG MC (DPEV) with root mean square deviation of 0.46 Å for corresponding Ca atoms ( Figure 4A and C), and that in Aa-FliG with 0.79 Å ( Figure 4B and C). These observations support the idea that the FliG M -FliG CN unit is a functionally relevant structure [28]. This is in good agreement with the previous mutational study showing that most of the known point mutations that affect FliMbinding [37] are located either on the bottom surface of the FliG M -FliG CN unit or on the interaction surface between FliG M and FliG CN (Figure 6A and C).

Discussion
The default direction of the wild-type flagellar motor of Salmonella enterica is CCW, and the binding of CheY-P to FliM and FliN increases the probability of CW rotation. CheY-P binding induces conformational changes in FliM and FliN that are presumably transmitted to FliG, which directly interacts with MotA to produce torque [1,2]. Mutations located in and around helix E FliG, which connects the FliG M and FliG C domains, generate a diversity of phenotype, including motors that are strongly CW biased, infrequent switchers, rapid switchers, and transiently or permanently paused, suggesting that helix E is directly involved in the switching of the flagellar motor [24]. However, it remains unclear how helix E affects the switch.
To investigate the switching mechanism, we characterized an extreme CW-biased S. enterica mutant in which an in-frame deletion of three residues, Pro-169, Ala-170, and Ala-171, in FliG caused an extreme CW-biased rotation even in the absence of CheY. Motors containing the FliG(DPAA) protein showed normal torque generation under a wide range of external-load conditions ( Figure 1 and Figure 1S). Thus, the conformational change in FliG induced by DPAA is presumably similar to one induced by CheY-P binding to FliM and FliN. Limited proteolysis revealed that DPAA induces conformational changes in the hinge between FliG M and helix E (Figure 2). This result is in agreement with the crystal structure of Tm-FliG MC (DPEV), which shows that the orientation of helix E relative to FliG M has changed significantly compared to wild-type FliG (Figure 3).
FliG forms a ring on the cytoplasmic face of the MS ring [17,18]. In vivo disulfide cross-linking experiments using Cyssubstituted FliG proteins have suggested that helix A is close to the D-E loop of the adjacent FliG molecule in the FliG ring [21]. Both a conserved EHPQR motif in FliG M and a conserved surfaceexposed hydrophobic patch of FliG CN are important for the interactions with FliM [21]. Because the conserved charged residues on helix M in FliG CC are responsible for its interaction with MotA [4,5,25], which is embedded in the cytoplasmic membrane, helix M must lie on top of FliG CC [21,28]. Considering those facts in light of the crystal structure of Tm-FliG MC (DPEV) described here, we propose a new model for arrangement of FliG subunits in the motor (Figures 6 and 7).
In the proposed model, the conserved charged residues on helix M are located on the top of the FliG M -FliG C unit and the EHPQR  [21]. In fact, these residues are very close to each other in our model in positions in which disulfide-crosslinking should occur. Moreover, the position of Cys residues that do not participate in disulfide cross-linking are far from each other in the model ( Figure 6D).
Our model can also explain the results of mutational studies of CW and CCW-biased fliG mutants [37,38]. The mutation sites are widely distributed from helix A to the H-I loop. Most of them are localized in three regions in our model ( Figure 6A and B). In the first region, the CCW-biased mutations, which are located on helix A, affect residues close to residues targeted by CW-biased mutations, which are on a segment between helix D and E of the adjacent subunit ( Figure 6A and B, 1). Because these residues are distributed on the interaction surface between the neighboring subunits, they presumably affect cooperative changes in subunit conformation. A second cluster of residues targeted by CW-biased mutations is located on the C-terminal half of helix B and the E-F loop ( Figure 6A and B, 2). These mutations may change the orientation of the E-F loop and probably alter the orientation of helix E, resulting in unusual switching behavior. The third cluster of residues affected by mutations causing a CW switching bias is located near the loop between helices H and I ( Figure 6A and B,  3). This region determines the relative orientation of FliG CC to the FliGM-FliG CN unit, and therefore the mutations may change the orientation of FliG CC to cause anomalous switching behavior.
Helix E is directly involved in the switching mechanism, but how does the structure of helix E affect the orientation of the FliG M -FliG C unit? Since the D-E loop and helix E interact with FliG M in the neighboring subunit, we propose that a hinge motion of helix E may directly change the orientation of the neighboring FliG M domain ( Figure 7A). This mechanism could explain the cooperative switching of the motor. The conformational changes of FliM induced by association or dissociation of CheY-P may trigger conformational changes in the FliG M -FliG C unit that it contacts, leading to a large change in the interaction between FliG CC and MotA. The conformational change in one unit is probably accompanied by a conformational change in the loop between FliG M and helix E. This change could influence the orientation of the neighboring subunit through the interaction between helix E and FliG M of the neighbor, thereby propagating the conformational change to the neighboring subunit ( Figure 7A).
If helix E actually contacts the more-distant FliG CN in the crystal structure, an alternative interaction could be responsible for the cooperative switching ( Figure 7B). However, the same general mechanism involving changes in the conformation of helix E would still be responsible for the cooperative switching.
Recently, Lee et al. have proposed a model for FliG arrangement and switching based on the structural differences in Aa-FliG and Tm-FliG MC [28]. In the crystal structure of Aa-FliG, the hydrophobic patch in FliG M is covered by the Nterminal hydrophobic residues of helix E (closed conformation),  Figure 3. The region of the three-amino-acid deletion is shown by the magenta bar. The charged residues essential for the motor function are highlighted in cyan. The EHPQR motif is highlighted in green, and the other residues thought to be related to FliM-binding are shaded highlighted in yellow [21,37]. In vivo cross-linking experiments using various Cys-substitution mutants of FliG M have shown that residues indicated by blue arrows are located near the residues indicated by red ones. The Cys-substitution sites that did not show any cross-linked products are indicated by green arrows [22]. Blue and red boxes indicate point mutations that bias the motor rotation to CCW and CW, respectively [38]. The residues within magenta boxes can give rise to CCW or CW-biased mutants, depending on the substitutions. The numbers under the boxes represent the number of the cluster to which the indicated residues belong.  whereas the patch is exposed in Tm-FliG MC (open conformation). Because mutations that may disturb the hydrophobic interaction result in strong CW-bias in motor rotation [38], the structures of Aa-FliG and Tm-FliG MC are proposed to be in the CCW and CW states, respectively [28]. The hydrophobic patch is also exposed in the Tm-FliG MC (DPEV) structure, although the conformation of helix E is different from that of Tm-FliG MC . Since DPAA in S. enterica FliG (DPEV in T. maritima) caused an extreme CW-bias, it is possible that the dissociation of helix E from FliG M leads to CW rotation. In our model, however, the hydrophobic patch of the FliG M is covered by the hydrophobic residues in the C-terminal half of helix E of the adjacent subunit. This arrangement raises the possibility that the closed conformation of helix E found in the Aa-FliG structure is an artifact of crystal packing.
Lee et al. assume that the FliG M -FliG C unit is present in the rotor ring, and hence is in agreement with the results of most of mutational studies. However, the arrangement of the subunits and the mechanism of switching are different than in our model. In their model, dynamic motion of helix E and helix n induces a large conformational change of the FliG M -FliG C unit, including the rotation of FliG M -FliG CN unit and relative to the FliG CC to the unit, leading to a change in the arrangement of the charged residues on helix M ( Figure 7C) [28]. Cooperative switching is explained by the strong interaction between FliG CN of one subunit and FliG CC of the adjacent subunit. However, helix A of one subunit and the D-E loop of the adjacent subunit are always at a considerable distance in both the CW and CCW states. Hence, their model cannot explain the in vivo disulfide cross-linking experiments ( Figure 7C) [21]. Since our new model can explain the cross-linking data, it appears to be more plausible than the model proposed by Lee et al. [28].
Although our model is consistent with most of the previous experimental data, it still contains ambiguity. The available density map of the basal body obtained by electron cryomicroscopy is not high enough to allow fitting of the atomic model. Thus, a higher-resolution rotor-ring structure will be required to build a more precise model to explain the molecular mechanism of directional switching.

Materials and Methods
Bacterial Strains, Plasmids, and Media S. enterica strains and plasmids used in this study are listed in Table 1. L-broth, soft agar plates, and motility media were prepared as described [39,40]. Ampicillin was added to a final concentration of 100 mg/ml.

Motility Assay
Fresh colonies were inoculated on soft tryptone agar plates and incubated at 30uC.

Bead Assay for Motor Rotation
Bead assays were carried out using polystyrene beads with diameters of 0.8, 1.0, and 1.5 mm (Invitrogen), as described before [8]. Torque calculation was carried out as described [8].

Preparation of Whole Cell Proteins and Immunoblotting
Cultures of S. enterica cells grown at 30uC were centrifuged to obtain cell pellets. The cell pellets were resuspended in SDSloading buffer, normalized in cell density to give a constant amount of cells. Immunoblotting with polyclonal anti-FliG antibody was carried out as described [41].

Purification of His-FliG and His-FliG(DPAA) and Limited Proteolysis
His-FliG and His-FliG(DPAA) were purified by Ni-NTA affinity chromatography as described before [39]. His-FliG and its mutant variant (0.5 mg/ml) were incubated with trypsin (Roche Diagnostics) at a protein to protease ratio of 300:1 (w/w) in 50 mM K 2 HPO 4 -NaH 2 PO 4 pH 7.4 at room temperature. Aliquots were collected at 0, 5, 15, 30, 60, 90, and 120 min and trichloroacetic acid was added to a final concentration of 10%. Molecular mass of proteolytic cleavage products was analyzed by a mass spectrometer (Voyager DE/PRO, Applied Biosystems) as described [42]. N-terminal amino acid sequence was done as described before [42].
Purification, Crystallization, Data Collection, and Structure Determination of Tm-FliG MC (DPEV) Tm-FliG MC (DPEV) was purified as described previously [23]. Crystals of Tm-FliG MC (DPEV) were grown at 4uC using the hanging-drop vapor-diffusion method by mixing 1 ml of protein solution with 1 ml of reservoir solution containing 0.1 M sodium phosphate-citrate buffer pH 4.2-4.4, 36%-50% PEG200, and 200 mM NaCl. Initially, we tried to solve the structure by the molecular replacement method using Tm-FliG MC structure (PDB ID: 1 lkv) as a search model. However, no significant solution was obtained, even though individual domains were used as search models. Therefore, we prepared heavy-atom derivative crystals and determined the structure using the anomalous diffraction data from the derivatives.
Derivative crystals were prepared by soaking in a reservoir solution containing K 2 OsCl 6 at 50% (v/v) saturation for one day. Crystals of Tm-FliG MC (DPEV) and its Os derivatives were soaked in a solution containing 90%(v/v) of the reservoir solution and 10%(v/v) 2-Methyl-2,4-pentanediol for a few seconds, then immediately transferred into liquid nitrogen for freezing. All the X-ray diffraction data were collected at 100 K under nitrogen gas  [43] and scaled with SCALA [44]. Phase calculation was performed with SOLVE [45] using the anomalous diffraction data from Os-derivative crystals. The best electron-density map was obtained from MAD phases followed by density modification with DM [44]. The model was constructed with Coot [46] and was refined against the native crystal data to 2.3 Å using the program CNS [47]. About 5% of the data were excluded from the data for the R-free calculation. During the refinement process, iterative manual modifications were performed using ''omit map.'' Data collection and refinement statistics are summarized in Tables S1 and S2, respectively.