The role of conserved charged residues in the bidirectional rotation of the bacterial flagellar motor

Abstract Many bacteria rotate their flagella both counterclockwise (CCW) and clockwise (CW) to achieve swimming toward attractants or away from repellents. Highly conserved charged residues are important for that motility, which suggests that electrostatic interactions are crucial for the rotor–stator function. It remains unclear if those residues contribute equally to rotation in the CCW and CW directions. To address this uncertainty, in this study, we expressed chimeric rotors and stators from Vibrio alginolyticus and Escherichia coli in E. coli, and measured the rotational speed of each motor in both directions using a tethered‐cell assay. In wild‐type cells, the rotational speeds in both directions were equal, as demonstrated previously. Some charge‐neutralizing residue replacements in the stator decreased the rotational speed in both directions to the same extent. However, mutations in two charged residues in the rotor decreased the rotational speed only in the CCW direction. Subsequent analysis and previous results suggest that these amino acid residues are involved in supporting the conformation of the rotor, which is important for proper torque generation in the CCW direction.

Second, the torque-speed relationship of various flagellar motors is similar. This relationship has been determined for E. coli (Chen & Berg, 2000) and V. alginolyticus (Sowa, Hotta, Homma, & Ishijima, 2003), and for motors driven by the chimeric stator, PomA/PotB in E. coli (Inoue et al., 2008). The absolute values of maximum torque and maximum speed are different for each motor, but the overall trend is the same.
Finally, the functionally important charged residues needed for rotation are highly conserved across bacterial species (Figures 1b,c, F I G U R E 1 The stator and the rotor of the flagellar motor. (a) The flagellar motor at the bacterial membrane. PomA (MotA), indicated in white, and PomB (MotB), indicated in light gray, form the stator complex. PomA (MotA) interacts with the rotor FliG (indicated in dark gray), which forms the membrane-proximal part of the C ring, to rotate the flagellum. CM, PG, and OM represent cytoplasmic membrane, peptidoglycan layer, and outer membrane, respectively. (b) Sequence alignment of amino acid residues of PomA (MotA), the component of the stator from Vibrio alginolyticus and Escherichia coli. Important charged residues for motility are indicated by filled arrowheads and shown in black boxes. The numbering of the amino acid residues on the tops of the sequences is from V. alginolyticus. (c) Sequence alignment of amino acid residues of FliG, the component of the rotor from V. alginolyticus and E. coli. Important charged residues for motility are indicated as described in (b). The position of the junction in the FliG EV chimera is indicated by an arrow at the top of the sequences. (d) Interface of the rotor-stator interaction. Highly conserved charged residues that are important for motility are indicated. The numbering of the amino acid residues is from V. alginolyticus. Black lines indicate interactions required for motor function, and gray lines indicate additional interactions reported in Vibrio. (e) Structure of the C-terminal domain of FliG from Thermotoga maritima (Protein Data Bank code: 1LKV) (Brown et al., 2002). The corresponding positions of the mutated residues in this study are shown as balls and labeled as the amino acid residues in V. alginolyticus. In the chimeric rotor protein FliG EV , the portions from E. coli and V. alginolyticus are shown in black and gray, respectively and S1). The electrostatic interactions between the charged residues in the C-terminal domain of FliG and the cytoplasmic domain of MotA (or PomA) are crucial for rotor-stator function (Attmannspacher, Scharf, & Schmitt, 2005;Morimoto, Nakamura, Hiraoka, Namba, & Minamino, 2013;Takekawa, Kojima, & Homma, 2014;Yakushi, Yang, Fukuoka, Homma, & Blair, 2006;Zhou et al., 1998). Specifically, in E. coli, R90 (R88 in V. alginolyticus) and E98 (E96) in MotA, and R281 (R301), D288 (D308), and D289 (D309) in FliG are the residues of primary importance, whereas E150 (D128) in MotA, and K264 (K284) and R297 (R317) in FliG are of secondary importance (Lloyd & Blair, 1997;Zhou & Blair, 1997). Interactions between R90 (R88) in MotA and D289 (D309) in FliG and between E98 (E96) in MotA and R281 (R301) in FliG are important for motor function (Zhou et al., 1998). Furthermore, these interactions have distinct functions in Salmonella enterica  are important for motor function (Takahashi & Ito, 2014). In V. alginolyticus, additional charged residues collectively participate in torque generation (Takekawa et al., 2014).
This model for the rotational switch implies that, after the rotational switch, the interface of the rotor-stator changes and the contribution of each charged residue that is important for rotation in the CCW direction is different from that in the CW direction. However, most studies have focused on the rotational switch itself (Bai et al., 2010;Wang, Yuan, & Berg, 2014) rather than on the mechanisms of CCW versus CW rotation, although it is known that the flagellar motor rotates stepwise similarly in both directions (Nakamura, Kami-ike, Yokota, Minamino, & Namba, 2010), and that the overall trend of the torque-speed relationship depends on the rotational direction (Yuan, Fahrner, Turner, & Berg, 2010).
In this study, we used site-directed mutagenesis to investigate if the functional roles of these conserved charged residues are symmetric (i.e., that their functions are independent of the rotational direction) or asymmetric (i.e., that their functions are dependent on the rotational direction). Our results show that some substitutions of charged residues in the rotor significantly affect rotation in only one direction, and we discuss possible differences in the roles of those residues in the generation of rotation.
Bacteria were cultured in LB medium, 1% (w/v) Bacto ™ Tryptone, 0.5% (w/v) yeast extract, and 0.5% (w/v) NaCl, overnight at 37°C from freezer stocks. These cultures were diluted 100-fold into fresh TB medium, 1% (w/v) Bacto ™ Tryptone and 0.5% (w/v) NaCl, with 0.02% (w/v) arabinose. After culture for another 4 h at 30°C, these bacteria were used for further analyses. We first observed the rotation of bacteria for ~30 s under unstimulated conditions. Then, we infused 50 μl Rotation Buffer with 10% (w/v) glycerol to cause cells to rotate in the CW direction and observed them for ~90 s. Next, we infused 50 μl Rotation Buffer with 10 mM serine and 10% (w/v) glycerol to cause them to rotate in the CCW direction and observed them for another ~90 s. These observations were recorded on a PC using a WV-1550 CCD camera (National) and PowerDirector ® software (CyberLink) at 30 frames/s.

| Data analysis
The rotational speeds of the cells were measured by replaying the movies using Move-tr/2D software (Library Co.). We selected cells that rotated in both directions continuously, counted the number of frames in which the cells rotated for 10 (or 5) rounds, and calculated their rotational speeds. We fitted the CCW-CW speed data to a straight line that passes through the origin. The slope indicates the average ratio of the rotational speed in the CW direction to the rotational speed in the CCW direction.

| Multiple sequence alignment
Multiple amino acid sequence alignment was performed using ClustalW. In the case of the alignment of the stator, some sequences around a gap were modified manually.

| Motile fraction and swimming speed
Cultured bacteria were diluted in TB medium at 100-fold and observed using a dark-field microscope (BH-2, Olympus) with a 10× objective (DPlan 10, numerical aperture 0.25, Olympus). Movies were recorded as described above. The motile fractions and swimming speeds of bacteria were analyzed using Move-tr/2D software (Library Co.

| Symmetric rotation in both directions
Our ultimate goal is to understand the rotational mechanism of the Vibrio flagellar motor. Unfortunately, however, it is difficult to observe its rotation directly with traditional methods because Vibrio cell has a single polar flagellum at its cell pole (Allen & Baumann, 1971). Therefore, we used PomB and the C-terminal region of E. coli MotB (Figure 2a), which can function with V. alginolyticus PomA as a Na + -driven stator in E. coli (Asai et al., 2003). When these chimeric proteins were coexpressed in E. coli, the bacteria spread on soft-agar plates (Figure 2b). Observing individual cells in a liquid environment under an optical microscope, we found that the fraction of motile cells was similar to that of cells driven by E. coli motors, whereas the swimming speed was higher (Figure 2c,d).
To examine the functions of the conserved charged residues in the stator and rotor components of the flagellar motor, we analyzed the rotational dynamics in both directions using a tethered-cell assay. We used glycerol and serine as a repellent and an attractant, respectively, to control the flagellar rotational direction (Mesibov & Adler, 1972;Oosawa & Imae, 1983). After exchanging buffers in the observation chamber from repellent (10% glycerol) to attractant (10 mM serine), we measured the rotational speeds in the CW and CCW directions for each cell. As shown in Figure 3, the rotational speeds in each direction vary widely because of differences in cell size and the position of the tethering flagellum along the cell cylinder. If the torque is constant in each direction and its rotational frictional drag coefficient is the same in both directions, the plots in the CCW-CW curve must be linear. Thus, we analyzed the ratio of CW speed to CCW speed by fitting those data to a straight line.
Variations from this line are presumably due to loose tethering, which allows for different rotational geometries in the two directions during the solution exchange. The correlation coefficient was relatively high, which supported the fitting (Table 1)

| Mutations in the stator
We introduced charge-neutralizing mutations into the stator to examine the importance of the charged residues. If the mutations affected the rotational speed, there would be two possible explanations. In the first case, the slope of the CCW-CW speed curve would be close to 1, as with the wild-type motor. This would mean that the mutation af-  Figure 4a. For all six mutants, the slope of the fitted line is almost one, and is similar to that of the wild-type motor (Table 1). Thus, the motors containing each charge-neutralized residue substitution function similarly in both rotational directions. In other words, the functions of the charged residues of the stator are symmetric. To distinguish if those mutations affected the rotational function equally in both rotational directions or had no effect, we analyzed the rotational speeds of these bacteria. All of the strains with single charge-neutralized residue substitutions in the stator rotated at a comparable speed in both directions as cells expressing the original chimera (Table 1). This result suggests that none of the single mutants of the stator affected rotation.
PomA, the stator component from Vibrio species, has additional charged residues compared to MotA from E. coli; we had previously suggested that these charged residues function redundantly (Takekawa  (Table 1). For the double and triple mutants, the slope of the CCW-CW speed plot was close to 1 (Figure 4b and Table 1). Additionally, the average rotational speed of the AA mutant was reduced slightly in both rotational directions (Table 1). Taken together, these results confirm that the charged residues are redundant for flagellar rotation and contribute equally to torque generation in both rotational directions.
The positions of the residue substitutions are conserved charged residues in V. alginolyticus. R301, D308, and D309 are located in an alpha helix in the C-terminal domain of FliG, and K284 and R317 are located beside the helix and close to each other (Figure 1e).
For bacteria with 3 of those five mutations in the rotor, the slope of the CCW-CW speed curve was the same as for wild-type cells ( Figure 5a and Table 1). Bacteria with the other two mutations, K284A and R317A, had a steeper slope than the wild-type cells. The rotational speeds of those two mutants were reduced significantly only in the CCW direction compared to the speeds of the wild-type strain and other mutants. These results suggest that the K284A and R317A mutations selectively impaired the rotational function in the CCW direction.
We then investigated the effect of charge-reversing mutations on K284 and R317 of FliG. As determined by tethered-cell assay, cells expressing the K284E and R317D mutations also exhibited a steeper CCW-CW speed slope and reduced average rotational speed in the CCW direction, but not in the CW direction ( Figure 5b and Table 1). Therefore, the positive charge in these residues is essential for rotational function selectively in the CCW direction.

| Dual mutations in both the stator and the rotor
We have comprehensively investigated the effects of dual mutations in the stator and the rotor of Vibrio on motility, and found several interactions between the stator and the rotor (Takekawa  (Figure 5a,b). In both dual mutants, we observed steep slopes and reduced rotational speeds in the CCW direction in comparison with wild-type motors (Figure 6a,b and Table 1), as we found for the single mutations in the rotor ( Figure 6c).

| DISCUSSION
In this study, we investigated the effects of the charged residues in the rotor and stator components of the flagellar motor, taking into account their effects on rotation in the CCW and CW directions (Figures 4, 5, and Table 1). Those residues are highly conserved and have been suggested to be important for the motility of many bacteria.
We  In the current model of the rotational switch of the flagellar motor, it is thought that FliG changes its conformation to orient the position of its C-terminal domain (Lam et al., 2012;Lee et al., 2010;Lloyd et al., 1999;Minamino et al., 2011). FliG relative to the base of the C ring is supported by a cross-linking experiment (Paul, Brunstetter, et al., 2011. We found that residue substitutions in two residues, K284 and R317, of the rotor protein FliG impaired rotational function selectively in the CCW direction. These results strongly suggest that the rotor, FliG, changes the interactionsite with the stator depending on the rotational direction. Two residues, K284 and R317, in the rotor have strong asymmetric effects on rotation. Moreover, the interaction of these rotor residues with the stator does not affect their asymmetric function ( Figure 6).
We interpret that to mean that K284 and R317 do not directly associate with torque generation. K284 of FliG appears in two conformations in the crystal structure of FliG. In one conformation, the side chain of K284 is exposed to the solvent and, in the other conformation, it is less exposed, as previously pointed out by Brown, Hill, and Blair (2002). In the crystal structure of FliG from T. maritima, the side chain of K266 (the equivalent of K284 in V. alginolyticus) is hydrogenbonded to the backbone carbonyl groups of residues L292 and G296 of FliG (Brown et al., 2002;Lloyd et al., 1999). In the structure of the complex of FliG and FliM from T. maritima, the side chain of K266 is hydrogen-bonded to the carbonyl group of L292 but not to that of G296 (Vartanian, Paz, Fortgang, Abramson, & Dahlquist, 2012). We predict that this subtle conformational change occurs during the rotational switch, which is consistent with our data. The side chain of K284A cannot be hydrogen-bonded to the main chain and alters the conformation required for CCW rotation, resulting in reduced torque generation in the CCW direction. Furthermore, we recently found that CW-biased rotation of a mutant of FliG, A282T, is caused by an additional hydrogen-bond to the main chain in the C-terminal domain of FliG (Miyanoiri et al., 2017). This result also implies that intramolecular interaction in FliG is important for rotational function. In contrast, the side chain of R317 is exposed freely to solvent in the crystal structure. However, substitution of cysteine for this residue induces crosslinking in the motor, implying that the residues in adjacent subunits can approach close enough to form a disulfide cross-link (Lowder, Duyvesteyn, & Blair, 2005). Thus, we propose that this residue has a role in stabilizing the conformation of FliG in the C ring of the motor in the CCW direction.
The results of this study raise two interesting questions. Why does neutralization of the charged residues in the stator not have an asymmetric effect on rotation, and why is the CW rotation unaffected by residue substitution of any of the charged residues?
Principally, charged residues must be involved in torque generation in both rotational directions, as discussed above. One possibility is that there are other, as yet unidentified, charged residues in the stator and/or the rotor that participate in torque generation. The R232E mutation in the chimeric stator, PomA/PotB, reduces the swimming ability of E. coli (Inoue et al., 2008;Yakushi et al., 2006). R232 is located in the cytoplasmic C-terminal domain of PomA, where there are many charged residues that are important for motor function (Obara, Yakushi, Kojima, & Homma, 2008). Although it is difficult to interpret the role of the charge of a histidine residue in proteinprotein interactions because its pK a value is near neutral pH, the residue substitution H136Y of PomA from V. alginolyticus impairs its rotational function (Fukuoka, Yakushi, & Homma, 2004;Takekawa et al., 2013). These residues might be candidates for charged residues that directly participate in torque generation.
Another possibility is that electrostatic interactions play a minor role in torque generation, especially in the CW direction. The motor from R. sphaeroides alternates between rotation in the CCW direction and cessation of rotation, even though its stator and rotor retain all of the highly conserved charged residues (Pilizota et al., 2009). On the other hand, S. meliloti has right-handed rather than left-handed helical flagellar filaments that rotate only CW (Götz & Schmitt, 1987), even though they change speed. One group reported that the motor of Caulobacter crescentus generates more torque in CCW rotation than in CW rotation (Lele, Roland, Shrivastava, Chen, & Howard, 2016). Interestingly, these three bacteria belong to Alphaproteobacteria. Therefore, despite the conservation of the charged residues across a wide range of species, the exact mechanisms that support CCW and CW rotation remain to be determined; they must differ among species and be related to their evolutionary histories.