Mutations in the stator protein PomA affect switching of rotational direction in bacterial flagellar motor

The flagellar motor rotates bi-directionally in counter-clockwise (CCW) and clockwise (CW) directions. The motor consists of a stator and a rotor. Recent structural studies have revealed that the stator is composed of a pentameric ring of A subunits and a dimer axis of B subunits. Highly conserved charged and neighboring residues of the A subunit interacts with the rotor, generating torque through a gear-like mechanism. The rotational direction is controlled by chemotaxis signaling transmitted to the rotor, with less evidence for the stator being involved. In this study, we report novel mutations that affect the switching of the rotational direction at the putative interaction site of the stator to generate rotational force. Our results highlight an aspect of flagellar motor function that appropriate switching of the interaction states between the stator and rotor is critical for controlling the rotational direction.

www.nature.com/scientificreports/ F92, and L93 residues of PomA are located close to FliG (Fig. 1a, Fig. S1) 7 . The region containing these residues was located at the external and membrane-distal portion of the A subunit. This region is not directly involved in the ion conduction. The direction of rotation is controlled by the C-ring (also called the switch complex). The chemotaxis signaling protein, CheY, is phosphorylated via intracellular signaling pathways from chemoreceptors 23,24 . The phosphorylated form of CheY (CheY-P) binds to a well-conserved motif in the N-terminal region of FliM (FliM N ) [25][26][27][28][29] . CheY-P binding induces structural changes in the middle domain of FliM 30 . As a result, a structural change occurs at the FliG-stator interaction site, thereby switching the direction of rotation of the flagellar motor from CCW to CW 31 . After the dissociation of CheY-P from the C-ring, the rotational direction return to CCW. The number of stator units incorporating into the motor is dependent on the external mechanical loads 32,33 . Increase in the stator incorporation at the viscous load induces conformational shifts in FliM/N, resulting in modulating the binding affinity for CheY-P to the switch complex 34 . An increase in external load increases the binding affinity of the C-ring for CheY-P, suggesting stator-rotor interaction contributes to the switching of rotation direction of the flagellar motor. In V. alginolyticus, a single transmembrane protein called ZomB is also necessary for the CW rotation, depending on CheY 35 .
Cryo-electron tomography of intact flagellar motors in Borrelia burgdorferi and V. alginolyticus showed that the diameter of the C-ring expands during CW rotation 8,36 . The C-terminal region of FliG, at the top of the C-ring close to the membrane, appears to interact with the stator on its inner side facing the rotor axis in CCW rotation, whereas it appears to interact with the stator on its outer side away from the rotor axis in CW rotation (Fig. 1b). Therefore, the rotation direction of the flagellar motor is determined by changing the portion of the stator interacting with the rotor. Previous genetic studies have reported that mutations in the stator confer Mot − phenotype (non-motile) in many cases. In some cases, mutations in the stator confer Che − phenotype (no chemotaxis or inhibited switching of rotational direction) [37][38][39] . For example, MotA T21I/MotB A39V in E. coli confers CWbiased rotation and PomA R88A/K89A/E96Q/E97Q/E99Q in V. alginolyticus confers CCW-biased rotation. Therefore, the rotational direction is probably affected by conformational change in the stator or stator-rotor interface though it is mainly determined by the CheY-P binding or the conformational change of the C-ring.
In this study, we focused on the F92 and L93 residues of PomA that have been identified to be very close to FliG 7 . We also focused on the PomA-L95 residue, which is highly conserved and is located next to the conserved glutamate residue, an important charged residue involved in torque generation. The substitution of these three residues to a charged residue decreases the switching frequency of the motor and is biased toward CCW rotation. Furthermore, when combined with a CW-biased mutation in FliG, they inhibited the switching of the motor and stabilized the CW rotation, indicating that L93 and L95 of PomA are involved in efficient and robust switching of flagellar motor rotation, providing novel insights into the gear rotation model of the flagellar motor.

Residues of PomA mutations are involved in stator-rotor interaction.
A previous report showed that F92 and L93 of PomA are located very close to the C-terminal region of FliG in the rotor and suggested that these residues are involved in the interaction between the stator and rotor in addition to conserved charged residues 7 . These two hydrophobic residues are highly conserved among orthologs of stator A subunits, implying functional significance of hydrophobicity at this position (Fig. 1a, Fig. S1). Furthermore, PomA-L95 is also a highly conserved leucine residue (or the similar residue of isoleucine) and is located next to PomA-E96, which is a conserved and important charged residue for the motor function. Thus, we substituted these residues with alanine, glutamate, arginine, phenylalanine or threonine and analyzed motility of cells expressing the mutant PomA proteins. We first examined colony expansion in soft agar plates (Fig. 2a). Alanine or glutamate substitutions of PomA-F92 did not affect motility, but arginine substitution slightly affects motility, suggesting that hydrophobicity at this position is not required for motility. On the other hand, the glutamate-substituted mutant at the position of PomA-L93 severely decreased the size of swimming ring in the soft agar plate, and the argininesubstituted mutant moderately decreased the ring size suggesting that the charged side chain at this position reduces motility. In the substituted mutants at the position of PomA-L95, the alanine and phenylalanine mutants retained motility although it was not as the wild-type level, the threonine mutation moderately reduced the motility, and the arginine and glutamate mutants showed no motility in the soft agar plate. These results suggest that the motility is affected by the charged or hydrophilic side chains at this position.
We examined the protein expression levels of the mutant PomA using immunoblotting for anti-PomA and anti-PomB antibodies (Fig. 2b). The PomA-L95E and L95R mutant proteins were detected at a reduced level than the wild-type PomA produced from plasmid pHFAB induced by 0.02% arabinose. Since the chromosomal endogenous expression level of PomA can hardly detected by the anti-PomA antibody used in this study, we concluded that the mutant proteins expressed from plasmid were sufficent for the function in the Vibrio cells (Fig. 2b). This result suggested that the motility defects of the mutants in the soft agar plate were not due to the decrease in the expression level of the mutant proteins.
We next examined the effects of temperature on motility. Cells expressing wild-type PomA, L93E, or L95R were incubated at 20 °C, 30 °C or 40 °C in the soft agar plate (Fig. S2). Wild-type PomA conferred a larger swimming ring at 40 °C than at 30 °C. On the other hand, the L93E and L95R mutants conferred a smaller swimming ring at 40 °C than at 30 °C. At 20 °C, it takes longer time to detect the swimming abilities than at 40 °C and 30 °C probably because of the growth effects. The relative swimming abilities of wild-type PomA and mutant PomA were similar under each temperature condition. These results implied that the mutant PomA protein is more sensitive to high temperatures than the wild-type protein. www.nature.com/scientificreports/ www.nature.com/scientificreports/ PomA-L93X and L95X mutants conferred a rotationally direction-biased phenotype. We next characterized the motility of the cells expressing PomA-L93E, L93R, L95E, or L95R, as these mutants severely affected motility in the soft agar plate (Fig. 2a). We captured cell motion under the microscope, and then analyzed motility properties: motile fraction, swimming speed, switching frequency, and ratio of rotational direction (CCW/CW). These four pomA mutants showed a significant decreased motile fraction compared to wildtype PomA (Fig. 3a), suggesting that L93 and L95 may contribute to the stable interacion of stator-rotor in the motor. However, the swimming speed of these pomA mutants was not significantly different from that of wild-type PomA (Fig. 3b), indicating that the L93E, L93R, L95E and L95R does not affect torque generation by the flagellar motor. On the other hand, the ratio of the rotational direction of the PomA-L93E, L93R, and L95R mutants, but not of the L95E mutant, significantly biased in the CCW direction compared to that of wild-type PomA (Fig. 3c). Moreover, the switching frequency of all mutants was significantly decreased compared to the wild-type (Fig. 3d). Therefore, these results indicates that these mutations affect the switching event of the flagellar motor. However, the che function was not completely lost by these mutations because the response of phenol to the CW rotation of flagella remained active (Fig. 4). Therefore, we found novel mutations in the stator protein conferring the Che − phenotype. This result indicates that the mutations in the stator affect the switching of the rotational direction.
PomA CCW-biased mutants combined with CW-locked or CW-biased FliG mutant. PomA L93 and L95 mutants, as described above, exhibited motility properties biased toward the CCW direction. Next, we examined whether PomA mutations are dominant for switching the rotational direction in combination with CW-locked FliG-G215A mutation or CW-biased FliG-Q147H mutation. When PomA L93E or L95R was combined with FliG-G215A, these two mutations did not showed a CW-locked phenotype of the FliG-G215A mutant (Fig. 5), suggesting that the switching of the rotational direction is predominantly determined by the rotor. The FliG-Q147H mutant showed the CW biased rotation with CCW:CW = 1:9, but a few switching events occurred. In contrast, when PomA-L93E or L95R mutation was combined with FliG-Q147H, these two mutations resulted in more CW-biased rotation and a much lower switching frequency (Fig. 5). This result suggests that the mutation at the position of PomA L93 or L95 biases the rotational direction of the motor by reducing the switching frequency.

Discussion
The rotational direction of the flagellar motor is mainly controlled by CheY-P and the C-ring proteins, FliG, FliM, and FliN. In this study, we found that mutations in the stator protein PomA affect the switching frequency. These mutations were the residues involved in the stator-rotor interaction (PomA-L93) and the highly conserved hydrophobic residue (PomA-L95) adjacent to the important charged residue (PomA-E96). Since the positions of PomA-L93 and L95 are conserved as hydrophobic residues among various species, hydrophobicity may be important for motor function. In fact, PomA L93E, L93R, L95E, and L95R showed a significant decrease in the ring expansion by swimming in the soft-agar plate, indicating that these mutations affect motor function. Further microscopic observation of these pomA mutants revealed that these substitutions result in a significant decrease in the switching frequency, thereby causing a CCW-biased phenotype. This suggests that L93 and L95 of PomA are critical for efficient switching of rotational direction of the flagellar motor. Interestingly, these pomA mutations significantly reduced the fraction of swimming cells compared to wild-type cells, suggesting that L93 and L95 contribute to the stable interaction of stator-rotor in the motor. On the other hand, the L93E, L93R, L95E, and L95R mutations did not decrease swimming speed, suggesting that L93 and L95 are not directly involved in the torque generation. It is thought that the rotational direction of the flagellar motor is determined by the structural change of the C-ring. Cryo-electron tomography has revealed that a diameter of the top of the C-ring expands when the rotational direction is changed from CCW to CW 8,36 . By this structural change of the C-ring, it was hypothesized that the rotor interacts with the inner side of the stator facing the rotor axis during CCW rotation, whereas it interacts with the outer side of the stator during CW rotation (Fig. 1b). Atomic-resolution structures of the stator are represented as a barrel-like shape of the pentamer of the A subunits surrounding the dimer of the B subunits, proposing that the pentamer of the A subunit rotates in the CW direction around the dimer of the B subunits and the rotation is transmitted to the rotor C-ring 5,6 . Recent studies have speculated that stator-rotor interactions resemble a mechanical mechanism, similar to a spur gear [5][6][7][8]36 . In other words, the A subunits and the C-ring function like a small gear and a large gear, respectively. The teeth of the small gear (A subunits of the stator) located on the outer side and inner side of the motor engage during CCW rotation and CW rotation, respectively (Fig. 1b). The rotational direction is determined by which side of the stator gear, outer side or inner side, engages according to the change in the size of the large gear (the rotor).
In this study, we propose that the stator also plays some roles in rotational switching of the flagellar motor. The mutations at the position of PomA-L93 or L95 decreased the switching frequency and made CCW bias of the WT motor and CW bias of a CW-biased mutant motor stronger, indicating the suppression of the stator in switching its interaction surface of the rotor ring between the external and internal sides. This hypothesis is supported by the motility property exhibited by a combination of switching frequency-decreased PomA mutation and CW-biased FliG-Q147H mutation. The resultant mutants showed more CW-biased rotation by lowering the switching frequency than the FliG-Q147H mutant. This suggests that the mutation at the position of PomA-L93 or L95 suppresses the shift in the stator position to switch its rotor-interaction side between the inner and outer sides. It has been proposed in a recent report that the increase of force between stator and rotor modulates the binding affinity of the rotor against CheY-P 34 . Based on the proposal, the evidence, that the PomA mutants showed CCW-biased rotation and phenol significantly affected the switching frequency of the flagellar motor, www.nature.com/scientificreports/   www.nature.com/scientificreports/ implies that the PomA mutations modulate the force of the stator-rotor interaction, therefore, the binding affinity of the rotor for CheY-P may change.
How are PomA-L93 and L95 involved in switching rotor gears? When PomA-L93 was substituted with charged residues, the affected switching implies that the charged residues may participate in the electrostatic interaction with FliG in the rotor, together with PomA-R88, K89, and E96, which are thought to be the residues involved in the stator-rotor interaction. Because the position of PomA-L95 is located inside the stator structure, the charged residues around this position may change the arrangement of PomA R88 and E96. Interestingly, our previous study showed that PomA R88A/K89A/E96Q/E97Q/E99Q conferred CCW-biased rotation to the motor 39 . These results suggest that the proper number and location of the charged residues, which may define the stator-rotor interaction, compensates proper switching of the rotational direction. Moreover, to achieve the proper electrostatic interaction, the conserved hydrophobic residues may act as effective insulators.
We tried to obtaine swimming mutants in the soft agar plate from the cells expressing PomA-L93E, L95E, or L95R, to identify the factors involved in switching frequency of motor rotation (Figs. S3, S4, S5). Although we had expected that a suppressor mutation is caused in fliG, fliM or fliN, which are the components of the C-ring or rotor, we could not obtained such suppressor mutants. Instead, putative suppressor mutations were caused in genes encoding F o F 1 -ATP synthase, a gene encoding ATPase of the Type II secretion system and so on (Table S2). In the previous studies, we have shown that mutations in uncB, which encodes the α-subunit of ATP synthase, induced increased motility in soft agar plates 40,41 . The defects in ATP synthase seem to increase the proton gradient generated by the respiratory chain and eventually to increase the sodium-motive force. Therefore, the swimming ring-restored mutants whose mutation was mapped in the component of ATP synthase might allow motility to be restored by increasing the ion motive force, but not by affecting directly the interaction between stator and rotor.
In summary, we provide a genetic evidence that PomA mutations located at a physical interface between stator and rotor inhibit the switching of flagellar motor rotation. Even though the electrostatic interaction becomes either too strong or too weak, shifting the interaction site between the stator and rotor may become difficult. This finding suggests that proper switching of the interaction between the stator and rotor is also important for controlling the rotational direction.

Materials and methods
Bacterial strains and plasmids. The bacterial strains and plasmids are listed in Table S1 Mutagenesis. Site-directed mutagenesis was performed using the QuikChange site-directed mutagenesis method, as described by Agilent Technologies (Santa Clara, USA). Transformation of V. alginolyticus by plasmids pHFAB or pYA303 was carried out by electroporation 42 . Transformation of V. alginolyticus by plasmid pNT1 was carried out by conjugational transfer from E. coli S17-1 43 .

Measurement of swimming speed and motile fraction. V. alginolyticus cells, cultured overnight,
were inoculated into fresh VPG broth with antibiotics and incubated by shaking at 30 °C for 4 h. The cells were suspended at a 100-fold dilution in V buffer (50 mM Tris-HCl pH 7.5, 300 mM NaCl, 5 mM MgCl 2 ), and then spotted onto glass slides for microscopic observation. The motile fraction was calculated by dividing the number of motile cells by the total number of cells. All experiments were performed three times, and the averages of the motile fraction and standard deviation (SD) were calculated. To measure the swimming speed of an individual cell, the cell was suspended at a 100-fold dilution in V buffer, and serine was added at a final concentration of 5 mM to allow the cell to swim straight. The motion of the cells was captured at 30 frames/s for 5 s under a microscope. The swimming speed of individual 10 cells was measured using ImageJ software (Rasband W.S., Bethesda, USA), and the average swimming speed and SD were calculated. Measurement of switching frequency and ratio of CCW/CW rotation. V. alginolyticus cells, cultured overnight, were inoculated into fresh VPG broth with antibiotics and incubated by shaking at 30 °C for 4 h. The cells in the 200 μL culture were collected by centrifugation at 2000×g for 5 min. The precipitated cells were resuspended in 200 μL of V buffer. The cells were then spotted onto glass slides for high-intensity dark-field microscopic observation. The motion of the cells was captured at 30 frames/s for 10 s under a microscope. The number of frames of either CCW or CW rotation was recorded, and the ratio of CCW/CW rotation and switching events was calculated for 1 s using ImageJ software.