Precise Measurement of the Stoichiometry of the Adaptive Bacterial Flagellar Switch

ABSTRACT The cytoplasmic ring (C-ring) of the bacterial flagellar motor controls the motor rotation direction, thereby controlling bacterial run-and-tumble behavior. The C-ring has been shown to undergo adaptive remodeling in response to changes in motor directional bias. However, the stoichiometry and arrangement of the C-ring is still unclear due to contradiction between the results from fluorescence studies and cryo-electron microscopy (cryo-EM) structural analysis. Here, by using the copy number of FliG molecules (34) in the C-ring as a reference, we precisely measured the copy numbers of FliM molecules in motors rotating exclusively counterclockwise (CCW) and clockwise (CW). We surprisingly found that there are on average 45 and 58 FliM molecules in CW and CCW rotating motors, respectively, which are much higher than previous estimates. Our results suggested a new mechanism of C-ring adaptation, that is, extra FliM molecules could be bound to the primary C-ring with probability depending on the motor rotational direction. We further confirmed that all of the FliM molecules in the C-ring function in chemotaxis signaling transduction because all of them could be bound by the chemotactic response regulator CheY-P. Our measurements provided new insights into the structure and arrangement of the flagellar switch.

molecules with alterable numbers of FliM and FliN molecules. Here, we investigated the stoichiometry and chemotactic function of FliM in CCW-and CW-locked motors of E. coli by fluorescence labeling, obtaining an unexpected arrangement of the C-ring subunit.

RESULTS
Robustness of the FliG copy number in functional motors. The strains and plasmids used in this study are listed in Table 1 (see Materials and Methods for details). By using the isopropyl-b-D-thiogalactopyranoside (mEGFP)-inducible plasmid pTrc99a, monomeric enhanced green fluorescent protein (mEGFP) (i.e., EGFP A206K ) was genetically fused to the N terminus of FliG, yielding mEGFP-FliG. The strain AT1 (DfliC, fliG) carrying two plasmids pAT3 (pTrc99a-megfp-fliG) and pKAF131 (pACYC184-fliC sticky ) was used for tethered-cell experiments (Fig. 1A). For the maturation of mEGFP molecules, motile cells were preincubated at 37°C for 45 min in motility buffer to minimize the additional expression of mEGFP molecules and restrict cell growth (36). Therefore, the concentration of mEGFP molecules should be roughly constant during incubation. The motor fluorescence was recorded under total internal reflection fluorescence (TIRF) illumination for 2 s, after which the intracellular fluorescence of the same cell was captured under epifluorescence illumination. The fluorescence intensity of individual motors for each frame was calculated, and the background fluorescence intensity was subtracted (see Materials and Methods for details). The final motor intensity was calculated as the mean value for the initial 5 frames (Fig. 1B). The intracellular fluorescence density, defined as the ratio of the net intracellular fluorescence to the volume of the cell, has been confirmed to be approximatively proportional to the intracellular concentration of fluorescent proteins in individual cells (37). The motor fluorescence intensity was nearly constant regardless of the changes in the intracellular fluorescence density (Fig. 1C), confirming the robustness of the FliG copy number in functional motors at various FliG expression levels. Therefore, the FliG copy number was restricted by the intrinsic structure of the motor, namely, the sequential assembly of the C-ring around the MS-ring (containing 34 FliF) rather than the intracellular concentration of FliG molecules. Our measurements were consistent with the previous finding that the stoichiometry of FliG in a functional flagellar motor has a mean value of 34, the same as the copy number of FliF.
FliM copy numbers in CCW-and CW-locked motors. We aimed to measure the FliM copy numbers in individual motors rotating exclusively CCW or CW by utilizing the fluorescence intensity of the motor composed of mEGFP-FliG as a reference. Monomeric EGFP was genetically fused to the C terminus of FliM, yielding FliM-mEGFP. The strain GL2 (DfliC, fliM, cheY) carrying pAT2 (pTrc99a-fliM-megfp) and pKAF131 was used for observing CCW-locked motors of tethered cells. For CW-locked motors, we constructed the plasmid pAT4 to simultaneously express the sticky phenotype of FliC and the mutant CheY 13DK106YW that was constitutively active even without phosphorylation (38). GL2 was transformed with pAT2 and pAT4, and excess arabinose was added to induce the overexpression of the CheY 13DK106YW mutant for CW-locked motors. First, by performing tethered-cell experiments, the photobleaching profile of individual motors containing mEGFP-FliG molecules was fitted to a single exponential function (39) as follows: where I(t) is the fluorescence intensity of a motor at laser exposure time t, k is the photobleaching rate constant, and I 0 is the initial intensity of the motor. The distribution of initial intensities (I 0 ) for the mEGFP-FliG motors was fitted to a Gaussian curve to obtain the average intensity (I G ), which corresponded to the overall brightness emitted from all of the mEGFP-FliG molecules in the C-ring. Subsequently, under identical illumination Stoichiometry of the Bacterial Flagellar Switch mBio and imaging conditions, the fluorescence intensities of CCW-and CW-locked motors composed of FliM-mEGFP molecules were recorded. For each photobleaching profile, the number of FliM molecules (N FliM ) was estimated by where I 0 is the fitted initial intensity, and 34 is the estimated average number of FliG subunits in the C-ring. Surprisingly, the copy numbers of FliM-mEGFP in CW-and CCWlocked motors exhibited a Gaussian distribution peaking at 45 6 11 and 58 6 11 (errors denote standard deviation [SD]), respectively (Fig. 2), which were both larger than the copy number of FliG molecules in the C-ring. The ratio of the FliM copy numbers in CCW-versus CW-locked motors was ;1.29, which was in good agreement with the previous measurement (20). We noted that the previously measured FliM copy numbers in the CCW-and CW-locked motors were ;44 and 34, respectively, on the Stoichiometry of the Bacterial Flagellar Switch mBio assumption that FliM in CW-locked motors had a copy number of 34 (20). CW-locked motors contain fewer FliM molecules than CCW-locked motors because of the adaptive remodeling of the motor (21). We further confirmed that the brightness emitted from a mEGFP cluster was comparable whether fused to the N terminus or the C terminus of the target protein by comparing the mean intracellular fluorescence density for mEGFP-FliG induced by 130 mM IPTG and FliM-mEGFP induced by 100 mM IPTG, obtaining a ratio in agreement with the value from Western blot analysis (further details in Text S1 in the supplemental material). In addition, using the strain GL1 (DfliC, fliM) transformed with pAT2 and pKAF131, we performed tethered-cell experiments to eliminate the possible dependence of the quantum efficiency of FliM-mEGFP molecules on the motor switch state (see Fig. S3 in the supplemental material).
Numbers of CheY-P molecules binding to a CW spinning motor. FliM molecules of a functional motor play a crucial role in the chemotaxis signaling pathway because binding of the response regulator CheY-P to FliM promotes CW rotation of the motor (14,40,41). Our measurements suggest that there are more FliM molecules in functional motors than previous estimation, raising the question of whether all of the FliM molecules could be bound by CheY-P molecules. Here, we used the strain XL2 (DfliC, cheY, cheB, cheZ) carrying pAT1 (pBAD33-cheY-megfp) and pFD313 (pBR322-fliC sticky ) to observe the binding of CheY-P-mEGFP to a rotating motor under identical illumination and imaging conditions with measurements of the motor fluorescence intensity of mEGFP-FliG and FliM-mEGFP. This strain promotes the phosphorylation of CheY and inhibits the dephosphorylation of CheY-P; hence, all of the intracellular CheY-mEGFP molecules are essentially phosphorylated with an analogous physiological function to wild-type (WT) CheY-P (42,43). The motor CW bias was analyzed with a 1-min time sequence of bright-field images. Then, the motor fluorescence of individual tethered cells was recorded under TIRF illumination for 5 s, after which the intracellular fluorescence of the same cell was captured under epifluorescence illumination. The profile of the CW bias versus intracellular fluorescence density fell onto a sigmoid curve with a Hill coefficient of ;9.6 ( Fig. 3A), demonstrating a similar ultrasensitivity to a previous measurement using a bead assay (42). The off-rate of CheY-P bound to FliM was estimated to be .15 s 21 (44), which was more than 750 times larger than the off-rate (;0.02 s 21 ) of FliM bound to the C-ring (20). Quickly exchanging components, such as CheY-P bound to the C-ring, are not suitable for the photobleaching fit method (18,22,39). Furthermore, the duration of a motor staying in the CW state is sometimes shorter than the photobleaching time of motor intensity, especially for motors with CW bias close to 0. Therefore, instead of the photobleaching fit using the CW frames, TIRF movies were manually analyzed by eye to find the initial CW interval longer than 5 frames, and the fluorescence intensity of CheY-mEGFP bound to the motor was calculated as the mean value for the 2nd to 4th frames of the CW interval. We performed a direct comparison between the intensities of motors containing mEGFP-FliG found by the two methods as follows: one is the photobleaching fit, and the other is the mean of the initial frames. As shown in Fig. S4 in the supplemental material, both methods lead to equivalent results. Moreover, we also detected the number of photons emitted from the three types of mEGFP fusions used in the present study at the single-molecule level. We found that individual mEGFP molecules emitted the same number of photons regardless of the type of fusion (see Text S1 and Fig. S5 in the supplemental material). This provided additional support for reliable counting by using different mEGFP fusions. Using the same reference as N FliM (equation 2), the motor fluorescence intensity was used to estimate the number of CheY-P molecules bound to the motor, which was then plotted versus the corresponding intracellular fluorescence density (Fig. 4B). Note that the intracellular fluorescence should be corrected by the fraction of bleached mEGFP during the TIRF illumination; however, applying this correction factor only results in a scale-up of horizontal coordinates that are originally in arbitrary units, thus neglecting that the correction should not affect the presented information of the profile. With the increase in CW bias from 0 to 1 as the intracellular fluorescence density increases, the number of CheY-P molecules bound to a CW spinning motor increases rapidly and nearly saturates when the intracellular fluorescence density falls into the region where the CW bias is 1. The FliM occupancy with CheY-P varies from 0.36 (21/58; CW bias, ;0) to 1 (45/45; CW bias, ;1) in CW rotating motors, indicating that not all of the FliM subunits are required to be bound by CheY-P molecules to induce CW rotation. This was consistent with the Monod-Wyman-Changeux (MWC) model (45,46) or the conformational spread model (47)(48)(49) of the flagellar switch. Therefore, all of the FliM molecules in the C-ring can bind CheY-P. Looking from a different perspective, the saturated number of CheY-P molecules bound to a CW rotating motor that we measured here provided an additional piece of evidence that the copy number of FliM in CW-locked motors is 45.
Function and expression levels of mEGFP-fused motor proteins. Flagellar motors composed of FliM molecules with C-terminal fusions of homologous fluorescent proteins have been confirmed to retain CW biases, switching rates, and rotating speeds similar to those of the wild-type motor (21,43). The functions of FliG molecules are partially affected by N-terminal fused GFP (50). In brief, GFP-FliG molecules are stable with negligible degradation and are functional, with motors exhibiting lower CW biases and switching frequencies than those of wild-type motors (50). We observed similar characteristics for mEGFP-fused molecules in this study. Therefore, both FliM-mEGFP and mEGFP-FliG are able to compose intact flagellar motors that are functional for further stoichiometry analysis.
For stoichiometry, the expression of FliM-mEGFP and mEGFP-FliG was induced to the wild-type levels according to quantitative Western blot analysis (Fig. 4). By using anti-FliM antibody, the amount of FliM-mEGFP induced by 100 mM IPTG was compared with the amount of FliM produced in the wild-type strain, obtaining a ratio (FliM-mEGFP/FliM) of 0.97 6 0.36 (mean 6 SD from 12 individual bands). Furthermore, the intracellular levels of FliG and FliM proteins in the wild-type strain were measured in a previous study, obtaining a ratio (FliG/FliM) of ;0.71 (51). In this work, the amount of mEGFP-FliG induced by 130 mM IPTG and the amount of FliM-mEGFP induced by 100 mM IPTG were compared by using anti-EGFP antibody, obtaining a ratio of 0.74 6 0.37 (mean 6 SD from 4 individual bands), which was in good agreement with the previous measurement (51).

DISCUSSION
The conundrum of symmetry mismatch between the MS-ring and the C-ring in E. coli and Salmonella motors seems resolved by recent cryo-EM analysis (23,24); that is, both the MS-ring and the C-ring have 34-fold symmetry. Furthermore, flagellar motors composed of FliF-FliG fusion proteins can assemble and rotate with normal-looking basal bodies (29)(30)(31)(32), and FliF requires FliG to facilitate MS-ring formation in the cytoplasmic membrane (52). FliF and FliG molecules in the motor are stabilized without protein exchange once a functional motor is fully assembled (19,34). In the present study, we demonstrated the robustness of the FliG copy number of the motor at various intracellular concentrations of FliG, suggesting that the assembly of FliG subunits was restricted by the sequential assembly of the C-ring around the MS-ring composed of 34 FliF molecules. This further confirmed that the mean value of the FliG copy number is ;34 in the C-ring of the E. coli flagellar motor. In contrast, previous measurements of motor fluorescence (with FliM-eGFP) as a function of cytoplasmic fluorescence showed that the number of FliM molecules in a motor increased as the cytoplasmic FliM concentration increased (53). Thereafter, we directly compared the fluorescence intensity of motors containing mEGFP-FliG and FliM-mEGFP under identical experimental conditions. By using an FliG stoichiometry of 34 as a reference, the distributions of FliM copy numbers peaked at 58 and 45 for the CCW-and CW-locked motors, respectively. Therefore, the ratio of FliM to FliG numbers in CW-locked motor was ;1.32, and the ratio of FliM numbers in CCW-locked to CW-locked motors was ;1.29. A previous study showed a similar ratio of FliM numbers in CCW-locked to CW-locked motors of ;1.3 and proposed that the FliM copy numbers in the CCW-and CW-locked motors were ;44 and 34, respectively, based on the assumption that FliM in CW-locked motors had a copy number of 34 (20) because the symmetry of the C-ring in the CW-locked Salmonella motor peaked at 34-fold in previous cryo-EM studies (27). Note that the symmetry of the C-ring, which was analyzed by looking at end-on view images of the ring, always peaked at 34-fold regardless of the rotational direction in recent cryo-EM analysis (24,32), suggesting that this symmetry probably corresponded to the copy number of FliG subunits in the C-ring or that it corresponded to the shrunken FliM ring (see below). In early protein structural analysis, the middle domain of FliG (FliG M ) and the C-terminal domain of FliG (FliG C ) were both supposed to interact with FliM, suggesting that there might be two FliM binding sites per FliG (54)(55)(56). However, the interaction between FliG M and neighboring FliG C , which played a crucial role in the organization and ring formation of FliG subunits in the motor, was proven in later reports (31,57,58). Therefore, the direct binding of FliM to FliG C , which would interrupt the array of FliG C -FliG M interactions, seemed unlikely (59). Alternatively, recent crystal structural analyses suggested that the middle domain of FliM (FliM M ) interacts with FliG M to form a complex with a stoichiometry of 1 FliG to 1 FliM in the basal body of the motor (32,(59)(60)(61). This raises the question of how the Cring would accommodate extra FliM subunits if the number of FliM binding sites is equal to the number of FliG molecules. Assuming each FliM needs to be bound to an FliG, the maximum copy number of FliM should be 34, probably corresponding to the CCW state. In this case, the copy number of FliM could be estimated to be ;26 in the CW state, which would result in 8 gaps in the FliM-FliN track. However, the possible gaps have not been observed by cryo-EM (24,27,32). Our measurements here showed that the C-ring contains more FliM subunits than FliG subunits for both CCW and CW states. A recent report proposed a "bridging" model in which each FliG M domain rests on a FliM M domain, and each FliG C domain stacks onto the FliG M domain of an adjacent subunit, while the segment linking FliG M to FliG C could be in either an extended or a helical conformation (31). The extended conformation would allow adjacent FliG subunits to bridge over an extra FliM subunit, which mainly interacts with the adjacent FliM subunits and does not interact strongly with FliG. The maximum copy number of FliM subunits could be estimated to be 68 if all of the linking segments exist in the extended conformation. Alternatively, other models envisioning a fixed-diameter C-ring could tolerate binding of "extra" FliM subunits that are not incorporated into the ring (62).
It was still unclear why the differences in stoichiometry and size of the C-ring between the CCW and CW states were not observed by cryo-EM. The symmetry and diameter of the C-ring are similar for the CCW and CW states in cryo-EM studies (24, 32), suggesting that there are possible changes in protein conformation rather than composition when the motor switches direction. Note that the MotA/MotB stator unit exchange took place in E. coli motors (62), and the stator complexes associated with the basal body seemed lost during sample purification for cryo-EM (59). A recent review proposed the possibility that during the purification of the basal body for cryo-EM, FliM/FliN subunits might leave the basal body without replacement by cytoplasmic FliM/FliN molecules (62). If so, cryo-EM images would reflect the native stable structure after ring shrinkage. Assuming that FliM molecules interacting strongly with FliG molecules are much more stable than those "extra" FliM molecules either below the extended "bridge" or not incorporated into the ring, the latter would leave the C-ring during purification, and FliG and FliM would exist in a 1:1 stoichiometry in the shrunken ring. In the present study, we found that both the CCW and CW FliM numbers were larger than the FliG number. This provided a possibility for ring shrinkage.
In previous studies, another component of the C-ring, FliN, has also been confirmed to undergo turnover and adaptive remodeling (19,22). The ratio of the mean FliN copy numbers in CCW-locked versus CW-locked motors is similar to the value found for FliM (20,22), suggesting that FliM and FliN might form a coexchange unit. Previous studies have proposed two models of the FliM-FliN complex. The first model with stoichiometry of FliM 1 :FliN 4 proposed that four FliN monomers form a tetramer at the base of the C-ring (15,63,64). The second model with FliM 1 :FliN 3 predicted that a FliM:FliN heterodimer and a FliN homodimer form a spiral base of the C-ring (65). Recent cryo-ET structural analyses agreed well with the FliM 1 :FliN 3 model because of the absence of additional density for a fourth FliN molecule (28,66). Our measurements of FliM copy numbers in the present study and the previously measured copy numbers of FliN in CCW-and CW-locked motors (22) suggest a composition of complex closer to FliM 1 :FliN 3 than FliM 1 :FliN 4 . Nevertheless, in situ structural analysis of the C-ring with higher resolution would be necessary to verify the model in future studies.
Here, we demonstrated that all of the FliM molecules in the C-ring could bind CheY-P. Note that the interaction between CheY-P and FliN also plays a critical role in motor switching (15). A recent report suggested that CheY interacts with FliM N , FliN, and FliM M during motor switching (41). In a proposed model, CheY would first be bound to a flexible high-affinity site of FliM (FliM N ), subsequently to FliN, and then to a separate, low-affinity active site on FliM (FliM M ) that would promote CW rotation of the motor (41). In the conformation spread model of the flagellar switch (48), the C-ring is modeled as a ring of a fixed number of subunits, each of which has a binding site for a CheY-P molecule with affinity depending on its conformational state. CheY-P binding changes the free energy level of the subunits. There is a coupling energy between adjacent subunits that favors the same conformations for adjacent subunits. The conformation spread model can well describe the ultrasensitivity of the flagellar switch but is not able to describe motor adaptive remodeling by simply varying the number of subunits. Our measurements here provided new insights to update the conformation spread model. The subunits in the model with a fixed number correspond to individual FliG molecules, and FliM molecules that interact strongly with FliG would bind CheY-P to modulate the energy level of the subunits. FliM N likely acts as a flexible "fishing line" that serves to bind and concentrate CheY-P (41). Therefore, more FliM molecules recruit larger numbers of CheY-P around the C-ring. Due to the flexibility of FliM N , CheY-P bound to FliM N of the extra FliM molecules might be able to interact with neighboring FliN/FliM M that interact strongly with FliG, thus increasing motor sensitivity to CheY-P concentration. This is consistent with previous experimental findings (21).
Cryo-ET was utilized to visualize in situ CCW and CW motor C-ring structures of Vibrio alginolyticus (28) and Borrelia burgdorferi (66). In V. alginolyticus and B. burgdorferi, the C-ring maintained 34-fold and 46-fold symmetries (in both CCW and CW states), respectively, suggesting that the C-ring protein composition probably remained constant for the two species, although there were some extra densities associated with the C-ring (28,66). These findings were different from previous fluorescence studies in which the FliM/FliN subunits underwent copy number changes in E. coli and Salmonella motors (20)(21)(22). However, there have been no reports of in situ cryo-ET studies of CCW and CW motor C-ring structures in E. coli or Salmonella to date. Therefore, it remains unclear whether C-ring protein exchange takes place in bacterial species other than E. coli or Salmonella, especially for those in which the stoichiometry of the C-ring protein composition remained constant in cryo-ET studies. It was interesting to note that E. coli and Salmonella stator complexes were not observed by in situ cryo-ET (3,67), probably because stator complexes in these two species were dynamic and the low viscous loads in sample preparation led to stator dissociation (62). However, some other species with large periplasmic structures, such as Vibrio, have shown welldefined and clearly imaged rings of stator units by in situ cryo-ET (28,67). This raises the possibility that the characteristics of stator unit exchange are different or that the stator units do not exchange in these species. A possible speculation is that motors with a stable composition should be crucial for generating continuous and sufficient torque, especially for bacterial species possessing a single polar flagellum such as Vibrio. In contrast, for E. coli and Salmonella, the run-and-tumble behavior depends on the "cooperation" of several flagella around the cell body (2). Stator unit exchange allows for rapid motor adaptation to changes in energy availability and load conditions, whereas FliM/FliN unit exchange would adjust the chemotactic sensitivity by adaptive remodeling of the C-ring. These different characteristics between bacterial species might correspond to different survival and evolutionary strategies.

MATERIALS AND METHODS
Strains and plasmids. The E. coli strains and plasmids used in this study are listed in Table 1 To estimate the background correction factor for motor intensity analysis, XL2 transformed with pAT5 and pFD313 was used. For the detection of photons emitted from single CheY-mEGFP molecules, HCB1357 (DflhC) transformed with pAT1 was used. For the detection of photons emitted from single FliM-mEGFP molecules, HCB1357 transformed with pAT2 was used. For the detection of photons emitted from single mEGFP-FliG molecules, HCB1357 transformed with pAT3 was used. The wild-type strain RP437 was used in Western blotting experiments as the control group to determine the expression level of FliM. Cell culture and preparation. All strains were grown in tryptone broth (1% tryptone, 0.5% NaCl) at 33°C to mid-log phase (optical density at 600 nm [OD 600 ] = 0.6). When needed, ampicillin and chloramphenicol were added to final concentrations of 100 and 25 mg/mL, respectively. For stoichiometry measurements, we grew the FliM-mEGFP and mEGFP-FliG strains with 100 and 130 mM IPTG to induce wild-type levels of protein expression, respectively. For CW-locked motors, an additional 0.1% arabinose was added to induce the expression of sticky FliC and excessive CheY 13DK106YW . For the strain XL2 carrying pAT1 and pFD313, 0.001 to 0.005% arabinose was added to cover the whole range of motor switching behaviors. For the strain XL2 carrying pAT5 and pFD313, 0.01% arabinose was added to induce mEGFP expression. For fluorescence observation, cells were washed twice with motility buffer (0.01 M potassium phosphate, 10 24 M EDTA, 0.067 M NaCl, pH 7.0), resuspended in lightproof tubes, and incubated in motility buffer at 37°C for 45 min. More than 90% of mEGFP molecules matured after incubation (36). To shorten flagellar filaments, resuspended cells were forced to pass through narrow polyethylene tubing (0.58-mm inner diameter) connected between two syringes more than 100 times. The sample chamber was precleaned and constructed as described previously (37). For tethered-cell experiments, cells with sheared filaments were harvested by centrifugation, resuspended in motility buffer, and loaded into the chamber for 15 min. Then, 400 mL motility buffer was gently injected through the chamber to remove unattached cells. The cell culture and sample preparation were kept in dark environments to avoid fluorescence excitation before data acquisition.
Microscopy and image acquisition. We used a Nikon Eclipse Ti2 TIRF microscope to observe the fluorescence emitted from mEGFP molecules. For the excitation of fluorescence, a fiber-pigtailed laser beam (Sapphire 488 FP; Coherent) was reflected by a primary dichroic mirror (Chroma) and focused on the back focal plane of the TIRF objective lens (100Â Apo TIRF, NA 1.49; Nikon). To measure the motor fluorescence of tethered cells, the angle of incidence was adjusted to generate an evanescent laser field with a characteristic decay length of 100 nm. A halogen lamp was used for bright-field illumination with a longpass glass filter (Thorlabs) to minimize the potential excitation of fluorescent molecules. The fluorescence emitted from the whole cell was captured under epifluorescence illumination. All of the fluorescence images of tethered cells were recorded at 65 nm per pixel with a back-illuminated scientific complementary metal oxide semiconductor (sCMOS) camera (Dhyana 400BSI; Tucsen Photonics) at 20 frames/s. The bright-field images for analyzing the CW bias were recorded at 64 frames/s. Intracellular fluorescence density. Fluorescence emitted from the whole cell was recorded under epifluorescence illumination. The net fluorescence was confirmed to be proportional to the quantity of mature fluorescent proteins in a previous study (36). Therefore, we defined the intracellular fluorescence density as the ratio of the net fluorescence to the cell volume, which has been confirmed to be approximatively proportional to the intracellular concentration of fluorescent proteins (37). The length of the cell was determined with the bright-field image, and the shape of an E. coli cell was approximated as a cylinder with a diameter of 0.8 mm with two hemispherical end caps to estimate the cell volume (37,69). The net fluorescence was calculated to capture most of the out-of-focus light (36) (see Fig. S2 in the supplemental material).
Motor intensity analysis. The initial 30 frames of fluorescence images were linearly overlapped, and the peak pixel intensity was used to estimate the motor centroid, usually in agreement with that based on the center of rotation of the cell body in the corresponding bright-field images. Two regions of interest (ROIs) were defined centering on the motor centroid. The raw intensity of the motor (I raw ) was calculated as the sum of all pixel intensities within the inner ROI of 7 Â 7 pixels (455 Â 455 nm). The background intensity (B) was defined as the mean pixel intensity within the outer ROI of 9 Â 9 pixels (585 Â 585 nm) but external to the inner ROI. However, the background intensity is not uniformly distributed throughout the cell due to rotation of the rod-shaped cell body (70). Therefore, the net motor intensity (I) should be calculated with the following equation: where n is the number of pixels within the inner ROI, and C is a correction factor for the estimation of background intensity in the motor region with the tethered-cell method under TIRF illumination. The mEGFP-producing strain XL2 carrying pAT5 (pBAD33-megfp) and pFD313 was used for estimating the correction factor C under experimental conditions identical to those used for motor intensity measurements (see Fig. S1A in the supplemental material). C was calculated as the ratio of I raw to (B Â n). As shown in Fig. S1B, the correction factor C remains nearly constant (;1.1) before total photobleaching of mEGFP molecules. Therefore, the correction factor (C) was set to 1.1 for the calculation of motor intensity. Quantitative Western blotting. Cells were grown in 20 mL tryptone broth at 33°C to mid-log phase (OD 600 = 0.6). A total of 15 mL of cell suspension was harvested by centrifugation, washed twice, and enriched in 1 mL of lysis buffer (25 mM Tris, 250 mM NaCl, 1 mM dithiothreitol [DTT], pH 7.5). Cell suspensions were lysed by sonication to prepare whole-cell extracts. Eighty microliters of whole-cell extracts was mixed with 20 mL of SDS-PAGE loading buffer (5Â), heated to 100°C for 6 min, and cooled down to room temperature. Ten microliters of each sample mixture was loaded onto a 10% SDS gradient polyacrylamide gel. After separation by electrophoresis, the proteins were transferred to a 0.2 mm nitrocellulose membrane (Cytiva). The membrane was blocked with 5% (wt/vol) skim milk in Tris-buffered saline with Tween 20 (TBST) buffer (20 mM Tris, 150 mM NaCl, pH 7.4, 0.1% Tween 20) for 4 h at room temperature. Next, the membrane was incubated overnight at 4°C with 1:2,500 diluted monoclonal mouse anti-FliM antibody (Sangon Biotech) or 1:1,500 diluted monoclonal mouse anti-EGFP antibody (Abcam) in TBST containing 5% bull serum albumin and then incubated for 1 h at room temperature with 1:1,000 diluted horseradish peroxidase (HRP)-conjugated goat anti-mouse antibody (Invitrogen). Finally, the chemiluminescence signal was generated by an ECL detection kit (Thermo Scientific) and scanned with a chemiluminescence imaging system (Tanon).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, DOCX file, 0.02 MB.