The Chemoreceptor Sensory Adaptation System Produces Coordinated Reversals of the Flagellar Motors on an Escherichia coli Cell

ABSTRACT

In isotropic environments, an Escherichia coli cell exhibits coordinated rotational switching of its flagellar motors, produced by fluctuations in the intracellular concentration of phosphorylated CheY (CheY-P) emanating from chemoreceptor signaling arrays. In this study, we show that these CheY-P fluctuations arise through modifications of chemoreceptors by two sensory adaptation enzymes: the methyltransferase CheR and the methylesterase CheB. A cell containing CheR, CheB, and the serine chemoreceptor Tsr exhibited motor synchrony, whereas a cell lacking CheR and CheB or containing enzymatically inactive forms did not. Tsr variants with different combinations of methylation-mimicking Q residues at the adaptation sites also failed to show coordinated motor switching in cells lacking CheR and CheB. Cells containing CheR, CheB, and Tsr [NDND], a variant in which the adaptation site residues are not substrates for CheR or CheB modifications, also lacked motor synchrony. TsrΔNWETF, which lacks a C-terminal pentapeptide-binding site for CheR and CheB, and the ribose-galactose receptor Trg, which natively lacks this motif, failed to produce coordinated motor switching, despite the presence of CheR and CheB. However, addition of the NWETF sequence to Trg enabled Trg-NWETF to produce motor synchrony, as the sole receptor type in cells containing CheR and CheB. Finally, CheBc, the catalytic domain of CheB, supported motor coordination in combination with CheR and Tsr. These results indicate that the coordination of motor switching requires CheR/CheB-mediated changes in receptor modification state. We conclude that the opposing receptor substrate-site preferences of CheR and CheB produce spontaneous blinking of the chemoreceptor array’s output activity.

IMPORTANCE Under steady-state conditions with no external stimuli, an Escherichia coli cell coordinately switches the rotational direction of its flagellar motors. Here, we demonstrate that the CheR and CheB enzymes of the chemoreceptor sensory adaptation system mediate this coordination. Stochastic fluctuations in receptor adaptation states trigger changes in signal output from the receptor array, and this array blinking generates fluctuations in CheY-P concentration that coordinate directional switching of the flagellar motors. Thus, in the absence of chemoeffector gradients, the sensory adaptation system controls run-tumble swimming of the cell, its optimal foraging strategy.

INTRODUCTION

Escherichia coli cells swim by rotating left-handed helical flagellar filaments. Counterclockwise (CCW) rotation produces forward swimming; clockwise (CW) rotation triggers random turning movements, called tumbles (1). E. coli cells swimming in a liquid environment track and navigate chemical gradients with high precision, a behavior termed chemotaxis (2, 3). To monitor attractant or repellent chemicals in its environment, E. coli uses transmembrane chemoreceptors known as methyl-accepting chemotaxis proteins (MCPs). E. coli has four canonical MCPs for sensing attractant nutrients: Tsr (serine), Tar (aspartate and maltose), Tap (dipeptides), and Trg (ribose and galactose). These chemoreceptors have a common functional architecture and transmit sensory messages to the flagellar motors through a two-component phosphorelay (4).

MCPs assemble large signaling arrays at the cell pole(s) through binding interactions with two cytoplasmic proteins (Fig. 1): a histidine autokinase CheA and a scaffolding protein CheW that couples CheA activity to receptor control (4–7). Receptor arrays modulate the autophosphorylation activity of CheA to control the flux of phosphoryl groups from CheA to the response regulator CheY. Phosphorylated CheY (CheY-P) is the intracellular messenger that binds to a flagellar motor to induce CW rotation (8–13). CheY-P molecules appear to reach the flagellar motors through intracellular diffusion (14). CheZ, a dedicated CheY-P phosphatase, degrades the CW signal, but it is located mainly at the polar receptor arrays through interaction with a variant form of CheA, CheA_short (15, 16). Thus, CheY-P is generated at receptor arrays, and the majority of CheY-P is degraded by CheZ before it can escape the array (14).

FIG 1.

An overview of the method used to follow steady-state switching of flagellar motors and the E. coli sensory adaptation system. A cell was stuck to a coverslip, and polystyrene beads (ϕ = 0.5 μm) were attached to the sticky flagellar stubs to observe the angular velocity of each motor from the position of its bead. The signaling activity of the polar chemoreceptor array within the cell modulates the autophosphorylation of CheA (A), which in turn donates its phosphoryl groups (P) to CheY (Y-P) and CheB (B-P). Phosphorylated CheY (CheY-P) interacts with the flagellar motors to promote clockwise (CW) rotation. CheB-P deamidates and demethylates adaptation site residues in the receptor signaling domain (in Tsr: Q297, E304, Q311, and E493); phosphorylation of CheB enhances these activities. CheB preferentially acts on receptors in the on state to shift their output toward the off state. CheR preferentially methylates available adaptation sites on receptors in the off state to shift their output toward the on state. CCD, charge-coupled device; CCW, counterclockwise.

Two MCP-specific chemotaxis proteins, CheR and CheB, mediate sensory adaptation in the E. coli chemotaxis system (Fig. 1). CheR is a methyltransferase that converts specific glutamate (E) residues in MCPs to methylated glutamate (Em) residues. CheB is a methylesterase that converts Em residues to E residues (17); that activity is enhanced upon phosphorylation of CheB by CheA. The serine chemoreceptor Tsr contains five methyl-accepting sites per subunit: Q297, E304, Q311, E493, and E502 (this fifth site is a minor one and was not manipulated in this study). Two of the residues are encoded as glutamine (Q), which is known to mimic the effect of an Em residue on MCP signaling activity (18–21). Under steady-state conditions, the Q residues in newly synthesized Tsr molecules are irreversibly converted to E residues by deamidase activity of CheB, yielding five E sites available for reversible modification by CheR and CheB (Fig. 1). CheR preferentially acts on receptors in the kinase-off state; methylation shifts their output toward the kinase-on state. In contrast, CheB preferentially acts on receptors in the kinase-on state and demethylation shifts their output toward the off state; therefore, the methylation level in MCP is feedback controlled according to its own activity. The interplay of CheR and CheB action comprises a feedback system of sensory adaptation control of receptor signaling activity (22–24).

Previously, we demonstrated that under steady-state conditions with no external chemotactic stimuli, two flagellar motors on the same E. coli cell coordinately switched their rotational direction through increases and decreases in CheY-P concentration. We proposed that fluctuations in CheY-P concentration might be generated by spontaneous activation and inactivation of signal activity in the receptor array (“array blinking”) (14, 25). This signal fluctuation depends on the signal-generating and signal-destroying reactions in the polar receptor array and on diffusion of the signaling molecule phospho-CheY between the array and the flagellar motors. In the present study, we investigated whether the sensory adaptation enzymes CheR and CheB might be responsible for blinking of receptor array activity. We found that the coordination of flagellar motor switching events required both CheR and CheB enzyme activities and modification-competent MCP substrates. From these results, we conclude that even in the absence of chemotactic stimuli, the sensory adaptation enzymes create transient fluctuations in the average modification states of receptors that underlie blinking of array activity, thereby producing fluctuations in CheY-P concentration that synchronize switching of flagellar motors on the cell.

RESULTS

CheR and CheB are required for synchronous motor switching.

To investigate the role of the sensory adaptation enzymes in coordinated rotational switching of flagellar motors on a cell, we used a strain (EFS073) with chromosomal deletions of the MCP, cheR, and cheB genes and supplied various combinations of Tsr, CheR, and CheB proteins with compatible expression plasmids, as described in the Materials and Methods section. We observed the rotation of two motors on the same cell with small beads attached to their flagellar filaments, as illustrated in Fig. 1 (see also Materials and Methods). The rotational motion of each bead was followed with a high-speed charge-coupled device (CCD) camera to obtain a time trace of the rotational velocity and directional switching of the two motors (14, 25).

A strain containing wild-type Tsr, CheR, and CheB appeared to switch motor rotation synchronously, as we previously reported (14, 25) (Fig. 2A). To assess switching coordination, a correlation analysis was performed of the rotational time traces for the two motors (Materials and Methods). The analysis showed a major peak near 0 s (Fig. 2B), indicating that the two motors switched in synchrony. This coordination prevailed in 14 of 15 cells that we observed (Fig. 2C, gray dotted lines) and was still apparent in the averaged correlation profile (Fig. 2C, red line). These results indicate that cells having CheR/CheB and only one type of receptor (Tsr) show coordinated switching in the absence of chemoeffector stimuli, as we found previously for cells with all-MCP arrays (14, 25). We note that the Tsr-only CheR/CheB cells were chemotactic in semisolid agar assays, indicating complementation activity of the plasmid-carried genes (Fig. S1).

FIG 2.

(A, left) Motor switching patterns of cells with and without CheR/CheB. Time traces of the rotational directions of two motors (blue, red) on an EFS073 cell carrying plasmids pPA114 (wild-type Tsr) and pFSRB1 (wild-type CheR and CheB). (Right) The time traces over a short time period. The gray areas mark episodes of clockwise (CW) rotation in the upper trace. (B) A cross-correlation profile for the time traces of the two motors in panel A. (C) Gray lines indicate individual correlations for 15 cells as in panel A; the red line shows the average trace of those correlation analyses. (D) Time traces of the rotational directions of two different motors on an EFS073 cell carrying plasmids pPA114 (wild-type Tsr) and pBAD24 (vector; no CheR and no CheB). (E) A cross-correlation profile for the time traces of the two motors in (D). (F) Individual (gray) and average (red) correlations for 14 cells as in panel D. All correlation analyses were performed for time traces over 60 to 120 s.

A strain containing wild-type Tsr, but neither CheR nor CheB, failed to switch its flagellar motors coordinately (Fig. 2D and 2E). Of 14 cells analyzed, all lacked motor synchrony (Fig. 2F). The Tsr protein in these cells carried a [QEQE] pattern at methylation sites 1 to 4, so in the absence of CheR and CheB, the two E sites would remain unmethylated, and the two Q sites would mimic Em sites. Evidently Tsr with an invariant [QEQE] modification pattern cannot support motor synchrony, but might Tsr with a different modification pattern behave differently? To explore this possibility, we examined the rotation patterns of cells expressing various Q/E variants of Tsr. In cells lacking CheR and CheB, no Tsr Q/E variants supported motor synchrony (Fig. S2B), whereas cells containing the CheR and CheB enzymes and the same Tsr variants showed coordinated motor switching (Fig. S2C). These results indicate that homogeneous arrays of receptors in any fixed modification state could not coordinate the switching of two motors on the same cell. Thus, one or both of the sensory adaptation enzymes is essential for coordinated motor switching.

We note that in all seven cell lines containing a single Tsr modification variant, the variance of CW bias between two motors was low in the presence of CheR and CheB due to synchronous switching (Fig. S3). In contrast, in the absence of CheR and CheB, the variance in CW bias between the two motors was much greater. Motor switching is known to show a steep sigmoidal dependence on CheY-P concentration, with reversal changes centered at ~3 μM CheY-P (26). In cells containing CheR and CheB, synchronous motor switching would require large changes in CheY-P concentration (e.g., from 1 to 6 μM). The absence of switching coordination in cells lacking CheR and CheB (Fig. 2) implies that their receptor array does not produce large changes in CheY-P concentration. However, if the CheY-P concentration in cells lacking the adaptation enzymes hovers near 3 μM, small fluctuations in CheY-P level would lead to large variations in CW bias between two motors in the same cell.

Both CheR and CheB activities are required for synchronous motor switching.

To investigate whether the methylation activity of CheR and/or the demethylation activity of CheB are required for coordinated motor switching, we investigated the coordination between motors in cells producing wild-type Tsr and mutants of CheR and CheB with catalytic defects. CheR-R53A and CheB-H190Y lack enzymatic activity, but both mutant proteins can still localize at receptor clusters (27, 28). An EFS073 cell containing wild-type Tsr receptors, CheR-R53A, and CheB-H190Y failed to show motor synchrony (Fig. 3A to 3C). This result held for cells that we observed (10/10 cells) (Fig. 3D). (The relative expression levels of the CheB and CheR variants are shown in Fig. S4). These results indicate that either the methylation activity of CheR, the demethylation activity of CheB, or both activities are required to coordinate the switching between flagellar motors on a cell.

FIG 3.

Motor switching patterns of cells containing enzymatically inactive CheR and CheB. (A) Schematic diagram of proteins in the cells. (B) Time traces of the rotational directions of two different motors on an EFS073 cell carrying plasmids pPA114 (wild-type Tsr) and pFSRB4 (CheR-R53A and CheB-H190Y). Portion of the time traces are shown here; see Fig. S5 for the full 60-s time trace. Gray areas indicate CW rotation episodes in the upper trace. (C) Cross-correlation profile for the two motors in panel B over 60 s. (D) Individual (gray) and average (red) correlations for 10 cells as in panel B.

CheR and CheB must covalently modify receptors to produce motor synchrony.

To determine whether CheR/CheB-promoted receptor modifications are essential for coordinated motor switching, we investigated switching patterns in EFS073 cells that contained wild-type CheR and CheB proteins, but in combination with a Tsr [NDND] variant receptor that has no suitable substrate sites for CheR or CheB modifications and has similar signaling property to Tsr-QEQE (19). Such a cell did not show coordinated switching events (Fig. 4A to 4C), a result that held for cells that we observed (12/12 cells) (Fig. 4D). The expression level of the Tsr [NDND] variant was comparable to that of wild-type Tsr (Fig. S4). We conclude that CheR and/or CheB must modify receptor molecules to cause synchronous switching of the cell’s flagellar motors.

FIG 4.

Motor switching patterns of cells containing receptors with unmodifiable adaptation site residues. (A) Schematic diagram of proteins in the cells. (B) Time traces of the rotational directions of two different motors on an EFS073 cell carrying plasmids pUCI52 (Tsr [NDND]) and pFSRB1 (wild-type CheR and CheB). Portions of the time traces are shown here; see Fig. S6 for the full 60-s time trace. Gray areas indicate CW rotation episodes in the upper trace. (C) Cross-correlation profile for the two motors in panel B over 60 s. (D) Individual (gray) and average (red) correlations for 11 cells as in panel B.

CheR and CheB must bind to receptors to produce motor synchrony.

CheR and CheB are known to bind to a C-terminal pentapeptide sequence (NWETF) present on Tsr and Tar molecules. Receptor binding substantially enhances the ability of CheR and CheB to encounter and modify adaptation site targets (29–31). To investigate whether the NWETF pentapeptide of Tsr is required for coordinated motor switching, we investigated the rotation patterns of cells containing wild-type CheR and CheB proteins in combination with mutant Tsr receptors lacking the NWETF motif (TsrΔNWETF).

We confirmed that CheR could not bind to TsrΔNWETF by imaging Venus-CheR in EFS073 cells containing wild-type Tsr or TsrΔNWETF. Wild-type Tsr cells exhibited polar localization of Venus-CheR (Fig. 5A), but TsrΔNWETF cells did not (Fig. 5B), suggesting that CheR cannot interact effectively with TsrΔNWETF. We also confirmed, as expected, that both wild-type and TsrΔNWETF formed polar receptor arrays by imaging green fluorescent protein (GFP)-CheW, an essential component of receptor signaling complexes (Fig. 5A and 5B).

FIG 5.

Motor switching patterns of cells containing receptors with or without a CheR/CheB tether (NWETF). (A) Fluorescence in EFS073 cells carrying plasmids pPA114 (wild-type [WT] Tsr) and pFSVnRB (Venus-CheR, wild-type CheB) and in EFS069 cells carrying plasmids pPA114 (wild-type Tsr) and pFSGW2 (green fluorescent protein [GFP]-CheW). (B) Fluorescence in EFS073 cells carrying plasmids pUCI30a (TsrΔNWETF) and pFSVnRB (Venus-CheR, wild-type CheB) and in EFS069 cells carrying plasmids pUCI30a (TsrΔNWETF) and pFSGW2 (GFP-CheW). (C) Rotational traces of two different motors on an EFS069 carrying plasmid pPA114 (wild-type Tsr). (D) Individual (gray) and average (red) correlations for 18 cells as in panel C. (E) Rotational traces of two different motors on an EFS069 cell carrying plasmid pUCI30a (TsrΔNWETF). (F) Individual (gray) and average (red) correlations for 13 cells as in panel E. (G) Fluorescence in EFS073 cells carrying plasmids pUCI23 (wild-type Trg) and pFSVnRB (Venus-CheR, wild-type CheB) and in EFS069 cells carrying plasmids pUCI23 (wild-type Trg) and pFSGW2 (GFP-CheW). (H) Fluorescence in EFS073 cells carrying plasmids pUCI24 (Trg-NWETF) and pFSVnRB (Venus-CheR, wild-type CheB) and in EFS069 cells carrying plasmids pUCI24 (Trg-NWETF) and pFSGW2 (GFP-CheW). (I) Rotational traces of two different motors on an EFS069 carrying plasmid pUCI23 (wild-type Trg). (J) Rotational traces of two different motors on an EFS069 cell carrying plasmid pUCI24 (Trg-NWETF). (K) Individual (gray) and average (red) correlations for 12 cells as in panel J. See Fig. S7 for the full 120-s time trace for motor switching patterns. All bars in fluorescence images are 1 μm.

Next, we compared the coordination of motor switching in cells containing plasmid-expressed wild-type Tsr or TsrΔNWETF. For these measurements, we used a strain (EFS069) devoid of all receptor genes but producing CheR and CheB from the wild-type chromosomal genes. We observed motor synchrony in 17 of 18 cells containing wild-type Tsr (Fig. 5C and 5D). However, 13 of 13 cells containing TsrΔNWETF did not show coordinated motor switching (Fig. 5E and 5F). These results indicate that the Tsr NWETF motif was essential for synchronous switching.

To further verify the requirement of the receptor NWETF pentapeptide for switching coordination, we measured the rotation patterns of cells containing the wild-type Trg receptor, which naturally lacks the NWETF motif and of cells containing a Trg with an appended NWETF pentapeptide of Tsr (Trg-NWETF), which enhances the ability of Trg to function in the absence of other NWETF-containing receptor types (32). Both Trg and Trg-NWETF formed polar clusters with the GFP-CheW reporter, demonstrating formation of signaling arrays (Fig. 5G and 5H). However, the Venus-CheR reporter did not bind to wild-type Trg clusters (Fig. 5G) but did bind to Trg-NWETF clusters (Fig. 5H). In 10 of 10 cells expressing wild-type Trg, we observed their motors to rotate only in the CCW direction (Fig. 5I). The low CW motor bias in cells with wild-type Trg as their only receptor type was previously reported by Feng et al. (32). In contrast, Trg-NWETF cells exhibited rotational switching of their flagellar motors (Fig. 5J), and switching of two motors on the same cell was coordinated in 9 of 12 cells that we observed (Fig. 5K). Thus, addition of the NWETF motif to Trg conferred coordinate motor switching behavior on the cells. The relative expression levels for all MCP variants are given in Fig. S4.

Phosphorylation of CheB is not essential for coordinated motor switching.

The wild-type CheB protein contains two domains: a C-terminal catalytic domain and an N-terminal regulatory domain that modulates CheB catalytic activity. In the on-output state, receptors promote CheA autophosphorylation, which in turn leads to phosphorylation of the CheB regulatory domain and a consequent increase in CheB activity (Fig. 6A, left). To investigate whether CheB phosphorylation is required for motor coordination, we measured the switching behaviors of EFS073 cells containing wild-type Tsr, wild-type CheR, and CheBc, a mutant CheB protein that lacks the regulatory domain. The CheBc protein has only the C-terminal catalytic domain, but it is able to promote sensory adaptation to a tumble-enhancing repellent stimulus (33) (Fig. 6A, right). We found that such a cell exhibited motor synchrony (Fig. 6B and C), and this result held for 12 of 12 cells that we observed (Fig. 6D). These results show that the demethylation activity of CheBc in combination with the methylation activity of CheR is sufficient to induce coordinated switching of a cell’s flagellar motors; CheB phosphorylation is not essential to this process.

FIG 6.

Motor switching patterns of cells containing CheB protein with or without the N-terminal regulatory domain. (A) Schematic diagram of proteins in the cells. (B) Time traces of the rotational directions of two different motors on an EFS073 cell carrying plasmids pPA114 (wild-type Tsr) and pUCI41 (wild-type CheR and CheBc). (C) Cross-correlation profile for the two motors in panel B. (D) Individual (gray) and average (red) correlations for 12 cells as in panel B.

DISCUSSION

We have investigated the mechanism of intracellular signaling by the E. coli chemotaxis machinery by observing the rotation of two different flagellar motors on the same cell (14, 25). In that work, we found that rotational reversals of a cell’s motors were highly correlated and consistent with spontaneous fluctuations in the level of CheY-P signaling molecules that emanate from the cell’s chemoreceptor array and diffuse to the flagellar motors. This blinking of receptor array activity occurs over time scales of several hundred milliseconds and under steady-state conditions in the absence of chemoeffector gradients. In the current study, we asked whether the receptor sensory adaptation system might be responsible for the blinking behavior of the chemoreceptor array.

Receptor adaptation enzymes and blinking array outputs.

We found that cells lacking CheR, the receptor methyltransferase, and CheB, the receptor demethylase/deamidase, did not blink: their flagellar motors reversed independently and asynchronously. Mutant forms of the CheR-R53A and CheB-H190Y proteins that lack enzymatic activity but that retain the ability to bind to the NWETF pentapeptide found at the C terminus of the Tsr and Tar chemoreceptors also could not support blinking. The pentapeptide tether is essential for efficient CheR and CheB modifications of the chemoreceptors (29, 30). Removal of the pentapeptide from the Tsr receptor abrogated blinking, whereas appending the pentapeptide to the Trg receptor, which normally lacks the NWETF tether, enabled it to promote blinking when it was the only receptor species in the cell. These results suggest that in homogeneous environments devoid of chemoeffector gradients, chemoreceptors probably undergo methylation and demethylation reactions that promote synchronous switching of the flagellar motors.

To test the idea that a precise receptor modification state might be responsible for synchronous motor reversals, we surveyed the switching patterns of cells containing Tsr Q/E variants with different combinations of methyl-mimicking Q residues at the adaptation sites. In cells lacking the adaptation enzymes, no Tsr variant, ranging from all-E to all-Q modification states, promoted coordinated motor reversals. In cells containing CheR and CheB, all of those Tsr Q/E variants promoted array blinking. These findings suggest that array blinking requires either a heterogeneous blend of receptor modification states in the cell or fluctuations in the average receptor modification state produced by stochastic activity of the CheR and CheB enzymes.

To ask whether CheR and CheB must act enzymatically at the receptor modification sites to cause array blinking, we examined motor synchrony in cells carrying Tsr [NDND], which has signaling properties comparable to wild-type Tsr [QEQE] (19) but is not a substrate for either adaptation enzyme. Tsr [NDND] did not promote coordinated switching of the flagellar motors in cells containing CheR and CheB, confirming that enzymatic modifications of the receptor molecules underlie array blinking behavior.

Working model for blinking output in receptor arrays.

Our study has demonstrated that the receptor methylation and demethylation activities of CheR and CheB are essential to produce coordinated switching of a cell’s flagellar motors. Phosphorylation of CheB, which substantially enhances its enzymatic activity, was not needed for array blinking. It seems that the interplay of CheB and CheR activities is sufficient to cause coordinated switching of the flagellar motors on a cell (Fig. 6). We conclude, therefore, that the adaptation enzymes cause fluctuations in receptor modification states that result in spontaneous blinking of array activity. Under steady-state conditions, the blinking of receptor array output generates fluctuations in CheY-P production that in turn elicit synchronous reversals of a cell’s flagellar motors.

Fig. 7 offers a simple mechanistic explanation for array blinking based on an integral feedback control model of Alon et al. (22) and Barkai and Leibler (23) and on Ising models of the receptor array (14, 34, 35). Although individual receptors may operate as two-state devices, the high cooperativity of signaling interactions in the chemosensory array effectively couples the output states of many receptors. Receptors in off-output domains are good substrates for CheR. The off-output domains would acquire methyl groups through CheR action until cooperatively shifting to on-output domains. Conversely, receptors in on-output domains are good substrates for CheB. The methyl groups in on-output domains will be removed through CheB action until triggering a cooperative shift to the off-output domain. In this scenario, the activity of the receptor domains blinks between on and off states through the sensory adaptation enzymes, leading to bursts of CheY-P production by the array, quickly followed by CheZ-mediated dephosphorylation.

FIG 7.

Model for spontaneous array blinking driven by the receptor sensory adaptation system. (A) Schematic of a receptor array in kinase-off and kinase-on-output states. CheR-mediated receptor methylation drives output toward the on state; CheB-mediated receptor demethylation drives output toward the off state. (B) The opposing receptor substrate preferences of CheR and CheB and the cooperative coupling between receptors could produce stochastic fluctuations in array activity. A predominantly off array would gain methyl groups through CheR action until cooperatively shifting to the on-output state. At that point, CheB activity would predominate, reducing array methylation until triggering a shift to the off-output state. Phospho-CheY, the CW flagellar rotation signal should track the blinking activity of the receptor array as it has a short half-life due to rapid turnover by its phosphatase CheZ. The abrupt changes in phospho-CheY level produce synchronous switching of the flagellar motors on the cell. In steady state without chemoeffectors, blinking arises from relatively small fluctuations in receptor methylation level due to the action of CheR and CheB. Upon attractant addition, the array shifts to off-output, triggering CheR-mediated receptor methylation. Upon reaching a methylation state that returns CheY-P to its prestimulus level, blinking of the now-adapted array resumes.

In steady-state conditions, the duration of CCW and CW rotation episodes is on the order of several seconds (Fig. 2A). Our model proposes that array fluctuations in receptor methylation levels occur on this time scale. This is the physiologically relevant time scale for gradient-tracking behavior because the cell can use only a several-second “memory” to assess temporal changes in chemoeffector level as it swims about; any longer and Brownian motion would randomize the cell’s heading (36). This time scale is considerably shorter than that required for the cell to adapt to a large, nonphysiological change in attractant concentration, which involves correspondingly large methylation increases over many minutes of experiment time (20).

Evolutionary implication of blinking receptor arrays.

A swimming E. coli cell coordinates rotational switching of its flagellar motors (37). This means that its characteristic run-and-tumble foraging behavior caused by cellular fluctuations in CheY-P concentration arises not only through stochastic motor reversals but also through motor switching synchrony. Turner et al. (1) reported that turn angles are greater when more flagella are unwound from the bundle. Thus, coordinated switching of the flagellar motors on a cell is probably responsible for producing larger changes in swimming direction from each tumbling episode. Perhaps array blinking represents a mechanistic strategy that enables E. coli to explore its environment more widely by optimizing its random walk movements.

MATERIALS AND METHODS

E. coli strains, plasmids, and cell growth conditions.

The strains and plasmids used in this study are listed in Table S1. All strains were derived from the K-12 strain RP437, which is wild type for chemotaxis (38). The replacement of the wild-type fliC gene with the fliC-sticky gene (39) was carried out using the λ red recombinase and tetracycline sensitivity selection method (40, 41).

Amino acid replacements at Tsr methylation sites for Q/E and N/D variants were introduced by QuikChange site-directed mutagenesis (Agilent, Santa Clara, CA) or NEBuilder (New England Biolabs, Ipswich, MA) using pPA114, which encodes wild-type tsr as the template. Missense mutations in cheR and cheB were also created by QuikChange site-directed mutagenesis using pFSRB1 encoding wild-type cheR and cheB as the template. Plasmid pUCI41 encoding wild-type cheR and cheBc, the CheB catalytic domain (residues 147 to 349), was constructed by inverse PCR. The plasmid pUCI30a encoding TsrΔNWETF (lacking the last 11 tsr codons) and the plasmid pUCI24 encoding Trg-NWETF (the final 19 tsr codons joined to the end of the trg coding region) were also constructed by inverse PCR.

LB broth (1% bactotryptone [BD, Sparks, MD], 0.5% yeast extract [BD, Sparks, MD], 0.5% NaCl [Nakarai, Kyoto, Japan]) was used for culture growth, transformations, and plasmid isolation. Tryptone broth (TB) (1% bactotryptone, 0.5% NaCl) was used to grow cells for measurements of motor rotation. For all measurement, the cells were suspended in 10NaMB (10 mM potassium phosphate buffer [Wako, Osaka, Japan], pH 7.0; 0.1 mM EDTA-2K [Wako, Osaka, Japan], pH 7.0; 10 mM NaCl, 75 mM KCl [Nakarai, Kyoto, Japan]). Plasmid-bearing derivatives of strains EFS073 and EFS069 were grown at 30°C for 5.25 h in TB containing 0.5 μM salicylate to induce receptors; 0.005% arabinose to induce CheR, CheB, GFP-CheW, or Venus-CheR; 25 μg/mL chloramphenicol; and 50 μg/mL ampicillin.

Motor rotation measurements.

The cells were prepared by a method similar to that described in our previous report (25). First, 1 mL of cell culture was centrifuged, and the pellet was suspended in 1 mL of 10NaMB. The cell suspension was centrifuged, and the pellet was suspended in ~300 μL of 10NaMB. The cell suspension was loaded into a sample chamber made from 18 × 18 and 24 × 50 mm coverslips with a spacer and incubated for 15 min to allow the cells to attach to the coverslip. The inside of the sample chamber was gently perfused with additional 10NaMB to remove the remaining unattached cells. A suspension of polystyrene beads with a diameter of 0.5 μm (Polysciences, 07307, PA), which was diluted 50-fold with 10NaMB immediately before loading the chamber, was injected. The mixture was incubated for 15 min to allow the beads to attach to the flagellar filaments. The space between coverslips was gently perfused again with additional 30 μL of 10NaMB to remove unattached beads.

The rotation of the beads was measured using a microscopic system similar to the one reported previously (42). The beads were observed under phase-contrast microscopy (IX70 or IX71; Olympus, Tokyo, Japan). Phase-contrast images of the beads were recorded through a UPlanFl 40× NA 0.75 Ph2 lens (Olympus, Tokyo, Japan) with a high-speed CCD camera (IPX-VGA210LMCN or CLB-B0620M-TC000; Imperx, Boca Raton, FL) at 1,000 or 1,255 frames/s. Each captured image was transferred via a frame-grabber card (PIXCI-EB1; EPIX, Buffalo Grove, IL) to a computer for image analysis. This high-speed CCD camera was controlled by real-time video nanometry (RTVN) software, which we developed using LabVIEW 2009 (National Instruments, Austin, TX). Phase-contrast images of the beads were fitted by a two-dimensional Gaussian function for every sampling frame, and the position of a bead was expressed as x and y coordinates of a peak of fitted Gaussian curve. The bead position was approximated by an ellipse function every 500 frames, and the bead position was corrected to approximate a perfect circle centered on the origin. The rotation angle was calculated for every two sampling frames, and time traces of the angular velocity, rotational velocity, and rotational direction were estimated by repeating this process every video frame.

Protein localization assays.

Cells were observed in microscope chambers like the ones used for motor rotation measurements. The cell suspension was loaded into the sample chamber and incubated for 15 min to allow the cells to attach to the coverslip. The inside of the sample chamber was gently perfused with 30 μL of 10NaMB to remove unattached cells. The fluorescence images of GFP-CheW and Venus-CheR were observed by IX71-based epifluorescence microscopy. A blue laser beam (Sapphire 488-20-SV; Coherent, Germany) was inserted into the microscope IX71. The blue laser beam was reflected by a primary dichroic mirror (FF495-Di02; Semrock, Lake Forest, IL) and focused on the back focal plane of the lens (APON 60XOTIRF NA1.49; Olympus, Tokyo, Japan). Fluorescence from the GFP or Venus fusion proteins was passed through the primary dichroic mirror and an emission filter (FF01-520/35, Semrock) and was focused on an electron multiplying charge-coupled device (EMCCD) camera (DU860D-CS0-BV; Andor Technology, UK). The fluorescent images were recorded at 0.5 frames/s, and EMgain was 100.

Correlation analysis.

To analyze the switching correlation between flagellar motors, the rotational velocity was classified into three states by the following procedure (25). The time trace of the rotational velocity was filtered by the Chug-Kennedy filtering algorithm (C-K filter) (43), using an analytical window of 100 data points and a weight of 10. Rotational velocities of more than +20 Hz, between +20 Hz and −20 Hz, and less than −20 Hz were assigned as CCW rotation (+1), pause (0), and CW rotation (−1), respectively. The correlation analysis was performed by applying Equation 1 to the time traces of the rotational directions between two flagellar motors:

$$Z\left(\tau \right)=\frac{\frac{1}{N}{{\displaystyle \sum }}_{t=1}^N\left[x\left(t\right)\cdot y\left(t\hspace{0.17em}+\hspace{0.17em}\tau \right)\hspace{0.17em}-\hspace{0.17em}\overline{x\left(t\right)}\cdot \overline{y\left(t\right)}\right]}{\sqrt{\frac{1}{N}{{\displaystyle \sum }}_{t=1}^N{\left[x\left(t\right)\hspace{0.17em}-\hspace{0.17em}\overline{x\left(t\right)}\right]}^2}\cdot \sqrt{\frac{1}{N}{{\displaystyle \sum }}_{t=1}^N{\left[y\left(t\right)\hspace{0.17em}-\hspace{0.17em}\overline{y\left(t\right)}\right]}^2}}$$ (1)

where Z is the function used for the correlation analysis, t is time, τ is the time difference, N is the total number of sampling points, and x(t) and y(t) are the time traces of the rotational directions of two motors, respectively. Correlations Z (τ) were calculated (−1 $\mathrm{\leqq }$ Z $\mathrm{\leqq }$ 1) by Equation 1 using the traces of rotational direction from 60 to 120 s.