An Ultrasensitive Bacterial Motor Revealed by Monitoring Signaling Proteins in Single Cells

See allHide authors and affiliations

Science  03 Mar 2000:
Vol. 287, Issue 5458, pp. 1652-1655
DOI: 10.1126/science.287.5458.1652


Understanding biology at the single-cell level requires simultaneous measurements of biochemical parameters and behavioral characteristics in individual cells. Here, the output of individual flagellar motors in Escherichia coli was measured as a function of the intracellular concentration of the chemotactic signaling protein. The concentration of this molecule, fused to green fluorescent protein, was monitored with fluorescence correlation spectroscopy. Motors from different bacteria exhibited an identical steep input-output relation, suggesting that they actively contribute to signal amplification in chemotaxis. This experimental approach can be extended to quantitative in vivo studies of other biochemical networks.

Biochemical networks have been the object of intensive experimental and theoretical studies. The understanding of their functioning relies mainly on data collected from populations rather than from single cells (1). When phenotypic variability is observed, however, single-cell measurements become indispensable (2). Here, we present an experimental method to correlate intracellular enzyme concentrations with behavioral characteristics in single cells. We adopted this approach to characterize, in Escherichia coli, the output device of the chemotactic network—an individual flagellar motor. Measuring the input-output characteristics of motors may be relevant for understanding the high amplification gain in the chemotactic sensory system (3). In measurements over bacterial populations (4–6), its input-output characteristic was found to be too mild to contribute substantially to the observed high amplification of the chemotaxis system. Several molecular mechanisms of amplification, such as clustering of receptors (7), were thus subsequently proposed.

E. coli is propelled by several flagella. Each flagellum rotates under the action of a rotary motor (8). When motors rotate counterclockwise (CCW), the flagella form a bundle and the bacterium swims smoothly (9); when motors rotate clockwise (CW), the bundle flies apart and the bacterium tumbles in an erratic fashion. Tumble events randomize the cell trajectory, and the modulation of their occurrence allows bacteria to perform chemotaxis by swimming toward attractants or away from repellents (10). Specific receptors detect changes of environmental chemical concentrations and send a signal through the chemotactic network to the flagellar motors. CheY-phosphate (CheY-P) is the output of the signal transduction network. CheY-P binds preferentially to the motor (11), and the CW bias of the flagellar motors, that is, the fraction of time that a single motor spends rotating in the CW direction, increases with CheY-P concentration [CheY-P] (4).

The experiment described here was designed to determine, in single cells, the bias of individual motors as a function of [CheY-P]. The intracellular concentration of chemotactic proteins fused to the green fluorescent protein (GFP) was measured directly in individual cells ofE. coli. To control the expression levels of CheY-P, we transformed the PS2001 strain of E. coli, lacking thecheY gene (12), with an inducible lac promoter plasmid expressing a cheY-gfp fusion gene (13). To our knowledge, there is no technique for measuring the phosphorylation levels of CheY-GFP in vivo. Therefore, a reliance on in vitro kinase activity measurements (12) reinforced the hypothesis that the entire pool of CheY molecules is phosphorylated in the transformed PS2001 strain (4).

Cell bodies were immobilized and specifically attached onto microscope slides so that some of the flagella were free to rotate. Rotating flagella were marked with latex microbeads to visualize their rotation with a dark-field illumination. The CW bias was obtained from analysis of video recordings (Fig. 1A).

Figure 1

(A) Schematic view of the experimental apparatus. We modified an inverted Zeiss microscope to perform FCS measurements on individual cells. The cell was specifically attached by its flagella onto a microscope slide. A 0.5-μm latex bead (Polyscience), attached to a flagellum with rabbit antibodies to flagellin, is used as a marker to visualize a free rotating flagellum. The CW bias was computed as the ratio of the time spent in CW to the total time duration. The FCS technique allowed us to measure GFP-tagged protein concentration in the same bacterium. The fluctuations of the total fluorescence intensity were processed in real time by a correlator (ALV-5000/E) that provided an autocorrelation function (14). CCD, charge-coupled device. (Inset) A dark-field illumination (red light) was used to record the rotation of a single flagellum of a bacterium attached to a cover slip. For clarity, only three images, 1/15 s apart, were superimposed to show the circular trajectory of the bead. [When a bead was attached to several flagella, its trajectory was no longer circular and it moved erratically. Here, the bead was rotating CCW, a state corresponding to smooth swimming (9)]. (B) Typical autocorrelation function measured for diffusing CheY-P–GFP molecules in a single cell. The amplitude of the autocorrelation function at the intercept with the vertical axis is inversely equal to the number of molecules (N) in the detection volume. We fit this function (continuous red line) withG(t) = 1/N[1 + (4Dt2)]−1, which describes two-dimensional translational diffusion (15). Dis the diffusion constant of the fluorescent molecules,t is the time variable, and 2ω = 0.3 μm is the diameter of the detection volume in our experimental configuration; one molecule in this volume represented a concentration of 44 nM. The autocorrelation functions were measured from acquisitions of 7 s. The average diffusion constant of the cytoplasmic CheY-GFP fusion, evaluated from this fit, was 4.6 ± 0.8 μm2s−1 (16). (Inset) A typical calibration curve, providing a linear relation between concentration of CheY-P–GFP and the fluorescent light intensity for five individual cells. Protein concentrations on this curve were obtained with the FCS technique. We then used the calibration curve to convert fluorescence intensity into GFP concentration in those cells whose flagellar rotation was monitored. This method reduced the photobleaching of GFP by measuring the fluorescence intensity for only 0.5 s (17).

An apparatus based on the fluorescence correlation spectroscopy (FCS) technique (14–18), mounted on an inverted microscope (Fig. 1A), allowed us to measure in vivo the concentration of proteins fused to GFP and to monitor behavioral cellular characteristics (CW bias). We focused the incident excitation laser beam onto a small volume of the cell and collected, in a confocal geometry, the fluorescence light emitted by the GFP molecules. Because the fluorescence intensity did not depend on the position of the illumination spot, we supposed that the expression of CheY-GFP was homogenous within a single cell. We obtained absolute concentration of CheY-GFP fusion, with less than 15% error in measured levels, by analyzing fluctuations of the fluorescence intensity (Fig. 1B).

Bacteria PS2001 strain did not tumble; motors were always in CCW state. Tumbling was restored when the CheY-GFP fusion was expressed from inducible plasmids (19). For a given concentration of inducer (isopropyl-β-d-thiogalactoside, IPTG), the cell-to-cell concentration [CheY-P] was widely distributed around a mean value (ranging from 0.8 to 6 μM), with the typical standard deviation of ∼24% of the mean. We used three IPTG concentrations (0, 5, and 10 μM) to cover the whole range of [CheY-P] to determine the motor characteristics.

When the CW bias for individual cells was plotted versus their internal [CheY-P] (Fig. 2A), we found that the CW bias measured from different cells, preinduced with various inducer levels, fell onto the same sigmoid curve. When an induction process was followed for an individual cell, the activity of CheY-P and the GFP fluorescence were also observed to correlate (20). Thus, within experimental resolution, individual motors were characterized by a uniform input-output relation (21). This remarkable uniformity of the motor characteristics also provided an internal control of the consistency of our measurements.

Figure 2

(A) Characteristic response of individual motors as a function of CheY-P concentration. Each data point describes a simultaneous measurement of the motor bias and the CheY-P concentration in an individual bacterium. The CW bias was computed by analyzing video recordings for at least 1 min. We introduced the cheY-gfp (13) fusion gene into the strain PS2001. It is believed that in this strain, all CheY molecules are phosphorylated (12). Cells were grown from an over- night culture in tryptone broth at 30°C and then harvested (absorbance = 0.5 at 595 nm). To cover the whole range of motor response, we grew cells with three different IPTG concentrations (0, 5, and 10 μM) and then washed and resuspended them in minimum medium •. The second set of experiments was also performed to check whether the folding kinetics of the GFP would affect the CheY-P activity under our experimental conditions. The expression of CheY-P–GFP fusion was monitored after the Luria-Bertoni (LB) medium was saturated with 10 mM IPTG. While the cells were expressing the CheY-P–GFP fusion, the motors' bias would increase and follow the same sigmoid curve (21). Time points correspond to 18, 28, and 33 min for ▾, to 60 and 69 min for ▪, and to 17, 23, and 26 min for ▴, after the IPTG was added. The dashed line shows the best fit obtained with a Hill function (Hill coefficientN H = 10.3 ± 1.1 andK M = 3.1 μM). Motors were locked in (CW) state for tested CheY-P concentrations ranging from ∼4.6 to 25 μM (27). (B) Switching frequency,F, measured from the same cells as in (A).F was defined as the number of times that a motor switched its direction of rotation divided by the duration of the recording. In agreement with previous observations, we observe that the data points for the switching frequency are more scattered than those obtained for the motor bias (5). The dashed line gives the first derivative of the Hill function [from (A)] with respect to [CheY-P]. It is interesting to note that F qualitatively behaves as F ∼ [∂(CWbias)/∂C], where C is [CheY-P].

The sigmoid characteristics of the flagellar motors cannot be well fitted by a Hill function in the whole range of concentrations (Fig. 2A). However, a Hill plot of our data for the bias values between 0.1 and 0.9 leads to an apparent slope of ∼10.3 ± 1.1 [with dissociation constant (K) = 3.1 μM/s] (19, 22). Previous experiments (4–6) reported much lower values for the Hill coefficient, ranging from 3.5 to 5.5. To explain this discrepancy, one should note that previous attempts to characterize flagellar motors were made by averaging the protein concentrations over cell populations. This averaging effectively smoothed out the characteristics of motors, leading to lower values of the Hill coefficient (23).

We measured independently the switching frequency between CW and CCW states (Fig. 2B). It was peaked strongly around [CheY-P] ∼3 μM, about the same concentration for which the CW bias is equal to 0.5.

The uniformity of the motor characteristic suggests that some of the structural features of the motors may be rather tightly regulated. Physiological measurements of the behavior of flagellar motors cannot provide information about their molecular architecture. However, to interact with the motor, CheY-P molecules bind to a cytoplasmic FliM protein ring, consisting of ∼30 binding sites (8,24). It may be that the steep input-output characteristic of motors is related to a cooperative binding process of the CheY-P molecules to the FliM subunits.

Why does the network signal transduction need such a sensitive readout to perform chemotaxis? According to the motor characteristic, a small variation in [CheY-P] leads to a large change in motor bias. Such a steep input-output relation, called ultrasensitivity (25), attributes to the motor the function of an amplifier. The high gain of the chemotactic signal transduction system may thus derive from functional properties of the motors (3). Furthermore, because of the steep input-output characteristic of the motors, the cell must adjust [CheY-P] around the operational value of 3 μM. It seems likely that an additional molecular mechanism to adjust [CheY-P] within the operational range of the motors (26) will exist. Such a mechanism would also help to reconcile the high amplification characteristic with a wide dynamic range of chemotactic sensitivity (3, 25).

The present experiments establish that it is possible, at a single-cell level, to correlate biochemical quantities, such as intracellular protein concentrations, with cell behavior. They should be of importance not only for the understanding of chemotaxis but also for quantitative studies of a wide range of biochemical and genetic networks.

  • * To whom correspondence should be addressed. E-mail: phcluzel{at}


View Abstract

Navigate This Article