Electrical Spiking in Escherichia coli Probed with a Fluorescent Voltage-Indicating Protein

See allHide authors and affiliations

Science  15 Jul 2011:
Vol. 333, Issue 6040, pp. 345-348
DOI: 10.1126/science.1204763

Introducing Bacterial Electrophysiology

Bacterial electrophysiology has been limited by the inability to measure the membrane potential of single cells. Kralj et al. (p. 345) engineered a class of voltage-sensitive fluorescent membrane proteins to perform electrophysiological measurements on individual intact bacteria. These measurements showed that Escherichia coli generate electrical spikes, reminiscent of action potentials in neurons. The response of electrical spiking in bacteria was assessed in response to a wide range of physical and chemical perturbations, and was correlated with efflux activity. In the future, the probe should be useful in determining the roles of membrane potential in a variety of medically, environmentally, and industrially important bacteria.


Bacteria have many voltage- and ligand-gated ion channels, and population-level measurements indicate that membrane potential is important for bacterial survival. However, it has not been possible to probe voltage dynamics in an intact bacterium. Here we developed a method to reveal electrical spiking in Escherichia coli. To probe bacterial membrane potential, we engineered a voltage-sensitive fluorescent protein based on green-absorbing proteorhodopsin. Expression of the proteorhodopsin optical proton sensor (PROPS) in E. coli revealed electrical spiking at up to 1 hertz. Spiking was sensitive to chemical and physical perturbations and coincided with rapid efflux of a small-molecule fluorophore, suggesting that bacterial efflux machinery may be electrically regulated.

Bacterial membrane potential provides a major component of the driving force for oxidative phosphorylation, membrane transport, and flagellar motion. Yet this voltage is inaccessible to techniques of conventional electrophysiology, owing to the small size of bacteria and the presence of a cell wall. Little is known about the electrophysiology of bacteria at the level of single cells (1).

We developed a genetically encoded optical indicator of membrane potential, Vm, in bacteria. The starting protein was green-absorbing proteorhodopsin (GPR), a light-driven proton pump found in bacteria in the ocean (24). In the wild, light-driven transport of a proton through GPR changes the color of the protein. We sought to run GPR backward: to use Vm to reposition a proton, and thereby to induce a color shift (5).

The dominant color-determining moiety in GPR is the Schiff base (SB), which links the retinal to the protein core. When the SB was protonated, the protein was pink and weakly fluorescent in the near infrared; when the SB was deprotonated, the protein was yellow and nonfluorescent (fig. S1). We hypothesized that a change in Vm could alter the local electrochemical potential of protons on the SB, and thereby tip the acid-base equilibrium between the fluorescent and nonfluorescent states (fig. S2) (6). However, the pKa (acid dissociation constant) of the SB was >12, indicating that protons were bound too tightly to be removed by physiological Vm. The mutant GPRD97N had a SB pKa of 9.6, showed pH-dependent fluorescence (Fig. 1A), and also lacked light-induced proton pumping (7), making it a promising candidate voltage sensor. We call GPRD97N a proteorhodopsin optical proton sensor (PROPS).

Fig. 1

PROPS is a fluorescent indicator of bacterial membrane potential. (A) Fluorescence spectra of purified PROPS as a function of pH, indicating titration of the Schiff base, with a pKa of 9.6 (versus >12 in wild-type GPR). Peak emission at λ = 710 nm, quantum yield 1.0 × 10−3. PROPS yielded 9.1 times as many photons per molecule before photobleaching as did the GFP (green fluorescent protein) homolog Venus (6). (Inset) Image of E. coli visualized via fluorescence of PROPS (scale bar, 5 μm). Cells were only fluorescent in the presence of retinal (fig. S11). Expression at ~28,000 copies per cell had a minor effect on growth rate in rich medium (6). Membrane fractionation yielded protein associated only with the inner membrane (6). (B) Spatially resolved change in fluorescence in a single cell subject to ITV. Scale bar, 2 μm. Electrodes not shown to scale; actual spacing, 1.6 mm. (C) Comparison of pH- and voltage-dependent fluorescence of PROPS. A change in membrane potential of 102 mV was equivalent to a change in pH of 1 unit in its effect on fluorescence. The sensitivity was ΔF/F = 150% per 100 mV (6). Error bars represent SEM. (D) Time course of the fluorescence response to positive and negative steps in membrane potential. Data in (B) to (D) represent the average of 20 voltage pulses.

Induced transmembrane voltage (ITV) (8) was used to calibrate fluorescence versus Vm in intact Escherichia coli expressing PROPS (6). Voltage pulses were applied to field-stimulation electrodes spanning a plate of cells. During the voltage pulse, the depolarized end of each cell became transiently bright, and the hyperpolarized end became transiently dark (Fig. 1B), indicating a cytoplasm-exposed SB (6). The fluorescence of PROPS was five times as bright at an induced Vm of +70 mV than at −170 mV (Fig. 1C). The response to a voltage step occurred with a time constant of 4.7 ms (Fig. 1D).

E. coli expressing PROPS were imaged at the interface of a coverslip and an agarose pad containing minimal medium. Unexpectedly, many cells exhibited quasi-periodic cell-wide blinks in fluorescence (Fig. 2 and movies S1 and S2). The blinks occurred simultaneously over an entire cell, to within the 10-ms time resolution of our imaging system. Blinks were uncorrelated between neighboring cells.

Fig. 2

E. coli expressing PROPS show transient flashes of fluorescence. (A) Dynamics of fluorescence intensity (I = 6.5 W/cm2) of five single cells in a freshly grown sample under an agarose pad containing minimal medium at pH 7.0. (B) Top: Simultaneous recording of PROPS and pHluorin fluorescence in E. coli treated with CCCP (50 μg/ml) during steps in pHo. Bottom: Intensities of PROPS and pHluorin in a single cell during a blink, in the absence of CCCP. (C) Simultaneous measurement of fluorescence (top) and rotation (bottom) of a cell (strain JY29 ΔcheY) tethered by its flagellum to a coverslip. Construction of the rotary kymograph is described in (6).

We observed a variety of blinking behaviors within a nominally homogeneous population of cells (Fig. 2A). Some cells were dark for many minutes, blinked once, and then returned to darkness; others had periods of quiescence punctuated by bursts of blinking. Blinks had durations from 1 to 40 s with rapid rise times followed by slower decays. The intensity of blinks varied within and between cells, but occasionally we observed periodic bursts of blinking to the same brightness. Some cells remained bright for many minutes, and some remained dark. Many individual cells exhibited different motifs at different times. Blinking cells continued to grow and divide when incubated in the dark at 35°C (fig. S3).

To determine whether blinks arose from fluctuations in internal pH (pHi), we coexpressed the cytoplasmic pH indicator super-ecliptic pHluorin (9) and PROPS, and simultaneously observed the fluorescence of both (Fig. 2B). The pHluorin indicated intracellular pH at the single-cell level with a response time of <1 s and a precision of better than 0.1 pH units (6). During a blink, pHi remained constant to within measurement precision (Fig. 2B and fig. S4). Thus, blinks did not arise from fluctuations in pHi.

To determine whether the blinks arose from electrical fluctuations, we used the torque of the flagellar motor as an indicator of the protonmotive force (PMF) (10). Cells of strain JY29 were adhered to a coverslip by a single flagellum, and we monitored the blinking of PROPS and the rate of rotation of the cell body simultaneously (6). During a blink, the rotation slowed or stopped, indicating that blinks coincided with a loss of PMF (Fig. 2C, fig. S5, and movie S3). The loss of PMF, but stable pHi, during a blink implied that blinks arose from electrical depolarization.

Lipophilic voltage-sensitive dyes (VSDs) did not label blinking cells (6) and thus were unable to provide independent confirmation that blinks were electrical. Within a field of view containing blinking and nonblinking cells, the VSDs only labeled nonblinking cells (fig. S6). The reason for this heterogeneous labeling is unclear. Previous efforts to use VSDs in E. coli were unsuccessful (11).

We examined the effect of metabolic state on blinking and pHi in cells coexpressing PROPS and super-ecliptic pHluorin. Interruption of aerobic respiration caused blinking to cease and all cells to become bright in the PROPS channel. This behavior was seen for inhibition of the electron-transport chain by exposure to intense violet light (Fig. 3A) (12) or by removal of oxygen (Fig. 3B). Inhibition of the F1–adenosine triphosphatase (ATPase) by sodium azide (10 mM, Fig. 3C) (13) and dissipation of the PMF by carbonyl cyanide m-chlorophenyl hydrazone (CCCP; 50 μg/ml, Fig. 3D) (14) induced similar responses (movies S4 to S7). None of these treatments affected the fluorescence of purified PROPS. Thus, E. coli needed to be alive and undergoing aerobic respiration to blink.

Fig. 3

Single-cell responses to metabolic perturbations. Fluorescence of pHluorin and PROPS were recorded simultaneously to indicate responses of pHi and Vm, respectively. (A) Violet light (100 W/cm2). Leakage of the violet light into the pHluorin and PROPS channels prevented acquisition of data during the perturbation. (B) Removal of oxygen. (C) Sodium azide (10 mM). (D) CCCP (50 μg/ml). Scale bars in (A) apply to data in (A) to (D). Colored bars indicate duration of the perturbation. (E) Blinking of E. coli expressing PROPS as a function of illumination intensity. (F) Blinks became more prevalent (blue) and shorter (red) at higher illumination intensity. Error bars represent SEM. Between 114 and 211 cells were analyzed at each intensity.

When the experiments above were performed at pHi ≈ pHo (external pH; corresponding to pHo = 8.3, fig. S7), the perturbations induced minimal change in pHi (Fig. 3). At other values of pHo, the perturbations caused gradual equilibration of pHi with pHo, indicating a failure of pH homeostasis (fig. S8).

The intensity of the red laser used to image PROPS affected the shape and frequency of the blinks. At higher illumination intensity the blinks were briefer and more intense, and came in more regular and higher-frequency pulse trains (Fig. 3, E and F). The mechanism by which illumination enhanced blinking is not known; but we note that E. coli contain endogenous chromophores in their electron transport chain with absorption throughout the visible spectrum (15). Heating by the imaging laser is expected to be negligible (6).

The enhancement of blinking by increased laser power suggested that blinking might form part of the stress response mechanism in E. coli. We thus tested whether blinking was associated with cationic efflux, another important mechanism of stress response.

We observed surprising dynamics of a cationic membrane-permeable dye, tetramethyl rhodamine methyl-ester (TMRM), in blinking E. coli. As expected for this Nernstian voltage indicator (16), TMRM gradually accumulated in the cytoplasm over ~10 min. However, blinks in PROPS fluorescence coincided with precipitous stepwise drops in TMRM fluorescence that showed little or no recovery after the blink (Fig. 4A). The duration of the step in TMRM fluorescence coincided with the duration of the blink: At moderate intensities of red illumination (I = 10 W/cm2) steps lasted less than 200 ms, whereas under little or no red illumination steps typically lasted several seconds (fig. S9). Stepwise disappearance of TMRM was also observed in cells without the PROPS plasmid, when only dim green illumination was used to image the TMRM (30 mW/cm2; Fig. 4B and fig. S10). The duration of these steps was comparable to that of steps in cells with PROPS under dim red illumination (2 W/cm2). The rapid disappearance of TMRM during a blink suggested an active-transport mechanism. Dissipation of Vm lowers the thermodynamic barrier to cationic efflux (Fig. 4C) (6). A concurrent dissipation of Vm and increase in membrane permeability would be sufficient to induce cationic efflux. PMF-dependent efflux of other cationic dyes has been observed in E. coli (17) in population-level assays that are insensitive to the dynamics of individual cells.

Fig. 4

Blinks are accompanied by efflux of a cationic dye. (A) Simultaneous measurement of fluorescence from PROPS and TMRM. Blinks in PROPS (red) coincided with sudden drops in concentration of TMRM (green). (B) Cells not expressing PROPS showed stepwise efflux of TMRM, suggesting that electrical spiking occurred in the absence of PROPS. (C) Model for voltage-regulated cationic efflux. The resting potential of −80 to −120 mV is necessary for metabolism, but opposes cationic efflux. Transient depolarization lowers the barrier to efflux.

Bacterial electrophysiology is likely to differ in several key regards from its eukaryotic version due to the comparatively small surface area, yet high surface-to-volume ratio found in bacteria. With a typical membrane capacitance between 10−14 and 10−13 F, a single ion channel with a conductivity of 100 pS can alter the membrane potential with a time constant of 10−3 to 10−4 s. In contrast, neurons only fire through the concerted action of a large number of ion channels. Thus, bacterial electrophysiology is likely to be dominated by stochastic opening of individual ion channels and pores. Additionally, the ionic composition of bacteria is less robust than that of eukaryotes. A bacterium with a volume of 1 fl and a cytoplasmic Na+ concentration of 10 mM contains only ~107 ions of Na+. A single ion channel passing a current of 2 pA can deplete this store in less than 1 s. These simple estimates show that some of the tenets of neuronal electrophysiology may need rethinking in the context of bacteria.

Correction (8 February 2017): The illumination intensity in the legend to Fig. 2A should be I = 6.5 W/cm2. This correction does not affect any of the findings or conclusions of the paper.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S11

Tables S1 to S6

References (1833)

Movies S1 to S7

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. Acknowledgments: We thank K. Rothschild for helpful discussions and for the GPR plasmid. We also thank H. Berg, R. Losick, J. Yuan, H. Inada, A. Garner, and D. Kahne for helpful discussions. G. Miesenbock provided the pHluorin plasmid. A. Fields, H. Bayraktar, V. Venkatachalam, and L. Bane helped with measurements. This work was supported by the Harvard Center for Brain Science, an Intelligence Community Postdoctoral Fellowship (J.M.K.), a National Science Foundation Graduate Fellowship (D.R.H.), a Helen Hay Whitney Postdoctoral Fellowship (A.D.D.), and a Charles A. King Trust Postdoctoral Fellowship (A.D.D). J.M.K., A.D.D., and A.E.C. have applied for a patent on PROPS as a sensor of membrane potential.
View Abstract

Navigate This Article