Bacterial Chemotaxis

What chemotaxis solves

A bacterium lives in a world dominated by viscosity and noise. At its scale, water feels as thick as honey: inertia is meaningless, and the moment a flagellar motor stops, the cell coasts less than the width of a hydrogen atom before halting. Worse, an E. coli cell is only about 2 µm long and 1 µm wide, so the difference in attractant concentration between its front and back end — even in a steep gradient — is utterly drowned out by thermal fluctuations. And because random Brownian torque rotates the cell by roughly 90° in about a second, it cannot even hold a heading long enough to aim. The puzzle of chemotaxis is therefore not "how does a cell steer toward food?" but "how does a blind, jittering, non-steerable swimmer still arrive at food on average?"

The answer, worked out by Howard Berg and Douglas Brown in 1972 by tracking single cells under a microscope, is a biased random walk. The cell does not navigate; it gambles, then adjusts the odds of the gamble based on whether things are getting better.

Run and tumble: the two-stroke engine

E. coli carries 4–10 helical flagella, each ~10 µm long, driven by an independent rotary motor embedded in the membrane. The motor is powered not by ATP directly but by the proton-motive force — protons flowing down their electrochemical gradient through the stator turn the rotor at up to ~100,000 rpm, one of the fastest rotary machines in biology. Crucially, the motor is reversible, and its direction sets the cell's behavior:

• Counterclockwise (CCW) rotation → a run. The left-handed helical filaments are swept into a single coherent bundle behind the cell, acting like a corkscrew propeller. The cell swims smoothly forward at ~20–30 µm/s for about a second, covering tens of body lengths.
• Clockwise (CW) rotation → a tumble. Even one motor reversing forces its filament through a polymorphic shape change (from the "normal" to a "curly" form). The bundle flies apart, thrust collapses, and the cell jerks and pivots in place for ~0.1 s, reorienting by a mean angle of about 68°.

When the motors switch back to CCW, the bundle reforms and a new run begins — in a new, roughly random direction. In a uniform environment the cell performs a true unbiased random walk: it explores, but goes nowhere on average. Chemotaxis emerges from a single, beautifully economical trick: the cell modulates run length, never tumble direction.

Biasing the walk: lengthen the good runs

Imagine the cell, by chance, starts a run that happens to point up the gradient. As it swims, the attractant concentration it experiences rises. The cell detects this improvement and suppresses the next tumble — the good run gets longer. A run that happens to point down the gradient brings falling concentration, no suppression occurs, and the cell tumbles on schedule to try a new direction. Up-gradient runs are extended; down-gradient and sideways runs are cut short at the baseline rate. Average enough of these biased steps together and the cell drifts steadily toward the source, even though every individual reorientation was random.

This is the whole logic of chemotaxis: make good news last longer. The mean run length in the absence of a gradient is about 1 s; a favorable gradient can stretch runs several-fold, which is more than enough to produce a measurable net drift velocity of a few µm/s up the gradient.

Sensing time, not space

Because spatial comparison across the cell body is impossible, the bacterium reads the gradient temporally. It effectively asks: "Is the concentration higher now than it was a couple of seconds ago?" To do this it needs a short-term memory — a set point that tracks the recent average — and a fast readout that responds to the instantaneous level. The difference between the two is the signal. Berg's classic test confirmed this: when cells were mixed into a spatially uniform but suddenly increased concentration of attractant, they all swam smoothly (suppressed tumbling) for a while even though there was no gradient at all — proof they were responding to the change over time, then adapting back to baseline.

The Che pathway: a two-component circuit

The molecular machinery is the most thoroughly understood signal-transduction system in biology, built from about ten proteins whose names begin with "Che" (chemotaxis). Receptors and kinase cluster into a large lattice at the cell pole, which gives the system both high sensitivity and signal amplification.

• MCPs (methyl-accepting chemotaxis proteins) are the transmembrane receptors — Tar, Tsr, Tap, Trg, Aer in E. coli — each tuned to particular ligands (Tar senses aspartate and maltose; Tsr senses serine). There are thousands per cell, clustered at the pole.
• CheW couples the receptors to the kinase.
• CheA is a histidine autokinase. When receptor activity is high (no attractant bound), CheA autophosphorylates rapidly.
• CheY is the diffusible response regulator. CheA hands its phosphate to CheY; CheY-P diffuses across the cell to the flagellar motor and binds the FliM switch protein, promoting CW rotation — a tumble.
• CheZ is a phosphatase that rapidly strips phosphate off CheY-P (half-life of CheY-P is only ~0.1 s), so the motor responds within tens of milliseconds.

The signal flow is inverted, which trips up newcomers: attractant binding turns the kinase OFF. Less CheA activity → less CheY-P → fewer tumbles → longer runs. So binding "good" molecules quiets the tumble signal and the cell keeps swimming forward. Repellents do the opposite — they raise CheA activity, flood the motor with CheY-P, and trigger frequent tumbling.

Methylation: the molecular memory and near-perfect adaptation

A readout that simply tracked occupancy would saturate: in a high background of attractant, all receptors would be bound, the kinase silenced, and the cell would lose the ability to detect any further increase. The methylation system solves this with a slow negative feedback that resets sensitivity. Two enzymes act on glutamate residues of the MCPs:

• CheR, a methyltransferase, slowly adds methyl groups (using S-adenosylmethionine). More methylation raises receptor/kinase activity.
• CheB, a methylesterase, removes them. CheB is itself activated by CheA, closing a feedback loop.

Attractant binding instantly lowers kinase activity (fast, seconds). Methylation then creeps up over the next ~10–30 s, pushing activity back toward its original set point. Because methylation lags binding, the receptor's current activity encodes the comparison between concentration now and concentration a few seconds ago — exactly the temporal derivative the cell needs. This negative-integral feedback produces near-perfect adaptation: after any step change, tumble frequency returns to precisely its pre-stimulus value, a robustness that is largely independent of protein concentrations (a result formalized by Barkai and Leibler in 1997). Adaptation lets a single cell respond to gradients spanning roughly five orders of magnitude in absolute concentration, from nanomolar to millimolar.

Comparison: chemotaxis across organisms

Feature	E. coli (peritrichous)	Rhodobacter sphaeroides / Vibrio (polar)	Eukaryotic chemotaxis (neutrophil, Dictyostelium)
Movement unit	Run and tumble	Run and stop / run and flick (Vibrio)	Crawling via actin-driven pseudopods
Reorientation	CW tumble, ~68° mean turn	Brownian reorientation while stopped, or 90° flick	Directed front-back polarization
Gradient sensing	Temporal (memory over ~1–4 s)	Largely temporal	Spatial — compares receptor occupancy across the cell
Speed	~20–30 µm/s	Tens of µm/s	~0.1–0.3 µm/s (much slower)
Motor	Proton-motive rotary flagellum, reversible	Often Na⁺-driven; some unidirectional	No flagellum; actin/myosin cytoskeleton
Adaptation	Receptor methylation (CheR/CheB)	Methylation + multiple Che paralogues	Receptor internalization, PI3K/PTEN feedback

Numbers and energetics

The whole system is cheap to run but exquisitely tuned. Each flagellar motor consumes roughly 1,000–1,200 protons per revolution; at full speed that is on the order of 10⁶ protons per second per motor, a tiny fraction of the cell's energy budget. The signaling pathway is fast where it must be — CheY-P turnover at ~0.1 s lets the motor track changes almost in real time — and slow where memory is needed — methylation over tens of seconds. The receptor array amplifies small ligand-occupancy changes into large kinase-activity changes (Hill coefficients well above 1 from cooperative receptor clustering), so a cell can respond to occupancy differences of a fraction of a percent and detect a single up-gradient run lasting only a second or two.

Why it matters: evolution, ecology, and medicine

• Foraging and survival. Chemotaxis lets bacteria find scarce nutrients in a dilute, fluctuating world and flee toxins — a clear and continuously tested fitness advantage.
• Pathogenesis. Helicobacter pylori uses chemotaxis toward urea and bicarbonate to penetrate stomach mucus and reach the epithelium; Vibrio cholerae and Campylobacter jejuni need chemotaxis to colonize the gut. Non-chemotactic mutants are frequently attenuated — chemotaxis is a genuine virulence factor.
• Biofilms. Chemotaxis guides the early surface-seeking and aggregation steps that precede biofilm formation, which underlies chronic and device-related infections.
• A model for information processing. The Che pathway is the textbook example of a two-component signaling system, of signal amplification by receptor clustering, and of robust, near-perfect adaptation — principles that recur from neurons to engineered genetic circuits and tumor-homing bacterial therapeutics.

Common misconceptions

• "The bacterium steers toward food." It cannot steer. It only biases when it tumbles; direction after a tumble is essentially random.
• "It compares concentration at its head and tail." Too small and too noisy. It compares concentration now versus a few seconds ago — temporal, not spatial, sensing.
• "Attractant binding excites the signaling pathway." The opposite — attractant binding turns the CheA kinase down, lowering CheY-P and suppressing tumbles.
• "Tumbles are the productive part." Runs do the traveling; tumbles only re-roll the dice. The bias is achieved by lengthening good runs, not by aiming tumbles.
• "Flagella are powered by ATP." The motor runs on the proton-motive force (or Na⁺ in some species), not directly on ATP hydrolysis.