Relationship between cellular response and behavioral variability in bacterial chemotaxis

Bacterial chemotaxis in Escherichia coli is a canonical system for the study of signal transduction. A remarkable feature of this system is the coexistence of precise adaptation in population with large fluctuating cellular behavior in single cells (Korobkova et al. 2004, Nature, 428, 574). Using a stochastic model, we found that the large behavioral variability experimentally observed in non-stimulated cells is a direct consequence of the architecture of this adaptive system. Reversible covalent modification cycles, in which methylation and demethylation reactions antagonistically regulate the activity of receptor-kinase complexes, operate outside the region of first-order kinetics. As a result, the receptor-kinase that governs cellular behavior exhibits a sigmoidal activation curve. This curve simultaneously amplifies the inherent stochastic fluctuations in the system and lengthens the relaxation time in response to stimulus. Because stochastic fluctuations cause large behavioral variability and the relaxation time governs the average duration of runs in response to small stimuli, cells with the greatest fluctuating behavior also display the largest chemotactic response. Finally, Large-scale simulations of digital bacteria suggest that the chemotaxis network is tuned to simultaneously optimize the random spread of cells in absence of nutrients and the cellular response to gradients of attractant.

💡 Research Summary

The paper investigates why individual Escherichia coli cells display large temporal fluctuations in their swimming behavior even when they are adapted to a homogeneous environment, while at the population level the chemotaxis system exhibits precise adaptation. The authors construct a stochastic model of the adaptation module, which consists of reversible covalent modification cycles: CheR methylates inactive receptor‑kinase complexes and CheB‑P demethylates active ones. Using a two‑state description of receptors (active/inactive) together with Michaelis–Menten kinetics for the (de)methylation steps, they derive equations for the average methylation level (Eq. 1) and for the conservation of total receptor complexes (Eq. 2). Solving these yields a steady‑state kinase activity that matches the classic Goldbeter‑Koshland formulation for a substrate modified by two antagonistic enzymes. Crucially, when the Michaelis constants K_r and K_b are smaller than one, the kinase activation curve becomes sigmoidal with an effective Hill coefficient H > 1. This non‑linearity makes the system highly sensitive to small changes in CheR or CheB‑P concentrations.

Applying a linear‑noise approximation, the authors separate fast fluctuations (ligand binding/unbinding) from slow fluctuations arising from the (de)methylation cycle. The slow component obeys a Langevin equation (Eq. 3) analogous to a mass‑spring system in a viscous fluid. The variance of the kinase activity noise is independent of the slope of the activation curve, but the relaxation time τ (the time over which the system returns to steady state after a perturbation) is proportional to the Hill coefficient (Eq. 5). Hence, a steeper activation curve both amplifies intrinsic stochastic fluctuations and lengthens τ. Numerical simulations (both linear‑noise calculations and full Gillespie stochastic simulations) confirm these analytical predictions.

Experimentally, the model reproduces three key observations: (i) cells expressing a constitutively active CheY mutant (CheY‑D13K) lose behavioral variability; (ii) mutants with fixed receptor methylation levels show reduced fluctuations; (iii) ΔcheR cells complemented with varying CheR levels display tunable variability, decreasing as CheR expression rises from one to four times wild‑type. The model predicts that increasing CheR (or decreasing CheB‑P) moves the system away from the steep region of the sigmoidal curve, reducing both noise amplitude and τ, which matches the observed reduction in run‑time variability.

Finally, large‑scale agent‑based simulations of “digital bacteria” demonstrate that the same architecture that generates noise also optimizes chemotactic performance. In nutrient‑free environments, the long τ and amplified fluctuations promote a broad random spread, enhancing the chance of encountering nutrients. When a shallow attractant gradient is present, the same long τ translates into prolonged runs toward higher attractant concentrations, improving chemotactic efficiency. Thus, the chemotaxis network appears tuned to simultaneously maximize exploratory diffusion in the absence of cues and directed movement when cues are present.

In summary, the study reveals that the chemotaxis adaptation module functions as a noise amplifier and temporal filter. Its sigmoidal response to enzyme activities links intrinsic molecular noise to behavioral variability, and the same parameters that increase variability also enhance chemotactic sensitivity. This unified mechanistic insight bridges single‑cell stochastic behavior with population‑level adaptation and suggests that similar design principles may underlie other signaling networks where robustness and responsiveness must coexist.