Figures and data
Phenotypic evolution of bacterial chemotaxis in disordered landscapes.
(a) Schematic of the experimental evolution protocol for selecting rapid chemotactic navigation in semi-solid agar gels. (b) Chemotactic navigation of the evolving populations as a function of selection cycle for the two agar concentrations. (c) The average intrinsic free run time τf of evolved populations, measured in liquid medium, converges to distinct, agar concentration-dependent values. Populations evolved in denser gels (0.3% agar) adapt a shorter optimal τf .
Non-monotonic navigation speed reveals an optimal run time.
(a) Design and validation of the τf-titratable strain. Left: Genetic circuit for anhydrotetracycline (aTc)-inducible control of cheZ expression. Right: The mean free run time (τf) of the engineered strain increases linearly with the logarithm of the aTc concentration. (dashed red line is the linear fit). More than 8000 cells were tracked at each aTc concentration, the standard error of the mean (SEM) of τf is smaller than marker size in all conditions. (b) Chemotactic navigation exhibits a non-monotonic dependence on the intrinsic free run time. Three replicates were performed for each condition with error smaller than marker size. The optimal free run time that maximizes chemotactic navigation is larger in 0.3% agar than in 0.2% agar.
Single-cell analysis of bacterial motility.
(a) Fluorescent labeling of bacterial flagella is achieved by incorporating an unnatural amino acid (UAA) into the FliC protein, followed by conjugation with AFDye-tetrazine, enables visualization of flagellar structure and dynamics. (b) Representative fluorescence time-laps micrograph of an E. coli cell embedded in a 0.2% agar gel, showing three distinct behavioral states: Run (cell in motion with bundled flagella); Trap (cell stopped with bundled flagella); Tumble (cell stopped with split flagella). (c) Left: Representative image frames with cells automatically identified and classified as having bundled or split flagella using an in-house trained YOLOv5 detection model. Right: Cells were tracked across frames to reconstruct trajectories, and velocities were calculated and annotated with corresponding flagellar states. The normalized swimming speed (normalized to the 95th percentile speed of each individual trajectory) is displayed as probability density distributions for each flagellar configuration: bundled (running or trapped) and split (tumbling), in both liquid medium (red curves) and in 0.2% agar gel (blue curves), revealing distinct motility dynamics across environments. (d) Probability density functions (PDF) of reorientation angles following tumbling (red) and trapping (blue) events in agar (see Methods).
A minimal model reveals a diffusion-bias trade-off.
(a) Schematic of the stochastic model combining intrinsic run-and-tumble dynamics with extrinsic trapping by the environment. (b) Predicted diffusion coefficient D as a function of the intrinsic mean free run time τf for different mean trap intervals τt. In the absence of a chemo-attractant gradient, D increases monotonically with τf but is suppressed as the trap density increases (red line: low trap concentration, τt = 0.6s ; blue line: high trap concentration, τt = 0.2s). (c) Modeling chemotaxis: cells extend runs (
Schematic illustration of bacterial chemotaxis in a porous agar gel environment.
This diagram depicts the movement of E. coli through a network of pores in agar gel, highlighting the three primary behavioral states: run (straight swimming with bundled flagella), tumble (reorientation via flagellar unbundling), and trap (a newly identified state where cells become temporarily immobilized due to physical confinement). The mean run duration between successive tumbles (τf) and mean run duration between successive traps (τt) are indicated by red arrows, representing key parameters governing motility dynamics. The orange lines represent the pore structure of the gel, while the gray background denote chemoattractant gradients. This spatially constrained environment imposes selective pressures on motility strategies of bacterial and raises questions on the optimal chemotactic navigation strategy to maximize the chemotactic navigation Vd in porous agar gel.
Evolutionary dynamics of motility and growth parameters in liquid culture across selection cycles.
Time courses of key phenotypic traits as measured in evolved E. coli populations over 40 selection cycles under two agar concentrations (0.2% and 0.3%). (a) Growth rates remain stable throughout the selection process, where error bars represent std of 3 independent measurements. (b) Mean run length declines slightly in the 0.3% agar line. (c, d) Tumble duration and mean run speed are maintained at a consistent level across cycles. (e) Tumble bias increases steadily in the 0.3% agar line, while remaining relatively constant in the 0.2% line. Motility related data represent averages from more than 4,800 individual cell tracks and over 100,000 run or tumble events per condition, with standard errors of the mean (SEM) smaller than the symbol size.
Distributions of key motility parameters in evolved strains compared to the ancestral population.
Probability density functions (PDFs) depict five fundamental motility traits measured for the ancestral strain (black lines) and two independently evolved lines selected under 0.2% (blue lines) and 0.3% (red lines) agar concentrations, with data collected from over 100,000 run or tumble events per condition. Panels (a-c) illustrate that distributions of run times, run lengths, and tumble durations all exhibit approximately exponential decay across all strains, indicating consistent stochastic processes underlying these traits. In contrast, panels (d) and (e) show that tumble bias and mean run speed are unimodally distributed, suggesting selective pressures lead to more uniform adaptations in these parameters
Competitive fitness assay reveals environment-dependent selection of optimal run duration (τf).
(a) Schematic representation of two genetically engineered E. coli strains, each with distinct inducible control over the mean run duration (τf), achieved via independent expression of CheY from the tetR and lacI systems using aTc and iPTG, respectively. The strains are fluorescently labeled (green and red) for spatial tracking during competition. (b,c) Competitive range expansion assays on 0.2% agar (b) and 0.3% agar (c), where both strains were co-inoculated at equal initial density and allowed to expand overnight at 37 °C. Fluorescence imaging reveals the spatial distribution of each strain across the expanding colony. On 0.2% agar (b) the green one dominates the outer edge of the colony, indicating superior dispersal in less confined environments. In contrast, on 0.3% agar (c), this red strain is enriched toward the center expands outward, demonstrating that shorter run durations are favored under higher physical confinement.
Motility behavior of E. coli in agar gel.
(a) Representative trajectory of a single bacterial cell moving through a 0.2% agar gel, with automatically detected behavioral states annotated: runs (blue line), tumbles (red dots), and traps (green dots). The trajectory reveals frequent reorientations and prolonged pauses indicative of physical confinement and interaction with the gel matrix. (b) Probability density functions (PDFs) of the duration for run, tumble, and trap events in 0.2% agar gel, showing distinct temporal signatures. Runs exhibit a broad exponential decay, consistent with stochastic motility, while tumbles are brief and sharply peaked. Trap durations are longer and more variable, reflecting transient immobilization due to pore entrapment.
Prediction of the model with complete chemotaxis pathway.
Simulations incorporating the full bacterial chemotaxis network reveal how key motility metrics depend on the mean trap intervals τt and intrinsic run duration τf. (a) Effective drift velocity in a chemoattractant gradient peak at intermediate values of τf, with optimal chemotaxis occurring when τf is tuned relative to τt(black line). (b) Contour plot of chemotactic ability (χ) across a range of τf and τt, revealing an increasing trend of 
