Fig. 2: Slower-growing resistant mutants are stabilized at a quasi-constant equilibrium width, reshaping the acquisition and effect of compensatory mutations.

a Experimental data analysis pipeline. Time series of colony images are acquired via fluorescence microscopy and analyzed via a machine-learning-based image segmentation and processing pipeline to extract clone trajectories. b Single-lineage trajectories (1 to 9 days of growth) for one representative colony (of n = 48 total number of colonies), sorted for trajectory length and type. Line width is proportional to the clone width at the colony edge. Not all clones extinct after one day are shown (see broken axis). Compare to Supplementary Fig. 3 for trajectories of no-rescue control. Inset: Probability P_comp for null model simulations (n₀ = 10,000 clones) and experiments (n₀ = 2898 ± 126 clones from n = 18 independent colonies) (see main text). c Width of compensated (blue) and uncompensated (red) clones. Connected points represent the median width with error bars/shaded areas indicating interquartile ranges. Circles represent mean values. n = 18 independent colonies with a mutation rate μ = 2.65 ± 0.25 × 10⁻⁴ μm⁻¹ (4 nM β-estradiol). d, e Estimated probability densities for the width of uncompensated (red) or compensated (blue) clones, respectively (see “Methods” for details). Hue indicates increasing colony radius over time. f Clone survival probabilities at the front (Eq. (1)) for experiments with compensatory mutations (with-rescue, μ = 2.65 ± 0.25 × 10⁻⁴ μm⁻¹, 4 nM β-estradiol) and minimal mutations (no-rescue, μ = 5.6 ± 3.5 × 10⁻⁶ μm⁻¹, 0 nM β-estradiol) (points), and null model (doted lines). n₀ = 2898 ± 126 clones in n = 18 independent colonies for both with-rescue and no-rescue experiments. See Supplementary Fig. 4 for survival probabilities with different mutation rates. Error bars/shaded area indicate Poisson distribution SD (vertical axis) and SD of the mean (horizontal axis). Arrows indicate treatment times as in Fig. 3. g Efficacy of compensatory mutations (Eq. (2)). Shaded areas indicate propagated SDs. Gray box represents the window of inefficacy, during which efficacy remains zero within errors. Source data are provided as a Source data file.