Figure 5. A simple model linking single cell expression levels to population fitness.

(A) In our model, the expression level $E$of individual cells is randomly drawn from a normal distribution $\mathcal{𝒩}\left({\mu }_E,{{\sigma }_E}^2\right)$. ${\sigma }_E$ is lower for a genotype with low expression noise (top, green line) and higher for a genotype with high expression noise (bottom, orange line). (B) The doubling time $DT$ of individual cells is directly determined from their expression level using a function $DT=f\left(E\right)$. (C) The growth of a cell population is simulated by drawing new values of expression converted into doubling time after each cell division. In this example, doubling time is more variable among cells for the population showing the highest level of expression noise. (D) Population growth is stopped after a certain amount of time (1000 minutes in our simulations) and competitive fitness is calculated from the total number of cells produced by the tested genotype relative to the number of cells in a reference genotype with ${\mu }_E=1$ and ${\sigma }_E=0.1$. In this example, fitness is lower for the genotype with higher expression noise (bottom) because it produced less cells than the genotype with lower expression noise (top).

Figure 5—figure supplement 1. Single-cell division rates estimated using time-lapse microscopy.

(A) The image series shows selected frames from a time-lapse movie that captured a picture of cells every six minutes in which we have zoomed on one initial cell to illustrate how doubling times (DTs) were determined within a cell lineage. DT was calculated as the time separating the appearance of two consecutive buds. In this example, the DT of mother cell 101 was measured as 96 min because this cell was first seen budding at 12 min and then seen budding again at 108 min. The doubling time of daughter cell 102 was measured as 126 min because it first appeared at 12 min and was observed starting to bud at 138 min. By repeating this cell tracking procedure for eight initial cells in four movies recording 480 min of growth each from four replicate samples from each genotype, the doubling times of at least 362 cells were scored for each strain assayed. (B–E) Histograms showing the distributions of single-cell DTs for one pair of genotypes with median TDH3 expression level far from optimum (B: strain YPW2879, C: strain YPW2868) and for one pair with expression close to optimum (D: strain YPW3064, E: strain YPW3047). In each pair, one genotype had a low level of TDH3 expression noise (green) (B,D) and one genotype had a high level of TDH3 expression noise (orange) (C,E). In panels B-E, arrows represent one standard deviation around the mean DT for each genotype. Permutation tests showed significant differences in standard deviation of doubling times between (B) and (C) (p=0.017) and between (D) and (E) (p=0.020), indicating that genotypes with lower expression noise displayed lower variability of single-cell doubling times. In addition, mean DT was significantly higher in genotypes with expression far from optimum (B,C) than in genotypes close to optimum (D,E) (permutation test, p=10⁻⁵), consistent with lower TDH3 expression levels decreasing fitness. Data and statistics are available in Figure 5—figure supplements 1—source data 1.