Fig 1.
lac operon induction lags are bimodally distributed and controlled by LacI activity before the switch (S1 Data, https://git.io/JTS5A).
(A) Summary of the lac gene circuit. We use a LacZ-GFPmut2 translational fusion integrated at the native genetic locus to monitor lac expression in single cells. (B) Time courses of cell length and number of LacZ-GFP molecules (on log scales; calibration as described in [6]) for a subset of cells from a typical experiment where cells are grown in a DIMM (S1 Fig) and exposed to 2 consecutive 4-h lactose episodes interspersed by a glucose period (which is 8 h in this experiment); for clarity, only the 4 cells near the closed end in one representative growth channel are shown using random colors to distinguish cells, and LacZ-GFP levels are offset by 100 molecules. (C) Histogram of single-cell induction lags for the lac operon at the first lactose exposure (“short” lags under 50 min and “long” lags above 50 min); lac induction lags were defined as the delay after the switch until cells increase their LacZ-GFP by 200 molecules and were estimated from time series of LacZ-GFP expression (shown in B) for 1633 cells in 9 independent replicates (S2 and S3 Figs). (D) A scatter of growth lags versus lac induction lags in individual cells shows that these two lags are tightly correlated (the solid line is a guide for the eye with y = x). (E) Violin plots of the distribution of single-cell induction lags for populations of cells whose LacI activity during growth in glucose was perturbed either by overexpression from a plasmid or by titration with sub-induction concentrations of IPTG. The fraction of cells with short lags is reported for each treatment; each contour corresponds to an independent replicate (with ≈150 cells), and the dotted line corresponds to 50 min separating long and short lags. DIMM, dual input Mother Machine; GFP, green fluorescent protein.
Fig 2.
Transcriptional memory analysis indicates that the critical lac expression threshold for inducing the lac operon is on the order of a single molecule (S1 Data, https://git.io/JTS5A).
(A) Distributions of initial fluorescence in cells with short (orange) and long (blue) lac induction lags; the gray line is a control showing the distribution of autofluorescence in a wild-type strain without GFP (fluorescence is centered per experiment, see Estimating LacZ-GFP levels before the switch). Note that the standard deviation of autofluorescence intensities corresponds to approximately 20 LacZ-GFP molecules, while the shift in fluorescence between cells with short and long lags corresponds to approximately 5 LacZ-GFP molecules only. (B) Reverse cumulative distributions of lac induction lags for cells that first grew in lactose and fully induced their lac operon and then grew in glucose for different lengths of time (shown in colors; 1 or 2 replicates per condition); the distribution for naive cells is shown as a reference (dark gray, corresponding to the histogram of Fig 1C) along with a control where cells are first exposed to lactose after 26 h in the microfluidic chip (light gray). (C) Fraction of short lags (<50 min) for cells preexposed to lactose as a function of the estimated number of inherited LacZ-GFP molecules. The dashed horizontal line shows the overall fraction of short lags in naive cells as a reference and semitransparent gray rectangles show the mean ± SE for each replicate. GFP, green fluorescent protein; SE, standard error.
Fig 3.
Fluorescence lifetime analysis supports that the critical lac expression threshold for inducing the lac operon is on the order of a single molecule (data from https://git.io/JTSdM).
(A) Distribution of decay times measured on induced bacteria (LacZ-GFPmut2) and on bacteria without GFP (autofluorescence in GFP channel). The peak at 2.5 ns corresponds to the laser pulse; hence, the fluorescence lifetime is the delay after this peak. (B) Fraction of cells where LacZ-GFP is detected by FLCS in a liquid culture of uninduced cells (n = 150). The fraction of cells with short lags in microfluidic experiment (Fig 1C) is shown for comparison. (C) Distribution of lac induction lags during experiments where the presence of LacZ-GFP before the switch (shown in color) is assessed by FLCS (66 cells from 3 independent replicates). The dotted vertical line indicates the threshold between short and long lags. FLCS, fluorescence lifetime correlation spectroscopy; GFP, green fluorescent protein.
Fig 4.
The response of single cells to lactose is context-dependent and shapes the population fitness after the switch (S1 Data, https://git.io/JTS5A).
(A) Cartoon depicting how the lac induction of a given cell is determined by the stochastic expression in the repressed state and hence depends on the conditions before the switch. Green dots depict Lac proteins. (B) Violin plots of the single-cell lag distributions for switches to lactose from growth conditions in which the lac operon is slightly less repressed than in pure glucose, i.e., 0.4% glycerol (blue) and 0.2% glucose + 0.2% lactose (green). Each contour corresponds to a separate replicate with ≈150 cells; the fraction of short lags (<50 min, indicated by the dotted line) is indicated below the plots; the lag distribution measured in the glucose to lactose switch (as Fig 1C) is shown in red for comparison. (C) Simulated population growth curves for different distributions of single-cell lags, as observed for different switches (different colors; with semi-log scale). The simulations assume deterministic exponential growth of each cell before and after the switch, and complete growth arrest during the lag. The pink and light blue curves show the population growth that would be obtained with only short and long lags, respectively. (D) Population growth lags as inferred from the simulated population growth shown in panel C as a function of the fraction of short lags. The lag is measured as a delay compared to a population with no arrest (i.e., where all cells grow immediately at their maximum growth rate on lactose; yellow) and reported in number of doublings in lactose (i.e., the number of divisions lost due to growth arrest). For example, the vertical arrow indicates the lag for the light blue condition and corresponds to the delay shown by the horizontal arrow in panel C. Note that “short only” (light blue) and “long only” (pink) correspond to hypothetical populations where either all cells have long lags, or all cells have short lags.
Fig 5.
Single-molecule triggering is observed across a diverse set of E. coli strains (data from https://git.io/JTSd5).
(A) Estimation of the population growth lag from diauxie experiments. Population growth curves of strain ASC662 during diauxie experiments with 0.02% lactose and 0.005% glucose (blue) and controls where the lac operon is expressed throughout with supplemented ITPG (gray). Each line corresponds to 1 replicate. The delay between the blue growth curve and gray control to reach a given optical density features a long plateau, which corresponds to the population lag (inset). (B) Population lag in diauxie experiments with medium (0.02%, blue) and high (0.2%, orange) lactose concentrations for both lab strains and a diverse set of natural E. coli isolates. Points and error bars correspond to mean and standard error over at least 3 replicates with 0.005% glucose; the circled dot correspond to the growth curves shown in panel A. Note that the lag is reduced in all strains when more lactose is provided.
Fig 6.
Sensorless cells are expected to occur for many two-component systems in E. coli.
Distributions of average expression levels of sensor kinases across 28 conditions for 9 of the 28 annotated two-component systems in E. coli. Assuming Poisson fluctuations in expression, at least 5% of the cells are expected to be sensorless when the average kinase level is 3 proteins per cell or below (vertical dotted line). Note that 4 out of 9 two-component systems (highlighted in orange) feature sensorless cells in 6 conditions or more (data from [18]).