Research ArticleEVOLUTIONARY BIOLOGY

Temporal scaling of aging as an adaptive strategy of Escherichia coli

See allHide authors and affiliations

Science Advances  29 May 2019:
Vol. 5, no. 5, eaaw2069
DOI: 10.1126/sciadv.aaw2069
  • Fig. 1 Measuring E. coli lifespan distributions at single-cell resolution.

    (A) A three-dimensional (3D) model of microfluidic devices used to trap and isolate large number of single cells. Upper half of the picture with the arrows represents the main flow channel, where fresh, carbon-source–free media are supplied. The array of light blue rods represents E. coli single cells trapped in a 2D array of cell-sized dead-end chambers. (B) Fluorescence microscopy image of the microfluidic device loaded with E. coli cells. Each fluorescent dot corresponds to a single cell. Cyan, constitutively expressed fluorescent protein signal; red, PI viability staining. Horizontal scale bar, 10 µm; inset, 2× magnification. (C) Sample time-lapse PI images of mortality events, from early (top) to late (bottom), deaths. (D) A heatmap of all single-cell PI signal time series in an experimental cohort. The color values of each row correspond to the time series of each cell. Cells are sorted according to lifespan (top to bottom), from the shortest living to the longest living. See Materials and Methods for data processing.

  • Fig. 2 E. coli lifespan distribution can be characterized by exponential increases in mortality rates, i.e., the Gompertz law of mortality.

    (A) Binomial estimators of the hazard rates, shown in log scale. Error bars on the y axis are the 95% CI. Cell deaths are binned into discrete time intervals, which are marked by the x-axis error bars. The time bins are chosen so that the Nelson-Aalen (N-A) CIs in (B) do not overlap with each other. Shading covers regions that deviate from the exponential regime. (B) Hazard dynamics in the phase plane, where instantaneous hazards are plotted against the cumulative hazards. Data, binning intervals, and y axis are the same as those in (A). Error bars on the x axis are 95% CIs of N-A estimator. Survivorships equivalent to the exponentials of negative cumulative hazards are shown in the top axis. The brown dashed lines in both (A) and (B) represent the same naive linear fit to the Gompertz regime, while the red straight line represents maximum likelihood parametric estimations using the Gamma-Gompertz-Makeham (GGM) model. The inset provides a zoom-out view of the whole data range, while the main figure zooms in on the first 63% of cell deaths. The vertical dashed line inside the inset approximates the end of the exponential regime.

  • Fig. 3 The GSR of E. coli modulates aging rate.

    (A) A scheme representing relevant regulatory features of the GSR and, in particular, the functions of the genes rpoS and rssB. (B) Experimental and GGM model survivorship. The lifespan distributions for the wild-type (wt), ΔrpoS (lacking GSR), and ΔrssB (overexpressing the GSR) strains are measured multiple times by independent microfluidic experiments. Representing the experimental survivorship, color bands are the 95% CI of the Kaplan-Meier estimators. Colored dashed lines are GGM models whose parameters are estimated from maximum-likelihood (ML) methods. (C) Hazard rates estimated using only cell deaths within discrete time intervals (error bar markers), and GGM hazard models estimated from the whole dataset using ML methods. Error bars are similar to those in Fig. 2A. Parametric comparisons and GLM-based models (see “Experimental design and experimental variation analysis” section) indicate that ΔrssB prolongs longevity entirely by reducing aging rate and that ΔrpoS both accelerates aging and increases basal mortality rate. (D) Aging rates for each strain, estimated by three independent experiments and GLMs. Error bars represent 95% CI.

  • Fig. 4 Trade-offs between growth and maintenance mediated by the rpoS pathway and its fitness consequences.

    (A) Scheme of the ecological processes (dashed line) and regulatory relationships (solid line) involved in the trade-offs mediated by rpoS. (B) Experimentally measured fitness of ΔrpoS (red), wild-type (blue), and ΔrssB (green) strains as functions of time spent in feast (top) and famine (bottom) conditions. Fitness is defined as the logarithmic change of population sizes (for measurements, see Materials and Methods). Ranges of environmental conditions could be modeled by two transition rates between feast and famine, k1 and k2. (C) Fitness simulation results identifying environmental regimes favoring faster and slower aging strategies, exemplified by the three strains measured in (B). The color-coded regions identify environmental conditions under which one strain dominates over the other two. Environmental conditions are parameterized by lifestyle ratio k1/k2, the ratio between time spent in famine and feast conditions, and the average duration of famine episodes 1/k2 (see Materials and Methods for simulation details).

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/5/5/eaaw2069/DC1

    Fig. S1. The feast-and-famine life cycle of E. coli.

    Fig. S2. Time-of-death determination from PI fluorescence time series.

    Fig. S3. Gompertz law of mortality informs and constraints biochemical models of aging and mortality in growth-arrested E. coli cells.

    Fig. S4. Mortality statistics and goodness-of-fit of three independent experimental replicates for the three strains under study.

    Fig. S5. Experimental repeatability and variances.

    Fig. S6. Fabrication and implementation of the two-layer microfluidic chip.

    Fig. S7. Noise removal from focusing fluctuations using regularized estimation of PI signal.

    Fig. S8. Visualization and nonparametric statistics of variability among imaging positions within one experimental cohort.

    Table S1. AIC-based model selection among GLMs for the three experimental replication cohorts (i = 1, 2, 3) of wild-type strain.

    Table S2. AIC-based model selection among GLMs for the three experimental replication cohorts (i = 1, 2, 3) of ΔrssB strain.

    Table S3. AIC-based model selection among GLMs for the three experimental replication cohorts (i = 1, 2, 3) of ΔrpoS strain.

    Table S4. Testing for aging rate differences among the three strains using AIC and GLM.

    Table S5. List of plasticwares that leach carbon-supplying contaminants and their nonleaching replacements.

  • Supplementary Materials

    This PDF file includes:

    • Fig. S1. The feast-and-famine life cycle of E. coli.
    • Fig. S2. Time-of-death determination from PI fluorescence time series.
    • Fig. S3. Gompertz law of mortality informs and constraints biochemical models of aging and mortality in growth-arrested E. coli cells.
    • Fig. S4. Mortality statistics and goodness-of-fit of three independent experimental replicates for the three strains under study.
    • Fig. S5. Experimental repeatability and variances.
    • Fig. S6. Fabrication and implementation of the two-layer microfluidic chip.
    • Fig. S7. Noise removal from focusing fluctuations using regularized estimation of PI signal.
    • Fig. S8. Visualization and nonparametric statistics of variability among imaging positions within one experimental cohort.
    • Table S1. AIC-based model selection among GLMs for the three experimental replication cohorts (i = 1, 2, 3) of wild-type strain.
    • Table S2. AIC-based model selection among GLMs for the three experimental replication cohorts (i = 1, 2, 3) of ΔrssB strain.
    • Table S3. AIC-based model selection among GLMs for the three experimental replication cohorts (i = 1, 2, 3) of ΔrpoS strain.
    • Table S4. Testing for aging rate differences among the three strains using AIC and GLM.
    • Table S5. List of plasticwares that leach carbon-supplying contaminants and their nonleaching replacements.

    Download PDF

    Files in this Data Supplement:

Stay Connected to Science Advances

Navigate This Article