Growth resumption from stationary phase reveals memory in Escherichia coli cultures

Abstract

Frequent changes in nutrient availability often result in repeated cycles of bacterial growth and dormancy. The timing of growth resumption can differ among isogenic cells and delayed growth resumption can lead to antibiotic tolerant persisters. Here we describe a correlation between the timing of entry into stationary phase and resuming growth in the next period of cell proliferation. E. coli cells can follow a last in first out rule: the last ones to shut down their metabolism in the beginning of stationary phase are the first to recover in response to nutrients. This memory effect can last for several days in stationary phase and is not influenced by environmental changes. We observe that the speed and heterogeneity of growth resumption depends on the carbon source. A good carbon source (glucose) can promote rapid growth resumption even at low concentrations, and is seen to act more like a signal than a growth substrate. Heterogeneous growth resumption can protect the population from adverse effect of stress, investigated here using heat-shock, because the stress-resilient dormant cells are always present.

Figure 1. Heterogeneous growth resumption from stationary phase.

(a) Assay principle. Cells with Crimson expression induced are grown until they reach stationary phase. Then Crimson inducer is removed and GFP inducer together with a new carbon source is added. Cells resuming growth at first become GFP-positive and later dilute Crimson by cell division. (b–e) Growth resumption in response to different carbon sourses. Cells were grown in MOPS 0.1% glycerol for three days and growth resumption was initiated by adding carbon sources indicated. Pseudocolours indicate the amount of cells.

Figure 2. Growth resumption order depends on the entrance into stationary phase.

(a) Experimental scheme to mark differences at the entrance into stationary phase. B and C refer to subsequent panels. (b) Crimson expression profile of stationary phase culture induced at the end of the growth phase. Two subpopulations are clearly distinguished from each other and the whole distribution is bimodal (Hartigan’s dip test for unimodality, D = 0.0061, p < 2.2 × 10⁻¹⁶). (c) Flow cytometry analysis of cells resuming growth. Crimson positive and Crimson negative subpopulations are gated separately for further analysis. (d) Growth resumption of cells from two different (Crimson-positive and Crimson-negative) subpopulation. Number of dormant (GFP-negative) cells is plotted at every timepoint. Values are an average from three independent experiments and error bars indicate s.e.m. Time-courses were analysed using linear regression t-test in Graphpad software package (GraphPad Software, USA) and found to be significantly different from each other (p = 0.0114).

Figure 3. Microscopy analysis of growth resumption.

(a) Examples of microcolonies exhibiting heterogeneous Crimson staining. (b) Growth resumption in microcolony after addition of gluconate and GFP inducer (IPTG). (c) Distribution of Crimson expression levels at 0 h timepoint. Black line separates low and high Crimson populations. (d) Growth resumption of cells with low and high Crimson content. Cells were grouped according to their Crimson expression levels at 0 h timepoint. Values are an average from three independent experiments and error bars indicate s.e.m. Time-courses were analysed using linear regression t-test in Graphpad software package and found to be significantly different from each other (p < 0.0001).

Figure 4. Growth resumption of strains with altered (p)ppGpp levels.

(a) Stationary phase distribution of Crimson expression (0 h) and growth resumption in ΔrelA strain. The distribution of Crimson expression in stationary phase is bimodal (Hartigan’s dip test for unimodality, D = 0.0104, p < 2.2 × 10⁻¹⁶). (b) Growth resumption timing correlates with Crimson expression levels in ΔrelA strain. Time-courses were analysed using linear regression t-test in Graphpad software package and found to be significantly different from each other (p < 0.0001). (c) Stationary phase distribution of Crimson expression (0 h) and growth resumption in ΔrelA ΔspoT strain. The distribution of Crimson expression in stationary phase is bimodal (Hartigan’s dip test for unimodality, D = 0.0175, p < 2.2 × 10⁻¹⁶). (d) Growth resumption timing correlates with Crimson expression levels in ΔrelA ΔspoT strain. Time-courses were analysed using linear regression t-test in Graphpad software package and found to be significantly different from each other (p < 0.0001). Values are an average from three independent experiments and error bars indicate s.e.m.

Figure 5. Change of enironment or heat-shock during the stationary phase does not change the order of growth resumption.

(a) Growth resumption of cells in response to gluconate with or without replacing the stationary phase medium with water. Time-courses were analysed using linear regression t-test in Graphpad software package and found to be significantly different from each other (p = 0.0088 for control, p = 0.0402 for water). (b) Growth resumption in response to glucose with or without heat-shock. Time-courses were analysed using linear regression t-test in Graphpad software package and found to be significantly different from each other (p = 0.0012 for control, p < 0.0001 for heat-shock). Values are an average from three independent experiments and error bars indicate s.e.m.

Figure 6. Heterogeneous growth resumption avoids periods of stress vulnerability.

(a) Growth resumption is homogeneous in response to glucose and heterogeneous in response to gluconate and glycerol. Pseudocolours indicate cell density. (b) Heat sensitivity of cultures resuming growth. A small aliquote was removed from culture, heat-shocked (HS) for 5 minutes at 55 degrees and plated for cfu determination at indicated times. Samples without heat-shock were plated as controls at the same time. Values are an average from three independent experiments and error bars indicate s.e.m.