A link between aging and persistence

Abstract

Despite the various strategies that microorganisms have evolved to resist antibiotics, survival to drug treatments can be driven by subpopulations of susceptible bacteria in a transient state of dormancy. This phenotype, known as bacterial persistence, arises due to a natural and ubiquitous heterogeneity of growth states in bacterial populations. Nonetheless, the unifying mechanism of persistence remains unknown, with several pathways being able to trigger the phenotype. Here, we show that asymmetric damage partitioning, a form of cellular aging, produces the underlying phenotypic heterogeneity upon which persistence is triggered. Using single-cell microscopy and microfluidic devices, we demonstrate that deterministic asymmetry in exponential phase populations leads to a state of growth stability, which prevents the spontaneous formation of persisters. However, as populations approach stationary phase, aging bacteria-those inheriting more damage upon division-exhibit a sharper growth rate decline, increased probability of growth arrest, and higher persistence rates. These results indicate that persistence triggers are biased by bacterial asymmetry, thus acting upon the deterministic heterogeneity produced by cellular aging. This work suggests unifying mechanisms for persistence and offers new perspectives on the treatment of recalcitrant infections.

Keywords: Escherichia coli; aging; microfluidics; persistence.

Fig 1

Cellular aging drives phenotypic transitions in bacteria. (A) Bacterial aging is a function of cell pole inheritance. Whenever a cell divides, it gives to its daughters a new pole, generated from the fission site, and an old pole that carries accumulated damage. The inheritance of either pole upon the next generation can be used to differentiate old daughters from new daughters. (B and C) Phase planes depicting the transition between growth states dictated by aging and rejuvenation, combining mathematical modeling (32) and empirical parameterization (33). Model details are provided in the supplemental methods. (B) Due to damage inheritance, old daughters take longer to divide. Cells consecutively inheriting either new or old poles over generations stabilize around new or old growth equilibria. Cells age or rejuvenate as they move between equilibria. (C) Under oxidative stress, damage accumulation leads old lineages toward increasingly long doubling times, until division arrest. Through rejuvenation, new daughters continue to proliferate under the same conditions.

Fig 2

Absence of persistence during exponential phase. (A) Unstressed bacterial populations had stable elongation rates over time, without signs of transient growth arrest (n = 11,496 cells). Data were normalized to combine populations from distinct microfluidic designs. (B) Old daughters displayed longer doubling times (1.024 ± 0.078 min, n = 5,670) than new daughters (0.978 ± 0.064, n = 6,195; one-tailed t test, t = 34.505, df = 11,037, P < 0.001). These subpopulations reached distinct points of predicted physiological equilibrium, in which the doubling time of the mother equals that of a daughter. (C) Both new and old equilibria were stable, suggesting that these lineages are not subject to stochastic growth arrest (see Materials and Methods and Fig. S2 for details). (D and E) Once cells reached this state of stability, no persisters were observed upon exposure to 100 µg/mL ampicillin. (F) On the other hand, when the antibiotic was introduced before cells reached stability, non-growing individuals (as indicated by cell area) carried over from the previous stationary phase persisted the treatment.

Fig 3

Old daughters favored by persistence triggering. Triggered persistence from exponential phase populations. (A) Pre-treatment with photo-oxidative stress increased the difference between new and old daughter growth physiology (shown as elongation rates from birth to division or lysis). Upon exposure to 100 µg/mL streptomycin and subsequent recovery, no persisters were observed. Detail: lifespan of cells that arrested growth during treatment, followed from birth until lysis. (B) Exposure to 1 µg/mL streptomycin produced higher phenotypic heterogeneity (Fig. S4) prior to treatment with 30 µg/mL nalidixic acid (NA). No individuals resumed growth, suggesting that heterogeneity alone does not trigger persistence. (C) Using a known method to trigger persistence (41), we induced bacteriostasis through a short period of exposure to 50 µg/mL tetracycline (dashed lines), followed by a treatment with 100 µg/mL ampicillin (gray area). Whereas all cells arrested division through the bacteriostatic treatment, old daughter showed recovery once the treatment was removed. (D) Lifespan of cells that arrested growth, followed from birth until lysis. Persisters, the only cells that resumed growth without lysing, are indicated by arrows. At the population level, the diagram (E) shows the total persistence triggered by tetracycline followed by ampicillin treatment.

Fig 4

Old daughters are more sensitive to growth arrest triggered by stationary phase. (A) To induce a transition from exponential to stationary phase, we gradually starved cells of nutrients (dashed lines indicate transition interval), followed by ampicillin exposure and recovery. (B) During starvation, old daughters continued to display slower elongation rates (0.0072 ± 0.0032 min⁻¹, n = 196) than new daughters (0.0078 ± 0.0028 min⁻¹, n = 183; one-tailed t test, t = 1.88, df = 374.34, P = 0.030). More importantly, old daughters displayed a steeper decline than new daughters, as indicated by linear models (solid lines). (C) There was an increase in doubling times and slopes between exponential (detail) and stationary phase (large phase plane). (D) As populations approached stationary phase, old lineages lost equilibrium (β = 0.502, σ₁ = 0.857) and arrested growth, whereas new daughters continued to proliferate (β = 0.447, σ₁ = 0.414). Error bars = 95% confidence interval.

Fig 5

Stationary aging cells show higher persistence rates. (A) Damage accumulation was induced through overnight exposure to 1 µg/mL streptomycin. Cells pre-treated exhibited larger aggregates (0.244 ± 0.171 a.u., n = 358) than non-treated cells (0.118 ± 0.126 a.u., n = 459; one-way ANOVA, F = 88.8, P < 0.001). (B) These aggregates occupied a larger area of the cell, thus increasing the anchoring of damage at old poles. This allowed us for using IbpA-YFP-labeled aggregates as a marker for old daughters. (C) Heatmap and phase contrast image of cells expressing IbpA-YFP-labeled aggregates after overnight streptomycin treatment, visualized immediately after inoculation onto agarose pads. (D) Cultures pre-treated with 1 µg/mL streptomycin were loaded onto agarose pads containing 100 µg/mL ampicillin and followed until lysis. A segmented fit to the killing curve indicated three break points (1.95, 3.05, and 7.48 h). Cells that survived to s3 were characterized as persisters. The detail shows separate killing curves according to the presence of aging markers. (E) The damage-free subpopulation had higher death rates from 1.95 to 3.05 h. (F) The frequency of persisters was higher among aging cells. Error bars = 95% CI.