Bacterial Evolution in High-Osmolarity Environments

Abstract

Bacteria must maintain a cytosolic osmolarity higher than that of their environment in order to take up water. High-osmolarity environments therefore present formidable stress to bacteria. To explore the evolutionary mechanisms by which bacteria adapt to high-osmolarity environments, we selected Escherichia coli in media with a variety of osmolytes and concentrations for 250 generations. Adaptation was osmolyte dependent, with sorbitol stress generally resulting in increased fitness under conditions with higher osmolarity, while selection in high concentrations of proline resulted in increased fitness specifically on proline. Consistent with these phenotypes, sequencing of the evolved populations showed that passaging in proline resulted in specific mutations in an associated metabolic pathway that increased the ability to utilize proline for growth, while evolution in sorbitol resulted in mutations in many different genes that generally resulted in improved growth under high-osmolarity conditions at the expense of growth at low osmolarity. High osmolarity decreased the growth rate but increased the mean cell volume compared with growth on proline as the sole carbon source, demonstrating that osmolarity-induced changes in growth rate and cell size follow an orthogonal relationship from the classical Growth Law relating cell size and nutrient quality. Isolates from a sorbitol-evolved population that captured the likely temporal sequence of mutations revealed by metagenomic sequencing demonstrated a trade-off between growth at high osmolarity and growth at low osmolarity. Our report highlights the utility of experimental evolution for dissecting complex cellular networks and environmental interactions, particularly in the case of behaviors that can involve both specific and general metabolic stressors.IMPORTANCE For bacteria, maintaining higher internal solute concentrations than those present in the environment allows cells to take up water. As a result, survival is challenging in high-osmolarity environments. To investigate how bacteria adapt to high-osmolarity environments, we maintained Escherichia coli in a variety of high-osmolarity solutions for hundreds of generations. We found that the evolved populations adopted different strategies to improve their growth rates depending on the osmotic passaging condition, either generally adapting to high-osmolarity conditions or better metabolizing the osmolyte as a carbon source. Single-cell imaging demonstrated that enhanced fitness was coupled to faster growth, and metagenomic sequencing revealed mutations that reflected growth trade-offs across osmolarities. Our study demonstrated the utility of long-term evolution experiments for probing adaptation occurring during environmental stress.

Keywords: cell morphology; cell shape; osmolytes; osmotic adaptation; proline; sorbitol; stress response; sucrose.

FIG 1

Schematic of evolution experiment across passaging conditions that result in a wide range of ancestral growth behaviors. (A) The ancestral strain (TC1407) was passaged daily with 1:100 dilutions as four independent populations in each of 11 environments of DM25 medium supplemented with osmolytes as denoted on the plate. Empty wells were used as negative controls. Gl, glycine betaine; Na, sodium chloride; Pr, proline; So, sorbitol; Su, sucrose. (B to F) Higher osmolarity generally inhibits growth. TC1407 cells were grown in the passaging concentrations of glycine betaine (B), sorbitol (C), NaCl (D), sucrose (E), and proline (F). For all osmolytes, growth during the first 10 h was inhibited more at higher osmolyte concentrations. Glycine betaine and NaCl also decreased the carrying capacity, signifying less-efficient use of glucose. The carrying capacity was greatly increased in sorbitol and in 0.5 M proline due to utilization of the osmolyte as a carbon source. Growth in sucrose was relatively unaffected by osmolarity. Growth curves represent averages of results from n = 4 replicates, and shaded regions represent standard errors of the mean (SEM). The SEM is small at some time points and is thus partially obscured by the lines.

FIG 2

All populations evolved in 0.5 M proline or sorbitol grew better than the ancestor in their passaging medium. (A and B) The populations evolved in 0.5 sorbitol (A) or 0.5 M proline (B) grew faster and to a higher OD after 24 h than the ancestor in their passaging medium. For both osmolytes, there was a range of growth behaviors of the four populations, suggesting different adaptations. Growth curves represent averages of results from n = 3 replicates for each of the evolved populations and averages of results from n = 6 replicates for the ancestor, with shaded regions representing standard errors of the mean (SEM). The black arrowhead in panel B indicates the time point at which the cultures likely shifted from glucose to proline utilization. The SEM is small at some time points and is thus partially obscured by the lines. (C and D) To account for the different growth kinetics of the ancestor and evolved populations in panels A and B, we analyzed growth rate as a function of OD₆₀₀. For sorbitol (C) and proline (D), the growth rate was consistently higher in the evolved populations. The black arrowhead marks the same time point as that in panel B, when growth rate briefly decreased. The SEM is small at some time points and is thus partially obscured by the lines.

FIG 3

Sorbitol-evolved populations showed increased fitness at high osmolarity, while proline-evolved populations showed increased fitness specifically in proline. (A) Schematic of calculation of levels of relative fitness from growth curves (see Fig. S2 in the supplemental material), which we define as the ratio of half the saturation OD of an evolved population (${\text{OD}}_{\text{max}}^{\text{ev}}$) to the OD of the ancestor (OD^anc) at the same time point (T_1/2). max, maximum. (B) The So0.5-1 population had increased and monotonically increasing fitness relative to the ancestor in both sorbitol (purple) and sucrose (pink) for concentrations of ≥0.25 M and slightly reduced fitness in DM25 or DM25 + 0.125 M osmolyte, indicating that the population had adapted to higher osmolarity conditions (n = 4). (C and D) The maximum growth rates of the ancestor and the sorbitol-evolved populations in sorbitol (C) and sucrose (D) decreased steadily with increasing osmolarity, with a faster decrease in the ancestor leading to larger growth rate differences at higher osmolarities (n = 4). (E) The Pr0.5-1 population exhibited monotonically increasing fitness relative to the ancestor at all concentrations of proline (blue) and neutral (i.e., unchanged) fitness without proline. Growth rates in sucrose (pink) were virtually identical to those of the ancestor at all concentrations (n = 4).

FIG 4

Fitness advantages relative to the ancestor are correlated with increases in steady-state growth rates of the sorbitol- and proline-evolved populations. (A) Cells from the So0.5-1 population grew slower than cells from the ancestor population in DM2500 under steady-state conditions (P = 8.1 × 10⁻⁴, t test), and Pr0.5-1 cells exhibited growth rates similar to those of the ancestor (P = 0.61, t test), consistent with fitness measurements presented in Fig. 3B and E, respectively. The growth rate for each cell was defined as its mean instantaneous growth rate measured over 15 min of imaging (n > 77 cells for the ancestor, n > 97 cells for the sorbitol-evolved population, and n > 218 cells for the proline-evolved population). Horizontal lines represent means across cells. (B) So0.5-1 cells grew more quickly at steady-state in DM2500 + 0.5 M sucrose than ancestor cells (P < 10⁻⁴, t test) and Pr0.5-1 cells (P < 10⁻⁴, t test), which had similar growth rates, which was again consistent with the fitness measurements. The growth rate for each cell was defined as its mean instantaneous growth rate over 21 min of imaging (n > 405 cells for the ancestor, n > 125 cells for the sorbitol-evolved population, and n > 291 cells for the proline-evolved population). Horizontal lines represent means. (C) Pr0.5-1 cells grew more quickly at steady state in DM + 0.125 M proline than ancestor cells (P < 10⁻⁴, t test), demonstrating that passaging in proline led to an enhanced ability to metabolize proline. The growth rate for each cell was defined as its mean instantaneous growth rate over 21 min of imaging (n > 159 cells for the ancestor population and n > 148 cells for the proline-evolved population). Horizontal lines represent means.

FIG 5

Cell size does not scale with growth rate across media with different osmolytes. (A) Typical images of single cells during exponential growth from the ancestor and So0.5-1 and Pr0.5-1 populations in DM2500, DM2500 + 0.5 M sucrose, and DM + 0.125 M proline (n > 160 cells were imaged for each population). (B to D) Normalized histograms of cell volumes of the ancestor (B), So0.5-1 (C), and Pr0.5-1 (D) populations grown in DM2500, DM2500 + 0.5 M sucrose, and DM + 0.125 M proline. All three populations were larger in DM2500 than in DM2500 + 0.5 M sucrose, as expected from the Growth Law based on growth rates (Fig. 4A and B), but the ancestor and Pr0.5-1 populations were smaller in DM + 0.125 M proline than in DM2500 + 0.5 M sucrose despite having a higher growth rate with proline as a carbon source (Fig. 4B and C). Vertical lines represent medians (n > 160 cells were imaged for each population). (E) Comparison of steady-state growth rates and mean cell volumes across populations and growth conditions, demonstrating the lack of agreement with the Growth Law across media or population comparisons, which would predict that the points would fall along a single line. Error bars represent the standard errors of the mean. Some error bars are not visible because the error is smaller than the size of the marker (n > 77 cells were imaged for each population).

FIG 6

Isolates of the sorbitol-evolved population exhibited fitness trade-offs consistent with continued adaptation to high osmolarity. (A and B) An isolate that contained a mutation in topA alone (red) and an isolate that also had a mutation in prc (green) had faster growth than the ancestor in DM25 + 0.5 M sorbitol (A) but slower growth in DM2500 (B). Lines represent averages of n = 6 growth curves, with shaded regions representing standard errors of the mean (SEM). The growth curve of the prc topA isolate was consistently below that of the topA isolate. The SEM is small at some time points and is thus partially obscured by the lines. (C) Time-lapse imaging of outgrowth in DM2500 of stationary-phase prc topA cells previously grown in DM2500 revealed cells that had already lysed when imaging commenced (blue arrowhead), along with normal cells (white arrowheads) and rounded cells (orange arrowheads) that eventually lysed (red arrowheads). The SEM is small at some time points and is thus partially obscured by the lines. (D and E) The cultures in panels A and B were passaged one additional time in the same media to quantify the changes in growth behaviors when the stationary phase prior to growth curve measurement was reached in a growth environment more similar to the passage conditions. They were grown in either DM25 + 0.5 M sorbitol (D) or DM2500 (E). Lines represent averages of n = 6 growth curves, with shaded regions representing SEM. By contrast to panel A, the growth curve of the prc topA strain was consistently above that of the topA strain in DM25 + 0.5 M sorbitol (D) but was substantially lower after an additional passage in DM2500 (E) than in panel B. (F and G) Genes with mutations identified at >5% prevalence in one of the four So0.5 and Pr0.5 evolved populations were classified by PANTHER Protein Class using PantherDB (http://pantherdb.org) (54). The contribution of each mutation to the histogram was scaled by its prevalence in the population; the orange bars in panel G represent the subset of mutations in the put operon. The mutations in the sorbitol-evolved populations represented a wide range of functional categories (F), while the mutations in the proline-evolved populations were predominantly related to proline utilization. In the four So0.5 populations, 19 of the 44 genes did not have a classification; scaled by prevalence, those genes accounted for 40% of the total. In the four Pr0.5 populations, 25 of the 50 genes did not have a classification; scaled by prevalence, those genes accounted for 34.3% of the total.