There is considerable understanding about how laboratory populations respond to predictable (constant or deteriorating environment) selection for single environmental variables such as temperature or pH. However, such insights may not apply when selection environments comprise multiple variables that fluctuate unpredictably, as is common in nature. To address this issue, we grew replicate laboratory populations of Escherichia coli in nutrient broth whose pH and concentrations of salt (NaCl) and hydrogen peroxide (H 2 O 2 ) were randomly changed daily. After~170 generations, the fitness of the selected populations had not increased in any of the three selection environments. However, these selected populations had significantly greater fitness in four novel environments which have no known fitness-correlation with tolerance to pH, NaCl or H 2 O 2 . Interestingly, contrary to expectations, hypermutators did not evolve. Instead, the selected populations evolved an increased ability for energy-dependent efflux activity that might enable them to throw out toxins, including antibiotics, from the cell at a faster rate. This provides an alternate mechanism for how evolvability can evolve in bacteria and potentially lead to broad-spectrum antibiotic resistance, even in the absence of prior antibiotic exposure. Given that environmental variability is increasing in nature, this might have serious consequences for public health.

Periodic bottlenecks play a major role in shaping the adaptive dynamics of natural and laboratory populations of asexual microbes. Here we study how they affect the 'Extent of Adaptation' (EoA), in such populations. EoA, the average fitness gain relative to the ancestor, is the quantity of interest in a large number of microbial experimental-evolution studies which assume that for any given bottleneck size (N0) and number of generations between bottlenecks (g), the harmonic mean size (HM=N0g) will predict the ensuing evolutionary dynamics. However, there are no theoretical or empirical validations for HM being a good predictor of EoA. Using experimentalevolution with Escherichia coli and individual-based simulations, we show that HM fails to predict EoA (i.e., higher N0g does not lead to higher EoA). This is because although higher g allows populations to arrive at superior benefits by entailing increased variation, it also reduces the efficacy of selection, which lowers EoA. We show that EoA can be maximized in evolution experiments by either maximizing N0 and/or minimizing g. We also conjecture that N0/g is a better predictor of EoA than N0g. Our results call for a re-evaluation of the role of population size in predicting fitness trajectories. They also aid in predicting adaptation in asexual populations, which has important evolutionary, epidemiological and economic implications.

Evolutionary studies over the last several decades have invoked fitness trade-offs to explain why species prefer some environments to others. However, the effects of population size on trade-offs and ecological specialization remain largely unknown. To complicate matters, trade-offs themselves have been visualized in multiple ways in the literature. Thus, it is not clear how population size can affect the various aspects of trade-offs. To address these issues, we conducted experimental evolution with Escherichia coli populations of two different sizes in two nutritionally limited environments, and studied fitness trade-offs from three different perspectives. We found that larger populations evolved greater fitness trade-offs, regardless of how trade-offs are conceptualized. Moreover, although larger populations adapted more to their selection conditions, they also became more maladapted to other environments, ultimately paying heavier costs of adaptation. To enhance the generalizability of our results, we further investigated the evolution of ecological specialization across six different environmental pairs, and found that larger populations specialized more frequently and evolved consistently steeper reaction norms of fitness. This is the first study to demonstrate a relationship between population size and fitness trade-offs, and the results are important in understanding the population genetics of ecological specialization and vulnerability to environmental changes.

There is considerable understanding about how laboratory populations respond to predictable (constant or deteriorating-environment) selection for single environmental variables like temperature or pH. However, such insights may not apply when selection environments comprise multiple variables that fluctuate unpredictably, as is common in nature. To address this issue, we grew replicate laboratory populations of E. coli in nutrient broth whose pH and concentrations of salt (NaCl) and hydrogen peroxide (H 2 O 2 ) were randomly changed daily. After ~170 generations, the fitness of the selected populations had not increased in any of the three selection environments. However, these selected populations had significantly greater fitness in four novel environments which have no known fitness-correlation with tolerance to pH, NaCl or H 2 O 2 . Interestingly, contrary to expectations, hypermutators did not evolve. Instead, the selected populations evolved an increased ability for energy dependent efflux activity that might enable them to throw out toxins, including antibiotics, from the cell at a faster rate. This provides an alternate mechanism for how evolvability can evolve in bacteria and potentially lead to broadspectrum antibiotic resistance, even in the absence of prior antibiotic exposure. Given that environmental variability is increasing in nature, this might have serious consequences for public-health.

Theoretical models of ecological specialisation commonly assume that adaptation to one environment leads to fitness reductions (costs) in others. However, experiments often fail to detect such costs. We addressed this conundrum using experimental evolution with Escherichia coli in several constant and fluctuating environments at multiple population sizes. We found that in fluctuating environments, smaller populations paid significant costs, but larger ones avoided them altogether. Contrastingly, in constant environments, larger populations paid more costs than the smaller ones. Overall, large population sizes and fluctuating environments led to cost avoidance only when present together. Mutational frequency distributions obtained from whole‐genome whole‐population sequencing revealed that the primary mechanism of cost avoidance was the enrichment of multiple beneficial mutations within the same lineage. Since the conditions revealed by our study for avoiding costs are widespread, it provides a novel explanation of the conundrum of why the costs expected in theory are rarely detected in experiments.