Introduction
Fluctuating environments pose several challenges to evolving populations. While some potential adaptations might confer benefits across all relevant selection regimes (Bennett and Lenski 1999; Buckling et al. 2000; Buckling et al. 2007; Kassen and Bell 1998; Satterwhite and Cooper 2015), others will confer benefits in some and costs in others (Bailey and Kassen 2012; Jasmin and Kassen 2007; Lee and Marx 2012; McGee et al. 2015; Roemhild et al. 2015). Indeed, even if unconditionally beneficial mutations are initially available, they are likely to become less common over time (Martin and Lenormand 2015; Satterwhite and Cooper 2015; Schick et al. 2015). When a mutation that confers a benefit in one environment, and a cost in another, fixes in a population, it creates selection for subsequent mutations that compensate for that cost (Maisnier-Patin et al. 2002; Moore et al. 2000; Moura de Sousa et al. 2017; Wood et al. 2013).
Compensatory mutations have long been used in molecular genetic studies as a tool to identify physical and genetic changes that can suppress the effects of a focal mutation, thereby identifying interacting components (Blank et al. 2014; Jarvik and Botstein 1975; Kacar et al. 2017; Manson 2000; Ponmani and Munavar 2014; van Leeuwen et al. 2016). Increasingly, they are also recognized as being important in broadening the scope of evolutionary trajectories a population can follow (Szamecz et al. 2014; Zee et al. 2014), allowing adaptations to be selected that might otherwise prove to be evolutionary dead-ends (Covert et al. 2013; Harrison et al. 2015), and influencing the ability of populations to simultaneously adapt to multiple environments (Melnyk et al. 2017).
Mutation interactions arising following selection in fluctuating environments have been extensively studied in the context of the evolution of antibiotic resistance. Antibiotic resistance mutations often confer a cost to bacteria in antibiotic free environments (Moura de Sousa et al. 2017; Nilsson et al. 2003; Rozen et al. 2007). This cost generates selection for subsequent compensatory mutations that relieve the cost (Levin et al. 2000; Björkman et al. 1998; Björkman et al. 2000; Nagaev et al. 2001). Compensation allows resistance mutations to be maintained when they would normally be selected against, influencing short-term evolutionary outcomes and perhaps longer-term potential. Similar patterns of compensatory mutations depending on earlier resistance mutations for their benefit have been seen in studies of bacterial resistance to bacteriophage (Lenski 1988; Wielgloss et al. 2016). Compensation during evolution has also been found to occur to overcome the loss of essential genes (Blank et al. 2014), the negative effects of synonymous (Knöppel et al. 2016) and gene deletion (Szamecz et al. 2014) mutations, and to restore a social trait (Zee et al. 2014).
In contrast to studies that have focused on mutations that compensate for a specific genetic perturbation, few studies have examined compensation during more general adaptation to an environment, especially when this adaptation involves repeating rounds of selection in contrasting environments. This distinction might be important. Compensation to a specific genetic perturbation, as to an antibiotic resistance mutation or a deletion of a focal gene, is thought to generally act locally to reverse the costly effect (Brandis et al. 2012; Filteau et al. 2015; Szamecz et al. 2014), although it can also arise in pathways unrelated to the perturbation (Blank et al. 2014). At least in experimentally evolving populations, adaptation often involves mutations in regulatory genes that are likely to have highly pleiotropic consequences (Cooper et al. 2003; Kurlandzka et al. 1991; MacLean et al. 2004; Rosenzweig et al. 1994). If costs of such pleiotropic mutations are revealed following an environmental change, it is not clear how subsequent compensatory mutations might affect fitness in the original environment. Indeed, it is easy to imagine that compensation causing a reduction in the cost of a focal mutation in a new environment might be associated with a reduction of the original benefit. In that case, reversion to the original environment might select for reversal of the effects of the compensatory mutation. This could occur through its direct reversion or through a second compensatory mutation, creating potentially complex patterns of environmentally dependent epistatic interactions between selected mutations.
The particular nature of the environmental fluctuations a population is exposed to is expected to play a major role in the selection of compensatory mutations. In a rapidly changing environment, mutations that increase in frequency are likely to confer a net benefit across the different environments (Buckling et al. 2007; Melnyk et al. 2017; Turner and Elena 2000). With this limitation, such mutations can only confer at most relatively small costs in any one environmental component so that the strength of selection for compensation may be small (Poon and Chao 2005a; Poon and Chao 2006). In a more slowly fluctuating environment, mutations selected in one component might fix before the population experiences a second component in which they might confer substantial costs (Bennett and Lenski 2007; Kassen and Bell 1998; Phillips et al. 2016). Such differential costs are consistent with the generally higher between-environment trade-offs seen in populations selected in slowly compared to quickly fluctuating environments (Bono et al. 2017; Satterwhite and Cooper 2015; Schick et al. 2015).
We examine compensation to an adaptive mutation selected in a series of experimentally evolved populations selected in environments that contained either lactose or glucose alone or a combination of lactose and glucose fluctuating daily or every 2000 generations (Cooper and Lenski 2010). Mutational inactivation of the LacI repressor was rapidly selected in many of the replicate bacterial populations that were selected in the presence of lactose (Quan et al. 2012). Loss of LacI causes the lac operon, a set of genes required for utilization of lactose, to be constitutively expressed (Markiewicz et al. 1994; Quan et al. 2012). When engineered into the ancestor of the evolution experiment, constitutive expression of the lac operon provided a benefit of ~9% during growth in lactose by shortening the lag time before resumption of growth following transfer into fresh medium (Quan et al. 2012). It also conferred a cost of ~3% in an environment containing glucose as the sole resource, probably due to some combination of the energetic cost of expressing unnecessary genes and toxicity of the LacY permease (Dekel and Alon 2005; Stoebel et al. 2008; Quan et al. 2012; Eames and Kortemme 2012). In populations selected in environments containing both lactose and glucose, this trade-off in the effect of lacI- mutations creates potential for subsequent mutations to provide a fitness benefit by compensating for the cost of the mutation in glucose.
We test whether evolved populations that fixed the lacI-mutation, and therefore constitutively express the lacoperon, evolved mechanisms that alleviate the cost of this expression during growth in glucose and, if so, whether this compensation comes at a cost of the benefit conferred by lacI- in lactose. To do this, we isolated strains from populations evolved for 8,000 generations in lactose, glucose, and combinations of both fluctuating daily and every 2,000 generations. The lacI- mutation was reverted in those strains that had substituted it, and its effect on fitness determined. We found that the fitness cost of lacI - in glucose was variable, including, in one strain, becoming beneficial, but did not differ consistently between populations evolved in lactose only, where compensation is not expected to be selected, and in environments containing glucose, where it is. Similarly, strains varied in their relationship between the fitness benefit conferred by the lacI - mutation in lactose and costs in glucose, but this variation did not depend on their selection environment. Together, these results demonstrate the potential for the action of compensatory mutations to influence costs of adaptation but indicate that their effects may either be idiosyncratic or be overwhelmed by the effects of additionally selected mutations.