Results
Evolution of compensatory mutations. In the ancestor to our evolution experiment, mutations that inactivate the LacI repressor, and lead to constitutive expression of the lac operon, confer a fitness cost of approximately 2.2% during growth in glucose (Quan et al. 2012). To determine whether this cost is compensated during evolution in environments containing both glucose and lactose resources, and, if so, whether this compensation differed depending on the presentation of the two resources, we reverted evolved lacI-mutations in nine clones isolated from populations selected in lactose alone (Lac, three populations), long-term switching of glucose and lactose (G_L, one population), or daily switching of glucose and lactose (G/L, five populations). We found that the cost of lacI - in glucose was significantly changed only in evolved clone G/L4 where the mutation had become beneficial (Fig. 2; Dunnett’s Test, P< 0.001). This result indicates the presence of sign epistasis, in that the lacI- mutation changed from being deleterious in the ancestor to being beneficial in the evolved background. No other difference in evolved lacI- costs was detectable among other evolved clones (χ 2 = 8.68, df = 9, P = 0.467).
Effect of environment on compensatory mutations. Although the cost of lacI- was significantly changed in only one clone when compared to the ancestor, there may be changes in costs apparent when grouping clones based on their evolution environment. To test this, we compared the effect of the lacI- mutation on fitness in glucose of each evolution environment group (strains evolved in Lac, G/L, or G_L). We expect clones evolved in the presence of fluctuations of glucose and lactose (i.e., G/L and G_L selection environments) to have a reduced fitness cost of the lacI- mutation in glucose because compensation for the original cost would provide an advantage. By contrast, selection for compensatory mutations was expected to be reduced or absent in the lactose only environment. We found no significant difference in cost of the lacI- mutation measured in the glucose environment among clones isolated from different evolution environments whether or not the outlier clone, G/L4, was included (ANOVA with G/L4: F2,106 = 2.65, P = 0.75; without G/L4: F2,100 = 1.29, P = 0.28).
Pleiotropic effect of lacI- compensation. Given that compensation for the cost of constitutive lac expression evidently can occur, one explanation for the low frequency at which it does occur is that it imposes a correlated cost in lactose. For example, it might be that compensation to constitutive expression of the lac operon involves a reduction in the maximum level of lac expression, perhaps reducing fitness in lactose and thereby causing compensation to be selectively disfavored. To test this possibility, we examined the fitness effect of lacI - mutations across glucose and lactose environments. To determine if there was any trend of a lower cost oflacI - in glucose corresponding to a lower benefit in lactose, we determined the relationship between the fitness effect of lacI - mutations in glucose and lactose across all strains. We found a marginally significant positive correlation between fitness in the two environments, indicating that a low cost of the lacI- mutation in glucose was, if anything, associated with an increased benefit in lactose (Fig. 3; Spearman’s rank correlation, rho = 0.72, P = 0.02). The significance of the correlation is dependent on the G/L4 evolved clone that compensated for the lacI- mutations cost in glucose. When that clone was omitted from the analysis, the correlation was no longer significant, though was still positive (rho = 0.63,P = 0.07). Focusing on the G/L4 clone revealed that thelacI - mutation is beneficial in glucose and its effect in lactose is significantly higher than in three of the evolved clones tested, as well as the ancestor (G_L1, G/L1, and G/L2; Dunnett’s test, P< 0.05). Together, these results indicate that there is no trade-off with fitness in lactose that limits selection for compensation of lacI - costs in glucose.
Mechanisms of lowered costs. To determine if lac operon expression is associated with changes in the fitness effects of thelacI - mutation, we measured lac operon expression in glucose using a reporter that is controlled by the promoter region, Plac, that drives expression of the lacoperon (Fig. 4). Expression of the lac operon contributes to the cost of constitutive expression, so we expected a negative relationship, such that clones that had higher lac operon expression would have lower fitness in the glucose environment (i.e., a higher cost) (Dekel and Alon 2005, Stoebel et al. 2008). In fact, there was no correlation (Fig. 5A; Spearman’s rank correlation, rho = 0.12, P = 0.78). This is especially surprising because half of the evolved strains had significantly higher lac expression in glucose than the ancestorlacI -, so that an effect of lac expression on fitness could have been detected (Dunnett’s test: G_L1, G_L3, G/L4, Lac3, Lac4, Lac6 P < 0.05). That increased expression was not associated with any fitness cost might indicate the action of compensation to some portion of the cost that would otherwise be associated with increased lac expression. Alternatively, there could be a limit to the cost associated with constitutive lacoperon expression (Eames and Kortemme 2012), although the model most analogous to the situation prevailing in our experiments predicts exponentially increasing costs with increasing expression (Dekel and Alon 2005).
All clones except G/L4 had similar lacI- associated fitness costs when compared to the ancestor, but clones varied when it came to differences in expression compared to the lacI - ancestor. G/L4 had equal lowest lac expression in glucose (Dunnett’s test,P < 0.05; except for G/L3, P = 0.90; and G/L6,P = 0.46), and all G/L clones had lower expression in glucose when compared to all Lac clones. Together these results indicate that there was some differential evolution of lac expression based on environment, but that consequences do not consistently map to fitness effects: clones with similar expression levels in glucose have differentlacI- fitness effects (G/L4 compared to G/L3 and G/L6), and clones with similar lacI- fitness effects have different expression (G/L compared to Lac clones). Evidently, reduced lacoperon expression cannot explain all of the decreased costs in glucose and alleviation of the cost is dependent on other mutations in the evolved background.
Finally, our reporter strains allow us to address a trade-off betweenlac expression and fitness related to compensation. If compensation of costs due to constitutive lac operon expression in glucose reduce the benefit of high expression in the lactose environment, compensation might impose a net cost and not be selected. To test this, we compared fitness and expression effects of lacI-mutations in lactose. We expected that if higher lac expression was selected due to increasing fitness in lactose, there would be a positive correlation between lac expression and fitness. In fact, although all evolved clones had increased lac expression, changes were not correlated with fitness (Fig. 5B; Spearman correlation, rho = -0.18, P = 0.64). Moreover, although the lacI- mutation conferred one of the biggest benefits when added to the G/L4 clone,lac expression was not significantly different in this clone compared to other evolved clones (Dunnett’s test, P> 0.05). These results suggest that the benefit of higher expression depends on the genetic background in which it occurs and that most of the changes in maximum lac expression are not caused bylacI - itself but by differences in the broader genetic background.