Discussion
Few studies have focused on the influence of compensatory mutations during adaptation to a general environment rather than to a specific genetic perturbation. Our results demonstrate that compensation to deleterious consequences of an adaptive mutation is possible but is rare among evolved clones in this experiment. Of the nine clones tested, only one clone evolved a mechanism to alleviate the cost of constitutivelac operon expression in glucose (Fig. 2). No differences in compensation were evident comparing groups of clones evolved in different environments, including some that were and some that were not expected to select for compensation.
Compensation of deleterious mutation effects has been observed in many systems and contexts – including compensation of costs due to antibiotic resistance, adaptation, and gene deletion (Levin et al. 2000; Björkman et al. 1998; Björkman et al. 2000; Nagaev et al. 2001, Szamecz et al. 2014; Zee et al. 2014, Blank et al. 2014, Knöppel et al. 2016). A common theme is that compensation is readily selected, consistent with their being many available genetic mechanisms. For example, one study found 68% of single-gene deletion slow growth mutants could be compensated to restore growth to near wild type within ~400 generations (Szamecz et al. 2014). Indeed, compensation to the costs of the lacI- mutation apparent in glucose containing environments was possible in our experiment, but occurred in only one of the nine clones we examined. One explanation is that compensation is limited not by mutational opportunity but by selection for those mutations. Our populations evolved asexually so that beneficial mutations of large effect outcompete those of smaller effect (Gerrish and Lenski 1998). Thus, even if they arise, small benefit compensatory mutations might be outcompeted until mutations of large effect are exhausted. Our populations exhibit higher fitness gains at early versus later time points (Satterwhite and Cooper 2015), however, the latter gains may still be larger than the small ~2.2% cost of lacI - in glucose. Indeed, the G_L populations steadily increased in glucose fitness up to 6000 generations so compensatory mutations may not be sufficiently competitive to be selected (Satterwhite and Cooper 2015). By contrast, G/L populations had small or undetectable fitness change across glucose and lactose environments after 4000 generations, which would have allowed selection of small benefit mutations and may be why only a G/L clone was suggested to have compensatory mutations (Satterwhite and Cooper 2015).
Another possible explanation for the rarity of compensation is that multiple mutations are required to compensate the cost of lacI - in glucose without affecting its benefit in lactose (Poon and Chao 2005b). Multiple compensatory mutations will take longer to fix and will be rare because successful compensation can depend on the order in which each required mutation occurs (Gong et al. 2013), the presence/absence of other mutations in the genetic background (Shah et al. 2015; Lunzer et al. 2010), and a clone’s fitness at each step relative to others in the population which can subject an intermediate clone to being purged by purifying selection (Gerrish and Lenski 1998). Replacing lacI - with a functional LacI repressor in the G/L4 clone, in which compensation did occur, significantly reduces fitness in glucose and expression of the lac genes. In other words, this clone has somehow rewired the lac network such that lac operon expression is beneficial for growth in glucose. The basis of this rewiring is unknown but might depend on multiple mutations as is the case, for example, in the case of evolved citrate utilization selected in a population started from the same ancestor as the one used here (Blount et al. 2008, 2012).
We consider that LacY is the best candidate lac gene to be involved in some new compensatory interaction. Not only is it the likely source of the cost of constitutive lac expression (Eames and Kortemme 2012), but it can also support glucose uptake, though, as characterized, that requires a mutational change in the enzyme that did not occur in clone G/L4 (Sahin-Tóth et al. 2001, Gram and Brooker 1992; King and Wilson 1990). Although glucose transport is not likely to limit growth at the concentration used in this experiment it is possible that an alternative import mechanism could provide an advantage by reducing dependence on the phosphotransferase system (Jahreis et al. 2008; Ferenci 1996; Postma et al. 1993). This system is a common mutational target in populations evolved from the same ancestor and selected in a glucose environment, suggesting it is not optimized for growth in the selective environment prevailing in our experiment (Woods et al. 2006). Sequencing of a clone isolated from the G/L4 population identified two genes that were not mutated in the other populations we examined here or in lines that evolved only in glucose. Of these, one, sohB, is a candidate for interacting with LacY. SohB is a peptidase active, along with LacY, in the cell periplasm (Baird et al. 1991). Although the function of SohB is not well characterized, it has been shown to be able to compensate for loss of other peptidases involved in periplasm protein recycling, suggesting a possibility of affecting LacY function.
An alternative mechanism of lacI- compensation is a change inlacI regulation. This could occur in either of two ways. First, the 4 bp insertion frameshift mutation in lacI - of G/L4 results in a truncated 204 residue LacI repressor. Although non-functional at the lac operator, it is possible that the truncated LacI, which includes the DNA binding domain, interacts with another region in the genome to cause some beneficial effect that is lost when the ancestral lacI was restored (Platt et al. 1973). Second, an operator site for a different gene may have evolved to provide a benefit in the absence of LacI. The LacI repressor has homology to other repressors in E. coli (Weickert and Adhya 1992), and studies have shown that as few as two mutations are needed to increase the affinity of LacI to another operator (Lehming et al. 1987; Lehming et al. 1990; Salinas et al. 2005; Daber and Lewis 2009). We note, however, that the sequenced G/L4 clone does not have mutations in recognized lacI family operator sites.
In summary, compensation was rare and did not occur based on specific fluctuations of glucose and lactose. In the clone in which compensation of lacI - did occur, lac operon expression was reduced but not more so than other strains that had a similar cost in glucose compared to the ancestor. This indicates that costs of constitutive expression were overcome by epistatic interactions with other mutations in the evolved background, and that the reduction in cost was not due solely to a reduction in expression. Future studies will examine potential long-term tradeoffs of compensation and what makes compensation rare.
Acknowledgements. This work was supported by grant DEB-1253650 awarded to T.F.C. by the National Science Foundation and by grant 19-MAU-082 awarded to T.F.C. by the Royal Society of New Zealand Marsden fund.
Data availability. T.F.C will make the strains constructed in this study available to qualified recipients following completion of an institutional material transfer agreement. The results of competition experiments, summary input data, and analysis scripts that pertain to the experiments and analyses reported in this paper have been deposited at http://dx.doi.org/XX.XXX/ dryad.XXXX.