The wholesale reworking of the cell's fitness landscape due to rho
* illustrates its potential to open evolutionary paths that would not otherwise be accessible. rho
* provides both direct fitness effects and broadly varying (and often positive) epistatic relationships with perturbations at other loci, allowing it to provide benefits early in an evolutionary trajectory while at the same time providing a different, and frequently larger, profile of possible adaptive secondary mutations (see Tables S3
). The interaction between rho
* and rpsL
* described above represents one such case: rho
* itself provides a beneficial fitness effect in the presence of ethanol, and also exhibits positive epistasis with a mutation at the rpsL
locus. A more general schematic is shown in : the fitness effects of mutations throughout the genome are strongly influenced by the genotype at rho
(and presumably other core transcriptional proteins as well), making some secondary mutations more or less beneficial than they would be otherwise (, genotype B). Mutations such as rho
* can also both provide a fitness benefit relative to the wild type under common growth conditions, and reveal higher fitness genotypes upon exposure to stress conditions (, genotype C). rho
* is expected to exert its effects simply by altering transcription (in this case primarily by allowing expression of regions which would not otherwise be transcribed); we thus expect that mutations to other core components of the cell's transcriptional machinery, or to other broadly influential regulators, would show similar levels of evolutionary and phenotypic leverage.
Transformation of the fitness landscape caused by rho*.
In support of this view, mutations to rho
, RNA polymerase 
and DNA supercoiling proteins 
have frequently been observed in a variety of other recent directed evolution experiments. In a few cases, specific epistatic interactions involving these core transcriptional components were found to shape the future adaptive trajectory of populations. For example, Applebee and coworkers 
found that in a set of E. coli
populations evolved to grow efficiently in glycerol minimal media, RNA polymerase mutations arising earlier in the evolutionary trajectories showed positive epistasis with subsequent glpK
mutations (and possibly mutations to dapF
as well). Similarly, in analyzing populations from an extremely long-term evolution experiment, Woods et al.
found the presence of two variant topA
alleles in competition; of these, the allele present in the subsequently evolved strain had a less positive direct effect on fitness, but also showed positive epistasis with a secondary mutation at spoT
that yielded an overall higher fitness phenotype. In general, these previous studies have not, however, fully explored the full breadth of both direct phenotypic and epistatic effects of the housekeeping mutations that they identified.
Because the primary effect of a hypomorphic rho
allele such as rho
* is to allow expression of regions of the genome that would not typically be expressed (see above; also 
), we thus see that the impairment of a system setting baseline boundaries for gene expression can in fact bring forth beneficial, but normally hidden, phenotypes. The concept that robustness to the effects of mutations may facilitate adaptive evolution by allowing the accumulation of genetic diversity that can be subsequently released by a single perturbation, has been proposed repeatedly in the theoretical literature. Wagner 
discussed the “neutral space" of a biological system – a range of equivalent solutions to a given condition – and notes that the presence of diversity within the neutral space allows variation that may be useful under subsequently encountered conditions. Draghi et al.
illustrated precisely this phenomenon more quantitatively using a computational model, showing that intermediate levels of robustness (modeled as the probability of a given mutation being neutral) accelerated the adaptation of populations by providing a reservoir of phenotypically neutral genetic diversity, including variants that could be adaptive under changing conditions. More recently, in modeling tradeoffs involved in the regulation of translational readthrough, Rajon and Masel 
found bistable solutions which required either global regulation to reduce readthrough rates, or a combination of higher readthrough rates but reduced incidence of deleterious products upon readthrough; the high readthrough rate solution was found to be more evolvable by allowing the accumulation of non-deleterious genetic diversity downstream of translational stop sites, which can subsequently be incorporated through a single mutation to the stop codon.
The behavior of rho
* is also reminiscent of two phenomena related to the core translational machinery of yeast. Jarosz et al.
recently showed that the chaperone Hsp90 acts to suppress the effects of genetic variation occurring naturally between yeast strains; temperature stress or chemical inhibition of Hsp90 yielded a wide variety of phenotypic changes among ~100 different yeast strains under 100 low-level stress conditions, frequently with differing signs of effect on fitness for different strains under the same condition. Furthermore, the authors found that Hsp90 in fact shows epistatic interactions with 20% of naturally occurring genetic variations between the strains under consideration. Similar phenomena have been observed for the yeast prion state [PSI+]
, where (as with rho
*) an alteration in the behavior of a regulatory protein gives rise to a highly pleiotropic phenotype which may be harmful or beneficial under a variety of conditions, interacts strongly with the precise genetic background of the cell in question, and appears to exert its effects by causing ectopic expression of sequences which are generally silent. The comparison between both mechanisms in yeast and rho
* must not be taken too far, as there are also substantial differences, most notably in that Hsp90 and [PSI+]
act post-transcriptionally, [PSI+]
in particular represents an epigenetic rather than genetic mechanism, and both the prion states and Hsp90 relaxation have been shown to be encouraged by environmental stress 
, whereas no similar mechanism would be expected to mutate core housekeeping genes in stressed E. coli
cells preferentially. Nevertheless, the effects of both yeast mechanisms, and bacterial rho
mutations, illustrate that microorganisms possess the genetic potential to grow under a broader array of conditions than their regulatory logic allows, that some of the hidden potential may be unlocked through perturbations of core regulatory proteins, and that even a single such perturbation may unleash a wide variety of positive or negative effects and interactions with other loci throughout the genome.
Taken together, our findings illustrate that a single amino acid substitution in the global transcriptional terminator Rho leads to a wholly different regulatory and phenotypic state, in which gene expression is globally altered and cellular fitness in a broad variety of environments has changed. The same mutation also dramatically alters the fitness landscape with regard to other genetic variations, making accessible a number of beneficial secondary mutations that are otherwise neutral or deleterious. The set of states reachable through rho
* or other point mutations of core regulatory proteins comprise a previously underappreciated reservoir of additional phenotypes accessible to bacterial populations under selective conditions. These findings imply a role for mutations to regulators such as rho
both as evolutionary catalysts, by making a variety of secondary mutations more favorable than they would be in the parental strain, and as evolutionary capacitors 
, by allowing silently accumulating genetic diversity to take effect rapidly upon changes in gene regulation. The full extent to which this capacity of core housekeeping and regulatory proteins is used during evolutionary trajectories, and the identity of the complete set of genes showing such broadly influential behavior, are not yet clear. It is also intriguing to speculate that classical global regulators may also show similarly diverse effects, either upon genetic perturbation or as a response to environmental signals, given that the number of genes substantially perturbed by rho*
(~200) is comparable to the number directly or indirectly affected by each global regulator (e.g.
, CRP, IHF, or FNR)