Expression noise facilitates the evolution of gene regulation

Abstract

Although it is often tacitly assumed that gene regulatory interactions are finely tuned, how accurate gene regulation could evolve from a state without regulation is unclear. Moreover, gene expression noise would seem to impede the evolution of accurate gene regulation, and previous investigations have provided circumstantial evidence that natural selection has acted to lower noise levels. By evolving synthetic Escherichia coli promoters de novo, we here show that, contrary to expectations, promoters exhibit low noise by default. Instead, selection must have acted to increase the noise levels of highly regulated E. coli promoters. We present a general theory of the interplay between gene expression noise and gene regulation that explains these observations. The theory shows that propagation of expression noise from regulators to their targets is not an unwanted side-effect of regulation, but rather acts as a rudimentary form of regulation that facilitates the evolution of more accurate regulation.

Poor sensing maximises microbial fitness when few out of many signals are sensed

“…This fitness cost can be reduced by increased protein fluctuations, i.e. of σ 2 p + α 2 σ 2 s by bet-hedging or 'noisy sensing' strategies, in agreement with previous results [1,2,3]. This can also be shown by differentiating equation 4 with respect to the normalised internal variance and solving for its optimum value, division by ω 2 then gives,…”

“…Evolutionary theories about the fitness of phenotypic strategies [1,2,3,4,5,6] have made successful predictions of the outcomes of natural selection in fluctuating environments [3,4,7,8]. In those theories, the (geometric) fitness [4,5,6] is generally maximised.…”

“…When bet-hedging adaptation strategies are more evolutionarily successful than sensing strategies has been a long standing question in evolutionary biology. Bet-hedging has been predicted to be evolutionarily advantageous under at least two conditions [9]: In slowly-changing, mild environments where sensing machinery would rarely provide an evolutionary benefit for long periods of time, and in environments that change quickly into extinction-threatening states where responsive adaptation would be too slow [1,2,3,4,6,9]. In all other cases, responsive signalling strategies are favoured.…”

Noise propagation shapes condition-dependent gene expression noise inEscherichia coli

“…Bull, 1987;Haccou and Iwasa, 1995;Kussell and Leibler, 2005;Tanase-Nicola and ten Wolde, 2008;Ackermann et al , 2008. However, in recent years evidence has been accumulating that gene regulation and gene expression noise may be inherently and intimately linked. For example, several studies have shown that transcriptional noise varies significantly across genes and is to a substantial extent encoded in the promoter sequence of a gene Taniguchi et al , 2010;Silander et al , 2012;Carey et al , 2013;Jones et al , 2014;Wolf et al , 2015. Indeed, whereas genes that are transcribed at a constant rate will exhibit Poissonian fluctuations in mRNA levels, most genes exhibit significantly higher levels of transcriptional noise.…”

“…This increased transcriptional noise is generally understood to result, at least partially, from the fact that binding and unbinding of transcription factors (TFs) causes the promoter to stochastically switch between different states that are associated with different transcription initiation rates. In this way, fluctuations in both the expression levels of TFs and their binding to promoter regions are propagated to the their target genes and this noise propagation has long been recognized as an unavoidable side effect of regulation Thattai and van Oudenaarden, 2001;Pedraza, 2005;Lestas et al , 2010;Lehner and Kaneko, 2011;Bruggeman and Teusink, 2018. In a recent study we showed that, in E. coli, unregulated promoters have low expression noise by default, and that the more regulatory inputs a gene has, the more noisy its gene expression tends to be Wolf et al , 2015. A similar general association between highly regulated genes and high expression noise has also been observed in eukaryotes Blake et al , 2003;Newman et al , 2006, and it has also been observed that co-regulated genes show correlated gene expression fluctuations Junker and van Oudenaarden, 2012. All these observations suggest that noise propagation may be a key determinant of gene expression noise.…”