Noise in gene expression is coupled to growth rate

Proper control of gene expression is critical in nearly all biological processes. However, genetically identical cells exposed to the same environment display heterogeneity in gene expression (noise), with important phenotypic consequences (Grossman 1995; Rao et al. 2002; Blake et al. 2003; Balaban et al. 2004; Colman-Lerner et al. 2005; Kaern et al. 2005; Balazsi et al. 2011; Munsky et al. 2012; Lee et al. 2014). Variability in expression is anti-correlated to population average gene expression, which in turn is tightly coupled to growth rate (Tyson et al. 1979; Ingraham et al. 1983; Bar-Even et al. 2006; Newman et al. 2006; Brauer et al. 2008; Klumpp et al. 2009; Taniguchi et al. 2010; Keren et al. 2013). However, except for isolated examples (Guido et al. 2007), the effects of growth conditions on expression noise have not been systematically investigated. The expression noise of a gene in a clonal population is determined by intrinsic and extrinsic factors (Elowitz et al. 2002). Intrinsic noise describes the variation at the level of a single gene due to the stochastic nature of the transcriptional process, whereas extrinsic noise relates to the variability in expression shared across different genes due to population dynamics, global differences in cellular environment, and shared upstream components (Thattai and van Oudenaarden 2001; Elowitz et al. 2002; Blake et al. 2003; Raser and O'Shea 2004; Pedraza and van Oudenaarden 2005; Volfson et al. 2006; das Neves et al. 2010; Stewart-Ornstein et al. 2012; Schwabe and Bruggeman 2014). Although research, in particular at the theoretical level, has focused on stochastic, intrinsic noise (for review, see Raj and van Oudenaarden 2008; Balazsi et al. 2011; Sanchez and Golding 2013), in most organisms that have been studied, the majority of the variability in gene expression is extrinsic (Raser and O'Shea 2004; Acar et al. 2005; Colman-Lerner et al. 2005; Newman et al. 2006; Volfson et al. 2006; Raj and van Oudenaarden 2008; Schwabe and Bruggeman 2014). Intrinsic expression noise is tightly coupled to the mean expression of the population and generally decreases as mean expression increases (Bar-Even et al. 2006; Newman et al. 2006; Taniguchi et al. 2010), as depicted schematically in Figure 1A. At high expression levels, there is no longer a dependence on the mean, as global, extrinsic factors set a lower bound (extrinsic limit) for the overall variability (Bar-Even et al. 2006; Newman et al. 2006; Taniguchi et al. 2010). Deviations of genes from this trend are attributed to their specific regulatory architectures, often encoded by their promoter sequence, which may specifically result in either high or low levels of noise (Blake et al. 2003; Raser and O'Shea 2004; Carey et al. 2013; Dadiani et al. 2013; Sanchez and Golding 2013; Jones et al. 2014). Figure 1. Gene expression noise is higher at lower growth rates. (A) A schematic of the genome-wide relationship between mean expression (x-axis) and noise (CV2, y-axis) in a single condition as determined experimentally (Newman et al. 2006; Taniguchi et al. 2010 ... Parameters affecting gene expression variability change across conditions in a coordinated manner, e.g., growth rate (Tyson et al. 1979; Ingraham et al. 1983), mean expression (Berthoumieux et al. 2013; Gerosa et al. 2013; Keren et al. 2013), and concentration of RNA polymerases and ribosomes (Pedersen et al. 1978; Ingraham et al. 1983; Klumpp and Hwa 2008). Harsher conditions, which support slower growth, also display lower protein abundances for most genes and, specifically, lower abundances of RNA polymerases and ribosomes (Pedersen et al. 1978; Tyson et al. 1979; Ingraham et al. 1983; Klumpp and Hwa 2008, 2014; Klumpp et al. 2009; Berthoumieux et al. 2013; Gerosa et al. 2013; Keren et al. 2013). However, it is not clear whether noise levels are expected to change globally between conditions and in what direction because many of these changes may have opposing effects (Fig. 1B; Supplemental Note 1). To examine how noise changes across conditions, we measured the single-cell expression distributions of approximately 900 Saccharomyces cerevisiae promoters upstream of a fluorescent reporter across four environmental conditions using flow cytometry. We find a genome-wide increase in gene expression noise at lower growth rates, with most genes displaying elevated noise levels at slow growth. We examine the dependence of noise in expression on growth rate by modeling the noise that results only from changes in the composition of cell-cycle stages in the population at different growth rates. Consistent with our data, we find that this highly simplified model predicts a non-monotonic relationship between growth rate and noise, as well as overall higher variability in expression for cell-cycle–regulated genes in all conditions. Measurements of several strains grown in a chemostat further support our model and suggest that differential partitioning of the population between cell-cycle stages in different growth rates is a major determinant of extrinsic noise. Our work underscores the importance of growth rate–related effects in noise, showing that some conditions show elevated levels of expression variability genome-wide, with potential phenotypic consequences. Since cell-to-cell variability in gene expression underlies important phenotypic phenomena such as persistence (Balaban et al. 2004), competence (Maamar et al. 2007), latency (Dar et al. 2014), metastasis (Lee et al. 2014), responsiveness to fluctuating environments (Acar et al. 2005, 2008; Blake et al. 2006; Kaufmann et al. 2007; Levy et al. 2012; Vardi et al. 2013), and triggering of meiosis (Nachman et al. 2007), our results suggest that the probability and efficiency of these processes will be tightly coupled to environmental conditions, with potential evolutionary implications.