Promoter Architecture Dictates Variability in Gene Expression

Recent experimental studies have reported that cell-to-cell variability in gene expression is a universal function of mean gene expression levels, implying that knowledge of mean levels of gene expression alone is sufficient to determine the level of variability. This result seems at odds with the physical intuition underlying efforts in the literature to predict the level of variability for different promoter architectures (where we take “promoter architecture” to mean the numbers, positions, and strengths of transcription factor and RNA polymerase binding sites). For instance, we would expect slow TF binding/unbinding kinetics to yield a relatively higher level of variability at a given mean than fast kinetics, as the switching between distinct transcriptional states will yield a broad distribution in mRNA production. To address this apparent conundrum, we constructed a set of 18 constitutive (unregulated) promoters with expression levels spanning a range of 2.5 orders of magnitude, and found that the observed mRNA distributions were consistent with Poissonian production of mRNA. We then placed one of these promoters under regulation by the repressor transcription factor LacI, and measured the resulting mRNA distributions at a range of LacI concentrations. We found that, for the regulated promoter, the level of variability as a function of mean expression followed a distinct, promoter architecture dependent curve as compared with the unregulated promoter. These results demonstrate that the level of variability in gene expression depends on the specific promoter architecture in play