Noise-mean relationship in mutated promoters

Gene expression depends on the frequency by transcription events (burst frequency), and on the number of mRNA molecules made per event (burst size). Both processes are encoded in the promoter sequence, yet their relative dependence on different mutation types is poorly understood. Theory suggests that burst size and burst frequency can be distinguished by monitoring the stochastic variation (noise) in gene expression: Increasing burst size will increase mean expression without changing noise, while increasing burst frequency will increases mean expression and decrease noise. To reveal principles by which promoter sequence regulates burst size and burst frequency, we randomly mutated twenty-two yeast promoters chosen to span a range of expression and noise levels, generating libraries of hundreds of sequence variants. In each library, mean expression (m) and noise (coefficient of variation, η) varied together, defining a scaling curve: η=b/m+η_ext². This relation is expected if sequence mutations modulate burst frequency primarily. The estimated burst size (b) differed between promoters, being higher in promoter containing a TATA box and lacking a nucleosome-free region. The rare variants that significantly decreased b were fully explained by mutations in the TATA box, or by an insertion of an out-of-frame translation start site. When subjecting TATA-mutated promoters to a second round of mutagenesis, the noise-mean relationship defined a shifted scaling curve, consistent with a lower burst size. These TATA box mutations also modulated the responsiveness of gene expression to changing conditions. Our results suggest that burst size is a promoter-specific property that is relatively robust to sequence mutations but is strongly dependent on the interaction between the TATA box and promoter nucleosomes.