Feedback-Induced Variations of Distribution in a Representative Gene Model

Abstract

Gene expression is inherently noisy, implying that the number of mRNAs or proteins is not invariant rather than follows a distribution. This distribution can not only provide the exact information on the dynamics of gene expression but also describe cell-to-cell variability in a genetically identical cell population. Here, we systematically investigate a two-state model of gene expression, a model paradigm used to study expression dynamics, focusing on the effect of feedback on the type of mRNA or protein distribution. If there is no feedback, then the distribution may be bimodal, power-law tailed, or Poisson-like, depending on gene switching rates. However, we find that feedback can tune or change the type of the distribution in each case and tends to unimodalize the distribution as its strength increases. Specifically, positive feedback can change not only a power-law tailed distribution into a bimodal or Poisson-like distribution but also a bimodal distribution into a Poisson-like distribution (implying that stochastic bifurcation can take place). In addition, it can make a Poisson-like distribution become more peaked but does not change the type of this distribution. In contrast to positive feedback, negative feedback has less influence on the shape of the distributions except for the bimodal case. In all cases, the noise-feedback curve used extensively in previous studies cannot well reflect the feedback-induced changes in the shape of distributions. Feedback-induced variations in distribution would be important for cell survival in fluctuating environments.