In vivo and in vitro consistency of thermodynamic models for transcription regulation

Abstract

The ability of cells to respond to external stimuli, by up and down regulation of genes, is a key strategy for survival in changing environmental conditions. Models based on equilibrium statistical mechanics have been able to successfully predict in vivo fold changes in gene expression by computing weights of configurational states of the promoter where genes can be expressed. These models are based on the same, perhaps unintuitive, assumption that transcription initiation - an inherently nonequilibrium process - can indeed be effectively described by the equilibrium binding of transcription factors (TFs) and polymerases to the promoter. The few earlier studies that independently test this assumption [P. H. von Hippel, Proc. Natl. Acad. Sci. USA 71, 4808 (1974)10.1073/pnas.71.12.4808], were published before much of our modern day understanding of molecular biology was established, and their models fail to explain more recent experimental results. As such, it is not obvious that the original conclusions remain valid. Equilibrium models depend on fitted free energy differences between binding of TFs to specific operator sites versus nonspecific DNA. In this article we compare the fitted binding free energy of the well-studied LacI repressor to equilibrium binding constants measured in independent in vitro experiments. To make this comparison we take into account the distribution of binding energies of the transcription factor to the nonspecific DNA, and we adjust LacI binding constants to a common set of physiological conditions. We find that the fitted binding energies of the LacI repressor in vivo indeed agree with in vitro measured equilibrium binding constants, reestablishing the idea that equilibrium statistical mechanical models of transcriptional regulation should be viewed not merely as mathematical tools, but also as informative physical representations of underlying TF-DNA interactions.