The evolutionary influence of binding site organisation on gene regulatory networks.

Gene regulatory networks are shaped by selection for advantageous gene expression patterns. Can we use this fact to predict and explain the structure and properties of gene regulatory networks? Here we address this question with evolutionary simulations of small (two to four genes) transcriptional regulatory networks. Each modeled network is tested for the frequency with which it evolves to produce a bimodal spatial expression pattern of a target gene (the output), in response to a linear trigger gradient (the input). By including network features such as the organisation of binding sites that do not evolve in the model, we can compare the relative chances of evolutionary success between networks differing only in these features. Specifically, we show that networks with competitive binding sites (where binding of one transcription factor excludes another) are more likely to evolve bimodal patterns of gene repression than are networks with independent binding sites (where binding of multiple transcription factors can occur simultaneously). These predictions have implications for the likely structure of gene regulatory networks carrying out bimodal (including bistable) gene expression functions in vivo. The capacity to predict the evolution of structure-function relationships in gene regulatory networks is constrained by gaps in current understanding such as the unknown prior probabilities of the network features, and the quantitative nature of the molecular interactions involved in gene expression. Methods for the circumvention of these constraints, and the potential of the evolutionary modeling approach, are discussed.