31 May 2013
Günter Wagner
Department of Ecology and Evolutionary Biology
Yale University
A basic premise of systems biology is the assumption that once we know the basic mechanisms of molecular biology, we should be able to scale up our inquiry to encompass complete biolgical systems like cells and eventually even whole organisms. The vulnerability of this approach lies in the question what exactly are the basic functional units of biological systems, or what are our units of analysis. A few years ago a plausible answer was that the basic units of gene regulatory networks are genes with their gene products, e.g. transcription factors and co-factors, spliced and modified, and the cis-regulatory elements that bind transcription factor proteins. One stream of discoveries that undermines the neatness of this picture are the increasing diversity of functional species of non-coding RNAs. In this discussion, however, I want to discuss another potentially more insidious threat to the standard model of gene regulatory networks that arises from our work on the evolution of transcription factor proteins.
The basic question we asked is this: we know from a variety of examples that transcription factor proteins change their functional role in evolution, in particular associated with major re-organizations of the gene regulatory network. What are the molecular mechanisms that change the transcriptional activity of a transcription factor protein in evolution?
Our paradigm to investigate this question is the evolution of pregnancy, or more precisely the evolution of gene regulation in the decidual cells of the endometrial lining of the uterus. We found a variety of mechanisms, but the most interesting is the following: in two examples, HoxA11 and C/EBP-beta, we can demonstrate that the evolutionary change to the protein is not the acquisition of a novel (derived) biochemical activity (e.g. cofactor binding), but a change to the context dependent performance of ancestrally present biochemical capabilities. One implication of these results is that it is hard to conceptualize the transcription factor protein as a molecular entity that has an intrinsic functional/molecular role and thus is not the unit of function. Instead I propose that it is more appropriate to consider a transcription factor protein as a minimal cognitive unit which should be described in the same way as behavioral biologists describe animal behavior: determine the behavioral inventory (what the protein can do) and the reaction norm that determines which molecular behaviors are shown in which situations. Based on this proposal I want to discuss what this idea implies for modeling gene regulatory networks as well as for the logical structure of gene ontology annotations.