Sensing the fly embryo's transcription factors*

Transcription factor (TF) concentrations can be seen as signals that need to be sensed by organisms in order for them to express their genes as precisely as is required during development or during adjustment to different conditions. The low concentration of all the relevant molecules means that these measurements will be noisy, setting the maximum that cells can extract.  In embryonic development, as an example, there is a minimum information that the organism needs in order to generate a complex body plan, reproducibly. In the fruit fly embryo, there is evidence that these two information bounds are close to one another, so there is a premium on extracting the most useful or meaningful bits.  In previous work we have shown that the information bottleneck provides a natural formulation for this problem. It may thus be possible to view the complex enhancer logic that "reads" the transcription factor concentrations as implementing a solution to an information theoretical optimization problem. Here we explore in more detail what happens when individual enhancers have low information capacity, so that the overall measurement of TF concentrations must be split or shared among multiple elements. Driven by direct measurements of signal and noise in the relevant TFs, we show that these multiple elements must be sensitive to combinations of their inputs, both in abstract and more microscopically realistic models.