Assessing modularity of developmental enhancers in Drosophila melanogaster

Abstract

Gene expression is critical for animal development as cells divide and differentiate into multiple cell types. Cell-type specific gene expression is controlled by enhancers, DNA sequences that can direct expression of a target gene from hundreds of kilobases away. Gene promoters contact at least two enhancers on average, and enhancers may also contact each other. A key question is therefore how enhancers operate in this complex regulatory DNA context. It has long been assumed that enhancers act as independent modules based on their ability to drive gene expression when isolated in reporter constructs. To test assumptions of enhancer modularity, I probed interactions between two developmental enhancers from the even-skipped locus in Drosophila melanogaster blastoderm embryos. My results contradict the classic definition of enhancers; I found that the arrangement of enhancers relative to one another and the promoter influences levels of gene expression while not affecting its spatial pattern within the embryo. These results are described in Chapter 2. However, these enhancers are modular in one aspect: when fused directly together, they still direct their distinct spatial expression patterns. In Chapter 3 I describe a collaboration with Md Abul Hassan Samee in Saurabh Sinha's group at the University of Illinois Urbana-Champaign to apply computational sequence-to-expression models to my data. We found that a mechanistic model describing interactions between transcription factors was unable to fit our data well; in contrast, a phenomenological model that finds active sequences fits the data much better. These results indicate that to predict gene expression from sequence we will need to learn how enhancer boundaries are defined. In summary, I present evidence that the organization of enhancers within a locus impacts expression of the target gene. This finding overturns assumptions about enhancer modularity and emphasizes the importance of considering higher level interactions across a locus. Structural variation is common in natural populations, and our results highlight a novel way in which these sequence variants may alter gene expression. To realize the long-standing goal to predict gene expression directly from sequence we must investigate how enhancers interact within a complex locus.