Developmental regulatory networks comprise a large number of signal-activated, cell type-specific and more widely expressed transcription factors (TFs) that act in combination to control the expression of genes that, in turn, determine the specification and differentiation of cells to form specific organs. Such TFs can be identified by gene expression profiling of the relevant cells, as well as by computational approaches that identify the DNA binding sequence motifs in relevant enhancers to which the regulatory TFs bind, thereby facilitating the identification of the cognate TFs from appropriate databases. We have used both approaches to identify a large set of TFs that are expressed in the Drosophila embryonic mesoderm, and we have used a combination of classical genetics, RNAi knockdown, gene overexpresion and transgenic mesodermal reporter assays to assess the functions of many of these TFs in muscle and heart development. We initially concentrated on the potential role of Ets, POU homeodomain (POUHD) and T-box motifs that were uncovered as being enriched by a machine learning classifier in muscle founder cell (FC) enhancers. Each of these motifs is over-represented in both individual FC enhancers and their orthologous sequences when compared to controls. Mutagenesis of each class of motifs in a known FC-specific enhancer affected activity of the reporter as compared to wild-type versions of the enhancer. Moreover, in the case of the T-box TF, both RNAi and targeted gene overexpression led us to conclude that this protein is a previously unrecognized determinant of muscle founder cell (FC) identity. In another example, a family of Forkhead TFs was found to regulate the expression of a single enhancer that is associated with the Nidogen (Ndg) gene and that controls the entire spectrum of Ndg mesodermal expression sites, including somatic muscles, visceral muscles and several cardiac cell types at multiple stages of embryogenesis. More specifically, the various spatiotemporal activities of the Ndg enhancer are governed by a trans mechanism that involves the binding of unique Fkh TFs to the same or different sequence motifs within the broadly active Ndg cis-regulatory element. Moreover, protein binding microarray analysis demonstrated that the Fkh binding sites in this one enhancer represent two distinctively different DNA sequence specificities, thereby revealing an additional, cis-acting level of complexity for how a single gene is activated in multiple related but unique cell types of the developing embryo. In addition, a genome-wide computational scan revealed that both classes of Fkh binding sites are statistically over-represented in putative heart enhancers in combination with the motifs for five other known cardiac regulatory proteins, thereby providing evidence for the existence of a complex 7-way AND code of TFs that mediates the transcriptional control of a significant subset of heart genes. The finding that Fkh TFs play a key role in cardiogenesis was independently validated by a machine learning strategy that uncovered a significant enrichment of Fkh DNA binding motifs within a large set of known heart enhancers that was used as a training set to build a computational classifier. Further functional analyses of Fkh TFs led us to discover that two members of this family that are expressed in the cardiac mesoderm, jumeau (jumu) and Checkpoint suppressor homolog (CHES-1-like), together regulate the symmetric and asymmetric division of cardiac progenitor cells, thereby ensuring that the correct numbers and types of cells are generated during the early stages of organogenesis. Moreover, we determined that one cardiogenic pathway controlled by jumu and CHES-1-like directly involves Polo, a kinase that functions in various steps of mitosis. However, our genetic studies revealed that these two genes also act in a partially redundant manner to regulate other, Polo-independent aspects of cardiogenesis, a finding that we are currently pursuing by gene expression profiling of flow-sorted cardiac mesodermal cells from wild-type embryos as well as from jumu and CHES-1-like gain- and loss-of-function embryos, including the double mutants. It is anticipated that this approach, combined with genetic and RNAi-based functional investigations of putative Fkh target genes in the heart, will reveal the molecular basis of the related cardiogenic activities and also the synergistic interactions between these two Fkh TFs that we previously identified solely by genetic studies. Machine learning of a large training set of cardiac enhancers revealed that Myb is likely to be an additional heart TF, and functional experiments showed that it indeed acts in concert with the two Fkh TFs (jumu and CHES-1-like) and Polo kinase to regulate cardiac progenitor cell divisions. Finally, differential motif enrichment and cis-trans genetic studies revealed that the Notch signaling pathway TF Suppressor of Hairless (Su(H)) discriminates PC and CC enhancer activities. Collectively, these studies elucidate molecular pathways used in the regulatory decisions for proliferation and differentiation of cardiac progenitor cells, and document the utility of both enhancer computational modeling and functional validation of computational predictions in uncovering developmental regulatory subnetworks.
|Ahmad, Shaad M; Busser, Brian W; Huang, Di et al. (2014) Machine learning classification of cell-specific cardiac enhancers uncovers developmental subnetworks regulating progenitor cell division and cell fate specification. Development 141:878-88|