Affiliations: 1) Stowers Institute for Medical Research, Kansas City, MO; 2) The University of Kansas Medical Center, Kansas City, KS
Keywords: e. enhancers; n. other (chromatin accessibility )
Enhancers, or cis-regulatory DNA sequences that control spatiotemporal gene expression programs, have intrinsically high nucleosome barriers that prevent the nonspecific binding of transcription factors (TFs). For an enhancer to become active, TFs must overcome this barrier, access their binding motifs, and evict the nucleosome, but the mechanisms by which these steps occur are not clear. Current models suggest that specialized TFs called pioneer factors can access their motifs in the presence of nucleosomes and foment nucleosome depletion through cooperativity with additional TFs. Even still, how pioneer and non-pioneer TFs cooperate to generate chromatin accessibility at enhancers is not yet known. To understand how TFs overcome the nucleosome barrier, we are using Zelda, a pioneer TF for the Drosophila maternal-to-zygotic transition, and a set of TFs that drive pattern formation in the early embryo (patterning TFs) as a model system. To dissect Zelda’s cooperativity with the patterning TFs, we have integrated ChIP-nexus TF binding data with our novel deep learning model, BPNet, which learns the relationship between DNA sequence and TF binding in an inherently combinatorial way. BPNet reliably predicts the genome-wide binding of Zelda and the patterning TFs, and successfully discovers known and novel motifs at well-studied enhancers. Through employing the trained BPNet model as an in silico oracle, we reveal that Zelda and the patterning TFs exert directional cooperativity within nucleosome-range distances. Using time course ATAC-seq measurements, we find evidence that Zelda alone is insufficient to generate accessibility, rather chromatin accessibility in the early embryo relies on cooperativity between Zelda and the patterning TFs. Taken together, these results suggest a nucleosomal basis for cooperativity between Zelda and the patterning TFs and present a model where Zelda establishes a more permissive chromatin state for TF binding but relies on cooperative interactions with the patterning TFs to bestow accessibility. To further dissect the hierarchical nature of TF cooperativity for chromatin accessibility, we are training a BPNet model on our time course ATAC-seq data, which will learn the relationship between DNA sequence and accessibility and will identify the sequence features that are predictive of chromatin accessibility.