Chromatin accessibility is a two-tier process regulated by transcription factor pioneering and enhancer activation

Abstract

Chromatin accessibility is integral to the process by which transcription factors (TFs) read out cis-regulatory DNA sequences, but it is difficult to differentiate between TFs that drive accessibility and those that do not. Deep learning models that learn complex sequence rules provide an unprecedented opportunity to dissect this problem. Using zygotic genome activation in the Drosophila embryo as a model, we generated high-resolution TF binding and chromatin accessibility data, analyzed the data with interpretable deep learning, and performed genetic experiments for validation. We uncover a clear hierarchical relationship between the pioneer TF Zelda and the TFs involved in axis patterning. Zelda consistently pioneers chromatin accessibility proportional to motif affinity, while patterning TFs augment chromatin accessibility in sequence contexts in which they mediate enhancer activation. We conclude that chromatin accessibility occurs in two phases: one through pioneering, which makes enhancers accessible but not necessarily active, and a second when the correct combination of transcription factors leads to enhancer activation.

Interpretable deep learning reveals the sequence rules of Hippo signaling

“…In this type of TF cooperativity, binding enhancement occurs when the motifs are spaced within ~150 bp and are strongest the closer the motifs are. Such sequence rules point to nucleosome-mediated cooperativity 15,20,26,[117][118][119][120][121] and may reflect the likelihood by which two motifs are covered by the same nucleosome. Although the mechanism is not well understood, it does not require specific interactions between TFs, which explains how signaling TFs can receive input from a wide variety of TFs in different cell types.…”

“…Here, we hypothesized that this TF cooperativity is DNA sequence-driven and thus can be studied by measuring the binding of TFs on DNA and identifying the underlying sequence rules using interpretable deep learning. During training, deep learning models accurately learn sequence rules within genomic regions in an inherently combinatorial manner de novo until they can predict the data from sequence alone [15][16][17][18][19][20][21] . The key step is then to interrogate the model and extract the learned sequence rules using interpretation tools 15 .…”

“…On the other hand, strictly spaced motifs are not frequently observed in the genome, raising the question of whether TF binding cooperativity may also occur through more flexible motif syntax [27][28][29] . For example, we previously found that a TF may enhance the binding of another TF through soft motif syntax, which occurs at variable motif distances within ~150 bp but is stronger at closer distances 15,20 .…”

“…To discover these potential sequence rules, we performed the TF binding experiments at the highest resolution and leveraged our previously developed deep learning model BPNet to predict the data at base resolution from genomic sequences of 1-kb 15,20,22,37,38 . This approach optimally resolves sequence rules between closely spaced motifs within enhancers 15 .…”

Protein-intrinsic properties and context-dependent effects regulate pioneer-factor binding and function

“…It is probable that additional cofactors, such as Zld itself, may be required for Twi occupancy at these regions. Prior work demonstrated an important role for Zld in promoting Twi binding in the early embryo 39,40 . Along with earlier studies, our data suggest that PF occupancy is regulated by tissue-intrinsic features, including levels of PF expression, the complement of cofactors expressed, and chromatin structure.…”