Quantitative decoding of transcriptional regulation in development and evolution

About This Grant

PROJECT SUMMARY/ABSTRACT Sequence-specific transcription factors (TFs) drive precise and timely control of gene expression during development. Changes in TF activity underlie much of the phenotypic divergence between species, and dysregulation is increasingly recognized as a cause of diverse human diseases. Despite the importance of TFs in cellular and organismal health and disease, most tools for studying TF function and conceptual models are binary in nature. The binary nature of TF models suggest an on-off switch, yet many disease-associated TFs operate in an analog fashion, where dosage is correlated with phenotypic outcome. Molecular tools to study and manipulate TF activities in an analog fashion are minimal, hindering quantitative modeling of TF function. This proposal seeks to fill this knowledge gap using an innovative chemical genetic approach that can precisely modulate TF dosage in human pluripotent stem cell (hPSC)-derived cell types. We will use this system to define how dosage of the key TFs SOX2, SOX9, and TWIST1 modulates chromatin state, gene expression, and cellular phenotypes and how this dosage-dependent regulation varies across cell types and evolutionary time. These TFs are ideal models to address our questions as they are associated with dosage-sensitive developmental disorders, function in multiple cell types, and have been associated with regulatory evolution. Our preliminary datasets validate our approach, revealing substantial variation in the dosage response of TF-dependent regulatory elements (REs) and genes, from highly sensitive to robust and buffered against quantitative TF dosage changes. We also recently applied cutting-edge deep learning models to learn sequence features underlying variation in RE dosage responses, revealing a surprising contribution of low-affinity binding motifs not apparent in unperturbed states. These results hint at a new layer of the cis-regulatory code, where certain sequence features are only apparent when studying transcriptional regulation in a quantitative regime. Our goal for this Program is to comprehensively characterize the underlying logic of these previously undefined additional regulatory layers and understand how they vary between cell types and species. To reach this goal, 3 directions will be pursued. 1) We will use multi-lineage differentiation of our unique hPSC reagents to investigate while certain cell types are sensitive to SOX2 and SOX9 dosage while others are buffered against quantitative changes. 2) We will apply chemical genetics and deep learning in chimpanzee PSCs to test whether the cis- regulatory logic defined in human cells generalizes across species and to distinguish cis from trans contributions. 3) Tool development, both experimentally where we will expand our system to allow bi-directional TF dosage control, and computationally, where we will develop improved deep learning models for predicting how REs respond to TF dosage. Together, these directions provide orthogonal but complementary approaches and tools that will converge on a unified model of how TF dosage impacts development and evolution, ultimately ushering our understanding of transcriptional regulation into a more mechanistic and quantitative realm.