Alladi Ramakrishnan Hall

Decoding fibroblast reprogramming through the lens of transcription factor stoichiometry and motif syntax at single cell resolution

Surag Nair

Stanford University

Ectopic expression of OCT4, SOX2, KLF4 and MYC (OSKM) in differentiated cells transforms them into induced pluripotent stem cells. Upon overexpression, the somatic program is silenced early, while pluripotency is acquired gradually in phases. These necessary transitions are accompanied by widespread transient chromatin and gene expression activity. A detailed understanding of the nature and functional role of transient activity in productive reprogramming is lacking. In this work, we characterize how continuous OSK stoichiometry variation shapes the chromatin and expression landscape of reprogramming at single-cell resolution. We leverage deep neural networks that predict cell-state specific ATAC-seq signal at base-resolution from underlying DNA sequence, decoupled from Tn5’s intrinsic sequence preference. Interpreting the networks generates a lexicon of context-specific regulatory motifs pertinent in progressive cell states. Initial diversification into a reprogramming trajectory, and non-reprogramming keratinocyte- and fibroblast-like fates stems from stochastic variation in the stoichiometry of OSKM. In cells with supraphysiological overexpression of all factors, OSK take over the chromatin landscape by extensively engaging and opening closed chromatin. A significant fraction of opened cis-regulatory elements (CREs) contain low-affinity motifs of OSK, including a novel partial POUS-HMG OCT-SOX motif. We observe that CREs with weaker motifs are progressively released as the concentration of OSK reduces, suggesting a simple mechanism for silencing of early transient CREs. Remarkably, the deep learning models reveal that single-cell ATAC-seq encodes differences in transcription factor (TF) footprint depths that correlate with TF stoichiometry and motif affinity. Finally, we elucidate how OSK stoichiometry affects somatic silencing by tuning the ability to sequester away somatic TFs such as AP1 to transient CREs. Together, our approach outlines a powerful paradigm for interrogating the sequence determinants of accessibility in trajectories with continuous variation in TF concentrations. Our findings connect TF stoichiometry over the course of reprogramming to diversification of trajectories, low-affinity motif binding, differential footprinting, CRE selection and somatic silencing.

Hybrid mode. Zoom link:
zoom.us/j/95038155052
Meeting ID: 950 3815 5052
Passcode: 272075