Thermodynamic principles link in vitro transcription factor affinities to single-molecule chromatin states in cells

Abstract

The molecular details governing transcription factor (TF) binding and the formation of accessible chromatin are not yet quantitatively understood-including how sequence context modulates affinity, how TFs search DNA, the kinetics of TF occupancy, and how motif grammars coordinate binding. To resolve these questions for a human TF, erythroid Krüppel-like factor (eKLF/KLF1), we quantitatively compare, in high throughput, in vitro TF binding rates and affinities with in vivo single-molecule TF and nucleosome occupancies and in vivo-derived deep learning models. We find that 40-fold flanking sequence effects on affinity are consistent with distal flanks tuning TF search parameters and captured by a linear energy model. Motif recognition probability, rather than time in the bound state, drives affinity changes, and in vitro and in nuclei measurements exhibit consistent, minutes-long TF residence times. Finally, in vitro biophysical parameters predict in vivo sequence preferences and single-molecule chromatin states for unseen motif grammars.

Keywords: biophysical models; chromatin accessibility; deep learning models; eKLF/KLF1; kinetics; single-molecule footprinting; target search; thermodynamics; transcription factor binding.