A biophysical approach to predicting intrinsic and extrinsic nucleosome positioning signals

Alexandre Morozov
Rutgers University

Nucleosomes are fundamental units of chromatin which in eukaryotic genomes function to compact DNA and to regulate access to it, both by simple physical occlusion and by providing the substrate for covalent epigenetic tags. While nucleosome positions in vitro are determined by the DNA sequence alone, in vivo competition with other DNA-binding factors and the action of chromatin remodeling enzymes play a role that needs to be quantified, with the goal of building better models of eukaryotic gene regulation that explicitly take chromatin structure into account.

To address these issues, we have developed a biophysical, DNA mechanics-based model for the sequence
dependence of DNA bending energies, and validated it against a collection of in vitro free energies
of nucleosome formation. We also successfully designed both strong and poor histone binding sequences ab initio, and tested them in the lab. For in vivo data from S.cerevisiae, the strongest positioning signal came from the competition with other factors. Based on sequence alone, our model predicts that functional transcription factor binding sites tend to be covered by nucleosomes, but are uncovered in vivo because functional sites cluster within a single nucleosome footprint and thus transcription factors bind cooperatively. Likewise, a weak enhancement of nucleosome binding in the TATA region becomes a strong depletion when the TATA-binding protein is included, in quantitative agreement with experiment.
Our model distinguishes multiple ways in which genomic sequence influences nucleosome positions,
and thus provides alternative explanations for several genome-wide experimental findings.
In the future our approach will be used to rationally alter gene expression levels in model
systems through a computational redesign of the nucleosome occupancy profile.

Presentation (PowerPoint File)

Back to Workshop IV: Search and Knowledge Building for Biological Datasets