Back

CpG islands act as topological sinks for transcription-induced DNA supercoiling

Naughton, C.; Bonato, A.; Chiang, M.; Corless, S.; Stocks, J.; Grimes, G. R.; Halliday, D.; Bentivoglio, A.; Brackley, C. A.; Marenduzzo, D.; Gilbert, N.

2026-08-25 molecular biology
10.64898/2026.08.24.746546 bioRxiv
Show abstract

Strong evolutionary selection has maintained CpG-dense islands (CGIs) at the promoters of constitutively expressed genes throughout the vertebrate genome, suggesting an important role in regulating DNA topology. Here, using Twist-seq, a psoralen-based approach for quantitative genome-wide profiling of DNA supercoiling, we reveal distinct topological states across human gene promoters. We show that CGI promoters accumulate elevated levels of negative supercoiling relative to non-CGI promoters and define localised topological domains at highly transcribed genes. Integrating genome-wide analyses with reaction-diffusion modelling and coarse-grained molecular dynamics simulations, we find that this behaviour is encoded by the intrinsic physical properties of CGI DNA. The GC-rich sequence context promotes nucleosome depletion and focuses torsional stress onto embedded AT-rich pockets, driving localised DNA melting and plectoneme-tip bubble formation within promoter-proximal nucleosome-free regions. This provides an energetically favourable pathway for redistributing transcription-induced torsional stress through transient strand separation and writhe, consistent with increased ssDNA formation at CGI promoters observed by ssDNA-seq. We propose that CGIs function as sequence-encoded topological sinks that buffer supercoiling while maintaining a promoter architecture permissive for transcription initiation, thereby preserving promoter integrity and genome stability.

Matching journals

The top 3 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.