Deep neural-network analysis of base-resolution transcription-factor binding data revealed that the pluripotency factor Nanog preferentially binds DNA when partner motif instances lie on the same face of the DNA double helix, producing a detectable ~10.5-bp periodicity in binding signals and in genomic inter-motif distances. Two competing biophysical explanations were proposed — cooperative protein-protein interactions between Nanog and partner factors, or preferred Nanog binding to nucleosomal DNA presented on the solvent-exposed face — but no experiment has distinguished them, and no later study has established the mechanism. The problem is to determine, by targeted biochemical, structural, or cell-based experiments, which mechanism (or which alternative) generates this helical binding preference, in a way that accounts for its specificity to Nanog, its independence of motif orientation, and its characteristic distance range.

Background

Transcription factors (TFs) are proteins that bind short DNA sequence motifs to regulate gene expression. In mammals, the pluripotency factors Oct4, Sox2, Nanog and Klf4 bind thousands of enhancers (short cis-regulatory DNA elements that activate gene transcription) in mouse embryonic stem cells (ESCs). How the arrangement of multiple motif instances on the DNA — their spacing, orientation and spacing-dependent cooperativity, often called motif syntax — shapes TF binding is a central open layer of the cis-regulatory code, i.e. of the rules mapping DNA sequence to regulatory output.

ChIP-nexus is a high-resolution genomic assay that maps where a TF binds DNA at near single-base precision, by sequencing the protected DNA ends around each bound protein. Avsec et al. (2021) introduced BPNet, a deep neural network trained on ChIP-nexus profiles of the four pluripotency factors to predict base-resolution binding from DNA sequence alone, and developed interpretation tools (contribution scores, which quantify each base's influence on the prediction, and in-silico motif interaction analysis, in which designed motif spacings are fed through the trained model) to extract predictive 'soft' syntax rules — weak, distributed spacing preferences rather than strict grammatical rules.

The most striking rule they found concerns Nanog, a homeodomain TF (a TF family with a conserved helix-turn-helix DNA-binding domain). Binding predictions, contribution scores, genomic inter-motif distances and the raw ChIP-nexus profiles all showed that Nanog binding is enhanced when the relevant partner motifs (Nanog itself, Sox2, the composite Oct4-Sox2 motif, and Zic3) are spaced at multiples of ~10.5 bp, matching one full turn of the DNA double helix, independently of motif orientation. The same spacing arrangement did not produce periodic binding predictions for other TFs, so the rule is Nanog-specific. CRISPR point mutations at one genomic locus validated the model's short-range predictions but said nothing about the mechanism of the periodicity. Because the preference means that Nanog effectively 'sees' partner factors located on the same side of the helix, the authors proposed two candidate mechanisms: (i) direct protein-protein interactions between Nanog and partner TFs on naked DNA, analogous to the cooperative binding of the bacterial lambda and lac repressors; or (ii) an indirect, chromatin-based mechanism in which Nanog preferentially binds nucleosomal DNA (DNA wrapped around the histone core) on the solvent-exposed face, as observed for some homeodomain TFs — since the nucleosome presents DNA in a fixed helical register, motifs on the exposed face would be periodically preferred. Distinguishing these requires experiments that were not performed in the original study and have not been performed since.

Problem Statement

Determine the biophysical mechanism that produces the ~10.5-bp helical periodicity of Nanog binding observed in mouse embryonic stem cells — the preference of Nanog for binding DNA regions whose Nanog, Sox2, Oct4-Sox2 or Zic3 motif instances are positioned on the same face of the DNA helix. Specifically, establish whether this periodic preference arises from (i) direct protein-protein interactions between Nanog and its partner factors (or Nanog self-interaction, for example the tryptophan-repeat-domain-mediated oligomerization proposed computationally by later work), analogous to lambda/lac repressor cooperativity on naked DNA, (ii) preferred binding of Nanog to nucleosomal DNA presented on the solvent-exposed face, exploiting the fixed helical register of DNA wrapped around histones, or (iii) some other mechanism. A complete answer must experimentally discriminate among these mechanisms — for example through binding assays on designed, helically phased motif substrates with matched naked-DNA and reconstituted-nucleosome templates, structural determination of the relevant Nanog-DNA or Nanog-nucleosome assembly, or targeted perturbation of the proposed interaction surfaces in cells — and must account for the established features of the phenomenon: its specificity to Nanog among the profiled factors, its independence of motif orientation, and its operation at protein-range and nucleosome-range distances up to roughly 150 bp, as well as consistency with the in-vivo ChIP-nexus observations. Since no discriminating experiment has yet been performed, a complete answer may be a concrete experimental design that would distinguish the candidate mechanisms by their distinct, testable predictions, rather than newly acquired experimental data; performing the experiment is not required.

References

  1. Žiga Avsec, Melanie Weilert, Avanti Shrikumar, Sabrina Krueger, Amr Alexandari, Khyati Dalal, Robin Fropf, Charles McAnany, Julien Gagneur, Anshul Kundaje, Julia Zeitlinger. 'Base-resolution models of transcription-factor binding reveal soft motif syntax.' Nature Genetics 53, 354–366 (2021). DOI: 10.1038/s41588-021-00782-6. https://doi.org/10.1038/s41588-021-00782-6
  2. Jieni Hu, Julianne Fetter, David R. Koes, Maria Chikina. 'InfluenceNet: Encoding Motif Influence for Interpretable Modeling of Cis-Regulatory Syntax.' bioRxiv 2025.11.20.689553 (2025). DOI: 10.1101/2025.11.20.689553. https://doi.org/10.1101/2025.11.20.689553
  3. Gökçe G. Agsu, Yuze Cao, Diyan Jain, Stanley E. Strawbridge, Sam Daly, Ziwei Zhang, Annabelle Wurmser, Kathryn Bowers, Igor Orsine De Almeida, Gergo Nikolenyi, Sebastian Berger, Mantas Jonaitis, Khalid Guma'a, Tekle Pauzaite, Smitha Maretvadakethope, Lawrence Bates, James A. Nathan, Brian D. Hendrich, Ruben Perez-Carrasco, Kevin J. Chalut, David Klenerman, Thorsten E. Boroviak, Steven F. Lee, David Holcman, Srinjan Basu. 'Protein-protein interactions drive differences in the spatiotemporal dynamics of transcription factors NANOG and SOX2 in naïve pluripotent cells.' bioRxiv 2025.12.03.691924 (2025). DOI: 10.64898/2025.12.03.691924. https://doi.org/10.64898/2025.12.03.691924
  4. Bing-Rui Zhou, Benjamin Orris, Ruifang Guan, Tengfei Lian, Yawen Bai. 'Structural insights into the recognition of native nucleosomes by pioneer transcription factors.' Current Opinion in Structural Biology 92, 103024 (2025). DOI: 10.1016/j.sbi.2025.103024. https://doi.org/10.1016/j.sbi.2025.103024
  5. Maria Tsompana, Patrick D. Wilson, Vijaya Murugaiyan, Christopher R. Handelmann, Michael J. Buck. 'Defining transcription factor nucleosome binding with Pioneer-seq.' PLoS Genetics 21, e1011813 (2025). DOI: 10.1371/journal.pgen.1011813. https://doi.org/10.1371/journal.pgen.1011813
  6. Seungsoo Kim, Joanna Wysocka. 'Deciphering the multi-scale, quantitative cis-regulatory code.' Molecular Cell 83, 373–392 (2023). DOI: 10.1016/j.molcel.2022.12.032. https://doi.org/10.1016/j.molcel.2022.12.032