Massively multiplex single-molecule oligonucleosome footprinting

Nour J. Abdulhay, Colin P McNally, Laura J Hsieh, Sivakanthan Kasinathan, Aidan Keith, Laurel S Estest, Mehran Karimzadeh, Jason G. Underwood, Hani Goodarzi, Geeta J. Narlikar, Vijay RamaniZenodo 2020 · DOI

TL;DR

This paper introduces SAMOSA, a novel single-molecule sequencing method that precisely maps nucleosome positions and their organization within chromatin fibers. By combining enzymatic footprinting with long-read sequencing, SAMOSA reveals a surprising diversity of nucleosome arrangements across the genome, challenging the notion of static chromatin structures, particularly in heterochromatin.

What Problem Does This Paper Address?

Our understanding of how DNA is packaged into chromatin, the fundamental unit of the genome, has largely relied on two types of approaches: imaging techniques that show the overall structure but lack molecular detail, and bulk biochemical methods that provide high resolution but average out information from many DNA molecules. This leaves a gap in our knowledge of how individual chromatin fibers are organized at the nucleosome level and how this organization varies across different genomic regions. Specifically, it's been difficult to determine the precise positioning and regularity of nucleosomes on individual DNA molecules across the entire genome, especially in a way that is both non-destructive and high-throughput. This limits our ability to understand how chromatin structure influences gene regulation and other crucial cellular processes.

Methods

The core innovation is the Single-molecule Adenine Methylated Oligonucleosome Sequencing Assay (SAMOSA). This method combines two key techniques:

  1. Adenine Methyltransferase Footprinting: An enzyme called EcoGII is used to methylate adenine bases in DNA. Crucially, this enzyme can only access accessible adenines. When nucleosomes are bound to DNA, they protect certain adenine bases from methylation. This creates a "footprint" that reveals where nucleosomes are located.
  2. Single-Molecule Real-Time (SMRT) Sequencing: The methylated DNA is then sequenced using PacBio's SMRT sequencing technology. This long-read sequencing method can detect the altered "interpulse duration" (IPD) that occurs when the polymerase encounters a methylated adenine, allowing for the direct detection of these modifications without damaging the DNA.

By combining these, SAMOSA can simultaneously capture information about nucleosome positioning (via protected DNA regions) and the accessibility of DNA (via methylation patterns) on individual chromatin molecules. The data is then processed using custom computational pipelines to call nucleosome positions and analyze patterns of regularity and nucleosome repeat length (NRL).

Datasets

The study utilized two main types of datasets:

  • In vitro assembled nonanucleosome arrays: These were used for initial proof-of-concept experiments to validate the SAMOSA method's ability to accurately determine nucleosome positions and NRLs.
  • Genomic DNA from K562 cells: This provided the primary dataset for genome-wide analysis of oligonucleosome patterns across different epigenomic domains.

The paper also references and compares its findings with data from existing epigenomic assays like ENCODE ChIP-seq, DNase-seq, and MNase-seq.

Key Insights and Ideas

  • Chromatin is surprisingly heterogeneous: Contrary to the idea of static, uniform chromatin structures, SAMOSA reveals a diverse landscape of nucleosome arrangements even within seemingly uniform regions like constitutive heterochromatin.
  • Oligonucleosome patterns are key to epigenomic domains: Different epigenomic states (e.g., active, silent, heterochromatic) are characterized by distinct distributions of these oligonucleosome patterns, suggesting that the precise organization of nucleosomes plays a crucial role in defining these domains.
  • Transcription factor binding sites exhibit dynamic accessibility: SAMOSA can capture transient nucleosome occupancy around transcription factor binding sites, revealing states of accessibility that may be modulated by TF binding and nucleosome dynamics.
  • A new "multi-omic" approach: SAMOSA represents a step towards integrating multiple biochemical signals on a single molecule, paving the way for more comprehensive studies of genome regulation.

Key Results

  • SAMOSA accurately recapitulates known nucleosome positions and nucleosome repeat lengths (NRLs) in in vitro assembled chromatin.
  • Genome-wide analysis of K562 cells revealed seven distinct "oligonucleosome patterns" based on regularity and NRL.
  • These patterns are not uniformly distributed; their relative abundance varies significantly across different epigenomic domains (e.g., euchromatin vs. heterochromatin).
  • Constitutive heterochromatin, previously thought to be static, shows a striking mixture of regular and irregular nucleosome patterns.
  • SAMOSA identified distinct nucleosome accessibility states around transcription factor binding sites, suggesting a dynamic model of TF-DNA interaction.

Limitations and Future Work

  • Fragment size limitation: The current SAMOSA protocol enriches for fragments between ~500 bp and ~2 kb. While PacBio can sequence much longer fragments, further optimization is needed to extend SAMOSA's reach to kilobase-domain-scale studies.
  • Transient interactions: The method requires solubilization of chromatin, making it less suitable for studying very weak or transient protein-DNA interactions.
  • Cell cycle contribution: The study used unsynchronized K562 cells, so the contribution of cell cycle dynamics to observed heterogeneity remains to be explored.
  • Reference map generation: While a proof-of-concept, SAMOSA could be routinely used to generate high-coverage reference maps of the epigenome with decreasing sequencing costs and improving enrichment technologies.
  • Future directions: The authors suggest further work to explore the biological significance of the observed heterogeneity, the interplay of transcription factors and nucleosome dynamics, and the role of chromatin remodeling enzymes. They also envision combining SAMOSA with cellular perturbation experiments.

Summary updated

Abstract

These are the intermediate data used in "Massively multiplex single-molecule oligonucleosome footprinting", where the nonspecific adenine methyltransferase EcoGII was used to footprint accessible chromatin, and the methylation was then read using the Pacific Biosciences sequencing platform. The files here are intermediate outputs that capture metrics about the inter-pulse distance values as well as predictions of methylation status. The .npy, .feather, and .pickle files are the output of extractIPD.py, and callNucPeaks.py. The .csv is the output of a cell in SAMOSA_analyses.ipynb. All of this code can be found at https://github.com/RamaniLab/SAMOSA, including SAMOSA_analyses.ipynb which contains all downstream analyses that were performed on this data. <strong>meanIPDinfoChrControls.csv</strong>: This contains the data used to generate Supplementary Figures 2 and 3. It contains various summary measurements of the IPD values in each molecule of the <em>in vitro</em> samples <strong>Files ending in _bingmm.npy:</strong> These contain the posterior probability of adenines being methylated for the <em>in vitro</em> data. The files beginning with pbrun3 were sequenced on the Sequel I, and the files beginning with pbrun4 or pbrun5 were sequenced on the Sequel II. Other than Supplementary Figures 2 and 3, the analysis in the paper was based on the Sequel II data. naked_neg and DNA_minusM are both negative controls. naked_methyl and DNA_plusM are positive controls. chromatin samples are <em>in vitro</em> assembled chromatin. <strong>pbrun4_gold_nuc47_chromatin_peaks.feather</strong>: The estimated nucleosome centers from the <em>in vitro</em> assembled chromatin, in a data frame. Each row is an individual nucleosome dyad prediction. <strong>Files ending in _onlyT_zmwinfo.pickle</strong>: These files each contain a pandas data frame containing information about each molecule sequenced in the <em>in vivo</em> samples. These should be read in in python using the pandas read_pickle function, and require the same namespace as was used when saving them, so pandas must be imported as pd, and numpy as np. The neg samples are deproteinated unmethylated molecules, the pos samples are deproteinated methylated molecules, and the chromatin samples are methylated chromatin. <strong>Files ending in _bingmm.pickle</strong>: These files contain the posterior probability of being methylated for each adenine in each DNA molecule in the sample. They each have a corresponding zmwinfo file described above, and similarly must be read in with numpy imported as np. Each file is a dictionary with the zero-mode waveguide (ZMW) hole number as a key, and the value a numpy array with length equal to the unaligned CCS of that molecule, with methylation posterior probabilities at each A/T base. The zmwinfo dataframe has a 'zmw' column that can be used to match the information in that file with the methylation information in this one.
BibTeX
@inproceedings{abdulhay2020massively,
  author = {Nour J. Abdulhay and Colin P McNally and Laura J Hsieh and Sivakanthan Kasinathan and Aidan Keith and Laurel S Estest and Mehran Karimzadeh and Jason G. Underwood and Hani Goodarzi and Geeta J. Narlikar and Vijay Ramani},
  title = {{Massively multiplex single-molecule oligonucleosome footprinting}},
  booktitle = {Zenodo},
  year = {2020},
  doi = {10.5281/zenodo.4289173},
}
Figure