Department of Chemistry, Massachusetts Institute of Technology, Cambridge, Massachusetts.
Department of Chemistry, Massachusetts Institute of Technology, Cambridge, Massachusetts.
Biophys J. 2019 May 21;116(10):2047-2056. doi: 10.1016/j.bpj.2019.04.006. Epub 2019 Apr 11.
Epigenetic modifications can extend over long genomic regions to form domain-level chromatin states that play critical roles in gene regulation. The molecular mechanism for the establishment and maintenance of these states is not fully understood and remains challenging to study with existing experimental techniques. Here, we took a data-driven approach and parameterized an information-theoretic model to infer the formation mechanism of domain-level chromatin states from genome-wide epigenetic modification profiles. This model reproduces statistical correlations among histone modifications and identifies well-known states. Importantly, it predicts drastically different mechanisms and kinetic pathways for the formation of euchromatin and heterochromatin. In particular, long, strong enhancer and promoter states grow gradually from short but stable regulatory elements via a multistep process. On the other hand, the formation of heterochromatin states is highly cooperative, and no intermediate states are found along the transition path. This cooperativity can arise from a chromatin looping-mediated spreading of histone methylation mark and supports collapsed, globular three-dimensional conformations rather than regular fibril structures for heterochromatin. We further validated these predictions using changes of epigenetic profiles along cell differentiation. Our study demonstrates that information-theoretic models can go beyond statistical analysis to derive insightful kinetic information that is otherwise difficult to access.
表观遗传修饰可以在长基因组区域上延伸，形成域级染色质状态，这些状态在基因调控中起着关键作用。这些状态的建立和维持的分子机制尚未完全理解，并且利用现有实验技术研究仍然具有挑战性。在这里，我们采用了一种数据驱动的方法，并对信息论模型进行了参数化，以便从全基因组表观遗传修饰谱中推断域级染色质状态的形成机制。该模型再现了组蛋白修饰之间的统计相关性，并确定了众所周知的状态。重要的是，它预测了常染色质和异染色质形成的截然不同的机制和动力学途径。特别是，长而强的增强子和启动子状态通过多步过程逐渐从短但稳定的调控元件中生长。另一方面，异染色质状态的形成具有高度的协同性，在过渡路径中没有发现中间状态。这种协同作用可能源于染色质环介导的组蛋白甲基化标记的扩散，并支持异染色质的塌陷、球形三维构象，而不是规则的纤维结构。我们进一步使用细胞分化过程中表观遗传谱的变化来验证这些预测。我们的研究表明，信息论模型可以超越统计分析，得出难以获得的有见地的动力学信息。
