Echo characterizes the desynchronization of gene expression and chromatin accessibility during cell-state transitions

Abstract

Cell-state transitions during differentiation and disease involve coordinated changes across gene expression and chromatin accessibility, but these modalities do not change in lockstep. For example, regulatory elements can be primed before their target genes are expressed or remain accessible after expression ceases. This desynchronization between changes in gene expression and chromatin accessibility can manifest at the level of cell states. Understanding the drivers of this desynchronization can give insights into the molecular mechanisms underlying cell-state progression. Here we introduce Echo, a statistical framework that identifies desynchronized cell states and the associated genes and regulatory elements from paired single-cell RNA and ATAC data. Echo States estimates cell-state density independently for each modality and compares them to determine which states are better resolved in RNA or ATAC. Echo Features then predicts feature values over each state space to identify the genes, regulatory loci, and transcription-factor motifs driving this desynchronization. Applying Echo to the developing human fetal retina, we find that desynchronization is pervasive across every major cell population. Expression of cell-cycle genes resolves multipotent progenitors in gene expression but not chromatin accessibility, while fate priming resolves cycling neurogenic precursors in chromatin accessibility before gene expression. By combining desynchronized states and features along the cone trajectory, we reconstructed the regulatory logic of cone fate specification from multipotent progenitors, revealing a tight coupling between multipotency exit, cell cycle and lineage specification. Applying Echo to human hematopoiesis, we identified that the balance between stem-cell quiescence and differentiation is resolved more strongly in chromatin accessibility than in gene expression. Our results establish desynchronization as a pervasive, structured feature of differentiating systems, and Echo as a framework for characterizing the interplay between gene expression and chromatin accessibility during cell-state transitions.