Universality of cell differentiation trajectories revealed by a reconstruction of transcriptional uncertainty landscapes from single-cell transcriptomic data

Abstract

We employed our previously-described single-cell gene expression analysis CALISTA (Clustering And Lineage Inference in Single-Cell Transcriptional Analysis) to evaluate transcriptional uncertainty at the single-cell level using a stochastic mechanistic model of gene expression. We reconstructed a transcriptional uncertainty landscape during cell differentiation by visualizing single-cell transcriptional uncertainty surface over a two dimensional representation of the single-cell gene expression data. The reconstruction of transcriptional uncertainty landscapes for ten publicly available single-cell gene expression datasets from cell differentiation processes with linear, single or multi-branching cell lineage, reveals universal features in the cell differentiation trajectory that include: (i) a peak in single-cell uncertainty during transition states, and in systems with bifurcating differentiation trajectories, each branching point represents a state of high transcriptional uncertainty; (ii) a positive correlation of transcriptional uncertainty with transcriptional burst size and frequency; (iii) an increase in RNA velocity preceeding the increase in the cell transcriptional uncertainty. Finally, we provided biological interpretations of the universal rise-then-fall profile of the transcriptional uncertainty landscape, including a link with the Waddington's epigenetic landscape, that is generalizable to every cell differentiation system. MAINCell differentiation is the process through which unspecialized stem cells become more specialized. Because of its importance in development, cellular repair, and organismal homeostasis, the molecular mechanisms of cell differentiation has been the subject of intense scrutiny. Since roughly 50 years ago -along with the promulgation of the central dogma of molecular biology by Francis Crick and the characterization of the lactose operon by François Jacob and Jacques Monod -the existence of a genetic program has become a prevailing explanation for the cell differentiation process. Although the details were originally not defined, at least not formally, such a genetic program purports a constellation of master genes (i.e., transcription factors) that orchestrate the transcription of downstream target genes in a precise spatiotemporal fashion, resulting in long-lasting alterations in the gene expression patterns 1-3 . A notable experimental evidence substantiating this view is the overexpression of myoD inducing a myogenic phenotype in seemingly naive cells 4 . Over the past few decades, the repertoire of such master genes across numerous stem cell systems, such as Nanog, Oct4, Sox2, BATF and MyoD, begin to coalesce 5-7 .Recent advances in single-cell technologies has revealed new aspects of the cell differentiation that are incompatible with the idea of ordered and programmed (i.e., deterministic) gene expression. More specifically, single-cell data paint a stochastic differentiation process that increases cell-to-cell variability of gene expression.

Global genome decompaction leads to stochastic activation of gene expression as a first step toward fate commitment in human hematopoietic stem cells

“…The 5h-to-48h period corresponds to the multilineage-primed stage of the CD34+ cells that precedes the emergence of the first signs of characteristic gene expression patterns accompanying differentiation [6]. It is a universal feature of the cells during the initial phases of the fate commitment process to progress through a transitional cell state marked by the rise-then-fall in transcriptional uncertainty and a concomitant rise-and-fall of cell-to-cell variability [12]. As reported here, the gene transcription in the CD34+ cells clearly follows the same pattern.…”

“…Therefore, the process of choice can be viewed as a continuing iterative process of constrained optimization of the cell phenotype over time, a kind of "learning process" that is accomplished by the cell through interactions and cooperation with the surrounding cells and environment. This way to frame the question of fate commitment has been theorized long ago [24][25][26], and single-cell studies in the recent years have provided more and more experimental support [3,6,12,27,28].…”

“…Note that each gene in each individual cell in a cluster can be characterized by a unique likelihood value (see Methods). Here, we use the negative logarithm of the gene likelihood value (NLL) as a metric of transcriptional uncertainty, using which we can probe into the intra-cluster cell-to-cell heterogeneity in the gene expression [10,12]. As shown in Fig.…”

Repeated Decision Stumping Distils Simple Rules from Single Cell Data

“…In addition to linear developmental processes, we are also able to characterise branching cell fate decisions [15] using ReDX. We illustrate this in the context of early Xenopus tropicalis embryogenesis [29].…”

“…stem cell and developmental biology, but also in immunology and cancer research. Changes at the transcriptome level are widely seen as indicative of, and even causal for, cell fate decision making and progression of cells along the differentiation path [14,15]. The riches contained in these data, do not reveal themselves easily, however, as a host of problems plague analysis and interpretation [12,2]: experimental (undesirable) noise interferes with the molecular noise (which holds signatures of molecular processes); the high dimensionality of the data poses considerable statistical challenges.…”