Single Cell Transcriptomics to Understand HSC Heterogeneity and Its Evolution upon Aging

Abstract

Single-cell transcriptomic technologies enable the uncovering and characterization of cellular heterogeneity and pave the way for studies aiming at understanding the origin and consequences of it. The hematopoietic system is in essence a very well adapted model system to benefit from this technological advance because it is characterized by different cellular states. Each cellular state, and its interconnection, may be defined by a specific location in the global transcriptional landscape sustained by a complex regulatory network. This transcriptomic signature is not fixed and evolved over time to give rise to less efficient hematopoietic stem cells (HSC), leading to a well-documented hematopoietic aging. Here, we review the advance of single-cell transcriptomic approaches for the understanding of HSC heterogeneity to grasp HSC deregulations upon aging. We also discuss the new bioinformatics tools developed for the analysis of the resulting large and complex datasets. Finally, since hematopoiesis is driven by fine-tuned and complex networks that must be interconnected to each other, we highlight how mathematical modeling is beneficial for doing such interconnection between multilayered information and to predict how HSC behave while aging.

Keywords: Boolean modeling; HSC aging; bioinformatics; single cell transcriptomic.

Figure A1

Example of Boolean network modeling for the three transcription factors Gata1, Gata2 and Zfpm1 in HSC erythroid-priming context. The influence graph represents the gene regulatory network whose dynamics is encoded with logical rules. Transitions between Boolean states (representative of cells at different times of differentiation) can be analyzed by asynchronously updating the components according to logical rules.

Figure 1

Cellular and molecular changes of aged HSCs. On the left, the zoom into young (top) and old (bottom) HSCs summarizes four biological processes that have been proposed to be involved in HSC aging: metabolism (ROS, proteostasis), DNA damage, epigenetics (green and red circles: active and repressive histone marks) and signaling. Interconnections between these processes lead to an altered function revealed by changes in transcriptome signatures. On the right, hematopoiesis is schematized. Young hematopoiesis is characterized by a balanced differentiation, leading to accurate levels of myeloid and lymphoid cells. With aging, intrinsic changes in HSCs occur, resulting in a myeloid bias and immunosenescence.

Figure 2

Evolution of single-cell transcriptomic technologies and their application.

Figure 3

scRNA-seq workflow for studying hematopoiesis in mice. (a) Isolation, sorting, capture of cells of interest and preparation of libraries for a pool of mice. Example of a droplet-based technology, R1: Read 1 biological, BC: cell barcode, BE: sample barcode. Primary data processing after sequencing: Demultiplexing of binary base call (BCL) files in FASTQ files that are aligned to the reference genome, then transcript counts per cell are quantified using the unique molecular identifiers (UMIs). In this example, the expression of 30,000 genes is detected for 8000 cells. (b) Quality control (QC) of the cells and filtering of lowly expressed genes. In this example 10,000 genes expressed in 7800 cells are conserved. Normalization of counts and some supervised analyses (cell cycle scoring/phase assignment, supervised cell type annotation) can be performed. (c) Highly variable genes (HVGs) are selected for dimension reduction and cell clustering. Confounding factors (cell cycle, percentage of mitochondrial transcripts, etc.) can be regressed out during the scaling of the HVG expression. (d) A first linear dimension reduction with a PCA to summarize the information. The most informative principal components (PCs) are kept regarding the drop in the percentage of explained variance. (e) A clustering and a visualization with UMAP (or tSNE) are conducted on these PCs. In addition, a pseudo-trajectory can be inferred with the selected top PCs. Finally, differentially expressed gene (DEG) analyses between clusters/conditions or along pseudotime (pseudoT) are usually performed on normalized expression data.

Figure 4

Boolean Network inference. Workflow of Boolean network inference from scRNA-seq data. (a) scRNA-seq data can be used to infer transcriptional interactions between TFs and complement influence graph constructed from prior knowledge of the searched BN. (b) The pseudo trajectories issued from the scRNA-seq data, can be discretized and translated in discrete observations of the searched BN. (c) Given these inputs, constraint programming can be used to infer the logical rules of the BN.