Figure 5. Characterization of erythropoietin (EPO)-exposed hematopoietic stem and progenitor cells (HSPCs) by single-cell RNA sequencing (scRNAseq).

HSPCs were sorted, barcoded, and cultured ex vivo with or without 1000 ng/ml EPO for 16 hr and analyzed by scRNAseq using the 10X Genomics platform. 1706 cells from control and 1595 cells from the EPO group passed quality control. (a) Volcano plot of log2 fold change of the differentially expressed genes between control and EPO-exposed cells versus the adjusted p-value. Genes of interest are annotated. Differentially expressed genes were used to define an EPO response signature. (b) UMAP visualization of the EPO-exposed and control HSPCs. (c) The level of expression in the EPO-exposed HSPCs of the genes in the EPO response signature (top), and definition of the EPO responder and nonresponder subgroups using the 90th percentile expression of the EPO response signature from (c) (bottom). (d) The expression of the indicated genes in the control, EPO responder, and nonresponder subgroups as defined in (c). Genes that are significantly upregulated in the EPO responder group when compared to the control and nonresponder groups. Differential expression was assessed using a logistic regression testing approach, as implemented in Seurat. Figure supplements correspond to one 10× experiment of a pool of eight mice.

Figure 5—figure supplement 1. Single-cell RNA sequencing (scRNAseq) analysis of control and erythropoietin (EPO)-exposed hematopoietic stem and progenitor cells (HSPCs).

(a) UMAP projection of the scRNAseq dataset from Dahlin et al., 2018 annotated with flt3, CD150, and gata1 gene expression. (b) Projecting of erythroid-biased progenitors from Tusi et al., 2018 on UMAP projection of (a). (c) Robustness of the UMAP visualization and unsupervised clustering of the data in Figure 4c.The amount of variance explained by each principle component (left) and UMAP-based visualization using 10 principal component analysis (PCA) (right) for different number of genes. (d) The expression of genes encoding known EPO receptors in each subgroup. (e) Overview of the reference map using supervised cell-type annotation of the dataset from Dahlin et al., 2018. On the right-hand side, we overlay the MPP4 signature defined by Pietras et al., 2015 onto our reference map to facilitate cell-type annotation. (f) HSPCs were sorted, barcoded, and cultured ex vivo with or without 1000 ng/ml EPO for 16 hr and analyzed by scRNAseq using the 10X Genomics platform. Mapping of the transcriptomes of the 1706 cells from control and 1595 cells from the EPO group obtained after quality control onto the reference map using a k-nearest-neighbors mapping approach.

Figure 5—figure supplement 2. Unsupervised clustering of control and erythropoietin (EPO)-exposed hematopoietic stem and progenitor cells (HSPCs).

(a) Cluster stability analysis varying the resolution parameter of the Seurat clustering method. The significant variable genes using 10 principal component analysis (PCA) were used as input. (b) UMAP visualization of the data in (a) using a clustering resolution of 0.1, with the proportion of each cluster as in the control, EPO responder, and nonresponder subgroups. (c) Expression of published signatures of established cell types used to annotate our clusters. All comparisons in signature expression between clusters were statistically significant (adjusted p-value<0.05) as determined by a Kruskal–Wallis test and Dunn’s post-hoc analysis. (d) Nearest-neighbor mapping of unsupervised clusters from (b) onto the reference map.