Scalable in vitro production of defined mouse erythroblasts

Abstract

Mouse embryonic stem cells (mESCs) can be manipulated in vitro to recapitulate the process of erythropoiesis, during which multipotent cells undergo lineage specification, differentiation and maturation to produce erythroid cells. Although useful for identifying specific progenitors and precursors, this system has not been fully exploited as a source of cells to analyse erythropoiesis. Here, we establish a protocol in which characterised erythroblasts can be isolated in a scalable manner from differentiated embryoid bodies (EBs). Using transcriptional and epigenetic analysis, we demonstrate that this system faithfully recapitulates normal primitive erythropoiesis and fully reproduces the effects of natural and engineered mutations seen in primary cells obtained from mouse models. We anticipate this system to be of great value in reducing the time and costs of generating and maintaining mouse lines in a number of research scenarios.

Fig 1. In vitro differentiated EBs as a source of erythroid cells.

A) RT-PCR of mature mouse globin transcripts in mESCs and EBs between days 4–7 of differentiation, normalized to the 18S housekeeping gene. Levels are shown relative to the maximal detected expression for each gene. Bars represent mean values from three independent differentiations; error bars represent standard deviation of the mean. B) Cell counts for immunophenotypically-defined populations using antibodies for CD71 and Ter119 cell surface markers through days 4–7 of EB differentiation. Data shown are from a representative differentiation in 10 cm dish format. C) Fluorescence levels of an α-globin-YFP tag in mESCs and day 7 EBs. Top: a schematic of the tag shows the insertion site after the final exon of the gene with a 2A self-cleaving peptide sequence in yellow, the mVenus coding region in green, nuclear localization signal (NLS) repeats in blue, the STOP codon in red and the untranslated regions (UTRs) in pink. Bottom: brightfield images of mESCs and a single EB are overlaid with YFP fluorescence signal (left panels). Flow cytometry histograms for YFP fluorescence demonstrate the presence of an α-globin-positive population (green peak) in day 7 EBs (middle panels). Flow cytometry plots for the erythroid markers CD71 and Ter119 show the overlap of YFP+ population from the histogram with the CD71+ cell populations (YFP+ cells labelled green as in the histogram) in day 7 EBs (right panels). D) Protocol summary for the generation of EB-derived CD71+ erythrocytes in vitro. Example data for column-based CD71+ cell purification, starting with brightfield images of cultured mESCs (top panel), to whole EB (middle panel), to CD71-separated populations shown as cell pellet images for CD71+ (red pellet) and CD71- (clear pellet) fractions. A stained (modified Wright stain) cytospin preparation is shown for the purified CD71+ erythroid population (bottom panel) with an inset (black square) highlighting (black arrows) specifically mature primitive erythroid cells and their distinclive morphology; large nucleated hemoglobinised cells. E) Flow cytometry plots for CD71 and Ter119 markers are shown for populations at each step of the protocol. F) An overlay of CD71 histograms from all day 7 EB-derived populations (as indicated by colour) highlighting the varying intensities of CD71 expression at each step of the protocol. Note the highest CD71 intensity marking the CD71+ fraction retained by the LS column. Stained CD71- fraction (histogram and FACS plot) shows low CD71 expressing fraction unretained by the magnetic column.

Fig 2. The EB-derived erythroid population resembles embryonic primary red cells by chromatin assays.

A) ATAC-seq tracks for the α- and β-globin loci (chr11:32,131,800–32,204,799 and chr7:110,952,000–111,027,999, respectively) from mESCs and erythroid cells from day 7 EBs, E10.5 embryonic blood and Ter119+ erythrocytes from adult APH-treated spleen. Tracks are RPKM-normalized and averaged over three biological replicates. Embryo-specific globin genes are highlighted in orange; adult globin genes are highlighted in pink. Other visible peaks represent erythroid-specific enhancer elements. B) PCA plot of genome-wide ATAC-seq peaks from the tissues shown above. EB-derived cells cluster with E10.5 primary erythroid cells (highlighted by the dashed circle), away from mESCs and spleen-derived erythrocytes. C) ChIP-seq tracks for CTCF and histone marks (H3K27ac and H3K4me1 for enhancers and H3K4me3 for promoters) in EB-derived erythrocytes at the α-and β-globin loci. ChIP-seq data are shown as average tracks across two biological replicates, normalized by RPKM.

Fig 3. EB-derived erythrocytes are uniformly of the primitive lineage.

A) Single-cell RT-PCR (Biomark System, Fluidigm) from FACS-sorted primary and EB-derived erythroid cells as indicated in the colour coded groups for embryonic (Hba-x and Hbb-y) and adult (Hba-a1/2 and Hbb-b1/2) globin genes. B) PCA of expression data from a panel of 40 probes used for single-cell RT-PCR (Biomark System, Fluidigm) to distinguish primitive and definitive lineages of specific differentiation stages. EB-derived erythroid cells (dark and light blue dots) most closely resemble E9.5 primary cells (orange circles with a black outline) as indicated with the black dotted circle.

Fig 4. EB-derived ΔR1 erythroid cells recapitulate the molecular phenotype of their in vivo mouse-derived ΔR1 erythrocytes.

A) Expression data for indicated genes based on mature transcripts from enhancer R1 knock out mESCs (ΔR1) day 7 EB-derived erythroid cells, normalized to the embryonic β-globin genes. Levels are shown relative to wildtype day 7 EB-derived erythroid cells (WT). Bars represent mean values from at least six independent differentiations; error bars represent standard deviation of the mean. Student’s t-test *P <0.001. B) Expression data for indicated genes based on mature transcripts from CTCF (HS38-39) knock out mESCs (D3839) day 7 EB-derived erythroid cells, normalized to the embryonic β-globin genes. Levels are shown relative to equivalent wildtype cells (WT). Bars represent mean values from at least six independent differentiations; error bars represent standard deviation of the mean. Student’s t-test *P <0.001. C) RPKM-normalized ATAC-seq and CTCF ChIP-seq tracks averaged for three replicates of wildtype and D3839 erythroid cells, both derived from day 7 EBs. D) Differential interactions (by NG-Capture-C) of α-globin regulatory regions and flanking genes between WT and D3839 d7 EB-derived erythroid cells. Capture-C data for the indicated viewpoints (black asterisks) in WT and D3839 erythroid cells are shown. Data representing at least 3 independent differentiation for two independently generated clones were used. Differential tracks show a subtraction (D3839-WT) of the mean number of normalized meaningful interactions per restriction fragment.

Fig 5. EB differentiation scaled down to a 96-well format.

A) A schematic comparison of protocols for 10 cm-dish format and miniaturized 96-well plate EB differentiation. B) Characterization of the erythroid output from miniaturized EB differentiation by flow cytometry for CD71 and Ter119 erythroid markers, gross-scale microscopy of whole EBs and cellular MGG staining. Whole EB images show a single 96-well and a sample of a 10 cm dish, respectively. C) Total and CD71+ cell counts from day 7 EBs for a range of mESCs plating densities in 200 μl differentiation media in 96-well format. Optimal CD71+ output is at around 300 cells/well. D) PCA plot of genome-wide ATAC-seq peaks, as in Fig 2B, with the addition of CD71+ cells from 96-well EBs (96W d7EB). 96-well miniaturized EB culture clusters with bulk day 7 EB-derived cells (EBd7) and E10.5 embryonic erythrocytes away from mESCs and spleen-derived erythrocytes. E) Median YFP fluorescence readings from erythroid cells from 1–6 combined wells of a 96-well plate containing EBs for three genetically engineered α-globin YFP-tagged mESC lines: wildtype (WT), enhancer R1 knockout (ΔR1) and enhancers R1, R2 knockout (ΔR1R2). Solid lines indicate the average median YFP fluorescence for each cell line. Points highlighted with a black outline represent the 95% confidence interval (CI) values for each dataset. Combining 3 wells is enough to reduce output variability/genotype.