Myelo-lymphoid lineage restriction occurs in the human haematopoietic stem cell compartment before lymphoid-primed multipotent progenitors

Abstract

Capturing where and how multipotency is lost is crucial to understand how blood formation is controlled. Blood lineage specification is currently thought to occur downstream of multipotent haematopoietic stem cells (HSC). Here we show that, in human, the first lineage restriction events occur within the CD19^-CD34⁺CD38^-CD45RA^-CD49f⁺CD90⁺ (49f⁺) HSC compartment to generate myelo-lymphoid committed cells with no erythroid differentiation capacity. At single-cell resolution, we observe a continuous but polarised organisation of the 49f⁺ compartment, where transcriptional programmes and lineage potential progressively change along a gradient of opposing cell surface expression of CLEC9A and CD34. CLEC9A^hiCD34^lo cells contain long-term repopulating multipotent HSCs with slow quiescence exit kinetics, whereas CLEC9A^loCD34^hi cells are restricted to myelo-lymphoid differentiation and display infrequent but durable repopulation capacity. We thus propose that human HSCs gradually transition to a discrete lymphoid-primed state, distinct from lymphoid-primed multipotent progenitors, representing the earliest entry point into lymphoid commitment.

Fig. 1

A continuum of differentiation outputs and transcriptomes within the human HSC/MPP pool and 49f⁺ HSCs compartments. a, d Percentage of colonies of the indicated type derived from HSC/MPP pool single cells (n = 435 colonies from six independent CBs) (a) and from 49f⁺ HSCs (n = 819 colonies from four independent CBs) (d); mean ± SEM is shown. b, c, e, f, Principal component analysis (PCA) of the surface marker expression at the time of sort of single HSC/MPP pool cells (n = 435 colonies from six independent CBs) (b, c) and of 49f⁺ HSCs (n = 714 colonies from four independent CBs) (e, f). Colours indicate the type of differentiated colony produced by each single-cell after culture. Density plots (top panel) and boxplots (bottom panel) of PC1 (b, e), and PC2 values (right panel, e) from single cells producing indicated types of differentiated colonies; *p < 0.05 by Kruskal–Wallis test with multiple comparisons. c, f Cell surface markers with the highest PC loadings are shown. g Mean time to first division of single cells producing colonies containing the indicated cell types (EC₅₀ of non-linear fit of cumulative first division kinetics); n = 4 independent experiments each with an independent CB (with respectively 19, 98, 34, 230 single cells per experiment), ***p < 0.001 by two-sided paired t-test. h Normalised cell surface expression of CLEC9A (left) and CD34 (right) at the time of sort (same HSC/MPP pool cells as in b) for single cells producing the indicated types of differentiated colonies. Of note, in e and h, My/NK/Ery includes My/NK/Ery and My/NK/Ery/Meg colonies. **p < 0.01 by Kruskal–Wallis test with multiple comparisons. i–k Single-cell RNA-seq analysis of 49f⁺ HSCs (n = 169 single cells that passed quality control). i, j Cell surface expression of CLEC9A (i) and CD34 (j) overlaid on the tSNE representation of 49f⁺ HSCs. k Cell surface expression levels of CLEC9A (left) and CD34 (right) in the 49f⁺ HSCs of the indicated ICGS clusters; ***p < 0.001 by two-sided unpaired t-test. All boxplots show median, interquartile and 5–95 percentile

Fig. 2

Prospective purification of HSC/MPP pool and 49f⁺ HSC cells with distinct in vitro differentiation capacity. a Representative examples of the gating strategy derived from in silico analysis used to isolate the indicated populations. b–e Single cells were plated in the same assay as in Fig. 1a–f. Percentage of colonies containing differentiated cells of the indicated lineages, generated by single Subset1 (red) or Subset2 cells (blue) (b) and by single 49f⁺ Subset1 (red) or 49f⁺ Subset2 cells (blue) (c). Percentage of colonies of the indicated types derived from Subset1 and Subset2 (d) and from 49f⁺ Subset1 and 49f⁺ Subset2 single cells (e). n = 4 independent CBs, n = 270 colonies from Subset1, n = 171 colonies from Subset2 (b, d). n = 5 independent CBs, n = 628 colonies from 49f⁺ Subset1, n = 522 from 49f⁺ Subset2 (c, e). Mean ± SEM is shown *p < 0.05, **p < 0.01, ***p < 0.001 by two-sided paired t-test (b–e). f Representative example of cumulative first division kinetics of 49f⁺ Subset1 (red) and 49f⁺ Subset2 (blue) single cells. Curves are a non-linear least squares ordinary fit (non-log); ***p < 0.001 by extra sum of squares F-test. g Mean time to first division of 49f⁺ Subset1 (red) and 49f⁺ Subset2 (blue) (EC₅₀ of non-linear fit of cumulative first division kinetics as in f); n = 5 independent experiments each with an independent CB (with, respectively, 384, 384, 360, 150 and 192 single cells per experiment); ***p < 0.001 by two-sided paired t-test. h Cells were plated in an assay supporting differentiation of My, B and NK cells. Percentage of colonies generated by Subset1 (red) or Subset2 (blue) single cells and containing differentiated cells of the indicated lineages; mean ± SEM is shown, n = 3 independent CBs (Subset1 138 colonies, Subset2 142 colonies), *p < 0.05, **p < 0.01 by two-sided paired t-test. i Number of colonies/100 cells from either Subset1 or Subset2 populations plated in a CFU assay. The type of colony: erythroid (E), granulocyte and myeloid (GM) or a combination of both (mix) is shown, (n = 2 independent CBs); mean ± SEM is shown

Fig. 3

Polarisation of transcriptional profiles and establishment of lineage-priming programmes in the 49f⁺ HSC compartment. a–e Single-cell RNA-seq analysis of 49f⁺ HSCs (n = 169), 49f⁺ Subset1 (n = 78) and 49f⁺ Subset2 (n = 75) single cells. a tSNE representation of single 49f⁺ HSCs (grey), 49f⁺ Subset1 (red) and 49f⁺ Subset2 (blue) cells. All tSNE analyses were performed on highly variable genes (2420 genes computed as in ref. ). b, c ICGS analysis. b Heatmap of expression of guide genes selected by ICGS. Genes are represented in rows, with selected genes annotated on the side. Columns represent single cells, the phenotype of which is indicated by the uppermost bar of the top panel. Bottom bar of the top panel shows major clusters identified by ICGS. c Percentage of 49f⁺ Subset1 and 49f⁺ Subset2 single cells found in the indicated ICGS clusters; ***p < 0.001 by two-tailed Fisher test. d Population-specific signatures from, and lineage-priming gene modules derived from significantly enriched (Benjamini-Hochberg adjusted p-value < 0.05 by pre-ranked GSEA) in 49f⁺ Subset1 and 49f⁺ Subset2 cells. e Log₁₀ normalised expression of selected genes differentially expressed between 49f⁺ Subset1 (red) and 49f⁺ Subset2 (blue) cells (FDR < 0.05 by DESeq2). f–h 20-cell RNA-seq of Subset1 and Subset2. Significantly enriched (Benjamini-Hochberg adjusted p-value < 0.05 by pre-ranked GSEA) population-specific signatures from (f), lineage-priming gene modules derived from (g), and gene ontology terms and MSigDB genesets (h) in Subset1 and Subset2 cells. NES normalised enrichment score

Fig. 4

49f⁺ Subset1 cells generate 49f⁺ Subset2 cells in vitro. a Sorting strategy used to isolate 49f⁺ Subset1 and 49f⁺ Subset2 populations at day 0 (top panel) and their derived populations after 5 days in culture (bottom panel). b Single cells from each of the indicated gates were sorted. Colours indicate the type of colony produced by each cell after 3 weeks of culture. n = 576 cells plated from three independent CBs for 49f⁺ Subset1 derived populations, n = 576 cells plated from three independent CBs for 49f⁺ Subset2 derived populations. c Percentage of colonies derived from single cells from the indicated day 5 populations. n = 3 independent CBs (except for Diff2 where n = 1). Mean ± SEM is shown. The total number of colonies analysed is shown for each population on the top of each bar

Fig. 5

Distinct long-term repopulation and differentiation capacities of Subset1 and Subset2 cells in vivo. a Left: Representative flow cytometry plots from the injected femur of mice engrafted with CB CD34⁺ cells and injected with seven doses of PBS (control) or EPO (20 units/injection). Right: relative Ery engraftment (percentage of total human engrafted cells), shown is mean ± SEM, *p < 0.05 by unpaired Mann–Whitney test. PBS: n = 4 mice, EPO: n = 11 mice. b–d Percentage of human engraftment (%CD45⁺⁺ + %GlyA⁺) in the injected femur of mice transplanted with Subset1 (red) or Subset2 (blue) cells at 2 weeks (b, n = 14 transplanted mice per population), 8 weeks (c, n = 8 transplanted mice per population) and 20 weeks (d, n = 25 transplanted mice for Subset1 and n = 20 transplanted mice for Subset2) after transplantation. Dashed line: threshold of engraftment (%CD45⁺⁺ + %GlyA⁺) ≥ 0.01 % and at least 30 cells recorded. Non-engrafted mice shown below dashed line. CD45⁺⁺ indicates cells positive for two distinct CD45 antibodies. e–g Distribution of differentiated cell types from each indicated lineage in the human graft of individual mice (injected femur, each bar represents one mouse) engrafted with Subset1 or Subset2 cells at 2 weeks (e), 8 weeks (f) and 20 weeks (g) after transplantation. My lineage: CD33⁺ cells; Ly lineage: CD19⁺⁺ cells (positive for two distinct CD19 antibodies); Ery lineage: GlyA⁺ cells. e Subset1: n = 3, Subset2: n = 4. f Subset1: n = 5, Subset2: n = 2. g Subset1: n = 19, Subset2: n = 6. h Estimation of frequency of long-term repopulating cells within 49f⁺ Subset1 and 49f⁺ Subset2 cells at 20 weeks post-transplantation by ELDA. i Percentage of human engraftment (%CD45⁺⁺ + %GlyA⁺) in the injected femur of mice injected with 49f⁺ Subset1 (red, n = 24 mice transplanted) or 49f⁺ Subset2 (blue, n = 21 mice transplanted) cells at 20 weeks after transplantation. j Distribution of differentiated cell types belonging to each indicated lineage in the human graft of individual mice (injected femur, each bar represents one mouse) engrafted with 49f⁺ Subset1 or 49f⁺ Subset2 cells at 20 weeks after transplantation. 49f⁺ Subset1: n = 19, 49f⁺ Subset2: n = 3. Arrows indicate mice with unusually low levels of Ly engraftment

Fig. 6

Molecular and functional differences between CD49f⁺ Subset2, Subset2 cells, LMPPs and MLPs. a–c Single-cell RNA-seq analysis of 49f⁺ Subset2 (n = 119), Subset2 (n = 100), LMPP (n = 140) and MLP (n = 134) single cells. a Left: 3D tSNE representation of indicated populations performed on highly variable genes (5831 genes). Right: density plot of the distribution of indicated populations along the first tSNE component. p < 0.01 by Kruskal–Wallis test with multiple comparison. b Number of significant differentially expressed genes between LMPP and the indicated populations (FDR < 0.05 by DESeq2). c Log₁₀ normalised expression of selected genes differentially expressed both between 49f⁺ Subset2 and LMPP and Subset2 and LMPP in single cells from indicated populations. d Cell surface expression of KIT on Subset2 and LMPP cells. Left: representative flow cytometry histogram, Right: quantification (n = 3 independent CBs, *p < 0.05 and **p < 0.01 by one-way ANOVA with Tukey’s multiple comparison). e–g In vitro differentiation assay of Subset2 and LMPP single cells (n = 240 single cells from two independent CBs plated per population) in conditions supporting differentiation of My, B and NK cells. e Clonogenic efficiency of Subset2 and LMPP cells (n = 2 independent CBs). f Percentage of colonies of the indicated type derived from Subset 2 (n = 153 colonies from two independent CBs) and LMPP (n = 72 colonies from two independent CBs) single cells, *p < 0.05 by two-tailed paired t-test. d–f Shown is mean ± SEM. g Boxplots of the size of colonies generated by Subset2 (n = 153 colonies from two independent CBs) and LMPP (n = 72 colonies from two independent CBs) single cells, **p < 0.01 by two-tailed Mann–Whitney test. Boxplots show median, interquartile and range. h Graphical representation of the single-cell structure of the 49f⁺ HSC compartment proposed here