Coordinated regulation of self-renewal and cell cycle during human lympho-myeloid lineage restriction

Abstract

Recent studies indicate the human lympho-myeloid restriction process to be a different and more heterogeneous one than historically inferred. Here we describe the development of bulk and clonal culture systems that efficiently support early B-lymphoid differentiation and its use to elucidate the biological and molecular changes that accompany their initial restriction from subsets of CD34+ human cord blood cells with lympho-myeloid-limited potential. Analyses of these changes revealed that the acquisition of B-lymphoid- and neutrophil/monocyte (NM)-restricted properties are accompanied by a concomitantly accelerated and lineage-shared cell cycling activity and loss of self-renewal potential. Single-cell transcriptome analysis identified reduced expression of multiple self-renewal-associated genes and an accompanying heterogeneous activation of lineage-regulatory modules during the production of B, NM, and dendritic cell precursors. By applying a novel culture system that supports early human lymphoid differentiation, we uncovered a shared mechanism of proliferation control, along with persistent biological and transcriptional heterogeneity in cells undergoing B- and NM-lineage restriction.

Graphical abstract

Figure 1.

New phenotypes identify cells at early stages of B and NM differentiation. (A) Experimental design used to generate phenotypically defined CD19⁺ (B lineage) and CD14/15⁺ (NM lineage) outputs in vitro from 1000 P-mix CB input cells in LEM cultures (LEM being composed of αMEM, FBS, and a StemSpan lymphoid expansion supplement; see “Methods” for additional details). (B) A representative flow cytometric profile of the surface marker expression of the cells present in the cultures described in panel A after 2 weeks (CD45RA [RA]; CLEC12A [C]). (C) Number of output cells per 1000 input P-mix cells analyzed weekly. The Lin^– phenotypes were gated within the CD45⁺14^–15^–10^–19^– subset. Each bar shows the mean ± standard error of the mean (SEM) of values pooled from 5 different experiments. “n.d.” represents signals below the limit of detection (10 cells). (D) Experimental design used to examine the lineage potentials of the 3 input phenotypes shown. (E) Fluorescence-activated cell sorting (FACS) gating used to identify 3 phenotypes within the CD34⁺ cells present in day 7 cultures of CD45⁺14^–15^–10^– cells generated from P-mix CB cells. (F) Percentages of output cell types (identified by their phenotypes) obtained in clonal cultures initiated with the 3 phenotypes shown in panel E. Each bar shows the mean ± SEM of values pooled from 4 different experiments with 150 to 350 total single cells tested per input subset. The statistical significance of differences in clone distribution between input phenotypes was assessed using t tests with Holm multitest correction (∗P < .05). (G) Reverse transcription quantitative polymerase chain reaction (PCR) analysis of mRNAs of various historically defined B-lineage (upper)– and NM-lineage (lower)–associated genes measured in sorted phenotypes generated from 2-week cultures initiated with P-mix CB cells. Each bar shows the mean ± SEM of values pooled from 3 to 4 different experiments. “n.d.” denotes signals below the limit of detection (PCR cycle, >40). Statistical differences in mRNA levels between the progenitor and all differentiated phenotypes were assessed using t tests with Holm correction (∗P < .05). GAPDH, glyceraldehyde-3-phosphate dehydrogenase; mRNA, messenger RNA.

Figure 2.

Preservation of in vitro determined early B and NM precursor phenotypes in CB-derived xenografts. (A) FACS gating used to define RA^–C^–, RA⁺C^– and C⁺ phenotypes in unmanipulated CD34⁺38med71^–10^– CB cells. (B) Clonal output frequencies of the 3 input phenotypes isolated from unmanipulated CD34⁺ CB cells. Each bar shows the mean ± SEM of values pooled from 4 different experiments with 150 to 340 single cells tested per input subset. The statistical significance of differences in clone distribution between input phenotypes was assessed using t tests with Holm correction (∗P < .05). (C) Experimental design used to analyze clonal outputs of RA^–C^–, RA⁺C^–, and C⁺ phenotypes obtained from the bone marrow (BM) of NRG-W41 mice transplanted with CD34⁺ CB cells (4 experiments). (D) FACS gating used to detect the phenotypes generated in panel C. (E) Clonal output frequencies of 3 input phenotypes isolated from the BM of engrafted NRG-W41 mice. Each bar shows the mean ± SEM of values pooled from all experiments with 260 to 550 total single cells tested per input subset. The statistical significance of differences in clone distribution between input phenotypes was assessed using t tests with Holm correction (∗P < .05). (F) Size of clones generated from single input cells isolated from the different sources described above. Clones (defined as ≥5 human cells) pooled from all experiments shown above were grouped by the input phenotype (left) or lineage output type (right). “n.d.” represents signals below the limit of detection (5 cells). (G) Experimental design for the OP9-DLL4 coculture assay of clonal pro-T– and NM-lineage outputs in sorted single cells from each of the 3 input phenotypes. (H) Frequency of clonal output types detected in the OP9-DLL4 clonal assays. Values shown are mean ± SEM pooled from 3 experiments, with 180 to 240 cells tested per input phenotype. The statistical significance of differences in clone distribution between input phenotypes was assessed using Wilcoxon tests with Holm correction (∗P < .05).

Figure 3.

Cells differentiating toward either B or NM lineages undergo rapid and simultaneous acceleration of their proliferative activity. (A) Design of the cell division tracking strategy used to analyze the progeny of individual CD34⁺RA^–C^– cells. (B) Number of output cells per 1000 input CD34⁺RA^–C^– cells tracked over an 8-day period in the design shown in panel A (data points show the mean ± SEM values pooled from 5 experiments). (C) CFSE and surface marker expression profiles of cells obtained at different times of cultures initiated with CFSE-labeled CD34⁺RA^–C^– cells. The number of completed divisions was determined by the fold dilution of CFSE fluorescence intensity. (D) Numbers of the 3 output phenotypes detected in successive CFSE fluorescence peaks indicative of completion of different numbers of cell divisions (generated per 1000 CD34⁺RA^–C^– input cells; data points show the mean ± SEM values pooled from 5 experiments). (E) Average cell cycle transit times (in hours) preceding the appearance of different output phenotypes at different time points (data points show the mean ± SEM values pooled from 5 experiments). (F) Average cell cycle transit times (mean ± SEM) of the immediate progeny of different sources of CFSE-labeled CD34⁺RA^–C^–, CD34⁺RA⁺C^–, and CD34⁺C⁺ phenotypes assessed after a 4-day culture in LEM. Input cells tested were isolated directly from unmanipulated CD34⁺ CB cells (n = 3 experiments), xenografts (n = 4 experiments), and cultures (n = 7 experiments). ∗P < .05; ∗∗P < .01 (via pairwise t tests post Holm adjustment). (G) Assignment of G0, G1, and S/G2/M phases based on the intensity of Ki67 (log10-transformed pixel values) and DAPI (10⁵ pixel values) measured by immunofluorescence. Data shown are representative profiles of primary CB CD34⁺38^– cells (left) and pooled week 2 outputs of CB P-mix cells (right). (H) Proportions of week 2 output cells of CB P-mix at each cell cycle phase within each phenotypic population (bars showing the mean ± SEM values of 3 experiments). ∗P < .05, determined by t tests post Holm adjustment. DAPI, 4′,6-diamidino-2-phenylindole.

Figure 4.

Clonally tracked B + NM-progenitor outputs reveal initial slow divisions are associated with delayed differentiation. (A) Experimental design used to track clonal changes in the surface expression of CD34, CD45RA, and CLEC12A. (B) Brightfield and immunofluorescence images of representative clones examined on day 12. Scale bar, 100 μm. (C) Percentages of clonal output types of early B and NM lineage cells (see detailed phenotypes defined in supplemental Table 6). Each bar shows the mean ± SEM of values pooled from 3 experiments with 944 total single cells examined. (D) Percentages of clones that maintained an exclusive CD34⁺RA^–C^– output phenotype (black line) and RA⁺C^–/C⁺ bilineage outputs (gray line) out of all bilineage clones that eventually produced RA⁺C^– and C⁺ outputs. Each point shows the mean ± SEM of values pooled from 3 experiments. The dotted line represents the time when the loss of CD34⁺RA⁺C^– clonal output reaches 50%. (E) Cumulative growth kinetics (locally estimated scatterplot smoothing [LOESS]-smoothed) of clones initiated from CD34⁺RA^–C^– cells (n = 201) according to the time of the first appearance of RA⁺C^– or C⁺ cells in them. (F) Growth curves of the geometric means of the clone sizes analyzed in panel E (points) and generated using a logistic model. (G) Growth parameters fitted to individual clones analyzed in panels D-E. Each data point represents a clone (center line, median; box limits, first and third quartiles; whiskers, 1.5× interquartile range). (H-I) Growth kinetics and fitted logistic models of clones with NM-restricted outputs (n = 350 clones pooled from 3 experiments). (J-K) Growth kinetics and fitted logistic models of clones with B-restricted outputs (n = 60 clones pooled from 3 experiments). (L) Cumulative size distribution of blast cell colonies on days 4, 8, 12, and 16 after plating from the ensemble of colonies that show the first lineage marker expression at days 2, 4, 6, and 8, respectively. Points show experimental data, and lines show a lognormal dependence: $\text{erfc}\left(\left(\ln n-\mu \right)/\sqrt{2}\sigma \right)/2$, in which $\mu =⟨\ln n⟩$ denotes the measured average logarithm of colony size and ${\sigma }^2=⟨{\left(\ln n-\mu \right)}^2⟩$ represents the corresponding variance (for details, see supplemental Mathematical Model). (M) Schematic showing the model of blast cell colony expansion. In this model, blast cells have a defined proliferative capacity at the time of plating. Blast cells expand at a rate ${\lambda }_B$ over a time $\tau$, where $\tau$ varies between blast colonies. At time $\tau$, blast cells transition near synchronously into lineage-restricted progenitor states, which have a strictly limited proliferative capacity that varies both within and between lineages. The colony growth characteristics and size are consistent with a model in which progenitor cells expand over a fixed period $T$, giving rise to subclones of size $e^N$, with $N$ drawn at random from a Gaussian distribution with fixed mean and variance, with the capacity of B-cell progenitors an order of magnitude smaller than NM progenitors (for details of the model and its fit to the data, see supplemental Mathematical Model).

Figure 5.

Evidence of a delayed B/NM differentiation process in CD34⁺RA^–C^– cells exhibiting a slower cycling behavior. (A) Experimental design used to compare the changing output phenotypes generated from slow- and fast-cycling CD34⁺RA^–C^– cells. (B) Output numbers (mean ± SEM) of different progeny phenotypes generated from slow- and fast-cycling input CD34⁺RA^–C^– cells, determined at different time points (n = 3 experiments). Slow- and fast-cycling input cells were defined by their average cell cycle transit time of ≥48 or ≤24 hours, respectively. ∗P < .05 via pairwise t tests post Holm adjustment. (C) Experimental design used to examine and compare the clonal outputs of individual slow- and fast-cycling CD34⁺RA^–C^– cells. (D) Percentages of clonal output types produced from CD34⁺RA^–C^– input cells. Slow- and fast-cycling input cells were defined by their average cell cycle transit time of ≥48 or ≤24 hours, respectively. Each bar shows the mean ± SEM of values pooled from 4 different experiments in which a total of 540 and 520 single cells were assessed from the slow- and fast-cycling progenitors, respectively. Clone size (E) and the number of CD34⁺ output per clone (F) generated from individual slow- or fast-cycling progenitors. Clones (defined as ≥5 human cells) pooled from all experiments shown above were grouped by the lineage output type. “n.d.” represents signals below the limit of detection (5 cells). ∗P < .05 via pairwise t tests.

Figure 6.

Timed changes in expression of self-renewal and lineage-associated genes across differentiation trajectories. (A) Experimental design used to track the sequential transcript outputs of in vitro–stimulated P-mix CB cells by single-cell CITE-seq analysis. For details, see supplemental Methods. (B) Hierarchical clustering of cell cycle states based on scaled gene expression values. Columns represent 16 269 individual cells colored by time point and cell cycle state. Rows represent genes expressed in specific cell cycle phases. (C) The percentages of cells in different cell cycle phases at each time point assessed. (D) UMAP presentation of transcriptome data (after cell cycle regression) combined from 12 181 single cells from the day 7, 10, and 13 time points. Cells are divided into A and B clusters based on the first UMAP dimension. (E-F) UMAP projection of 10 385 cells in the A cluster. Cells are colored by surface phenotype inferred from antibody-derived oligonucleotide tag signals (E) or assigned cell cycle phase (F). (G) Percentage of different cell cycle phases in cluster A cells separated by different phenotypic subsets. (H) Unsupervised modeling of lineage trajectories embedded in the UMAP manifold. (I) Distribution of TF activities in the UMAP space. Expression of each TF module is indicated by colored dots. Gray lines represent a density level of 0.006 of all cells in cluster A. (J) Top 30 most active TF motifs in each lineage trajectory. Columns represent individual cells ordered by pseudotime. Rows represent TFs that are detected in >20 target genes. Colors on the heat map denote presence (colored by lineage) or absence (blank) of the TF.

Figure 7.

Functional association between cell proliferation and lineage restriction. (A) Three-dimensional FLE plot of 10 385 cells from the A cluster in Figure 6D. Cells are color coded by their respective cell cycle phases, with transitions between phases indicated by dashed red loops. (B-C) Three-dimensional FLE plot, with cells color coded based on their surface phenotypes (B) and gene expression levels (C) (the scale represents log2-transformed values). (D-F) Three-dimensional FLE plot (from a distinct angle), with cells color coded by their cell cycle phases (D), phenotypes (E), and gene expression (F), respectively. (G) Experimental design for assessing cell cycle progression and differentiation outputs of CD34⁺RA^–C^– progenitors treated with 100 nM wortmannin (Sigma-Aldrich, catalog no. 681675). The inhibitor is replenished every 2 days to account for its short half-life. (H) Number of output cells per 100 input cells analyzed on day 6. Each bar represents the mean ± SEM, calculated from 6 experimental replicates derived from 2 independent CB pools. (I-J) Numbers of CD34⁺RA^–C^– (I) and CD34^– (J) output cells detected across successive CFSE fluorescence peaks, reflecting the number of completed cell divisions. Each point represents the mean ± SEM from the same experiments as shown in panel H. ∗P < .05, determined by t tests. (K) A proposed model of the process of cell fate restriction into lymphoid and myeloid lineages from bipotent human progenitors. Ctr, control; DMSO, dimethyl sulfoxide; FLE, force-directed layout.