Aging alters the epigenetic asymmetry of HSC division

Abstract

Hematopoietic stem cells (HSCs) balance self-renewal and differentiation to maintain homeostasis. With aging, the frequency of polar HSCs decreases. Cell polarity in HSCs is controlled by the activity of the small RhoGTPase cell division control protein 42 (Cdc42). Here we demonstrate-using a comprehensive set of paired daughter cell analyses that include single-cell 3D confocal imaging, single-cell transplants, single-cell RNA-seq, and single-cell transposase-accessible chromatin sequencing (ATAC-seq)-that the outcome of HSC divisions is strongly linked to the polarity status before mitosis, which is in turn determined by the level of the activity Cdc42 in stem cells. Aged apolar HSCs undergo preferentially self-renewing symmetric divisions, resulting in daughter stem cells with reduced regenerative capacity and lymphoid potential, while young polar HSCs undergo preferentially asymmetric divisions. Mathematical modeling in combination with experimental data implies a mechanistic role of the asymmetric sorting of Cdc42 in determining the potential of daughter cells via epigenetic mechanisms. Therefore, molecules that control HSC polarity might serve as modulators of the mode of stem cell division regulating the potential of daughter cells.

Fig 1. Young and aged CASIN-treated HSCs allocate Cdc42 asymmetrically to daughter cells, while aged and young Wnt5a-treated HSCs allocate it symmetrically.

(A) Schematic representation of the gating strategy and of the experimental setup (graphical sources: https://www.servier.de/medical-art and https://openclipart.org/). (B) Percentage of HSCs that divided over time (1st: first divisions; 2nd: second divisions). Percentage of dead cells is also plotted (indicated with †). (C) Representative confocal single z-stack (central xy plane; i–iii), 2D xy plane projection of all z-stacks (iv), 3D reconstruction of Cdc42 signal overlapped on the 2D projection of all z-stacks, (v) and rotation of the 3D Cdc42 reconstructed signal (vi–viii) of dividing (telophase) young, aged, aged treated with CASIN 5 μM, and young treated with Wnt5a 100 ng/mL HSCs. Panels show DAPI (nucleus, blue), Cdc42 (magenta), and tubulin (green). (D) Percentage of the ratio of Cdc42 (quantified as volume of the fluorescence signal) amount in the daughter cells (higher amount over the lower amount). Each dot represents a pair. Only telophases and late anaphases were considered for the analysis. Cartoon depicting the scoring for symmetric and asymmetric divisions according to the amount of inherited protein as analyzed by 3D confocal microscopy. *p < 0.05, **p < 0.01, ***p < 0.001; n = 2–3 biological repeats; 25 pairs for young, 26 pairs for aged, 26 pairs for aged + CASIN, and 14 pairs for young + Wnt5a. The primary data the figure is based on are provided in S1 Data. CASIN, Cdc42 activity specific inhibitor; Cdc42, cell division control protein 42; Ctrl, control; HSC, hematopoietic stem cell.

Fig 2. Young and aged CASIN-treated HSCs allocate H4K16ac asymmetrically to daughter cells, while aged and young Wnt5a-treated HSCs allocate it symmetrically.

(A) Representative confocal single z-stack (central xy plane [i–iii]), 2D xy plane projection of all z-stacks (iv), 3D reconstruction of H4K16ac signal overlapped on the 2D projection of all z-stacks (v) and rotation of the 3D H4K16ac reconstructed signal (vi–viii) of dividing (telophase) young, aged, aged treated with CASIN 5 μM, and young treated with Wnt5a 100 ng/mL HSCs. Panels show DAPI (nucleus, blue), H4K16ac (magenta), and tubulin (green). (B), Percentage of the ratio of H4K16ac (quantified as volume of the fluorescence signal) in the daughter cells (higher amount over the lower). Each dot represents a pair. Only telophases and late anaphases were considered for the analysis. *p < 0.05, **p < 0.01, ***p < 0.001; n = 2–3 biological repeats; 41 pairs for young, 37 pairs for aged, 40 pairs for aged + CASIN, and 26 pairs for young + Wnt5a. (C–F) Pie charts depicting the frequency of asymmetric/symmetric division in each sample based on the amount of Cdc42 in daughter cells. Any division with a ratio (see cartoon scheme in Fig 1D) equal to or below 75% was considered asymmetric. The primary data the figure is based on are provided in S1 Data. CASIN, Cdc42 activity specific inhibitor; Cdc42, cell division control protein 42; Ctr, control; HSC, hematopoietic stem cell.

Fig 3. Polarity is a major driver of asymmetric cell division, and Cdc42 allocation predicts potential.

(A) Pie charts in white/grey show the observed ratios of polar versus apolar HSCs in young, aged, Wnt5a-treated young, and CASIN-treated aged HSC samples, respectively [15, 16]. Pie charts in green show the observed modes of division for the respective cell populations (see Fig 2C–2E). The intermediate panel depicts the possible connections (arrows) between cell polarity status and mode of division. Application of a maximum likelihood approach on a simple transition model (with d_a being the probability for a polar HSC to undergo an asymmetric mode of division and d_s being the probability for an apolar HSC to undergo a symmetric mode of division) allows estimating the most probable configuration for d_a and d_s, thereby supporting the notion that the solid arrows represent the most prominent choices. (B) The bifurcation diagram illustrates the differentiation switch. For certain parameter values of k_c, the ODE for total Cdc42 concentration c_tot(t) allows for 2 stable steady states: (I) total Cdc42 is highly expressed (associated with the HSC state) and (II) total Cdc42 is lowly expressed (associated with the progenitor state). (C, D) The pseudo-potential (“stemness potential”) landscape for total Cdc42 concentration with two stable steady states (two “valleys,” representing HSCs and progenitors). For HSCs undergoing asymmetric cell division (blue line in panel C), one daughter receives lower Cdc42 concentration (“passes the hill”) and progresses into the progenitor state (dashed arrow), while the other daughter retains the stem cell state. Daughters of HSCs undergoing symmetric division (blue line in panel D) retain their configuration. (E) The panels summarise the mechanistic model results (red boxes) for populations of young and aged HSCs, respectively. The pie diagrams on the right-hand side show the predicted ratios of resulting cell types (HSCs versus progenitors) derived from a population of dividing young and aged HSCs, respectively. The primary data the figure is based on are provided in S1 Data. CASIN, Cdc42 activity specific inhibitor; Cdc42, cell division control protein 42; HSC, hematopoietic stem cell; ODE, ordinary differential equation.

Fig 4. Restoring the asymmetry of aged HSC divisions rejuvenates the function of daughter stem cells.

(A) Schematic representation of the experimental setup (graphical sources: https://www.servier.de/medical-art and https://openclipart.org/). (B) Representative asymmetric and symmetric division based on the contribution of the single daughter cell to PB. Each single daughter cell was injected into 1 recipient mouse and donor-derived Ly5.1⁺ cells (engraft) B220⁺, CD3⁺, and myeloid (Gr1⁺, Mac1⁺, and Gr1⁺Mac1⁺) cells among Ly5.1⁺ donor-derived cells were measured at 4, 12, 16, and 24 weeks after transplant. (C) Pie charts depicting the percentage of asymmetric and symmetric divisions of young, aged, aged treated with CASIN 5 μM and young treated with Wnt5a 100 ng/mL HSCs based on the profile of the reconstitution in PB of single daughter cell pairs. n = 12 pairs for young HSCs, n = 13 pairs for aged HSCs, n = 8 pairs for aged CASIN-treated HSCs, and n = 8 pairs for young Wnt5a-treated HSCs. (D–E) Percentage of donor-derived cells at 4 weeks and 24 weeks after transplant in PB of mice transplanted with single (retrospectively identified) daughter stem cells. *p < 0.05. The primary data the figure is based on are provided in S1 Data. CASIN, Cdc42 activity specific inhibitor; Cdc42, cell division control protein 42; Ctr, control; HSC, hematopoietic stem cell; PB, peripheral blood.

Fig 5. Kinetics of engraftment and lineage contribution of single daughter stem cells.

(A) Chimerism kinetics of overall engrafted donor-derived cells and of each donor-derived lineage (B cells, T cells, and myeloid cells). Shown are young, aged, aged treated with CASIN, and young treated with Wnt5a daughter cells that were retrospectively identified as daughter stem cells. The pie charts represent the relative contribution to B cells, T cells, and myeloid cells in PB at 24 weeks after transplant. (B) Percentages of donor-derived B cells, T cells, and myeloid cells in PB of recipient mice 24 weeks after transplant. n = 9–22. Shown are young, aged, aged treated with CASIN, and young treated with Wnt5a daughter cells that were retrospectively identified as daughter stem cells. *p < 0.05. The primary data the figure is based on are provided in S1 Data. CASIN, Cdc42 activity specific inhibitor; Cdc42, cell division control protein 42; Ctr, control; PB, peripheral blood.

Fig 6. HSCs are found in clusters in bone marrow in aged animals.

(A, B) Representative 3D confocal pictures of a whole-mount femur from a young (panel A) and an aged (panel B) mouse. HSCs are indicated by yellow arrowheads. Dashed lines in panel B highlight clustered HSCs. Lineage markers CD41 and CD48 are shown in white; CD150 in red and endothelial cells (IV injected VE-Cadherin CD144 and PECAM-1 CD31) are in magenta. (C) Distance between HSCs in young and aged femurs. (D) Percentage of HSCs located in 20 μm intervals from the closest HSC in young and aged femurs. (E) Pie charts depicting the percentage of HSCs found in clusters (2 cells or more adjacent to each other) or as single cell in young and aged mice. Imaging data refer to 27 young and 14 aged longitudinal shaved whole-mount cross-sections from 2 young and 2 aged mice; 144 young HSCs and 394 aged HSCs. (F, G) Representative 3D confocal pictures of whole-mount femur analyses from young recipient mice transplanted with young (panel A) and aged (panel B) RFP+ HSCs. Donor-derived RFP+ HSCs are marked by yellow arrowheads. Lineage markers CD41 and CD48 are shown in white, CD150 in green and endothelial cells (IV injected VE-Cadherin CD144 and PECAM-1 CD31) are in magenta. All donor-derived cells are RFP+ (red). (H) Distance between donor-derived RFP+ young and aged HSCs in the femurs of young recipient mice. (I) Percentage of donor-derived RFP+ young and aged HSCs located in 20 μm intervals from the closest donor-derived RFP+ HSC in the femurs of recipient mice. (J) Pie charts depicting the percentage of donor-derived RFP+ young and aged HSCs found in clusters (2 cells or more adjacent to each other) or as single cell in the femurs of young recipient mice. Imaging data refer to 8 young and 4 aged longitudinal shaved whole-mount cross-sections from 2 young and 2 aged mice; 75 young HSCs and 108 aged HSCs. The primary data the figure is based on are provided in S1 Data. HSC, hematopoietic stem cell; IV, intravenous; Lin, lineage; RFP, red fluorescent protein.

Fig 7. The transcriptome does not reflect the potential of daughter cells.

(A) Schematic representation of the experimental setup for the preparation of scRNA-seq libraries of daughters pairs from young, aged, aged treated with CASIN, and young treated with Wnt5a HSCs (graphical sources: https://www.servier.de/medical-art and https://openclipart.org/) (B) Correspondence based on BGA of the global expression profile of single daughter cell pairs after division. In total, 25–28 pairs for young, 31–39 pairs for aged, 19–23 pairs for aged CASIN-treated, and 18–18 pairs for young Wnt5a treated HSCs; n = 3–5 biological repeats. (C) Representative radar plots of the GSEA showing 1 similar and 1 dissimilar daughter pair. The analysis approach simultaneously interrogated 13 previously published HSC and polarity signatures, and each daughter cell pair was depicted in a radar plot where vertices in the plot correspond to each of the considered gene set. Pairs were then compared for significance of similarity/dissimilarity by a goodness of fit test across all significant gene sets. (D) Pie charts depicting the percentage of asymmetric and symmetric divisions of young, aged, aged treated with CASIN, and young treated with Wnt5a daughter cell pairs based on the GSEA. (E) Representative 3D-SOM analysis of the whole scRNA-seq transcriptome of 1 concordant and 1 discordant daughter pair. (F) Pie charts depicting the percentage of asymmetric and symmetric divisions of young, aged, aged treated with CASIN, and young treated with Wnt5a daughter cell pairs based on the 3D-SOM metagene analysis. To quantify the degree of similarity between SOMs of daughter pairs, we identified genes that showed statistically significant log fold change compared to the mean global expression profile (namely metagenes, corresponding to the hills and valleys in the 3D plots shown in panel E). This information was then used to compare daughter pairs and identify concordant genes, i.e., genes that show same direction of change. Eventually, we compared the ratio of concordant genes to the discordant ones and calculated a QCR. Each pair was then identified as concordant or discordant based on the magnitude and direction of the QCR value (tested for significance using Monte-Carlo simulation). (G) Venn diagram showing differential gene expression analysis for up-regulated genes. (H) Pie charts depicting the percentage of asymmetric and symmetric divisions of young, aged, aged treated with CASIN, and young treated with Wnt5a daughter cell pairs based on the 134 genes up-regulated in the daughter cells from young and aged CASIN-treated arms but not in the aged HSCs and young Wnt5a-treated ones. (I) Bar graph showing fold enrichment for biological processes in the 134 up-regulated genes in young and aged CASIN-treated daughter pairs but not in the aged HSCs and young Wnt5a-treated ones, and in the 6 genes down-regulated in young and aged CASIN-treated daughter pairs but not in the aged HSCs and young Wnt5a-treated ones. GO analyses for biological processes were done with Panther version 11.0. The primary data the figure is based on are provided in GEO GSE116712. BGA, between-group analysis; CASIN, cdc42 activity specific inhibitor; cdc42, cell division control protein 42; GO, gene ontology; GSEA, gene set enrichment analysis; HSC, hematopoietic stem cell; QCR, Quadrant Count Ratio; scRNA-seq, single-cell RNA sequencing; SOM, self-organizing maps.

Fig 8. scATAC-seq on daughter cells shows asymmetric division for young HSCs, while aged HSCs undergo mainly symmetric divisions.

(A) Schematic representation of the experimental setup for the preparation of scATAC-seq libraries from young and aged daughter cell pairs. Graphical sources: https://www.servier.de/medical-art and https://openclipart.org/. (B) Ratio of number of peaks of daughter pairs in young (green) and aged (blue) cells. For each pair, the ratios have been adjusted with the ratio of mapped reads in the pair (unadjusted ratios are shown in S11C Fig). A ratio of one represents an equal number of peaks in a given pair. The right panel represents the distribution of the ratio of peak counts. (C) Pie charts depicting the percentage of asymmetric and symmetric divisions of young and aged HSCs based on scATAC-seq analysis of daughter cells (see also S8 Table for statistics). For those pairs for which we have sufficient number of peaks for the reshuffling simulation test of the null hypothesis that cells in a pair show same peak counts in same genomic positions (column V in S8 Table), the asymmetry assignment is based on the significance of no overlap and on the ratio of peak counts (as a measure of global chromatin accessibility) greater than 2 (column P in S8 Table). (D) Radar plot based on 13 published “stemness” signatures. Signatures that show significant enrichment based on a hypergeometric test (adjusted p < 0.05) are shown by red and blue shades representing the 2 cells in a given pair. Signatures that do not extend beyond the shaded gray center are considered not significant (see S14 Fig. for the reference to each specific signature). (E) “Stemness” enrichment analysis-based pie charts showing the frequency of symmetric and asymmetric division of young and aged HSCs. A goodness-of-fit chi-squared test assessing the matches in significance between cells in a pair was used to assign cells either as symmetric or asymmetric. (F) Example of a Venn diagram depicting an asymmetric (left panel; significant difference in peak counts) and a symmetric (right panel; comparable peak counts) daughter cell pair from young HSCs based on the difference in peak counts (ratios) as a measure of the degree of chromatin accessibility. For symmetrically dividing HSCs, both daughter cells were considered daughter stem cells (supported by our sc-transplant data). For asymmetrically dividing HSCs, the daughter presenting with the lower amount of peak counts (lower degree of chromatin accessibility) was considered the daughter stem cell, while the other daughter cell was assigned to the daughter progenitor (i.e., not stem cell) group. (G, H) Daughter progenitor cells were subjected to Reactome GO analysis. The most frequently significantly enriched GOs among young (panel G) and aged (panel H) daughter progenitor cells are shown. (I, J) Daughter stem cells were subjected to Reactome GO analysis. The most frequently significantly enriched GOs among young (panel G) and aged (panel H) daughter stem cells are shown. The primary data the figure is based on are provided in GEO GSE116712. A, asymmetric; GO, gene ontology; HSC, hematopoietic stem cell; S, symmetric; scATAC-seq, single-cell transposase-accessible chromatin sequencing.