Hematopoietic Stem Cells Count and Remember Self-Renewal Divisions

Abstract

The ability of cells to count and remember their divisions could underlie many alterations that occur during development, aging, and disease. We tracked the cumulative divisional history of slow-cycling hematopoietic stem cells (HSCs) throughout adult life. This revealed a fraction of rarely dividing HSCs that contained all the long-term HSC (LT-HSC) activity within the aging HSC compartment. During adult life, this population asynchronously completes four traceable symmetric self-renewal divisions to expand its size before entering a state of dormancy. We show that the mechanism of expansion involves progressively lengthening periods between cell divisions, with long-term regenerative potential lost upon a fifth division. Our data also show that age-related phenotypic changes within the HSC compartment are divisional history dependent. These results suggest that HSCs accumulate discrete memory stages over their divisional history and provide evidence for the role of cellular memory in HSC aging.

Keywords: aging; cell division counting; cellular memory; dormancy; hematopoietic stem cells.

Figure 1. LR-HSCs Persist in BM Throughout Life and Contain All LT-HSC Activity in Aging BM

(A) Schematic of long-term dox treatments. 2–4 month old 34/H2BGFP mice were placed on dox for periods ranging from 3–22 months. At the end of dox chase BM was analyzed for the presence of LR-HSCs. (B) Histogram of LSKCD48⁻Flk2⁻CD150⁺ HSC compartment before and after 12 month dox chase. LR-HSCs were determined by gating above the background GFP levels of single transgenic TetO-H2BGFP HSCs. (C) Time course of label dilution after initiation of dox chase; n=2–15 mice per time point. (D) Percent of HSCs that are label-retaining after 10–22 months of dox chase (sLR-HSCs). n=42 mice from 12 independent experiments. (E–L) HSC populations were sorted from 19 month-old mice chased with dox for 15 months into Total, GFP^Hi, and GFP^Lo HSC populations. 200 cells from each population were competitively transplanted per mouse. (E) Gating strategy for Total, GFP^Hi, and GFP^Lo HSC fractions. (F–G) Blood chimerism of granulocytes (F), and total white blood cells (G) during primary and secondary transplants. (H–J) Analysis of donor-derived stem and progenitor cell compartments in recipient BM. Gating strategy (H) and quantification of donor-derived HSPCs in primary (I) and secondary (J) transplantations after 22 and 24 weeks respectively. (K–L) Lineage distribution of donor-derived peripheral blood in primary (K) and secondary (L) hosts at 22 and 24 weeks respectively. Transplant data are represented as mean ± SEM of 8–14 mice per group from two independent experiments. *P<0.05, **P<0.01, ***P<0.001 by Welch’s t test. See also Figure S1 and Figure S2.

Figure 2. CD41 Expression on Young, Aging, and LR-HSCs

FACS analysis and quantification of the primitive HSPC compartment from young and aging mice. (A) CD150⁺ cells marking the HSC compartment in young and aging BM. (B) Quantification of (A). (C) The same populations in (A) displayed as a function of CD150 and CD41 expression. (D) Quantification of (C). (E) CD41 expression on Total, GFP^Lo, and GFP^Hi HSCs. (F) Quantification of CD41 expression on Total HSCs. (G) Ratio of CD41⁻ to CD41⁺ HSCs found in Total, GFP^Lo, and GFP^Hi HSCs. Data are represented as mean ± SEM of 9–11 mice per group from 3 independent experiments. **P<0.01, ***P<0.001 by Welch’s t test. See also Figure S3 and Figure S4.

Figure 3. Clonal Analysis of the Aging HSC Compartment Based on CD41 Expression and Label-Retention

The HSC compartment was sorted from 19-month old mice chased with dox for 17 months into four populations based on CD41 expression and label retention, and transplanted at a dose of 15 cells per mouse. (A) Reconstitution curves of donor-derived (CD45.2⁺) total white blood cells, myeloid, B cells, and T cells for each transplanted mouse through 24 weeks in primary and secondary recipients. The transition from primary to secondary transplantation is marked by the x-axis break. The horizontal line marks the threshold of successful reconstitution. (B) Examples of the five reconstitution patterns observed. Definition of repopulation patterns: myeloid-restricted only repopulated myeloid cells; bipotent progenitors gave rise to myeloid and B cells; ST-HSCs showed transient repopulation of all three lineages, with donor chimerism of at least one lineage dropping below threshold by 24 weeks after primary transplantation; IT-HSCs repopulated all three lineages, but had at least one lineage drop below threshold by 24 weeks after secondary transplantation; and LT-HSCs maintained repopulation in all three lineages above threshold throughout both primary and secondary transplantation (C) Distribution of repopulating cell types found within each aging HSC compartment. The zoomed region represents 10% of the Total HSC compartment. (D) Heat maps displaying the regeneration of primitive BM populations by clonally transplanted aging HSCs after primary transplant. Transplanted cell populations are listed above each column—initially sorted cell (left panel) and retrospectively categorized repopulating cell (right panel). Regenerated HSPC types are listed to the right. The darker the chamber, the greater the proportion of reconstituted mice regenerated the given cell type. Numbers within each chamber represent the percentage in decimal format of reconstituted mice with each cell type. See also Figure S5.

Figure 4. HSCs Count Symmetric Self-Renewal Divisions Throughout Adult Life and Progress Towards Dormancy

(A–C) Analysis of H2BGFP subpopulations within the GFP^Hi LR-HSC compartment. (A) Histogram displaying the H2BGFP peaks 0–4 visible within the LR-HSC compartment of young (3–4 months on dox) and aging mice (14–22 months on dox). (B) Quantification of (A). n=21 and 13 mice from 6 independent experiments for young and aging mice, respectively. (C) Least squares fitting of single cell GFP intensity data collected from LR-HSCs found within each GFP peak of young mice. Observed experimental data are plotted as open circles, while predictions of a theoretical model in which H2BGFP concentration is reduced by a factor of 2 with each cell division is given by the dashed blue line. Experimental data were collected from 1568 single LR-HSCs from 6 independent experiments. (D) Percentage of LR-HSCs within the HSC compartment. (E) Absolute number of LR-HSCs per long bone in young and aging mice. Predicted LR-HSC numbers were generated by extrapolating the expansion of each young LR-HSC data point based on the distribution of cells found in peaks 0–4 for each mouse using the model in panel (F), then corrected based on the average distribution of cells found in aging mice in panel (B). Data are mean ± SEM of 17 and 10 mice from 5 and 3 independent experiments for young and aging mice respectively. (F) Symmetric self-renewal expansion model of LR-HSCs. As LR-HSCs slowly divide throughout adult life they transition from peak 0 to peak 4, symmetrically self-renewing to double their numbers with each cell division. Arrows in the histogram depict the expansion capacity of cells as they progressively divide to reach peak 4. Numbers displayed are the average numbers of LR-HSCs in each peak per long bone from 17 young mice in 5 independent experiments. Boxed in red is the summation of LR-HSCs predicted to accumulate in peak 4 with aging. (G) Mathematical modeling of cell cycle progression as a function of divisional history within the LR-HSC compartment. Five models were considered (see Methods for details). Displayed are representations of cell cycle time progressions for each model (red dashed lines), as well as the experimentally determined (open circles) and model predicted sLR-HSC numbers (dashed blue curves) found in each GFP peak of aging mice. Cell cycle times for the step function and super-exponential models are actual times predicted by the model. As the constant, linear, and exponential models do not fit the data well, their corresponding cell cycle times are only visual representations. (H–I) Distribution of LR-HSCs across each GFP peak (H), and quantification of LR-HSC absolute numbers after various lengths of dox chase (I). Legends refer to the length of dox chase. Data are representations of 2–6 independent experiments per group. (J) Cell cycle analysis of GFP Peak cells in young (5 months old, 3 month dox chase, n=3) and aging (11 months old, 9 month dox chase, n=2) mice. Each mouse represents an independent experiment. See also Figure S6.

Figure 5. Young CD41⁻ LR-HSCs Generate CD41⁺ sLR-HSCs in Aging Mice

(A) Distributions of CD41⁻ (left) and CD41⁺ (right) LR-HSCs in young and aging mice across GFP peaks 0–4. (B) Representative histograms directly comparing GFP levels in CD41⁻ and CD41⁺ LR-HSCs from young (left) and aging (right) mice. (C) GFP mean fluorescence intensity (MFI) of CD41⁻ and CD41⁺ LR-HSCs in young and aging mice. n=9–11 mice per group from 3 independent experiments. (D) Quantification of CD41⁻ and CD41⁺ LR-HSCs in young and aging mice. Predicted expansion capacity of young HSCs as predicted by the model in Figure 4F. n=6–7 mice from 2 independent experiments. (E) Mathematical modeling of CD41⁻ LR-HSC contribution to the CD41⁺ LR-HSC compartment. Predictions of a model in which the cell cycle time extends super-exponentially with the number of cell divisions (see Figure 4G) and in which CD41⁻ LR-HSCs gain CD41 expression with probability 1 − α each time they divide are given in blue (full model details in Methods). This model most accurately fits the data when α = 0.88, suggesting that approximately 1 in every 10 CD41⁻ LR-HSC divisions gives rise to a CD41⁺ daughter cell. Data are displayed as mean ± SEM *P<0.05, **P<0.01, ***P<0.001 by Welch’s t test or paired t test.

Figure 6. Self-Renewal Counting Model of Hematopoietic Stem Cell Aging

The HSC pool can be segregated into two populations based on label-retention. The LR-HSC pool contains all of the transplantable LT-HSC activity, while the non-LR-HSC pool is comprised of cells with minimal self-renewal and restricted regenerative capacity. With aging, both the LR- and non-LR-HSC pools expand. The LR-HSC pool asynchronously undergoes four traceable symmetric self-renewal events, increasing the functional stem cell pool size over time while simultaneously diluting the GFP label with each cell division. After a fourth traceable self-renewal event, LR-HSCs enter a state of dormancy—as a fifth cell division would result in complete loss of LT-HSC potential—indicating that the LR-HSC population counts their cell divisions throughout life. The fact that the LR-HSC pool exclusively undergoes symmetric cell divisions before entering a stably dormant state means that they contribute minimally, if at all, to homeostatic hematopoiesis, unless activated to divide again by stress. The non-LR-HSC pool represents the vast majority of the stem cell pool in aging mice. Within the non-LR-HSC pool CD41⁺ cells enriched for myeloid progenitor activity accumulate with time and dominate the aging HSC compartment. It is most likely that the non-LR-HSC pool maintains active hematopoiesis during steady state conditions. In the context of regeneration, we identified 5 types of stem and progenitor cells with regenerative potential after transplantation within the total HSC compartment. When analyzed as a total HSC population, the predominance of myeloid progenitors and cells with limited self-renewal potential contributes to increased myeloid representation in regenerated peripheral blood and reduced long-term engraftment.