SEARCH

SEARCH BY CITATION

Keywords:

  • aging hematopoiesis;
  • stem cell exhaustion;
  • functional heterogeneity;
  • serial transplantations;
  • mathematical modeling;
  • novel concepts of aging

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Author contributions
  8. References
  9. Supporting Information

Hematopoietic stem cells (HSCs) are the source for the life-long supply of functional cells in peripheral blood while they simultaneously maintain their own reserve pool. However, there is accumulating evidence that HSCs are themselves subject to quantitative and qualitative exhaustion. Although several processes linked to mitotic activity can potentially account for the observed aging phenomena (e.g., DNA damage, telomere shortening, epigenetic modification), a precise understanding of HSC exhaustion is still missing. It is particularly unclear how individual aging processes on the single-cell level translate on the phenotypic level of the overall tissue and whether there is a functional implication of an age-structured HSC population. We address these issues by applying a novel mathematical model of HSC organization in which division-specific, cumulative alterations of stem cell quality determine the phenotypic and functional appearance of the overall cell population. Adapting the model to a number of basic experimental findings, we quantify the level of additional heterogeneity that is introduced by a population of individually aging cells. Based on this model, we are able to conclude that division-dependent processes of cellular aging explain a wide range of phenomena on HSC exhaustion and that HSC aging needs to be considered as a highly heterogeneous process. We furthermore report that functional heterogeneity between young and old HSCs appears closely similar to the phenomena described for long- and short-term repopulating cells. We speculate whether differential, division-coupled stem cell aging introduces an intra-animal variability that also accounts for heterogeneity with respect to the repopulation ability of HSCs.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Author contributions
  8. References
  9. Supporting Information

Hematopoietic stem cells (HSCs) are the source for the life-long supply of functional cells in peripheral blood, while at the same time they maintain their own reserve pool. This maintenance process is generally referred to as stem cell self-renewal. Although the terminology implies that stem cells maintain (or even renew) their own population without loss of quality, there is also accumulating evidence that HSCs are themselves subject to quantitative and qualitative exhaustion over time. Such age-related alterations are assumed to be a causative factor for organismal aging and the occurrence of aging-related diseases.

As a critical experiment, it has been shown that serial transplantations of HSCs into lethally irradiated recipients can only be carried out for a restricted number of passages suggesting that the ability of stem cells to re-establish their own population has an upper limit (Ogden & Mickliem, 1976; Harrison et al., 1978; Ross et al., 1982; Zant et al., 1997; Kamminga et al., 2005). Furthermore, these experiments support the idea that qualitative exhaustion is linked to the artificial acceleration of HSC turnover, presumably induced by the cellular expansion after transplantation. The possible reasons for this aging effect, both on the molecular and on the cellular level, are manifold and controversially discussed. It seems to be a common observation that the engraftment ability of older HSCs is reduced compared to younger HSCs leading to a reduced overall repopulation potential of aged HSCs (Marley et al., 1999; Kamminga et al., 2005). Additionally, experimental data indicate, e.g., that the age-dependent development of HSC activity is highly strain specific. Whereas in some mouse strains the cycling activity of HSC has been reported to decrease with age (e.g., DBA mice), in others strains this effect is not detectable (e.g., B6 mice) (De Haan et al., 1997; De Haan & Zant, 1999).

These diverse phenomena observed on the tissue level necessarily have a correspondence on the level of individual cells. However, neither the aging-related alterations on the level of individual cells nor the translation from the cellular to the tissue level is completely understood. There are a number of processes linked to mitotic activity that can potentially account for the observed aging phenomena such as accumulating DNA damage (Hamilton et al., 2001; Lombard et al., 2005; von Zglinicki et al., 2005), progressive telomere shortening (Hayflick & Moorhead, 1961; Notaro et al., 1997; Matulic et al., 2007), or epigenetic modifications (Wilson & Jones, 1983; De Haan & Gerrits, 2007). However, none of these processes seems to be the sole reason for the complex aging processes in stem cell function and phenotype that manifest on the tissue level rather than on the level of individual cells.

Given the general idea that the accumulation of changes and damages within individual HSCs (i.e., cellular aging) leads to impaired self-renewal ability and final exhaustion of the functional tissue (i.e., tissue or organismal aging), we raise the question how this connection between changes in the individual cells and the tissue level is qualitatively and quantitatively characterized. In this context, it has been discussed whether the aging process acts equally on all HSCs leading to a continuous functional decrease in the overall population or whether the HSC population consists of certain subsets that are differentially susceptible to aging and thus change the composition of the overall tissue. It appears in numerous studies that HSCs posses an intrinsic level of heterogeneity with respect to functional criteria such as repopulation ability and lineage contribution as well as with respect to their phenotypic appearance (Sieburg et al., 2006; Wagers & Weissman, 2006; Dykstra et al., 2007; Roeder et al., 2008; Challen et al., 2010). Adding to this, the differential occurrence of aging-related processes, most likely linked to mitotic activity, imposes a further level of heterogeneity which is subject to the following analysis.

We have previously shown that the occasional and reversible activation of primarily quiescent HSCs into the cell cycle is sufficient to explain the balance between the sustained maintenance of the HSC pool and the continuous supply of differentiating cells (Roeder & Loeffler, 2002; Roeder et al., 2005; Glauche et al., 2009). In particular, this explanation is in very good agreement with available data on HSC turnover which became accessible using label dilution techniques (Kiel et al., 2007; Wilson et al., 2008). However, our interpretation implies that in an unperturbed situation, the activation of quiescent HSCs is best described by a stochastic process that can only be predicted in a statistical manner. This results in a wide distribution of individual turnover times for murine HSCs ranging from few days up to more than 200 days. Consequently, the population of HSCs at a certain time point consists of cells that underwent different numbers of activation cycles, i.e., cell divisions.

By taking these aspects into consideration, we apply an appropriate mathematical model including the aforementioned additional degree of heterogeneity. Using the model, we can track the divisional history of each cell including all prior divisions of its ancestry. We raise the hypothesis that the cellular aging process corresponds to a slight degradation of ‘stem cell quality’ that occurs as a consequence of mitotic activity and leads to an accumulating decline in repopulation ability. In that sense, the number of prior divisions of a cell’s ancestry (‘mitotic age’) not only characterizes the history but also the future fate of a HSC. Giving up the view that all HSCs are in principle identical and taking the additional aging process into account, it is the mitotic age of an individual stem cell that ultimately influences its future development (i.e., whether the HSC retains the stem cell functionality is further activated or differentiates). To put this in more simple words, the more often a HSC has divided the less it is able to maintain and regenerate the stem cell pool in the long run.

In the following, we employ a mathematical model of HSCs organization to study how such cell-specific, cumulative alterations determine the phenotypic appearance of the overall tissue. Making simple assumptions, we consider the case that with a probability pa, the daughter cells of each dividing stem cell randomly decrease their ‘capability’ to further act as a stem cell by a factor ka. Given this cell-specific ‘capability’, we are able to address the question how this cell-specific process introduces an additional level of age-dependent heterogeneity among the HSCs. Adapting the mathematical model to a number of basic experimental findings, we are also able to quantify the level of additional heterogeneity and to provide interpretations and predictions on its functional influence.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Author contributions
  8. References
  9. Supporting Information

Modeling cellular aging of HSCs

The modeling approach to cellular aging is embedded in a previously established mathematical model of hematopoietic stem cell organization (Roeder & Loeffler, 2002; Glauche et al., 2007). This agent-based model was successfully applied to explain a range of phenomena in hematopoiesis, including clonal competition (Roeder et al., 2005, 2008), individual cell fate decisions and lineage specification (Roeder & Glauche, 2006; Glauche et al., 2007), HSC turnover (Glauche et al., 2009), or leukemia development and treatment effects (Roeder et al., 2006; Horn et al., 2008).

For the model, we assume that HSCs reside in either of two signaling contexts, named A and Omega, and that they can reversibly change between them. Importantly, the signaling contexts impose different effects on the cellular development: whereas context A is inspired by the concept of a stem cell supporting niche and promotes cellular quiescence and regeneration, context Ω represents an escape of HSCs from the niche signals and promotes proliferation and differentiation. A cell’s tendency to switch from one context into the other is determined by the cell number in the target context (i.e., the ‘packing density’ given a certain niche-specific carrying capacity) and by a cell-specific affinity a to reside in context A. The affinity a is gradually lost in context Ω, but regained in A up to the maximum value amax. Therefore, the system is able to establish a dynamically stabilized equilibrium, balancing quiescent cells in A and proliferating cells in Ω.

If the cell-specific affinity a drops below a certain threshold amin, the cell lost the ability to changing back to context A and is committed to undergo further proliferation and differentiation. Hence, the affinity a can be interpreted as a measure of the long-term repopulation potential of an individual cell. Consequently, the residence in context A is necessary to prevent differentiation and, therefore, implicitly to maintain the HSC population. A sketch of the model is provided in Fig. 1A.

image

Figure 1.  Model setup. (A) The model setup is characterized by two different signal contexts, A and Ω between which the cells can reversibly change depending on the cell number within the target context and the cell-specific affinity a. Whereas activated cells in Ω undergo divisions and exponentially degrade their cell-specific affinity a (blue arrows), cells in A are quiescent and regain their affinity value (green arrows). Introducing the cell-specific aging effect, the upper limit for regeneration in A, termed acap, is reduced (red shift) with an ‘aging probability’pa at each division event by a factor ka (‘aging strength’). (B) Illustration of an activation and binding process of a single hematopoietic stem cell (HSC) in the affinity a vs. time t space. Upon division, the regeneration limit acap is reduced by factor ka (red). Color coding corresponds to the signal contexts as in (A). (C) Visualization of the corresponding developmental sequence in the model sketch. Encircled time points relate to (B).

Download figure to PowerPoint

Within the original model, the parameter amax describes an upper limit for the regeneration process of the HSCs residing in the signaling context A. We could show that this parameter is a critical regulator for the long time survival of individual cells and their progeny. To represent the aging effect of individual cells, we have chosen this parameter to account for the accumulation of unfavorable changes because of mitotic activity. In particular, we transformed the global variable amax in a cell-specific parameter indicated as acap. Upon division, the cell-specific regeneration limit acap is decreased by a factor ka (ka < 1) with probability pa and it is inherited to the daughter cells. We refer to the variable pa as ‘aging probability’ and to the factor ka as ‘aging strength’. These modifications are illustrated in Fig. 1B,C. The cell-specific regeneration limit acap characterizes the current maximum repopulation ability of an individual cell and reflects the divisional history of its ancestry.

The overall aging effect of the population is determined by the superposition of the aging probability pa and the aging strength ka. Throughout the paper, we compare three different, representative parameter settings for the aging probability pa and the aging strength ka. In particular, we study the case that a moderate aging effect (ka = 0.988) occurs at each division event (i.e., pa = 1, named deterministic aging). This case is compared to two other scenarios in which the aging effect only occurs with a certain probability (pa = 0.28 or pa = 0.45); however, the aging strength is increased compared to the deterministic case (ka = 0.9 or ka = 0.965, referred to as low and medium probability aging, respectively). Parameter values are chosen such that stem cell exhaustion occurs after the fourth to sixth sequential transplantation.

For the present analysis, the repopulation ability of a cell sample in (competitive) transplantation assays serves as the central read-out parameter to quantify the aging-related changes. Further details of the simulation procedure and parameter tables are provided in the Data S1.

Serial transplantations

Serial transplantations are an appropriate experimental setup to study declining repopulation ability within a stem cell population. Failure of repopulation after the fourth to sixth subsequent serial transplantation was commonly observed in a number of different experimental approaches (Ogden & Mickliem, 1976; Ross et al., 1982; Kamminga et al., 2005). Our originally proposed model of HSC organization (Roeder & Loeffler, 2002; Glauche et al., 2007) did not account for this experimentally indicated exhaustion of the repopulation ability, implying that serial transplants could be repeated indefinitely. However, by including a division-dependent, albeit stochastic decline of the stem cell’s repopulation potential (indicated by the newly introduced, cell-specific upper regeneration limit acap, see Materials and Methods), the model is now amended to reflect these findings.

Figure 2 documents stem cell exhaustion as a function of serial transplantations. The subfigures in the upper panel (Fig. 2A–C) show actual stem cell numbers over time. Technically, a primary system was initialized using 3000 stem cells each initialized with acap = 1.0. After about 3 months, the system reaches a stable equilibrium, in which the total number of HSCs is almost constant and a surplus of cells contributes to differentiation. After 6 months, 30 stem cells are randomly extracted from the stem cell pool and transplanted into a secondary empty system mimicking an irradiated recipient. The number of extracted stem cells roughly corresponds to the transplantation scheme used by Kamminga et al. (Kamminga et al., 2005) in which the authors transplanted 1500 Lin-, Sca-1+, c-Kit+ (LSK) cells with about 2% HSC content (De Haan et al., 2000). After about 2–3 months, the system did again reach a stable equilibrium, however, because of the accelerated division during repopulation the stem cell population declined in number. This effect is caused by an overall reduction in the cell-specific regenerations limits acap which in turn results in higher activation of HSCs and thus a reduction in the quiescent stem cell reservoir.

image

Figure 2.  Serial transplantations. Upper Panels: Stem cell numbers after serial transplantations for low (A), medium probability (B), and deterministic aging (C). After six simulated months, 30 hematopoietic stem cells (HSCs) are randomly selected and transferred in an empty model system mimicking the irradiated recipient. The green curve indicates the initial donor mouse and the red ones the engraftment in the subsequent recipients including a brief period of stable hematopoiesis. Lower Panels: Comparison with experimental data on the fraction of LSK cells (Kamminga et al., 2005) for low (D), medium probability (E), and deterministic aging (F). Simulated data from the upper panel are scaled such that the mean values of the number of LSK cells over four engraftments equal between model and experimental data. Standard deviation is indicated by the error bars (where available). Chi-square values are provided for each scenario.

Download figure to PowerPoint

For the simulation of serial transplantations, 30 randomly chosen stem cells are repeatedly extracted after 6-month simulation time and transplanted into an empty system mimicking the lethally irradiated tertiary and higher order recipient. This sequence is continued until no stable repopulation equilibrium could be established owing to the accumulating decline of stem cell quality in terms of their individual regenerations limits acap. Figure 2A–C compares three different parameter settings for the aging probability pa and the aging strength ka, which are referred to as low probability, medium probability, and deterministic aging (see previous section for details). The overall aging effect for all three scenarios has been adapted such that the system repopulation fails on average between the fourth and the sixth simulated transplantation.

The lower panels of Fig. 2(D–F) compare the simulated stem cell numbers to the experimentally observed decline of LSK cells in serial transplanted animals as reported in the publication of Kamminga et al. (Kamminga et al., 2005). As the relation between repopulating stem cells within the model and phenotypically defined LSK cells is not clearly defined, we use the ratio between the mean values (averaged over four repopulated systems) of the number of in silico HSCs and the percentage of experimentally detected LSK cells as an appropriate scaling factor. We can now calculate the χ2 statistics

  • image

as a measure of fitting quality for each of the three scenarios (ei– experimental result for ith transplantation; si– corresponding simulation result). The comparison demonstrates that our model is able to account for the observed relative loss in stem cell numbers (lower panels of Fig. 2). The low probability aging scenario (Fig. 2D) shows a slightly better fitting compared to the other two scenarios, as indicated by the χ2 values.

Consistent with earlier findings (Harrison & Astle, 1982), our simulation studies indicate that a decrease in the number of transplanted cells accelerates the exhaustion process, whereas an increase in cell numbers allows for more rounds of sequential transplantations. In this context, the increased survival of recipients after the fifth transplantation of unfractionated bone marrow cells instead of LSK cells, as observed by Kamminga et al. (Kamminga et al., 2005), could be well attributed to a slightly increased number of actually transplanted, repopulating stem cells as compared to the transplantation of the LSK cell pool.

Quantitative stem cell exhaustion

Using the above-stated assumptions, the model also predicts stem cell exhaustion even without challenging the system by serial transplants. Figure 3 shows the total number of modeled HSCs as well as their distribution in context A and Ω as a function of time for the three different parameter configurations outlined above (deterministic, medium, and low probability aging). Our simulations indicate that the model (under the given parameter configurations) predicts final exhaustion of HSCs to occur between 10 and 30 years, thus covering up to 15 times the average life expectancy of a mouse. This result supports and quantifies earlier findings (Harrison & Astle, 1982) indicating that a single transplantation corresponds to the accumulated aging of several mouse life times of normal function. Therefore, our model prediction is consistent with a hematopoietic system which is in general well equipped to robustly serve its duty over the time of a normal life span even with several milder challenges like infections or blood loss. This in turn would affirm the perception of organismal aging as a multifactorial process which simultaneously affects several other tissues besides the hematopoietic one.

image

Figure 3.  Stem cell exhaustion. Stem cell numbers are shown as functions of time [total amount of hematopoietic stem cells (HSCs) within the model (A), activated HSCs in signaling context Ω (B), dormant HSCs in context A (C)]. Colors indicate low probability aging (green), medium probability aging (red), and deterministic aging (blue).

Download figure to PowerPoint

Our model is built on the idea that there is a constant exchange between activated and quiescent stem cells. In other words, quiescent HSCs are activated occasionally (on average once in 70 days), divide and can then either return into quiescence or contribute to further differentiated progeny. Changing stem cell numbers in Fig. 3 indicate that this balance is shifted because of the aging effect. The overall decline of the cell-specific regeneration limit acap leads to accelerated activation. The shorter residence time in a state of cellular quiescence manifests as an overall reduction in cell numbers among the quiescent HSCs (Fig. 3C), which can only incompletely be compensated by the pool of activated cells (Fig. 3B). Our model simulations predict that aging-related loss of stem cell quality should be detectable first by a decline in the number of quiescent stem cells (about 18% within 2 years life time). As a phenotypic distinction between activated and dormant HSCs is still challenging, we argue that differential turnover kinetics (using, e.g., label dilution techniques with chromosomal markers such as BrdU or H2B-GFP) between young and old animals should be suited to reveal the same effect.

Clonal heterogeneity

Because of the specific divisional history of each cell and the additional stochastic component within the aging model (i.e., the aging probability pa), the distribution of the cell-specific upper regeneration limit acap gives rise to another intrinsic heterogeneity. This effect is illustrated in Fig. 4 in which the color coding reflects the frequency of cells as functions of acap on the x-axis and of the time after initialization on the y-axis. Starting from a homogeneous cell population (acap = 1.0) at time point t = 0, there is a clear trend that the upper regeneration limit acap is progressively decreasing over time in all the scenarios. However, there is a remarkable difference in the level of heterogeneity for the three scenarios. Whereas in the deterministic aging scenario all cells reduce their individual regeneration limit acap rather synchronized (Fig. 4C), there is a significant increase in this dispersion toward the low probability aging scenario (Fig. 4A). The insets of Fig. 4 show histograms of this dispersion in the acap levels for the time point t = 1.5 years (also indicated by the dashed line in the main figures). Whereas in the low probability aging scenario cells are centered around acap = 0.65, there is a small population of cells with significantly smaller values down to 0.3. In contrast, such cells with lower acap values are not detected for the medium probability aging and the deterministic aging scenarios in which the distribution is considerably more peaked around acap = 0.7–0.75.

image

Figure 4.  Aging induced heterogeneity of the regeneration limit acap. Density plots provide the frequency of cells with certain acap values (x-axis) as functions of time (y-axis) for the three different aging scenarios [(A) low probability aging (B), medium probability aging (C), deterministic aging]. The insets depict histograms of the acap values at 1.5 years (dashed line).

Download figure to PowerPoint

Competitive transplantations

Competitive co-transplantations of different cell populations are a sensitive assay to detect qualitative differences of the transplanted cell types especially with respect to their short- and long-term repopulation ability. However, it has been reported that some mouse strains compensate the decrease in functional abilities of aged HSCs by an increase in their HSC numbers (Morrison et al., 1996; Sudo et al., 2000; Chambers et al., 2007). To exclude such quantitative effects and to assess the qualitative difference between young and old HSCs, we suggest to co-transplant equal numbers of functional HSCs (instead of an equal amount of bone marrow) together with a primary recipient. Using a limiting dilution approach, it is in principle possible to determine the number of functional stem cells within a certain preselected HSC population, which is necessary for the suggested strategy.

Translating this strategy into our modeling approach, we use 30 cells randomly chosen among all HSCs of a 3-month-old simulated mouse and transplant them together with 30 cells taken from HSCs of an 18-month-old mouse based on the low probability aging scenario. It should be noted that each of these cells from either a young or an old system is in principle able to repopulate an empty model system on its own. However, assessing the engraftment levels of young vs. old cells 3 months post-transplantation, it is striking that the young cells quickly dominate the older ones (Fig. 5A). This advantage is attributed to the higher values of the regeneration limit acap in young cells. In contrast, if two aliquots with comparable numbers of distinguishable young (or old) donors, HSCs are transplanted together with a conditioned recipient, the ratio between the two donor cell types stabilizes at the ratio of the original mixture (see Data S1).

image

Figure 5.  Competitive transplantations: (A) Fraction of 3-month-young (green) vs. 18-month-old (red) hematopoietic stem cells (HSCs) after competitive retransplantation (30 randomly chosen cells from each HSC population, again based on the low probabilityaging scenario). (B) The same experiment has been repeated with ‘young’ (green) and ‘old’ (red) cells (with respect to mitotic age) from cells within one 1.5-year-old mouse (i.e., the two subpopulations correspond to the left and the right tail of the histograms in Fig. 4). (C) An ‘old’ stem cell clone with extensive repopulating ability [compare (D)] is rapidly outcompeted by a secondary ‘young’ clone transplanted 100 days after the initial transplant. (D) Repopulating ability of the ‘old’ clone without secondary transplant.

Download figure to PowerPoint

It is particularly interesting that a similar, dominated repopulation behavior is obtained when co-transplanting preselected HSCs from one model system that have distinct (i.e., low or high) acap states. This effect is most pronounced in the low probabilityaging scenario (compare Fig. 4A) that shows the widest distribution of acap values. Again, the cells with higher values of the regeneration limit acap outcompete the ones with lower values, which only contribute to the pool of differentiating cells on intermediate timescales (Fig. 5B). This observation indicates that a similar heterogeneity as observed between young and old HSCs (extracted from young and old animals) is already present within one model mouse. This heterogeneity results from the individual divisional history of each stem cell. Our model predicts that a robust population of HSCs always consists of a mixture of younger and older cells (in terms of the number of there previous divisions) with different functional abilities (intra-animal variability).

This interpretation is also supported by in vivo results observed by Kamminga et al. (Kamminga et al., 2005). The authors reported on an initially engrafted five times serially transplanted cell population that was subsequently replaced by freshly isolated cells, transplanted 100 days past the initial transplantation. Nevertheless, in a noncompetitive situation, the population of older HSCs was still able to sustain for extended time periods. Figure 5C,D show the results of the in silico adaptation of these experimental settings with qualitatively similar results. Within the model, the decreased repopulation ability (indicated by lower values of the regeneration limit acap) of the older cells makes them prone to fast dilution in the competitive situation with young cells. Although the simulation results do not perfectly match the fast outcompetition observed in vivo (about 100 days in the experimental situation), the model is well suited to explain the general phenomena observed in these experiments.

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Author contributions
  8. References
  9. Supporting Information

We complemented a previously established mathematical model of HSC organization by an essentially novel feature, i.e., the cellular aging of HSCs. The model describes aging as a cumulative, mitosis-related decrease of stem cell functionality on the single-cell level, and it is able to explain tissue failure as a consequence of cellular exhaustion. Our model is based on the assumptions that a dominantly quiescent population of HSCs is occasionally activated to undergo cell division. The resulting daughter cells can either return to quiescence or undergo further divisions and differentiate. Both the activation as well as the further decisions of the daughter cells are modeled as stochastic events. Because of this stochasticity, each model cell has an individual history, which is determined by the number of divisions of its ancestry (‘mitotic age’). Assuming that cellular aging is coupled to cell division, this individual divisional history translates into a stem cell-specific ‘quality’ that declines over time corresponding to the increasing number of forgoing division events. Within the model, this quality measure is associated with the cell-specific regeneration limit acap. We have shown that the stochasticity of the activation and decision events ultimately leads to a heterogeneous stem cell population with respect to this parameter (Fig. 4). Our model also predicts that this aging-related decrease can be quantified in a competitive transplantation setting with young vs. old HSCs under the condition that equal numbers of functional stem cells are co-transplanted besides a recipient’s background.

The simulation results furthermore predict that similar differences as between HSCs from young and old donors are also present within one animal. Extracting cells with low and high mitotic age (in terms of their individual regeneration limit acap) and co-transplanting them into the same host, the cells with lower mitotic age (i.e., higher regeneration limit acap) outcompete their competitors in the long run. The cells with higher mitotic age (i.e., lower regeneration limit acap) do only transiently populate the stem cell compartment and predominantly contribute to the pool of differentiating cells. We speculate that this population of transiently contributing cells with lower regeneration limit acap might be closely related to the population of short-term repopulating stem cells (STR-SCs) that is commonly identified in different experimental situations (Jones et al., 1989; Zijlmans et al., 1995; Lanzkron et al., 1999). In contrast to long-term repopulating stem cells (LTR-SCs), which provide long-term reconstitution of hematopoiesis even in multiple, consecutive transplants, STR-SCs have limited self-renewal capacity (Lanzkron et al., 1999; Curtis et al., 2004) and only reconstitute the hematopoietic system for a short timespan of 8–12 weeks (Christensen & Weissman, 2001). Our model suggests that a population-inherent heterogeneity, which is caused by the differential accumulation of aging-related damages in individual HSCs, is a possible origin for the observed different functional capabilities of long- and short-term repopulating stem cells. This interpretation implies that STR-SCs correspond to a population of normal HSCs that underwent more cell divisions, which in turn makes them more prone to further differentiation. Our computational approach supports a similar, conceptual view that has been suggested by van Zant et al. (Zant et al., 1997).

Under the assumption that aging is sufficiently described as a cumulative, mitosis-related decrease of individual stem cell functionality, our model furthermore predicts that the overall (organismal) aging should be accelerated by the administration of cell cycle-activating drugs such as G-CSF (Morrison et al., 1997) or IFN-alpha (Essers et al., 2009). As a sensitive readout, we again suggest competitive transplantation experiments to test whether the functional abilities of HSCs derived from mice under long-term stem cell-activating drugs differ from control animals of the same age.

The thorough and conceptually concise understanding of cellular and organismal aging is still in its infancy. It appears that aging is not just caused by one mechanism but by a whole spectrum of aging-related mechanisms. It has long been suggested that progressive telomere shortening is a restricting factor of stem cell replication limiting the number of subsequent stem cell divisions to about 50 (commonly known as Hayflick-limit) (Hayflick & Moorhead, 1961; Hayflick, 1965). However, it could be demonstrated that mice overexpressing telomerase (and therefore preserving telomere length) still undergo aging (Allsopp et al., 2003). There is accumulating evidence that, for a variety of reasons, stem cells cease their functionality to act as stem cells to prevent cancer but retain their phenotypic appearance (Boulanger & Smith, 2001; Youn et al., 2004). This phenomenon is referred to as senescence. Although senescent cells with HSCs phenotype are not considered as stem cells by definition, it is experimentally challenging to prospectively distinguish these cells from functional HSCs. Besides the stem cell-specific, potentially mitosis-related aging events, there is a range of phenomena related to age-dependent alteration of the supportive microenvironment. Experimental data imply that age-related changes in the quality of support, mediated by the hematopoietic niches, have a direct effect on HSC regulation (Mauch et al., 1982; Jiang et al., 1992; Geiger & Van Zant, 2002).

Neither aspects of an aging microenvironment nor of HSC conversion to senescence are currently represented within the described modeling approach. For the case of the aging microenvironment, the experimental evidence is weak. Especially considering the fact that serial transplantations fail even in the case of using young recipients, it appears challenging to derive justifiable model assumptions. Although the experimental confirmation of senescence is more substantial, there are still no quantitative results about this effect in populations of aging HSCs. Moreover, we argue that neglecting the senescent subpopulation only alters the phenotypic detection, but should not influence the functional results as long as the senescent cells do not occupy niches or outcompete and replace normal hematopoiesis. We furthermore point out that our model in the current form is only suited to evaluate aging with respect to changes in the repopulation ability of the HSCs but not with respect to changes in lineage contribution as recently suggested in (Sieburg et al., 2006; Dykstra et al., 2007; Challen et al., 2010). However, with respect to the findings of Muller-Sieburg and colleagues, it is interesting to note that the population of lymphoid-biased HSCs is lost during aging, while the frequency of myeloid-biased HSCs seems to increase (Sieburg et al., 2006; Cho et al., 2008; Muller-Sieburg & Sieburg, 2008). At the same time, the authors report that the lymphoid-biased HSCs outcompete the myeloid-biased HSCs in a competitive transplantation assay (Cho et al., 2008). Taken together, these findings indicate that the overall competitive repopulation ability of the HSC pool seems to decline over the life time of the individual. This is consistent with our findings that also suggest an age-related shift in the composition of HSCs toward cells with reduced repopulation potential.

From our modeling perspective, we conclude that cellular aging induces a level of HSC inherent heterogeneity that reflects the cells’ ancestral development. The accumulation of aging-related alterations leads to a functional diversity that also influences the future fate of the individual HSCs. Although the mechanisms of aging most likely act on the cellular level, we argue that an understanding of tissue failure is only possible on the population level. This view fits in the general perception of stemness as a concept characterizing the repopulation ability of a cell population rather than of a single cell. This also implies that a stem cell population does not necessarily consist of identical cells but allows for intrinsic diversity as long as the functional requirements for the overall tissue are fulfilled.

Acknowledgments

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Author contributions
  8. References
  9. Supporting Information

This research was supported by the European Commission project EuroSyStem (200270), by the German Research Council (DFG), grant RO3500/1-2, and by the German Ministry for Education and Research, BMBF-grant on Medical Systems Biology ‘HaematoSys’ (BMBF-FKZ 0315452).

Author contributions

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Author contributions
  8. References
  9. Supporting Information

Ingmar Glauche: Conception and design, data analysis and interpretation, manuscript writing. Lars Thielecke: Simulations, data analysis and interpretation, manuscript writing. Ingo Roeder: Conception and design, data analysis and interpretation, manuscript writing.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Author contributions
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Author contributions
  8. References
  9. Supporting Information

Data S1 Simulation algorithms and parameters.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
ACEL_692_sm_data.pdf288KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.