| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TISSUE-SPECIFIC STEM CELLS |
aInstitute for Medical Informatics, Statistics and Epidemiology, University of Leipzig, Leipzig, Germany;
bDivision of Hematology/Oncology, Interdisciplinary Center for Clinical Research, University of Leipzig, Leipzig, Germany
Key Words. Lineage specification • Hematopoiesis • Stem cells • Mathematical model • Simulation • Systems biology
Correspondence: Ingmar Glauche, Institute for Medical Informatics, Statistics and Epidemiology, University of Leipzig, Haertelstr. 16/18, 04107 Leipzig, Germany. Telephone: +49 341 97 16 112; Fax: +49 341 97 16 109; e-mail: ingmar.glauche{at}imise.uni-leipzig.de
Received January 10, 2007;
accepted for publication March 28, 2007.
First published online in STEM CELLS EXPRESS April 5, 2007.
 |
ABSTRACT |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionConclusionDisclosure of Potential...AcknowledgmentsReferences |
|---|
Disclosure of potential conflicts of interest is found at the end of this article.
|
INTRODUCTION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionDisclosure of Potential...AcknowledgmentsReferences |
|---|
The question remains whether this phenomenological view is consistent with the findings on the progressive restriction of lineage potential. Moreover, it is unclear whether the proposed molecular dynamics of lineage specification agree with the observable temporal pattern of differentiating stem and progenitor cells. Therefore, it is our objective to derive a generalized analytical framework to understand lineage specification as a temporally extended process. Within our approach, the role of complex cell-extrinsic signaling events that influence the lineage specification process (e.g., by cell-cell and cell-environment interactions) is approximated by two contrary control regimes. Particularly, a regressive control regime maintains an undifferentiated priming state, and a progressive control regime promotes the process of lineage commitment. In this framework, lineage specification emerges as the result of a sequence of small decision steps (approximating a continuous process) rather than a singular decision event. The decision sequence slowly shifts the probabilities for development into a particular lineage and passes this potential on to the daughter cells. Therefore, daughter cells are identical after mitosis but continue lineage specification independently from each other.
The proposed intracellular lineage specification dynamics are embedded in the model for hematopoietic stem cell organization recently proposed by Roeder and Loeffler [7, 8]. Within this model, stem cells have the ability to change between two signaling contexts that impose different effects on the cellular development. For the model extension described here, we assume that these signals also affect the intracellular lineage specification dynamics, therefore inducing a correlation between the regulation of self-renewal and lineage specification. The extended model comprises a whole new class of phenomena in full consistency with former results on stem cell self-renewal and clonal competition [8–12].
To verify the proposed theoretical model, we compare our simulation results with different sets of experimental data. Specifically, we apply literature data that describe the lineage contribution of single differentiating cells [13] as well as the lineage contribution within the progeny of two first generation daughter cells derived from a common parental cell (sibling analysis) [1, 13]. Furthermore, we show that the model is able to account for the typical kinetics of lineage development as observed for the differentiation of the (stem cell-like) FDCP-mix cell line, which have been measured in our lab.
|
MATERIALS AND METHODS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionDisclosure of Potential...AcknowledgmentsReferences |
|---|
Intracellular Dynamics
For a conceptual understanding of the complex molecular dynamics governing lineage specification, we assume that all regulators that are specific for one common lineage fate are summarized in a single generic measure called lineage propensity. The level of each lineage propensity represents the potential of a cell to develop into the corresponding lineage. Furthermore, we assume that lineage specification can be understood as a mutual competition process between such propensities, in which every gain (or loss) in a particular lineage propensity will lead to the reduction (or increase) of the remaining propensities. Particularly, we apply a stochastic competition process leading either to the maintenance of the low-level coexpression (priming) or to the dominance of one or the other lineage propensity (commitment).
Formally, these model assumptions are specified as follows:
Figure 1 outlines the intracellular lineage specification dynamics for a particular model cell. The horizontal dashed line indicates the propensity level xcom that is used for the phenotypic separation of undifferentiated (xmax < xcom) and committed (xmax > xcom) cells. The specification of xcom is used solely for the phenotypic mapping of model results to experimental data and does not imply the irreversibility of the commitment decision. However, the probability for a change of the dominant lineage decreases considerably with increasing values of the propensity xi.
|
), its position in the cell cycle (G1, S, G2, M, or G0), and its affinity a, which quantifies the propensity of a particular cell to reside in signaling context A. Cells in A are assumed to be nonproliferative, maintaining (e.g., in in vitro scenarios) or even regaining their affinity a up to an upper limit amax = 1 (e.g., in in vivo scenarios). In contrast, cells in signaling context are characterized as proliferative, accompanied by a gradual loss of their affinity a. It can be shown that the affinity a characterizes the cell's ability to realize long-term system repopulation. Accounting for the presumed underlying complexity, transitions between the two signaling contexts are described by a stochastic process. The probability of switching depends on the actual value of a as well as on the number of cells in the target signaling context. Transition from signaling contexts to A is impaired for cells with a < amin. These cells continue to divide throughout a proliferative phase, followed by a maturation phase without further amplification. Finally, mature cells are removed from the system to reflect their limited life span (Fig. 2A).
|
to impose contrary effects on the intracellular lineage specification dynamics of each individual cell. We link the progressive control regime to signaling context , thereby relating the processes of lineage specification with the loss of repopulation potential. Vice versa, we link signaling context A with the regressive control regime. Therefore, cells in A experience no lineage commitment and simultaneously maintain their ability to act as stem cells. Characteristic time courses for the intracellular lineage specification dynamics of individual cells are shown in Figure 2B–2E.
Model Comparison with Experimental Data
We compare our model with experimental data from three different types of experiments.
Differentiation of Unselected Progenitor Cells. Early quantitative approaches used hematopoietic spleen-derived mouse cells [1] and human umbilical cord blood cells [2] to examine the developmental fate of two daughter cells derived from one parent cell. The authors concluded from their results that lineage potential is progressively restricted by a sequence of stochastic commitment steps that take place at each cell division.
Differentiation of Enriched Cells. A set of results comparable in many respects with those presented by Suda et al. [1] was recently published by Takano et al. [13], using CD34– c-Kit+ Sca-1+ lin– (CD34– KSL) cells taken from adult mouse bone marrow. The lineage composition of the progeny of these cells was analyzed in vitro using single cell differentiation experiments. Furthermore, the variability of the lineage contribution of a parental cell was evaluated by following the fate of its two daughter cells. In these experiments, the authors studied the influence of the culture conditions within the initial division assay on the subsequent asymmetry of daughter cell development.
Differentiation of FDCP-Mix Cells. The FDCP-mix cell line is a well established example of a stable cell line, derived from murine, multipotent hematopoietic progenitors, which retains the capacity to self-renew in the presence of high concentrations of Interleukin-3 (IL-3) [15, 16]. When transferred to low concentrations of IL-3 combined with other hematopoietic growth factors or injected into experimental animals, FDCP-mix cells show an apparently normal progression of lineage commitment and differentiation. FDCP-mix cells maintained in Iscove's modified Dulbecco's medium (IMDM) containing 20% horse serum and 100 U/ml IL-3 were washed and transferred at a density of 4 x 104 cells per milliliter to IMDM containing 20% fetal calf serum and either myeloid (M) or erythroid (E) growth factors as previously described [16]. The combination of growth factors supports differentiation either into a mixture of granulocytes and macrophages (M) or into a predominantly erythroid population (E). On consecutive days up to day 9, cells were harvested from replicate cultures and cytospun. Following May-Grunwald staining, differential counts were performed blind on 100–200 cells per time point. This way, we obtained a temporal pattern of the differentiation process.
Simulation Strategy
Following the experimental situations outlined above, we used three simulation protocols for model verification.
Comparative Differentiation of Paired Daughter Cells (Fig. 3A).
|
. Cells used for transplantation into division assays are chosen randomly among a well defined subpopulation of the source assay (transfer pool), characterized by the range of the affinity parameter atrans. The boundaries of these transfer pools are the central parameters to fit the simulation results to the experimental data by Suda et al. [1] (pool S) and Takano et al. [13] (pool T). The division assay is represented by an empty model system that mimics the culture conditions for the division of the parent cell. For simulation efficacy, all transferred cells are under the governance of signaling context . The cell cycle position, the affinity a, and the lineage propensities xi are preserved. After division, both daughter cells are transferred into two separate empty model systems, in which the development of the progeny is observed for 240 hours (lineage assay). Finally, the number and the lineage of cells produced in each lineage assay are evaluated. Due to an expected deficiency of a properly functioning hematopoietic niche environment in cell cultures, we assume that, for all simulated in vitro assays, the signaling context A simply maintains the self-renewal ability of a cell (measured by its affinity a) but does not promote its regeneration.
Lineage Contribution of Single Differentiating Cells (Fig. 3B). The single cell differentiation experiments by Takano et al. [13] are incorporated into the simulation protocol with only minor adaptations. From transfer pool T of the source assay, randomly chosen cells are directly transferred into the lineage assay, where their lineage contribution is determined. All other parameters are left unchanged.
Lineage Specification in Differentiating Cell Cultures (Fig. 3C).
In order to reflect the usage of relatively homogenous cells from a cell line, the differentiation assay is initialized with a population of 250 cells with a well defined initial affinity, uniformly distributed in the range ainit 
[0.01,0.1]. The fraction of undifferentiated and committed cells is evaluated hourly for a period of 9 days. A balanced expression of the n = 3 lineage propensities x1 = x2 = x3 = 1/3 is assigned to the cells such that they are initially unbiased for the development in each of the three experimentally observed cell types: granulocytes, macrophages, and erythrocytes. The lineage specific rewards mi are adjusted to meet the observed development of the cultures under granulocyte/macrophage (M) or erythrocyte (E) stimulating conditions. Particular parameter values are given in the supplemental online data.
|
RESULTS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionDisclosure of Potential...AcknowledgmentsReferences |
|---|
[0.000001,0.99], as shown in Figure 4A. In close correspondence to the findings of Suda et al. [1], we also observed the case in which one daughter cell develops into up to five lineages, whereas the other daughter cell is restricted to just one or two. Furthermore, some simulations generated daughters that contribute to the same overall combination of lineages with considerably different proportions of the individual cell types among their progeny.
|
[0.012,0.99]) and marks the central difference to the setup of Suda et al. [1] (Fig. 3A). Changing no other parameters, the model reproduces the results of the experiments. Among the initial parental cells with complete lineage contribution (all four lineages), paired daughter cells with identical lineage development dominate over pairs with asymmetric development (Fig. 4B). In the simulation results, we observe additional minor contributions (0.1%–8.0%) to other combinations of lineages (data not shown). These are not described experimentally, which might be due to the limited number of observations. Takano et al. [13] also report that different combinations of cytokines in the in vitro division assay influence the lineage potential of the daughter cells and change the particular ratios of symmetric versus asymmetric development. Qualitatively similar phenomena can be observed by modifying the lineage specific rewards mi within the division assay of the in silico model (data not shown).
Lineage Contribution of Single Differentiating Cells
In single cell differentiation studies of bone marrow derived CD34– KSL cells, the majority (43%) of plated cells contributed to all four determined lineages, whereas other combinations are observed with lower frequency [13]. Applying the identical transfer pool T that has been used for the comparative differentiation of paired daughter cells and evaluating the lineage contribution of the single differentiating cells without an intermediate division step, these findings can be reproduced in silico. In particular, we find that, in the majority of cases (45.2%), the progeny contained all four lineages, whereas other combinations are reduced (Fig. 5A). It should be emphasized that this qualitative pattern is achieved even under the simplifying assumption of equal lineage potentials (i.e., equal rewards mi). However, the precise matching of the results is incomplete. The experimental data suggest that there is a correlation between neutrophil and macrophage differentiation (see lineage combination 14 in Fig. 5A). Introduction of a moderate, positive correlation between lineages 1 and 4 (neutrophils and macrophages, respectively) in the in silico model leads to a shift in the differentiation pattern similar to the experimental observations (Fig. 5B). Progeny of single cells containing neutrophil and macrophage cells are now significantly enhanced compared with other developments. Due to the complexity of these potentially weak correlations between certain lineages, a detailed quantification of this process is hardly possible on the basis of the available experimental data.
|
During erythroid development, erythrocytes mature from erythroblasts. Since it is possible to distinguish between these cell types morphologically, erythroid cells are subdivided for the phenotypic mapping such that the committed cell stage now comprises early committed cells (erythroblasts) and mature cells (erythrocytes).
Figure 6 shows that the simulation model is able to quantitatively account for the temporal development of the proportions of observed cell types in both M and E media. Although the lineage specific rewards mi have been adapted to meet these particular experimental results, the agreement of simulation and experiment can be regarded as a proof of principle that the proposed model is able to adequately account for differentiation kinetics on the population level.
|
|
DISCUSSION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionDisclosure of Potential...AcknowledgmentsReferences |
|---|
We showed that the model supports the idea of a progressive restriction in lineage potential in the course of differentiation. To make this idea more visible, we have studied the lineage contribution of single differentiating cells depending on their initial repopulation ability, characterized by the affinity parameter a for the example of a system with four possible lineages (Fig. 7). As expected, nearly all cells with high repopulation ability contribute to all four lineages, whereas tri- and bipotent cells are found mostly in the population with moderate repopulation capacity. The ultimate loss of repopulation potential is associated with commitment to a single lineage. It is essential to note that this hierarchic decrease of the lineage potential is predicted by the model as an emergent system property and is not intrinsically predefined in individual cells.
|
Another important aspect is the distinction between lineage contribution (being the lineages actually generated by the progeny of a particular cell) and lineage potential (being the lineages to which the same progeny could have contributed). Since a single cell can only differentiate once, the lineage potential cannot be determined experimentally. However, despite this inherent uncertainty, the notion of lineage potential is important to understand the organizational principles of cell populations and tissues as well as for the characterization of the stem cell properties. As we have exemplarily shown for the lineage specification kinetics of the FDCP-mix cell population, the fluctuations in lineage potential that occur on the single cell level average out on the population level. This means that, although the outcome on the population level is robust, the particular fate of a single cell can only be predicted in a probabilistic sense. Based on this understanding, our model predicts that heterogeneity of a progenitor population is inherently generated as a consequence of the autonomous development of individual cells.
The role of asymmetric cell divisions in the process of hematopoietic lineage specification is still controversial [17–19]. Although such divisions are reported for a number of other systems [20, 21], no evidence for (functional) asymmetric cell division has yet been found within the hematopoietic system. Although we cannot rule out the possibility of asymmetric cell divisions in hematopoietic cell differentiation, our model demonstrates that a consistent explanation of the heterogeneity among differentiated cells is possible without assuming an asymmetric division process. Technically, any simulated cell division is symmetric. Differences in the individual development of the daughter cells occur only due to their independent differentiation sequences after mitosis. Asymmetric development is thus interpreted as the asymmetry of cellular fates, not of the division process itself (see also [12]).
Our model concept supports the hypothesis that the experimentally observed priming behavior is a common molecular representation of the stem cell state (see also [14, 22, 23]), which is maintained under specific conditions (e.g., due to niche signals). Maintenance of the priming state could feasibly be achieved by the active epigenetic stabilization of chromatin structures that retain parallel developmental options. Changing microenvironmental signals destabilizes the priming state. Under these modified conditions, chromatin changes at key loci may result in a sequential shift of the expression state toward one or the other lineage specific expression pattern. This process represents a molecular view of the differentiation process with progressively decreasing probabilities for multipotent development. At an experimentally accessible level, our model predicts that targeted up- or downregulation of certain lineage specific genes upsets the balance at the priming level and, consequentially, supports or discriminates certain options in the subsequent differentiation process. A particular strength of the model is the foundation on the level of single cells. Alongside with the experimental tracing of individual cells in culture [24, 25], our model is able to identify critical phenomena of the molecular differentiation sequence (as there are asymmetric developments, the occurrence of lineage specific markers and their inheritance to the daughter cells, and the role of apoptosis and selection) and to link them to the population level.
The assumed temporal extension of losing multilineage potential is closely associated with a reversibility of the differentiation process. We model lineage specification as a process that favors a certain lineage development by progressively decreasing the probabilities for the competing options. Therefore, reversibility depends strongly on both the actual state of differentiation and on the influence of the microenvironment. The model predicts that reversibility of lineage specification is a rare event in a homeostatic system. However, it is expected to be more common in a disturbed situation with the need for system repopulation. Similar effects should be observable when cells that are primarily cultured in a particular differentiation promoting medium are transferred into a condition with different properties (e.g., promoting self-renewal or another differentiation program). The model predicts that the fraction of cells with "reverted" development is not an all-or-nothing decision but depends in the first place on the exposure time in the particular medium. A rigorous experimental test of this prediction would have to use molecular markers that are irreversibly switched on if a certain characteristic gene expression identifies a particular lineage commitment. The detection of such markers can elucidate to what extent early committed cells actually "reverse" their previous development under changing environmental conditions. The model predicts that the fraction of cells with reversible developments gradually decreases as the process of lineage specification continues.
The particular underlying mathematical process of the lineage specification dynamic was chosen because it resembles a number of desired characteristics, such as the low-level priming, the competitive lineage specification in which one lineage is favored at the expense of others, the controllability of lineage development on a predictable level involving stochastic elements, the temporal extension, and the capability for reversible events. However, the process is based on a number of simplifying assumptions that hamper the application to a directly measurable molecular process. For instance, lineage specification dynamics presumably require a set of many coregulated factors that have been summarized into one generic lineage propensity. This simplification neglects subsequent activation steps, mutual interactions between the members of each of the sets of coregulated factors, and the role of late signaling events. Similarly, the role of extrinsic signaling by cell-cell and cell-environment interactions is reduced to the influence of two antagonistic control regimes that govern the lineage specification process. Furthermore, the phenotypic mapping to classify cells as either undifferentiated or committed is only a rough approximation of the highly complex maturation process. Despite, or perhaps because of, this simplicity, the model proves sufficient to account for a considerable number of phenomena on the lineage specification of hematopoietic stem cells. Most notably, all these results are consistent with previous findings on self-renewal and clonal competition.
This is not to say that the explanation is either unique or complete. Indeed, a detailed quantitative understanding of lineage specification must eventually take account of the characteristics and interactions of a plethora of regulatory molecules, starting with the lineage-specific transcription factors. Modeling approaches to describing the sequential downregulation of lineage specific (transcription) factors during differentiation have previously been suggested by Preisler and Kauffman [26], Furusawa and Kaneko [27, 28], Cinquin and Demongeot [29], and Laslo et al. [30]. In a similar context, we recently showed that autostimulation and specific mutual inhibitions of the transcription factors PU.1 and Gata-1 are sufficient to explain a robust, switch-like behavior from a low-level coexpression of both transcription factors to different states of predominant expression of one of them [31]. This change in the system dynamics can be explained by alterations in transcription efficiency of the individual transcription factors and might provide a molecular basis for the differences between the two antagonistic control regimes in the model presented here.
|
CONCLUSION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionDisclosure of Potential...AcknowledgmentsReferences |
|---|
As outlined above, stem cell development and lineage specification are considered temporally extended processes of continuously changing cellular characteristics. This concept does not exclude certain preferred trends in the differentiation sequence, but it comprises the possibility of reversible developments for individual cells and, thus, allows the system to flexibly react to changing demands. In this sense, "stemness" is no longer understood as a cellular feature but as a system property, a perspective which has been proposed independently by us [7, 8, 12, 32, 33] and by other groups [34–39]. This concept is fundamentally different from approaches that describe stem cell organization as the consequence of a predefined, cell-intrinsic differentiation program. Such approaches assume discontinuous transitions from one confined stem cell or progenitor subpopulation to another in a predefined, strictly unidirectional differentiation sequence [40–43]. Clearly, the grouping of stem and progenitor cells according to features such as cell surface marker expression and functional characteristics remains useful for classification, selection, and enrichment, since it accurately reflects the behavior of a population under a certain set of conditions. Ultimately, however, our increasing awareness of heterogeneity, flexibility, and plasticity within stem and progenitor cell populations questions the validity of these strictly unidirectional concepts at the mechanistic level in single cells. It is here that the combination of experimental and modeling approaches, as the one presented here, may prove most productive.
|
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionDisclosure of Potential...AcknowledgmentsReferences |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionDisclosure of Potential...AcknowledgmentsReferences |
|---|
|
REFERENCES |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionConclusionDisclosure of Potential...AcknowledgmentsReferences |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| STEM CELLS | THE ONCOLOGIST | CME | ALPHAMED PRESS JOURNALS |
and 
, which depend on the affinity a and the number of cells in the target signaling context (for details see [
