Special Reviews

Origin of Stem Cells in Organogenesis

See allHide authors and affiliations

Science  05 Dec 2008:
Vol. 322, Issue 5907, pp. 1498-1501
DOI: 10.1126/science.1162782


The development of individual organs in animal embryos involves the formation of tissue-specific stem cells that sustain cell renewal of their own tissue for the lifetime of the organism. Although details of their origin are not always known, tissue-specific stem cells usually share the expression of key transcription factors with cells of the embryonic rudiment from which they arise, and are probably in a similar developmental state. On the other hand, the isolation of pluripotent stem cells from the postnatal organism has encouraged the formulation of models of embryonic and postnatal development that are at variance with the conventional ones. Possible explanations for the existence of such cells, and the issue of whether they also exist in vivo, are discussed.

The body of an animal or human contains various populations of stem cells that sustain cell turnover in their respective tissues. Where do these stem cells come from? They must arise from the embryo—but the early embryo, before the stage of organ formation, does not contain stem cells. This apparent paradox arises from the conventional definition of a stem cell as an undifferentiated cell that can self-renew indefinitely and can give rise to at least one type of differentiated progeny. This definition was propounded nearly 30 years ago by Lajtha (1) in a theoretical paper that identified most of the key features of stem cell behavior that are considered important today. These included the distinction between stem cells (which are self-renewing) and transit-amplifying cells (which are destined to differentiate after a certain number of divisions); the occurrence of asymmetric divisions; the requirement that stem cells should self-renew for the lifetime of the organism; and the existence of stem cell niches [microenvironments formed by other cells that maintain stem cells—a concept introduced by Schofield (2)].

These properties are displayed by the adult stem cells of mammals such as those of the epidermis, the hematopoietic system, or the intestinal epithelium (37). However, the early embryo does not contain stem cells in this sense, because all the cells in the vertebrate embryo before the onset of organogenesis are destined to change their character as development proceeds. Embryonic stem (ES) cells grown in culture do show indefinite self-renewal, but their normal precursors within the embryo do not, as they soon turn into cells of the three germ layers: ectoderm, mesoderm, and endoderm. One particular example—the origin of hematopoietic stem cells in the mouse embryo—has been studied in detail, and it is known that cells capable of long-term repopulation of irradiated host animals are not present in the early embryo and only arise after about 10.5 days of development (8).

Although the term “progenitor cell” has been less debated and less clearly defined, it carries the implication of cells in a transient state, destined eventually to become one or more differentiated types. By this definition, all mitotic cells found in the early embryo are progenitor cells, including the in vivo precursors of ES cells, which are located in the embryo's inner cell mass or epiblast. The transit-amplifying cells of postnatal tissues in which there is a high level of cell turnover [“renewal tissues” (9)] are also progenitor cells. However, differentiated cells that divide slowly and simply reproduce themselves—such as pancreatic β cells or hepatocytes—might not be considered progenitor cells, as they are not progenitors to anything other than themselves. Definitions are of course arbitrary, and some authors prefer a definition of “stem cell” that is compatible with only a short-term existence. However, such a definition has the disadvantage of abolishing the otherwise useful distinction between stem cells and progenitor cells.

Hierarchy in Development

The formation of tissue-specific stem cells is an aspect of organogenesis. So what is the relationship between the development of a specific tissue rudiment and the formation of the stem cells? Regional specification in embryonic development proceeds as a hierarchy. Starting from the blastula (or blastoderm or epiblast, depending on the species), any particular tissue rudiment or cell type is formed by a sequence of developmental decisions. This process is now understood in great detail. As an example, the progenitors of the β cells of the mammalian pancreas undergo a series of steps of developmental commitment to become definitive endoderm, foregut endoderm, dorsal or ventral pancreatic bud, endocrine precursor cell, and finally β cell (10) (Fig. 1). At each step, the production of a particular combination of transcription factors is activated or repressed in response to a particular extracellular signal, which may be composed of one or more substances (called “inducing factors” in an embryonic context). Each step leads to multiple pathways constituting a developmental “choice.” Different concentrations of the inducing factor will result in the activation or repression of genes encoding different transcription factors, and therefore the adoption of different developmental pathways. Repeated occurrence of this process enables an early embryo, consisting initially of a simple mass of similar cells, to develop autonomously into an organism with a very complex pattern of structures.

Fig. 1.

The developmental hierarchy. This shows the normally accepted mode of formation of pancreatic β cells, involving six developmental steps, each controlled by one or more inducing factors. The inducing factors (NODAL, WNT, FGF, BMP, Notch ligands) are shown in color. The final step distinguishes the insulin-producing β cells from other types of endocrine cell also present in pancreatic islets (α, δ, ϵ, PP).

It is now clear that some tissues arise from more than one separate embryonic rudiment. This process is again exemplified by the pancreas, which arises from two endodermal buds in mammals (three in birds) that have somewhat different transcription factor combinations at initial formation but become very similar once differentiated (10). This shows that there can be more than one route, in terms of intermediate developmental states, from the original progenitor cells to the final differentiated cell phenotype.

So how do stem cells fit into this hierarchical scheme? I have argued that the stem cell state arises as the final developmental decision involved in creating a particular tissue type. This implies that tissue-specific stem cells will be in a similar state of developmental commitment to the embryonic rudiment that produced them (11) (Fig. 2). This hypothesis is not yet proved. To prove or refute it will require detailed comparisons of transcriptional profiles from pure embryonic rudiments and pure populations of the corresponding tissue-specific stem cells. Collection of this material is very challenging. However, a number of studies on mice indicate that key developmental transcription factors are essential both to the formation of a specific rudiment and to the maintenance of the resulting stem cell population. For example, Sox2 is a transcription factor that determines the properties of both the embryonic neuroepithelium and the postnatal neuronal stem cells (12, 13). Runx1 is required for both the initial formation of the hematopoietic tissue in the dorsal aorta and the maintenance of adult hematopoietic stem cells (14). Cdx2 is required to specify the intestine in the embryo as well as to maintain intestinal stem cell zones in the adult (15). p63 appears when embryonic epithelia become stratified and remains critical for controlling the properties of squamous epithelia and of keratinocyte stem cells (16, 17). Sox9 has recently been shown to be expressed in a specific cell population of the epidermal placodes that form hair follicles in the embryo. It is also required for maintenance of the postnatal stem cells, which are located in the bulge region of the follicle (18). [In this case, data from cell lineage labeling experiments (18) and grafting experiments (19) indicate the distinct origin of hair follicle and interfollicular stem cells.] One recent counterexample concerns the development of the anterior pituitary gland, where a nestin-positive cell derives independently of the corresponding embryonic progenitors and has a different transcription factor profile (20). This example again illustrates that the same specific differentiated cell type may arise through more than one developmental pathway. Whether any other tissues have stem cell populations distinct from the main embryonic rudiment is an open question.

Fig. 2.

Formation of stem cells in a tissue rudiment. In this hypothetical model, a few cells permanently retain their embryonic qualities because they lie in stem cell niches. Green represents cells in the embryonic tissue rudiment developmental state. Some of these are preserved as tissue-specific stem cells because of signals provided by neighboring mesenchymal cells (orange). These signals define stem cell niches. The rest differentiate to become a simple cuboidal epithelium, which later becomes a stratified epithelium maintained by production of cells from the stem cells. In such a tissue, labeling of a single stem cell (for example, by retroviral insertion) will label a domain of the tissue consisting of all the progeny of that stem cell.

To settle the precise origin of each specific stem cell population, it will be necessary to obtain better cell lineage data by labeling specific cell populations at the tissue rudiment stage, and then determining whether the stem cells are labeled at later stages. Such studies would normally use the CreER-lox method (21), which requires construction of transgenic mice containing two components: a gene encoding a hormone-activated DNA recombinase (CreER) driven by a tissue-specific promoter, and a reporter gene whose expression is activated by Cre-mediated excision of an inhibitory DNA segment. When the synthetic estrogen tamoxifen is given to the mice, the CreER becomes active and, in the cells where it is present, excises a segment from the reporter transgene, thereby activating production of the reporter protein and labeling the cells and all their descendants in a permanent manner (Fig. 3). The cells labeled are only those that had the tissue-specific promoter active at the time of the tamoxifen administration, rather than before or after. This is a very useful method, but it relies on high fidelity and specificity of the promoter used to drive the CreER. It is likely that some completely new method of cell lineage tracing will be needed to finally resolve such complex problems as the origin of tissue-specific stem cells. It should also be borne in mind that some tissues (such as the heart, many glands, and connective tissue structures) show very little self-renewal in postnatal life (9), and in such cases there may not be a stem cell population at all.

Fig. 3.

The CreER labeling method (21). This requires the production of mice containing two transgenes, as shown in yellow. TSP represents a tissue-specific promoter, so the CreER protein is present in all cells where TSP is active (green). When the synthetic estrogen tamoxifen is given to the mice, the CreER becomes active and can excise DNA sequences lying between loxP sites. This has the effect of activating production of a reporter gene, which is driven by UP (a ubiquitous promoter). The reporter remains active in all progeny of the labeled cells. If stem cells are labeled, then this will result in the permanent labeling of the whole tissue that they maintain.

The Problem of Pluripotent Adult Stem Cells

The conventional view of development as a hierarchy of decisions (Fig. 1) is generally accepted by researchers working with well-characterized tissue-specific stem cells or with ES cells. But there is another group of stem cell biologists who, explicitly or implicitly, do not accept the developmental biology consensus. Instead, they hold a view that all or most tissues in the body are continuously turning over and that the source of cells is one or more pluripotent stem cell populations—located in the bone marrow or elsewhere—that can circulate around the body and turn into multiple cell types, depending on the local environment. This view was expressed by Zipori (22), who started from the presumption that there are pluripotent adult stem cells throughout the body and preferred to define stem cell status by plasticity rather than by long-term self-renewal. Comparable models have been proposed by other authors (23, 24), who have also tended to be sympathetic to the idea that tissue-committed stem cells could “transdifferentiate” into other tissue types upon transplantation. This has now been experimentally disproved, except perhaps in very rare instances (25), although some very remarkable cell fusion events do occur after bone marrow transplantation and can involve profound respecification of gene expression by the donor cell nucleus (26, 27). Here, my intent is not to revisit the “transdifferentiation” debate but rather to consider the logically distinct proposition of whether postnatal pluripotent cells exist.

There have been numerous reports of pluripotent cells isolated from the postnatal organism. Such cells are reported to have a developmental potency much wider than that of tissue-specific stem cells and in some cases approaching that of ES cells. For example, mesenchymal adult progenitor cells (MAPCs) are described as cells isolated from the mesenchymal stem cells (MSCs) of mouse, rat, or human bone marrow that can be expanded without limit and are able to turn into most cell types of the body (28). Marrow-isolated adult multilineage inducible (MIAMI) cells are a population from human bone marrow reported to form neurons and pancreatic islet–like structures as well as the usual mesenchymal derivatives (29). Pluripotent stem cells (PSCs) can be isolated from many tissues of mice (30) and are reported to turn into muscle, adipose tissue, and neurons under suitable conditions. Tissue-committed stem cells (TCSCs) from the bone marrow of mice and humans (31) are reported to be derived from a precursor population of pluripotent cells capable of forming most cell types. MSCs from human adipose tissue (32) are described as fibroblastic cells that adhere to plastic and have been reported to turn into a wide variety of cell types in vitro, including neurons, cardiomyocytes, hepatocytes, and pancreatic cells. Skin-derived precursors (SKPs) are isolated from dermis of rodents and humans and grow as floating spheres in neurosphere medium (33). They are described as differentiating into neurons, glia, smooth muscle, or adipocytes under appropriate culture conditions. In addition to pluripotent cells from postnatal organisms, there are also many reports of similar cells isolated from fetal sources such as placenta, umbilical cord, amniotic fluid, or fetal tissues (3438).

Although some of these cell types have proved impossible to isolate in laboratories other than those producing the original work, the collective body of work is substantial enough to indicate that cells of wide potency can be grown in vitro from postnatal organisms. Possible in vivo precursors include the pericytes that are found on the outer surface of blood vessels (39). However, at present there is no definitive evidence that such cells exist in vivo. To prove that they do, it would be necessary to label one cell in situ and show that it formed a clone colonizing many tissue types. This could perhaps be done by means of retroviral labeling in vivo, a technique that can permanently label clones of cells at random (40, 41).

If pluripotent cells do not exist in the normal postnatal organism, then they must arise during in vitro culture. This presumably means that they spontaneously undergo changes similar to those seen during the formation of induced pluripotent stem (iPS) cells. These are cells that have been reprogrammed to a state similar to that of ES cells by the introduction of a small number of specific transcription factors (42, 43). Under the right conditions, overexpression of transcription factors known to be important in controlling the pluripotent state of ES cells (such as Oct4, Sox2, and Nanog) can reprogram differentiated cells to pluripotency in a permanent manner. However, iPS colony formation is a rare event, with a frequency between 10–2 and 10–6 depending on the starting cells and the particular genes introduced. This suggests that an unusual combination of stochastic events is necessary in addition to overexpression of the necessary genes. In the absence of overexpression of the pluripotency-inducing transcription factors, production of a pluripotent cell would require even more events to occur spontaneously and simultaneously, which must be very rare indeed. It may well be that such cells could be isolated by using highly selective growth conditions over months of culture. However, some pluripotent cells—for example, the SKPs—can be isolated more quickly and reproducibly than this [e.g., (44)], which suggests that they do have some sort of in vivo counterpart.

Because the explicit or implicit beliefs associated with postnatal pluripotent stem cells pose a direct challenge to the orthodox models of current developmental and stem cell biology (Fig. 4), it is worth considering possible explanations for their existence. The pluripotent cells might be some type of neural crest cell, as the neural crest is an embryonic cell population that does seem to undergo a more stochastic type of differentiation than other embryonic progenitor cells. Alternatively, they might be some kind of “embryonic remnant” comprising pluripotent cells left over from the early embryo.

Fig. 4.

The hypothetical postnatal pluripotent stem cell. The upper part represents the conventional account of development of four tissues; the lower part represents an entirely different view based on the existence of pluripotent adult stem cells. Not shown is the spatial pattern of the embryo, which is accounted for by the first model but not by the second.

The contention that some pluripotent adult stem cells are actually neural crest cells is not new. Indeed, several groups have isolated “neural crest stem cells” from various tissues: fetal sciatic nerve (45) and adult gut (46), heart (47), and skin (33, 48). Some aspects of the development of the neural crest follow the model of Fig. 1, especially its initial formation and the control of differentiation by inducing factors. However, the range of differentiated cell types that can be formed is very wide, and there seems to be an element of stochasticity in the differentiation process that is not so apparent in other parts of the embryo. The in vivo potency of individual neural crest cells has been studied by clonal labeling (4951). This shows that there are many multipotent cells in the crest before migration. During migration, their numbers and degree of multipotency decline, although at least some degree of multipotency is retained by some cells throughout the process. Neural crest cells become programmed to differentiate as a result of exposure to various inducing factors during their migration (52), but the stochastic nature of responses is shown by that fact that neighboring cells become committed at different times and can do different things.

There is evidence of a neural crest origin for at least two classes of relevant cell: SKPs and bone marrow MSCs. In the head of the mouse embryo, SKPs can be labeled by wnt1-Cre, a marker of cranial crest (53), and they also retain expression of various crest-type genes. They are at their most numerous in the late embryo, and numbers fall postnatally. The skin normally contains various neural crest derivatives, including melanocytes and Merkel cells, so it is not surprising that some precursors of these cell types may persist into the postnatal period. A labeling study that used the neural crest markers sox1-Cre and P0-Cre suggests that at least some of the MSCs found in murine bone marrow are of neural crest origin (54). Again the labeled cells declined in frequency after birth, and it seems that very few of the MSCs in bone marrow of older animals are neural crest–derived.

These labeling studies are suggestive rather than conclusive. Because Cre is expressed whenever its promoter is active, the cells would become labeled even if they had come from a non–neural crest source, so long as the appropriate promoter was activated at some stage during their formation. More definitive results could be obtained with the use of a hormone-inducible Cre, as discussed above (Fig. 3), that can confine the labeling to a specific developmental period (21).

The concept of embryonic remnants has a long history and was originally proposed as an explanation for the origin of cancer (55). There are many of types of visible embryonic vestiges that may be found in some individuals, such as remnants of Rathke's pouch, the thyroglossal duct, the urachus, the notochord, and so on (56). Smaller, and therefore less easily visible, cell clusters or persistent individual cells might be even more common and need not generate visible pathology. It is possible that some epiblast cells may lie beyond the reach of the earliest inducing signals and persist in an ES-like state. These might become incorporated into the amnion or extraembryonic endoderm, later finding their way into the bloodstream and returning to the embryo proper. They need not occur in all individuals and could be very rare when they do occur. They could be impossible to identify by conventional microscopy, yet be isolatable when a high degree of selection is imposed by tissue culture. With time, such cells would probably lose their pluripotent character or die off.

This would explain some of the less publicized properties of postnatal pluripotent stem cells: They tend to be difficult to isolate, are more common in younger animals, and show a lesser potency than ES cells. One argument against this view comes from the anatomical distribution of germ cell tumors (including teratomas). This type of tumor might be expected to arise from misplaced pluripotent cells. However, they are mostly found in the gonads and along the embryonic migration routes of the primordial germ cells rather than elsewhere in the body (57). Having said this, there are also some germ cell–type tumors arising in the brain (pineal and neurohy-pophyseal regions) that cannot easily be explained as originating from primordial germ cells (58).

In summary, the well-characterized stem cells in the postnatal organism are the tissue-specific stem cells. These are believed to arise in their respective rudiments during embryonic development as the end product of a hierarchy of decisions controlled by inducing factors, and are defined by the expression of a particular combination of developmental transcription factors. But to fully understand the origin of tissue-specific stem cells, we need better cell lineage data. Pluripotent stem cells can be isolated from postnatal organisms but may not exist in vivo, and if they do, they are not necessarily involved in normal development or tissue turnover.

References and Notes

View Abstract

Navigate This Article