Both these criteria are, however, relative. There are unipotent self-renewing cells, most notably germline stem cells, which most scientists would argue are obvious stem cells. Hence, the existence of such unipotent stem cells disqualifies multipotency as an obligate stem cell criterion. Neither is the feature self-renewal absolute. Some stem cells may undergo self-renewal during a short window of time, whereas others may do so for the lifespan of the organism. A limited period of self-renewal is characteristic for transitory developmental stages, for example for inner cell mass cells in the blastocyst. These cells can be propagated indefinitely in vitro as embryonic stem cells, indicating that the limited self-renewal in vivo is not due to a cell intrinsic constrain, but rather to the loss of an environment that can support sustained self-renewal. In contrast to the inner cell mass cells, adult stem cells reside in stable niches in the mature tissue and may self-renew over the lifespan of the organism.

The problem with multipotency and self-renewal as defining characters of stem cells is not limited to these criteria being relative. Perhaps even more troublesome is that neither multipotency nor self-renewal is a feature unique to stem cells. First, stem cells often give rise to multipotent progenitor cells that have no or limited self-renewal. What distinguishes the multipotency of stem cells from that of progenitor cells? Second, unipotent, bipotent and multipotent progenitor cells can often self-renew for quite a number of times (Back et al, 2004; Trentin et al, 2004). Even differentiated cells, such as B and T lymphocytes in response to antigenic stimuli (Fearon et al, 2001) as well as pancreatic β-cells (Dor et al, 2004), can renew themselves.

Let us compare this view with the classical definition of a stem cell as a multipotent self-renewing cell. In the novel model, the multipotency and self-renewal features are not specific. For example, transit amplifying progenitor cells may be multipotent and self-renew, but they are not halted at an intermediate position, as they inevitably progress toward terminal differentiation. The unique feature of a stem cell in this view is the resistance to progress along the line of specialization, which cannot be reduced to the features multipotency or self-renewal capacity.

The tight interdependence between a stem cell and its immediate environment indicates that the key to the stem cell state may be found in the signals that are provided by the niche. Different niches could potentially share signals that maintain the stem cell's state preventing its further differentiation and allowing asymmetrical cell divisions to self-renew and generate progenitor cells in response to cues induced by a deregulated tissue homeostasis. Several extracellular signaling pathways such as members of the WNT, hedgehog, bone morphogenic protein (BMP) and NOTCH families indeed have important functions in multiple stem cell lineages (Fuchs et al, 2004).

The regulated accessibility of genes appears to be a critical feature in the stem cell state. The so-called ‘multilineage priming' model suggests that stem cells express low levels of a large number of genes that are highly expressed in its different committed progeny (Cross and Enver, 1997). Along the course of differentiation, the number of cell fate options available to the differentiating cell becomes more and more restricted as genes attributing to the selected lineage are activated and genes associated with other lineage are shut down (Figure 2). Comparison of the expression profiles of hematopoietic cells at different stages of differentiation demonstrated that stem cells express a large number of genes highly expressed by different specific hematopoietic lineages, whereas committed precursors only express the proportion of the genes that are related to their lineage (Akashi et al, 2003). In addition, hematopoietic stem cells express a relatively large number of non-hematopoiesis-associated genes, in particular neuronal genes. Although most of these neuronal genes have not been linked to the induction of neuronal differentiation, and their neuronal exclusiveness remains elusive, it illustrates the wide variety of genes expressed in hematopoietic stem cells. A similar situation seems to hold for embryonic stem cells. Embryonic stem cells, which have the tendency to acquire a neural identity, express several neuronal lineage-specific genes under normal culture conditions (Tropepe et al, 2001).

Thus, it appears that differentiation is accompanied by a successive restriction in the repertoire of genes that can be expressed, and that the stem cell state is characterized by accessibility to genes for downstream fates. This implies that there is a close relation between differentiation potential and chromatin accessibility, and that multipotency at the transcriptional level is dependent on having access to several differentiation programs.

Chromatin modifications maintain chromatin status through DNA replication and mitosis. In particular, methylation of the lysine residues in histone 3 appears to play a role in epigenetic memory. These histone modifications recruit polycomb or trithorax group protein complexes to maintain the silencing or activation status of genes, respectively (for review, see Ringrose and Paro, 2004). Since stem cells self-renew without losing any of their features, they should inherit the same epigenetic status in order to maintain their characteristics. From this viewpoint, it is not unexpected that polycomb group members, such as Bmi1, play a role in the maintenance of several stem cell populations and affect self-renewal (Park et al, 2003) and stem cell memory (Iwama et al, 2004).

Although there often are clear limitations in the power of the milieu over the cellular state, there are several examples of a remarkable reprogramming activity of the environment. One striking example is that of ectoderm-derived cranial neural crest cells delaminating from the neural tube, which are induced by local cues to form myocytes, cartilage and bone cells, indistinguishable from such cells derived from mesoderm (Trainor et al, 2002).

Cell transplantation experiments during development and in adults have delineated the determination of cells to specific lineages and the possibility of an altered environment to influence this. An alternative approach to investigate the relative role of the environment and the cell state is to change the molecular signals in a niche. A recent study demonstrated that an experimentally overactive WNT pathway induced progenitor cells in the lung to take on an intestinal phenotype (Okubo and Hogan, 2004).

In the last years, numerous reports have reported lineage plasticity of stem and progenitor cells (Lakshmipathy and Verfaillie, 2005) (Figure 3). However, the mechanisms responsible for the observed plastic behavior have remained a controversial issue. It appears that, dependent on the tissue and experimental settings, de- or transdifferentiation as well as cell fusion can explain the occurrence of transdetermination of stem cells (Frisén, 2002; Weimann et al, 2003; Wurmser et al, 2004). On the basis of the ‘multilineage priming' model, transdetermination would be by far simpler at an early stage of differentiation, when numerous genes belonging to several differentiation programs are readily available, than later on in terminally differentiated cells.