The only vertebrate from which pluripotential stem cells from early embryos that can form germ as well as all somatic cell types have been obtained is the mouse. Moreover, even in this species success has been achieved only with certain strains [29]. Cell lines derived from morulae or blastocysts of various other mammals have given somatic but not germline chimaerism following blastocyst injection or morula aggregation. Those from yet other species have so far altogether failed to give any discernible chimaerism [29]. This obvious heterogeneity in properties and potential of cells derived at similar stages in early development from different species raises questions about the appropriateness of assigning them the common status of ES cells. In the case of the mouse, it is clear that what may be termed ‘true’ ES cells originate from the epiblast, the progenitor tissue of both the foetus and various extra-embryonic structures [30]. Murine epiblast is permissive for ES cell derivation only during an extremely restricted interval, which seems to correspond rather precisely with the period during which its cells can colonize the blastocyst [31]. The basis of this restriction is not clear, especially since embryonal carcinoma cells obtained from ectopic grafts of more advanced epiblast can yield chimaeras following blastocyst injection [32]. Most intriguingly, even cells from adults can sometimes give generalized chimaerism by this route [15].

Although ES lines can be obtained efficiently from cleavage stages and morulae in the mouse, these do not differ discernibly in properties and developmental potential from those of blastocyst origin [33]. It seems, therefore, that regardless of the stage at which pre-implantation conceptuses are explanted, cells may have to attain the status of nascent epiblast in order for their continued division to be uncoupled from further differentiation. The generally low success rate in deriving human ES cell lines may relate, at least in part, to the unavoidable dependence as starting material on conceptuses developing in vitro. In vitro developed blastocysts are inferior to in vivo in yielding ES cells the mouse, [34], though the adverse effect of culture can be partially mitigated by sequential use of different media [35]. The latter may also aid the derivation of human ES cells [36].

How well do stem cells fare during prolonged passage in vitro? This is less easy to answer than is often supposed even for ES cells, although it is commonly asserted that they can be propagated indefinitely in culture without loss of potency. This assertion is made especially with regard to human ES cells that, largely on the basis of retention of a normal karyotype, are often held to withstand prolonged passage in vitro better than their murine counterparts do. Failure of murine ES cells to give chimaerism by blastocyst injection following prolonged passage has been attributed to acquisition of epigenetic changes that cannot be reversed by re-programming their nuclei via what has come to be known as ‘therapeutic’ cloning [37]. However, in this, as in many other studies, the cells were assumed to have retained euploidy on the basis of chromosome counts rather than detailed karyological analysis. According to our own recent experience with mouse ES cells, even lack of karyotypic change may not be a reliable indicator of normality. Thus, although strictly monoclonally derived as opposed to presumptively polyclonal mouse lines can remain entirely euploid for up to 40 or more passages, and show no discernible genomic rearrangements, they do not necessarily behave normally when assessed by blastocyst injection. Notably, a euploid monoclonal line derived from a blastocyst which, at low passage, yielded respectable chimaerism on blastocyst injection, not only failed to give chimaerism at high passage but disrupted the development of host embryos (F.A. Brook, T.J. Davies, E.P. Evans and R.L. Gardner, unpublished observations). Curiously, a few weak postnatal chimaeras were obtained if the number of cells injected into each blastocyst was reduced from the usual 10–15 to just 3. Since, this monoclonal line showed no genomic changes when analysed by representational oligonucleotide microarray analysis at a resolution of ca. 30 000 base pairs (Ira Hall, personal communication of unpublished observations) a likely, but as yet unproven, explanation for its altered behaviour is that it has indeed undergone stable deleterious epigenetic change [38]. Since this is not seen with all monoclonal lines at high passage, we are at present attempting to ascertain whether such adverse change correlates with restriction in the capacity of the cells to differentiate in vitro. This could help inform studies concerned with critical evaluation of human ES lines that are destined for clinical use in regenerative medicine for which, unless done interspecifically, testing by blastocyst injection is not an option. So far, however, the results of injecting human ES cells into mouse blastocysts have not proved very informative [39]. Moreover, enduring intra-specific chimaerism with ES cells in non-human primates has yet to be demonstrated [40].

The haematopoietic stem cell from adult bone marrow is still the most thoroughly characterized stem cell. Putative stem cells from various other adult tissues and organs need to be assessed against this yardstick before contemplating their use for regenerative purposes. This is because it is conceivable that cells that would never have cycled again if left in situ may be induced to engage in extensive proliferation following isolation and exposure to a cocktail of growth factors they would not have encountered in vivo. Consequently, sustained growth in vitro in the absence of overt differentiation may not be unique to stem cells. Nonetheless, the case has been made for having identified stem cells from a variety of tissues, and this has been aided by finding an increasing number of molecular markers that are shared between such cells.

A further important issue, regardless of whether the stem cells are of embryonic, foetal or adult stem origin, is establishing that cells into which they are induced to differentiate are normal. While it may be convenient to assess differentiation in terms of expression of specific marker molecules, this is no substitute for devising rigorous physiological assays. A particular compelling example of such an assay is the grafting of foetal hypothalamic tissue into the brains of adult mice that fail to mature sexually through homozygosity for deficiency of gonadotrophin releasing hormone. Such grafts prompted rapid sexual maturation and reversal of infertility [41,42]. Obviously, for human cells the scope is limited to assays based on in vitro culture versus xenografts, the latter posing the problem of choice of appropriate graft site [43,44], as well as host species.

As well as ascertaining that differentiated cells are physiologically normal, there is the further question of whether it is important to ensure that they are homogeneous – i.e. that only cells of the desired type are included in grafts. The need to eliminate residual stem cells from differentiated cultures of ES origin is generally accepted because of their exceptional tumorigenicity at low inoculum size [43,45]. While injection of undifferentiated ES cells into articular cartilage defects produced repair without any tumours being detectable eight weeks thereafter [46], presumed contaminating ES cells yielded tumours after a similar interval following sub-retinal grafting of ES-derived neural precursor cells [47]. It is possible that interspecific grafts used in the cartilage study are less susceptible to tumorigenesis than interspecific grafts used in the retinal one. This may also be true of allogeneic as opposed to syngeneic grafts. Murine ES cells transplanted locally into the brains of host mice showed down-regulation of expression of stem cell markers following multiplication and migration towards sites of experimentally-induced injury [48]. However, no evidence was provided that cells had differentiated appropriately, and the interval between transplantation and analysis was too short to exclude their participation in tumorigenesis.

Whether inclusion of some inappropriately differentiated cells as opposed to persisting stem cells in grafts really matters is not clear. It is interesting, in this context, that no prior differentiation or selection of cells was undertaken in the successful use of bulge stem cells from hair follicles to re-myelinate damaged sciatic nerve in rats [49]. If purity of cell populations for grafting is indeed necessary, it is questionable in view of public mistrust of genetic modification whether engineering cell type-specific expression of an antibiotic resistance gene will prove an acceptable approach for achieving this in a clinical situation. Use of flow cytometric methods offers an alternative providing, of course, that a suitable specific marker is available [50]. Yet another strategy is to exploit the differential sensitivity of ES cells relative to their differentiated derivatives to various agents such as ceramide analogues [45].

A further important issue is whether grafts are intended to make good loss of a population of cells that are terminally differentiated and thus entirely post-mitotic or are required to self-renew. In the latter case, appropriate stage stem cells need both to be present in the graft and to persist thereafter. The notion of dependence of stem cell on specific niches for their maintenance is now familiar [51], particularly as a result of elegant genetic analysis of germline stem cells of both sexes in Drosophila [52,53]. At present, rather less is known about conditions required for the maintenance of stem cells of somatic tissues in mammals. It is, however, encouraging that relatively simple conditions have been found to support sustained in vitro propagation of various types of stem cells in addition to ES cells. These include neural stem cells, and especially those of epidermis, hair follicles, and cornea [18,20,54]. In the case of the latter three types of stem cells, conditions of culture that are used presently for regenerative medical purposes are essentially as defined many years ago by Howard Green and his colleagues [55]. Interestingly, these include use of gamma-irradiated murine 3T3 feeder cells and bovine serum [18], components that many argue need to be avoided in culturing human ES cells lines intended for clinical use. Moves to the use of same species feeders for human ES cells have already been made (e.g., [56]).

As noted earlier, wherever practicable, autografts are the obvious choice, although there are at present rather few situations where a suitable source of stem cells is available from the patient. Obviously, if efficient re-programming of stem cells from skin or bone marrow directly rather than via nuclear transplantation proved feasible, the scope for autografting would be enormously enhanced. There is, however, evidence relating to skeletal myoblasts to suggest that cells may be altered by culture so as to render them susceptible to destruction by NK cells even when grafted syngenetically [57].

A further possible way of avoiding allografting, which has been much discussed, is to resort to what has rather prematurely been termed ‘therapeutic cloning’. This technique will unquestionably aid the study of human genetic diseases in vitro in the types of differentiated cells in which they are normally manifest (e.g., [58]). However, there is division of opinion within the biomedical community as to whether this is a realistic option clinically. Aside from the low efficiency of obtaining cloned blastocysts for deriving patient-specific ES cell lines and the consequent heavy demand for oocytes, questions remain regarding the likely normality of the resulting cells with respect to genomic imprinting. Hence, for the foreseeable future, the continued widespread use of allografting seems almost inevitable. Therefore, one pertinent consideration is whether tissues from some types of stem cells are likely to fare better as allografts than those from others. Another is the prospect of achieving graft-specific states of immunological tolerance.

A recent study in mice provided no support for the notion that the differentiated derivatives of ES are less susceptible to rejection than other types of grafts, even where there were only minor as opposed to major histocompatibility differences between donor and host (N. Robertson, F. Brook, R.L. Gardner, H. Waldmann and P. Fairchild, unpublished observations). However, the situation may be different for mesenchymal stem cells and, as noted earlier, also for limbal stem cells of the cornea. A growing number of studies attest to the possibility that mesenchymal stem cells enjoy a special status immunologically. This is suggested most cogently by the production of widespread enduring chimaerism in sheep following introduction of human mesenchymal cells into immunologically competent foetuses [59]. However, while numerous in vitro studies argue for active interference of these cells with the immune response rather than their failure to express histocompatibility antigens, there is still no general consensus on how this is achieved [60–64]. More significantly, contradictory results have been obtained regarding the lack of immunogenicity of these cells in in-vivo studies [65–67]. Most interestingly, successful allografting of developing metanephroi has been achieved where corresponding grafts of adult kidneys failed [25].

Further areas of active exploration to avoid recourse to immuno-suppression, are induction of tolerance to grafts or their protection from humoral and cell-mediated immune attack, either by their artificial encapsulation or by their location in immunologically privileged sites. Inducing tolerance may possibly prove easier with tissues derived from ES cells than from other sources (N. Robertson, F. Brook, R. L. Gardner, H. Waldmann and P. Fairchild, unpublished observations).

Grafts that are designed to restore specific endocrine functions can presumably be placed other than at the site of damage since, with few exceptions, their location is unimportant providing they become sufficiently well vascularised to work effectively. Introduction of dissociated cells directly into the vascular system may be sufficient to achieve the desired outcome where appropriate homing of transplanted cells can be effected by this route. Where grafts need to be placed specifically at the site of damage, e.g., in countering neurodegenerative diseases involving extracellular deposition of insoluble protein aggregates, there is the question of whether they will function any better than the native cells they are designed to replace unless a more normal environment is restored. Indications from several recent studies are encouraging with regard to the central nervous system in suggesting that transplanted stem cells may not only survive at sites of damage, but also actively migrate towards them [48]. The situation is less clear for damaged myocardium where it may be more pressing to obtain rapid neovascularization if grafts are to survive.

In some cases, the extracellular matrix of damaged tissue may remain sufficiently well ordered to provide the cues necessary to allow transplanted cells to become organized appropriately to restore normal function. Otherwise, as touched on earlier, it may be necessary to attempt to obtain organogenesis in vitro, which means achieving defined in vitro arrangements of cells in three dimensions rather than just two. This obviously depends on use of scaffolds for which a wide variety of materials of both biological and non-biological origin is currently being evaluated. Many challenges confront tissue engineering, a rapidly developing field that has encouraged very productive interdisciplinary collaborations between the physical and biological sciences.