The generation of endoderm derivatives, in particular pancreatic -cells and hepatocytes, has become the focus of many investigators in the field of ES cell biology. The interest in the efficient and reproducible development of these cell types derives from their clinical potential for the treatment of Type I diabetes and liver disease, respectively. Despite the interest in these lineages, progress in generating endoderm-derived cell types has been slow. The lack of progress in this area can be attributed to several different factors. First, several genes used as markers of definitive endoderm (Foxa2, Gata4, and Sox17) (Arceci et al. 1993; Monaghan et al. 1993; Sasaki and Hogan 1993; Laverriere et al. 1994; Kanai-Azuma et al. 2002), early liver (-fetoprotein and albumin) (Dziadek and Adamson 1978; Meehan et al. 1984; Sellem et al. 1984), and early pancreas (Pdx1 and insulin) (McGrath and Palis 1997) development are also expressed by visceral endoderm. Visceral endoderm, a population of extraembyonic endoderm, is an extraembryonic tissue that functions in a regulatory capacity but does not contribute directly to the formation of any adult organs (Fig. 1; Gardner and Rossant 1979; Thomas and Beddington 1996). Given the overlapping expression patterns, it can be difficult to distinguish definitive and extraembryonic endoderm in the ES cell differentiation cultures.
While ES cells do not contribute extensively to visceral endoderm in vivo following injection into blastocysts (Beddington and Robertson 1989), they do display some capacity to generate this population in culture (Doetschman et al. 1985; Soudais et al. 1995). The transcription factors nanog, GATA-4, and GATA-6 all appear to play some role in the regulation of visceral endoderm development from ES cells. ES cells lacking nanog (Mitsui et al. 2003) or those engineered to overexpress GATA-4 or GATA-6 (Fujikura et al. 2002) differentiate to visceral endoderm in the presence of LIF. Recently, Hamazaki et al. (2004) demonstrated that aggregation of ES cells in the presence of LIF is sufficient to down-regulate nanog expression and induce visceral endoderm development. With hES cell cultures, BMP2 has been shown to induce cells with visceral endoderm characteristics (Pera et al. 2004). The fact that ES cells can generate visceral endoderm in culture makes it imperative that one establish appropriate strategies for identifying definitive endoderm and its derivatives. Approaches to address this issue could include the identification of genes that are differentially expressed between the endoderm populations and the establishment of conditions that specifically support the development of definitive endoderm.
A second problem encountered in endoderm differentiation from ES cells is the lack of specific inducers of this lineage. In most studies to date, endoderm development has been analyzed in ES cell cultures differentiated in the presence of FCS. While FCS appears to efficiently induce mesoderm populations, particularly the hematopoietic, vascular, and cardiac lineages, findings from a recent study demonstrate that it is not the optimal inducer of endoderm (Kubo et al. 2004).
In spite of these obstacles, several studies have provided evidence for the generation of endoderm-derived cell types including populations that display characteristics of pancreatic islets (Colman 2004; Stoffel et al. 2004), hepatocytes (Hamazaki et al. 2001; Jones et al. 2002; Yamada et al. 2002a), thyrocytes (Lin et al. 2003), lung (Ali et al. 2002), and intestinal cells (Yamada et al. 2002b). This review focuses on pancreatic and hepatocyte development.
Pancreatic development from ES cells
Two general protocols have been used for the generation of pancreatic islet-like cells from ES cells in culture. The first, a five-step protocol, described by Lumelsky et al. (2001), is an adaptation of a protocol for neural cell differentiation and involves the transfer of serum-induced EBs to serum-free medium followed by treatment with FGF and factors that promote maturation of endocrine cells. Clusters resembling pancreatic islets developed in these cultures and cells within the clusters were shown to contain insulin, glucagon, or somatostatin by immunohistochemistry. While the insulin content of the cells was low, they were able to secrete insulin in response to glucose. However, when tested for their ability to function in vivo following transplantation into streptozotocin (STZ)-induced diabetic mice, these cells failed to correct the hyperglycemia of these animals. The inability to "cure" these animals could be due to the fact that the cells were too immature or that they were not islet cells.
Several other groups have modified this approach and reported improved development of -like cells in the cultures. The modifications included constitutive expression of the transcription factor Pax4 (Blyszczuk et al. 2003) during differentiation, inducible expression of Pdx1 (Miyazaki et al. 2004) during differentiation, or exposure of the cells to the inhibitor of phosphoinositide 3-kinase (PI3K), LY294002 (Hori et al. 2002), at the final maturation step. Pdx1 and Pax4 are essential for -cell development in vivo (Murtaugh and Melton 2003). Transplantation of the Pax4-induced and LY294002-treated cells into STZ-treated recipients did improve the hyperglycemia in these animals, suggesting that ES-derived cells were producing insulin. Of significance was the observation that removal of the LY294002-treated graft resulted in reversion to a diabetic state, indicating that the observed improvement was due to the implanted cells. While these findings are encouraging, several issues remain to be resolved regarding the types of cells generated and the amount and type of insulin produced in these cultures.
Rodents contain two nonallelic insulin genes, insulin I and insulin II (Melloul et al. 2002). While insulin I is expressed predominantly in -cells, insulin II is found in the yolk sac and developing brain in addition to the pancreas (Deltour et al. 1993; Devaskar et al. 1993; Giddings et al. 1994). Given that the ES cell cultures described above do contain significant numbers of neuronal cells, it is possible that some of the insulin detected is derived from these cells, as demonstrated by the recent study of Sipione et al. (2004). A second issue relates to the observation that ES-cell-derived cells can absorb insulin from the tissue culture medium and then release it when stressed or undergoing apoptosis (Rajagopal et al. 2003). It is unclear to what extent neuronal-derived insulin or the release of absorbed insulin is involved in the above studies as the cultures contain mixed populations of cells. Resolution of these issues and the development of large numbers of functional insulin-producing cells will require improved protocols for endoderm induction and -cell specification and maturation.
Some of the concerns relating to -cell development in ES cell cultures were addressed in a recent study of Ku et al. (2004). In this work, modification of the components of the serum-free culture and maturation of the differentiating cells in the presence of activin B, exendin-4, and nicotinamide, resulted in the development of cell populations that expressed insulin I, following the appearance of cells that expressed various genes indicative of endoderm differentiation. Immunostaining revealed that cells within the cultures also contained c-peptide, the cleavage by-product of proinsulin, indicating that at least some of the insulin detected is produced by the cells. The modification to the maturation cultures significantly increased the levels of insulin I mRNA as well as the frequency of cells that express insulin (to 2%–3% of the total) in the population. These findings are encouraging and offer hope that further modifications together with selection approaches will yield populations highly enriched for -cells.
A second approach for the generation of islet-like structures relies on their development in heterogeneous populations derived from ES cells following serum induction. Such approaches have yielded, at a low frequency, cells that display many characteristics of -cells, including the presence of c-peptide and insulin in discrete granules within the cells (Kahan et al. 2003). To enrich for -cells from heterogeneous serum-induced cultures, Soria et al. (2000) developed a selection strategy based on the insulin promoter driving expression of the selectable gene -geo. Addition of G418 to the cultures resulted in the isolation of an insulin-secreting clone that had insulin contents similar to that of normal islets. When injected into STZ diabetic mice, these cells corrected the hyperglycemia of these animals for up to 12 wk. Beyond this time, however, a significant number of animals reverted back to a hyperglycemic stage, indicating that the cells were unable to provide long-term -cell function.
As a strategy to select for early pancreatic cells as they develop in ES cell differentiation cultures, Micallef et al. (2005) targeted GFP to the pdx1 locus. Pdx1 is expressed in the earliest stages of pancreatic development as the organ rudiments are specified from the gut endoderm (Murtaugh and Melton 2003). Treatment of the developing EBs with retinoic acid in this study led to the development of a small GFP-pdx1+ population by day 8 of differentiation. Analysis of the isolated GFP-pdx1+ cells revealed that they expressed endoderm-specific genes but not genes associated with pancreatic maturation, suggesting that they represent an early stage of endoderm development. Alternatively, some of the cells within the population could be visceral endoderm, a tissue that also expresses pdx1 (McGrath and Palis 1997). Further studies will be required to demonstrate that this population does represent the earliest stages of pancreatic differentiation. The generation of the pdx1 selectable marker is a good strategy that will enable the quantitation and isolation of pancreatic lineage progenitors from cultures induced under optimal conditions for definitive endoderm development.
Insulin-expressing cells have also been detected in hES cell differentiation cultures (Assady et al. 2001). However, as with many of the studies with mouse cells, the frequency of these cells in the hES cell cultures is currently too low to allow detailed characterization and functional analysis.