Translational embryology: Using embryonic principles to generate pancreatic endocrine cells from embryonic stem cells


  • Jason R. Spence,

    1. Division of Developmental Biology, Cincinnati Children's Hospital Research Foundation, Cincinnati Ohio
    Search for more papers by this author
  • James M. Wells

    Corresponding author
    1. Division of Developmental Biology, Cincinnati Children's Hospital Research Foundation, Cincinnati Ohio
    • Division of Developmental Biology, Cincinnati Children's Hospital Research Foundation, 3333 Burnet Avenue, Cincinnati, OH 45229-3039
    Search for more papers by this author


Diseases that affect endodermally derived organs such as the lungs, liver, and pancreas include cystic fibrosis, chronic hepatitis, and diabetes, respectively. Despite the prevalence of these diseases, cures remain elusive. While several promising transplantation-based therapies exist for some diseases such as Type 1 diabetes, they are currently limited by the availability of donor-derived tissues. Embryonic stem cells are a promising and renewable source of tissue for transplantation; however, directing their differentiation into specific, adult cell lineages remains a significant challenge. In this review, we will focus on one endodermally derived organ, the pancreas, and discuss how studies of embryonic pancreas development have been used as the basis for the directed, step-wise differentiation of mouse and human embryonic stem cells into pancreatic endocrine cells that are capable of rescuing Type 1 diabetes in animal models. Developmental Dynamics 236:3218–3227, 2007. © 2007 Wiley-Liss, Inc.


The endoderm germ layer gives rise to the epithelial lining of the digestive and respiratory tracts, as well as organs such as the thyroid, lungs, liver, gall bladder, and pancreas. Diseases involving endodermally derived organs, particularly the lungs, liver, and pancreas include cystic fibrosis, chronic hepatitis, and diabetes, respectively, affect tens of millions of people in North America. An emerging approach to treating patients with a disease is to stimulate regeneration of damaged tissue in vivo or to generate replacement tissue ex vivo. This broad area of research is often referred to as regenerative medicine. Stem cells have effectively been used to treat diseases of mesodermally derived organs (the hematopoietic system), and ectodermally derived tissue (skin) has been grown and expanded in culture for the treatment of burn victims. Given these successes, it is likely that regenerative medicine will prove similarly effective to treat diseases of the endoderm.

Many existing transplantation-based therapies are currently limited by the availability of donor-derived tissues. One key goal of regenerative medicine is to identify a renewable source of cells and tissues for replacement therapies. Embryonic stem cells (ESCs) represent a promising source of material for transplantation because of their unique properties. ESCs are primary cells that can be expanded indefinitely in culture and, thus, are a renewable source of tissue. In addition, ESCs are capable of becoming all cell types of the body and, thus, could be used to generate cells for replacement therapy of any diseased or damaged cell. One significant challenge in using ESCs therapeutically is to identify methods that promote their differentiation into specific and functional adult lineages.

Over the past 2 decades, developmental biologists have made significant progress in identifying the molecular basis of pancreas development. Studies in model organisms, including Xenopus, zebrafish, chick, and mouse, have identified signaling pathways and downstream transcriptional networks that direct cell fate decisions throughout the formation of the pancreas. In this review, we will briefly highlight several key stages of pancreas development and several molecular pathways involved in these cell fate decisions (Fig. 1; for reviews that focus on endoderm and pancreas development, see Slack,1995; Wells and Melton,1999; Lewis and Tam,2006; Murtaugh,2007; Zorn and Wells,2007). We will then discuss how these embryonic principles have been used as the basis for the directed, step-wise differentiation of mouse and human embryonic stem cells into specific cell types in culture, in particular, pancreatic endocrine cells (Fig. 1). There has been great progress made over the past few years in directing ESCs into this lineage, and we will discuss some hurdles that remain in the way of creating fully functional pancreatic β-cells that could be used therapeutically.

Figure 1.

The lineage of the developing pancreas in vivo and in embryonic stem (ES) cultures. The left panels are mouse embryos at different stages of development (embryonic day 3.5 through 13.5 [E3.5–E13.5]), and the black bars on the right indicate the equivalent stages in human development. Embryos are oriented with anterior to the left and posterior to the right. The E13.5 stage shows the dissected stomach, pancreas, and duodenum with the stomach (s) and dorsal pancreas (dp) to the left and the ventral pancreas (vp) and duodenum (d) to the right. The LacZ staining in the E9.5 and 13.5 embryos shows expression of Pdx1 (Pdx1LacZ/+ animals were from Chris Wright at Vanderbilt University and were described in Offield et al.,1996). The lower left panel is a pancreatic islet showing glucagon-expressing α-cells (green) and insulin-expressing β-cells (red). The curved arrows highlight several signaling pathways involved pancreas development that have been used to direct ESCs into the pancreatic lineage. The middle panel indicates the lineage of the developing endocrine pancreas. The arrowheads in the cell lineage diagram indicate two separate roles of Nodal/Activin signaling: to initiate gastrulation and to promote endoderm vs. mesoderm fate in a dose-dependent manner.


Endoderm Formation

During vertebrate development, the morphogenetic process of gastrulation results in a group of undifferentiated cells forming the three principal germ layers: ectoderm, mesoderm, and endoderm. There is increasing evidence from studies in fish, frogs, and mice, that the endoderm and mesoderm share a common progenitor referred to as mesendoderm. Cell lineage analyses of early to mid-stage mouse gastrula indicate that mesendoderm progenitor cells are concentrated at the anterior primitive streak near the node and that they adopt either a mesoderm or endoderm fate as they migrate through the primitive streak and incorporate into their respective germ layers (Lawson and Pedersen,1987; Lawson et al.,1991). However, the exact nature of when and how endoderm cells acquire their identity during gastrulation is not well understood.

The node is the source of Nodal, a member of the transforming growth factor-beta (TGFβ) family of growth factors. The nodal-related growth factor signaling pathway is required for establishing the anterior–posterior axis by restricting gastrulation to the posterior of the embryo (Conlon et al.,1994; Brennan et al.,2001; Perea-Gomez et al.,2002). Nodal subsequently plays a role in promoting mesoderm and endoderm formation (for a review, see Ang and Constam,2004; Zorn and Wells,2007). Studies in frogs, fish, and mice indicate that the dose of nodal signaling directs a mesendoderm cell into either the endoderm or mesoderm lineage. For example, the use of hypomorphic nodal alleles in mouse revealed that high levels of nodal signaling are required for an endoderm fate whereas lower levels promote a mesoderm fate (Tremblay et al.,2000; Lowe et al.,2001; Vincent et al.,2003). These data suggest that nodal dose is important in endoderm vs. mesoderm fate. Consistent with this finding, endoderm cells arise from the anterior primitive streak that is close to the node, which is the principal source of nodal (Lawson and Pedersen,1987; Lawson et al.,1991). Presumptive endoderm cells may also be exposed to higher levels of nodal ligand as they traverse the primitive streak and come into closer proximity with the node during their anterior migration. Given that Nodal activity is additionally regulated through proteolytic processing, secreted antagonists, and by the presence or absence of receptor complexes (Bouwmeester et al.,1996; Ding et al.,1998; Roebroek et al.,1998; Constam and Robertson,2000; Perea-Gomez et al.,2002), it is clear that establishing the correct dose of Nodal is a critical step in specifying endoderm.

There is also evidence that the Wnt signaling pathway is involved in the mesoderm vs. endoderm cell fate choice. Embryos lacking β-catenin, a key effector of the canonical Wnt pathway, have ectopic mesoderm cells forming in the endoderm germ layer (Lickert et al.,2002). The transcription factors that act downstream of the nodal and Wnt signals to direct endoderm formation are also remarkably conserved across vertebrate species and include Mix-like homeodomain proteins, Gata zinc finger factors, Sox HMG (high mobility group) domain factors, and Fox forkhead domain factors. In particular, Sox17 paralogs are required for normal endoderm development in all vertebrates analyzed (for review, see Zorn and Wells,2007).

Foregut Development and Pancreas Specification

Once formed, the endoderm germ layer in mouse is a simple squamous epithelium (embryonic day [e] 7.5). Within 24 hr, morphogenetic movements in the anterior region of the endoderm result in the formation of the foregut and at this stage the endoderm cells are a cuboidal epithelium (e8.5 in mouse). The foregut will give rise to the thyroid, lungs, liver, stomach, and pancreas. The dynamic nature of foregut morphogenesis brings the endoderm into proximity with several mesodermal tissues that provide patterning signals to establish the presumptive organ domains within the foregut. Signaling pathways that have been implicated in foregut formation and patterning include the fibroblast growth factor (FGF), Wnt, retinoic acid (RA), and hedgehog pathways. These pathways are known to regulate the expression of key transcription factors, including Fox/HNF, ParaHox, and Hox factors, which are important mediators of cell fate (Grapin-Botton,2005).

Formation of the foregut depends on proper anterior–posterior (A-P) patterning of the endoderm (Wells and Melton,1999). Experiments in mouse and frog have demonstrated that gastrula stage endoderm cells are unspecified, but will adopt an A-P fate in response to signals from mesoderm (Wells and Melton,2000; Horb and Slack,2001). Subsequent studies identified that FGF and Wnt signaling pathways, which are restricted to the posterior at this stage of development, are key modulators of A-P patterning (Dessimoz et al.,2006; McLin et al.,2007). These pathways act to posteriorize endoderm at the gastrula stage and excluding FGF and Wnt signaling from the anterior is necessary for proper development of the foregut.

As morphogenesis of the foregut commences, it has been shown that additional signaling pathways are involved in establishing the pancreatic domain within the developing gut tube. These pathways include the FGF, RA, and hedgehog (HH) signaling pathways. It is suggested that RA signaling is important for the global patterning of the foregut endoderm, because alterations in RA signaling cause defects in numerous foregut derivatives, including the lung, the stomach, and the pancreas. In particular, development of posterior/dorsal foregut derivatives like the stomach and dorsal pancreas are severely perturbed in embryos lacking active RA signaling due to absence of the enzyme retinaldehyde dehydrogenase 2 (Raldh2 knockout; Martin et al.,2005; Molotkov et al.,2005; Wang et al.,2006b). RA signaling in the foregut was shown to act upstream of key transcription factors, including the Pancreatic duodenal homeobox factor 1 (Pdx1) and several Hox genes, again suggesting that RA-mediated signaling functions to pattern the foregut before specification of the pancreatic domain (Kumar et al.,2003). In addition to Pdx1, which is necessary for proper pancreas development (Jonsson et al.,1994; Offield et al.,1996), there are several other transcription factors involved in foregut patterning and early pancreas development, including HNF6 (onecut), HNF1β, Hlxb9, and Ptf1a/p48 (for review, see Jensen,2004; Murtaugh,2007).

Hedgehog signaling is involved in global patterning of the gut tube and in early pancreatic development (Hebrok et al.,1998,2000; Kim and Melton,1998; Ramalho-Santos et al.,2000). During the early stages of gut tube development (e8.5 to e9 in the mouse), Sonic hedgehog (Shh) is broadly expressed by the endoderm of the developing gut tube along the A-P axis. Coincident with the start of pancreas development, Shh expression is repressed in the dorsal and ventral pancreatic buds. The repression of Shh in the dorsal bud depends on signals from the notochord that include FGF2 and activin (Hebrok et al.,1998). Deletion of the notochord results in ectopic expression of Shh in the dorsal pancreatic bud and loss of pancreatic gene expression (Kim et al.,1997). Inhibition of hedgehog signaling with the alkaloid Cyclopamine causes ectopic pancreatic buds to form in cultured gut tube explants (Kim and Melton,1998).

Pancreatic, Endocrine, and Exocrine Progenitor Cells

After pancreatic specification and the onset of bud formation, there are continued signaling processes that regulate the proliferation and differentiation of progenitor cells for the endocrine and exocrine lineages. FGF10-mediated signaling is important for expansion of a pool of Pdx1+ pancreatic progenitor cells (Bhushan et al.,2001). Notch-Delta signaling results in the expression of the NGN3 transcription factor in a subset of pancreatic cells, and cell lineage experiments have demonstrated that these Ngn3-expressing cells are a multipotent progenitor that gives rise to all cells of the endocrine lineage (Apelqvist et al.,1999; Gu et al.,2002). Loss- and gain-of-function experiments confirm that Ngn3 is key for endocrine cell development (Gradwohl et al.,2000; Schwitzgebel et al.,2000; Grapin-Botton et al.,2001). Of interest, a null mutation in human Ngn3 does not result in overt glucose intolerance, suggesting the existence of an Ngn3-independent pathway to endocrine cell development not revealed by studies in mice (Wang et al.,2006a). Additional transcription factors, specifically Ptf1a/p48, have been shown to be required for development of multiple pancreatic lineages (Krapp et al.,1998) and recent studies indicate the β-catenin/Wnt signaling pathway is required for maintenance of Ptf1a/p48+ exocrine progenitor cells (Murtaugh et al.,2005; Wells et al.,2007). Microarray data indicate that there is a reduction in Ptf1a/p48+ levels that coincide with a premature burst of exocrine gene expression in β-catenin null pancreata at e14.5. One possibility is that there is premature differentiation of exocrine progenitor cells, which could explain why exocrine pancreas development appears to arrest these animals between e14.5 and e16.5 (Wells et al.,2007).

Specifying the Different Endocrine Lineages

There are five endocrine cell types in the mature islet: glucagon-expressing α-cells, insulin-expressing β-cells, somatostatin-expressing δ-cells, ghrelin-expressing ϵ-cells, and pancreatic polypeptide-expressing cells (PP; Fig. 1). In addition to Ngn3, which is required for specification of all endocrine lineages in mice, loss-of-function experiments have identified several other transcription factors that play a role in specifying multiple endocrine lineages in the developing mouse pancreas (for review, see Jensen,2004). The basic helix–loop–helix transcription factor NeuroD, the zinc-finger protein Insm1 and the LIM homeodomain protein Isl1 are all required for the development of multiple endocrine cell lineages and are thought to act downstream of Ngn3 (Ahlgren et al.,1997; Naya et al.,1997; Gierl et al.,2006).

Expression of additional factors in a combinatorial manner further restricts cells into distinct endocrine lineages, including the α/glucagon cells (Pax6, Nkx2.2, Foxa2, Arx), δ/somatostatin cells (Pax4, Pax6), PP cells (Nkx2.2), and β/insulin cells (Pax4, Pax6, Nkx2.2, Nkx6.1, Hlxb9, Pdx1) (Sosa-Pineda et al.,1997; St-Onge et al.,1997; Sussel et al.,1998; Li et al.,1999; Lee et al.,2002; Collombat et al.,2003,2007; Prado et al.,2004; Heller et al.,2005). Given the recent discovery of ϵ/ghrelin cells, it is not known what transcription factors are necessary for specification of this lineage in the pancreas, however, like the other endocrine lineages, these cells derive from a Ngn3 progenitor cell (Heller et al.,2005). At the phenotypic level, some of these transcription factors appear to act synergistically to promote one lineage, while at the same time act to antagonize the activity of other factors. For example, β-cells seem to be replaced by ϵ-cells in Nkx2.2 null animals, suggesting that Nkx2.2 positively promote β-cell fate, but also might be required to repress an ϵ-cell fate (Prado et al.,2004). At the biochemical level, there are relatively few studies that have shown how these factors synergistically regulate the transcription of cell-specific target genes during endocrine cell development (Edlund,1998; Jensen,2004). Also, surprisingly little is know about the signaling pathways that promote these specific endocrine lineages during embryonic development in vivo.

As mentioned above, cell lineage studies have been important in identifying the developmental lineage of pancreatic endocrine cells. For example, it is thought that all pancreatic cell types transition through a Pdx1+ and a Ptf1a/p48+ stage during development (Gu et al.,2002; Kawaguchi et al.,2002) and that all endocrine cells are derived from a Ngn3 progenitor cell (Gu et al.,2002). Equally important have been cell lineage analyses that show how the different hormone lineages are not derived. For example, cells that express both insulin and glucagon have been described in the developing pancreas, and it has long been thought that these cells represent a common progenitor for α- and β-cells. However, lineage-tracing experiments using the insulin and glucagon promoters to label developing α- and β-cells cells have demonstrated that mature β-cells have never expressed glucagon and mature α-cells have never expressed insulin (Herrera,2000). Therefore, embryonic cells that express both insulin and glucagon are not a bipotential progenitor but are perhaps a developmental dead end. Similarly, it had been reported that multipotent progenitor cells in the pancreas may express Nestin (Zulewski et al.,2001), but expression and cell lineage experiments have clearly shown that this is not the case (Selander and Edlund,2002; Treutelaar et al.,2003; Delacour et al.,2004). Below, we will discuss the importance of understanding the correct embryonic origin of the endocrine pancreas in attempts to generate endocrine cells from embryonic stem cells.


Differentiation of ESCs into tissues for replacement therapy is a challenging task. Several different protocols have been described for differentiation of mouse and human ESCs (mESCs or hESCs) into pancreatic cells. Successful attempts in efficiently differentiating ESCs into endoderm and pancreas have largely been due to the lessons learned from basic developmental biology research. We will focus on some recent breakthroughs, where ESCs have been efficiently differentiated into definitive endoderm and endocrine pancreas using the molecular pathways described above.

Efficient Differentiation of Definitive Endoderm From ESCs

Directing ESCs into the endoderm lineage is a prerequisite for generating therapeutic endoderm derivatives. As discussed above, nodal signaling is required for induction of mesoderm and endoderm in vertebrates. However, until recently, a commercial source of biologically active nodal protein was unavailable (Tada et al.,2005). Therefore, initial efforts to derive endoderm from both mouse and hESCs used activin, another member of the TGF-β superfamily (Kubo et al.,2004; D'Amour et al.,2005). Activin has been used to mimic nodal activity in studies of mesoderm and endoderm formation during amphibian development (Smith et al.,1990; Thomsen et al.,1990; Gamer and Wright,1995; Henry et al.,1996). Activin binds and activates the same receptors as nodal. However, activin does not require the coreceptor cripto for receptor binding, whereas nodal does (Gray et al.,2003).

The use of activin to differentiate ESCs into endoderm was first reported with mouse ESCs using embryoid body formation (aggregated ESCs grown in suspension) and then using hESCs differentiated as a monolayer (Kubo et al.,2004; D'Amour et al.,2005). In both of these studies, the effects of activin on ESC differentiation were nearly identical to its effect on naive ectoderm cells (animal caps) in frog; high levels of activin induced the formation of endoderm cells (Fig. 2B). As with endoderm formation in vivo, definitive endoderm (DE) cells in culture express a unique combination of transcription factors, including Sox17 and Foxa2. One difference between mESCs and hESCs is in the efficiency of differentiation using different methods. Mouse embryoid bodies consisted of ∼50% endoderm, as measured by expression of Foxa2, where as hESCs differentiated as monolayers consisted of >80% DE as measured by coexpression of Foxa2 and Sox17. In contrast, deriving definitive endoderm from hESCs by means of embryoid body formation is much less efficient, with efficiencies ranging from ∼5 to 19% (Kim et al.,2007). Indeed, microarray analysis of hESC-DE derived by means of monolayer cultures indicates that numerous well-documented markers of DE are up-regulated after 3 days of culture in activin (Fig. 3). Whereas this protocol is highly efficient in producing DE (>80%), it is important to point out that other cell lineages exist in these cultures at low levels including ectoderm, visceral endoderm, and mesoderm (Fig. 2B, TBra staining, and D'Amour et al.,2005). This finding suggests that activin is not sufficient to fully repress these other cell types.

Figure 2.

Efficient differentiation of human embryonic stem cells (hESCs) into definitive endoderm (DE). A: The ESCs are directed to become pancreatic endocrine cells over the course of 2–3 weeks. By adding various factors at different stages (see text and D'Amour et al.,2006), hESCs are first differentiated into a mesendodermal progenitor and then into DE. DE then transitions through stages reminiscent of primitive gut tube, foregut endoderm, pancreatic endoderm, and finally into the endocrine cell lineage. By analyzing different markers (shown under each cell stage), hESC differentiation is correlated with pancreas development in vivo. Figure modified from (D'Amour et al.,2006). B: hESC differentiation into DE. The hESC cell line H9 was cultured using standard techniques. To induce differentiation, ESCs were cultured in RPMI + Activin (100 ng/ml) and low serum (as described in D'Amour et al.,2005) for 3 days. Before differentiation, hESCs express Oct4 (image shows one ESC colony). After differentiation, >80% of cells are endoderm cells that coexpress Sox17 and FoxA2 (these cells are a monolayer). A small number of HESCs differentiate into mesoderm and express brachyury (TBra).

Figure 3.

Microarray analysis of human embryonic stem cells (hESCs) and human definitive endoderm (DE). A: RNA was isolated from undifferentiated hESCs and from cells differentiated in Activin (100) for 3 days and subjected to Affymetrix microarray analysis. All experiments were performed in quadruplicate. Over 100 mRNAs were either increased or decreased >3.5-fold during endoderm differentiation. B: The most highly up-regulated transcripts are well-documented DE markers. References for endoderm gene expression are Gata3 (Debacker et al.,1999), Sox17 (Kanai-Azuma et al.,2002; Zorn and Wells,2007), Isl1 (Lin et al.,2006), Foxa1 and Foxa2 (Friedman and Kaestner,2006), and Cxcr4 (D'Amour et al.,2005).

More recently, efficient differentiation of mouse ESCs into mesoderm and endoderm was accomplished using the tetracycline transactivator system to drive expression of nodal (Takenaga et al.,2007). Nodal expression induced a mesendodermal progenitor population in >70% of cells when compared with treatment with exogenous activin, which induced a progenitor population >50%. More efficient differentiation of endoderm induced by nodal-expressing cells may occur for several reasons. It is known that commercially available nodal protein is much less active than Activin, and this finding is thought to be due, in part, to inefficient processing of nodal. However, in the above experiments, the ESCs produce their own nodal, and presumably are exposed to a constant, uniform level of biologically active nodal protein. It will be interesting to see if Nodal-conditioned medium will similarly promote efficient differentiation of hESCs into definitive endoderm.

That neither nodal nor activin are fully capable of transforming all of the ESCs into endoderm cells suggests that other factors play a role in endoderm formation. Nodal is thought to initiate gastrulation in vivo, which results in expression of several Wnt and FGF ligands in the primitive streak. These ligands are known to regulate gastrulation movements and cell fate cell fate decisions. In ESC cultures, it is likely that activin/nodal is important for both initiating a gastrulation-like process and for directing mesendoderm progenitors into the endoderm lineage. In support of this, activin treatment of ESCs induces the expression of gastrulation factors including Wnt and FGF ligands in a manner that is temporally similar to gastrulation in vivo (D'Amour et al.,2005). Although in vivo it is not well understood how these factors act in combination to promote endoderm formation, ESC-derived endoderm cultures are a powerful tool to perform cellular, molecular, and biochemical analyses of this process.

It is well documented in studies of Xenopus and zebrafish embryos that a common mesendoderm progenitor cell exists that gives rise to both the mesoderm and the endoderm (for review, see Zorn and Wells,2007). It has not been conclusively demonstrated that mammals have a mesendoderm progenitor; however, one study in mouse suggests that such a cell exists in vivo (Lawson et al.,1991). In this cell lineage study, a population of cells in the anterior primitive streak was capable of giving rise to descendants that populate the mesoderm and endoderm germ layers, but not ectoderm. Recent studies have used mouse and hESC cultures as a basic research tool, and the data now support the idea that a common mesendoderm progenitor exists in mammals (Kubo et al.,2004; D'Amour et al.,2005; Takenaga et al.,2007). For example, Kubo et al. sorted activin-treated ESCs using the mesodermal marker Brachyury (Bra, also called T or TBra; Wilkinson et al.,1990) and found that high levels of activin induced this Brachyury+ population of cells to express the endoderm markers Foxa2, Sox17, Hhex, and Albumin, whereas low activin-induced mesoderm markers. This finding suggested that the Brachyury-sorted cells were a bipotential mesendoderm progenitor. Tada et al. used a similar cell-sorting approach, which supports the existence of a mesendoderm progenitor cell (Tada et al.,2005). Under differentiation conditions, a population of Goosecoid-positive (Gsc+), E-cadherin–positive (ECD+), platelet-derived growth factor receptor-alpha–positive (PDGFR-α+) cells is produced that will then give rise to two different populations of cells. One population is Gsc+, ECD+, PDGFR-α−, which will preferentially give rise to definitive endoderm. The other population of cells is Gsc+, ECD−, PDGFR-α+, and will differentiate into mesoderm (Tada et al.,2005). The hESCs were shown to go through a mesendoderm-like state early during the endoderm-differentiation process in which cells coexpress mesoderm and endoderm markers such as Brachyury and Sox17. In these cultures, Brachyury is rapidly down-regulated and markers of the DE are up-regulated, including FoxA2 and Sox17 (D'Amour et al.,2005). Although these experiments support the idea that some portion of the endoderm comes from a mesendoderm progenitor, they do not rule out the possibility that some endoderm cells are derived by means of a separate pathway as in Xenopus.

Differentiation of ESC-Derived Endoderm Into the Pancreatic Lineage

The aforementioned studies provide an important proof of principle that molecules that direct embryonic development in vivo can be used to direct the differentiation of ESCs into specific cell lineages in culture. While this remarkable finding highlights the utility of ESC cultures to study developmental biology, the therapeutic goal is to generate cells for replacement therapy. Because the DE derived from ESCs is theoretically capable of becoming any endoderm derivative (liver hepatocytes, lung alveolar cells, pancreatic β-cells, etc.), the challenge now is to determine protocols for the efficient generation of specific and functional cell types. We will focus on a recent study that has successfully used developmental paradigms to generate pancreatic endocrine cells from human ESCs. D'Amour et al. developed a protocol to generate pancreatic cells with 20% efficiency and insulin-producing cells with as high as 12% efficiency (D'Amour et al.,2006). This was accomplished by manipulating many of the signaling pathways discussed above to mimic the in vivo development of the endoderm, foregut, pancreatic progenitors, and endocrine progenitor cells (Fig. 2A). In this stepwise differentiation protocol, hESCs are first differentiated into DE with activin, then cells were treated with a combination of RA, FGF10, and cyclopamine (a hedgehog inhibitor). RA and cyclopamine are known to promote a posterior foregut/pancreatic fate in vivo, and in these cultures induces the expression of the posterior foregut markers Hnf6, Hnf1β, Hlxb9, and Pdx1. In vivo, FGF10 is known to promote the proliferation of Pdx1+ pancreatic progenitor cells and, in this context, appears to have a similar activity. In the later stages of this differentiation protocol, cells were cultured in the presence of extendin 4 (Ext4), insulin-like growth factor (IGF1), and hepatocyte growth factor (HGF). Although the role of these factors in pancreas development is not well defined, these factors have been shown to promote endocrine cell differentiation in various contexts (Anastasi et al.,2005; Tei et al.,2005). Although these cells have high insulin content, contain secretory granules, and secrete insulin in response to several secretagogues, they do not respond to glucose, suggesting that they are not mature β-cells.

Differentiation of ESCs into insulin+ cells is more efficient using monolayer cultures rather than embryoid body cultures. This finding is likely due to the fact that cells are more evenly exposed to the soluble factors in the medium, where as the three-dimensional structure of embryoid bodies will favor more cell–cell communication resulting in secondary induction of other cell types. It therefore seems that efficient differentiation protocols based on three-dimensional cultures will be challenging, as supported by mouse ES protocols that are based on embryoid body formation where only ∼1% of the total population of cells are insulin-positive after differentiation (Ku et al.,2004). While addition of factors such as activin increased the number of insulin-producing cells to almost 3%, this is still an inefficient process. One explanation for why embryoid body formation is an inefficient method was suggested by Mfopou et al., who showed that secondary inductions resulted in the expression of sonic hedgehog, a known inhibitor of pancreas development (Mfopou et al.,2005).

It is possible that a combination of two- and three-dimensional culturing approaches might work best to derive functional pancreatic β-cells. One such method was used to generate insulin-producing cells from hESCs (Jiang et al.,2007a), where hESC monolayers were exposed to low activin and sodium butyrate to generate Sox17/Foxa2/Cxcr4-positive DE, which was then scraped off the plate and grown in suspension. Over several weeks, FGF2, noggin, and endodermal growth factor were added to the medium to promote the formation of Pdx1-positive cells. This method resulted in up to 25% the cells becoming Pdx1-positive and 8% expressing insulin C-peptide. The C-peptide–positive clusters of cells also contained glucagon- and somatostatin-expressing cells, which is reminiscent of pancreatic islets. Additionally, these islet-like structures contained secretory granules when examined by electron microscopy and were able to secrete C-peptide in response to glucose stimulation (Jiang et al.,2007a). Taken together the above studies could suggest that monolayer cultures are best suited for efficient differentiation of hESCs into pancreatic and endocrine progenitor cells, whereas three-dimensional cultures might be more effective for maturation of progenitors into functional β-cells. In support of this statement, several reports have shown that grafting ESC-derived Pdx1+ cells in vivo promotes their maturation into β-cells, and this process will be discussed below.

Transplantation of Endoderm Tissue Derived From ESCs

Toward the goal of generating therapeutics, it is necessary that ESC-derived tissues are capable of ameliorating disease symptoms in animal models. There have now been several reports investigating the impact of ESC-derived insulin+ cells in mouse models of type-1 diabetes. Several groups have reported the derivation of insulin-producing cells from mouse ESCs using a method based on selection of nestin+ progenitor cells (Lumelsky et al.,2001; Hori et al.,2002; Blyszczuk et al.,2003). One group transplanted these cells into streptozocin (STZ)-induced diabetic mice and observed the maintenance of insulin-expressing cells over different periods of time, without restoring euglycemia in STZ-treated animals (Lumelsky et al.,2001). Further characterization of the insulin-producing cells derived from this method showed that these cells were not pancreatic β-cells but were neuronal. These insulin-producing cells rarely expressed C-peptide, were devoid of β-cell secretory granules, and were unable to rescue STZ-induced hyperglycemia (Sipione et al.,2004). In some cases, insulin-expressing cells derived using the nestin selection method were able to rescue STZ-induced diabetes, however, this rescue ultimately failed because the transplanted cells formed teratomas (Fujikawa et al.,2005). These findings support the idea that, to generate functional β-cells (and other tissues), protocols that recapitulate development are essential. While the nestin selection protocol is not useful for generating β-cells, it has played an important role in our understanding of how (not) to experimentally approach ESC differentiation.

Recent studies support the conclusion that human ESCs differentiated into endoderm and pancreatic lineages may have therapeutic potential. D'Amour et al. have demonstrated that grafting human ESC-DE under the kidney capsule of mice promotes their further differentiation into more mature endodermal derivatives as measured by expression of markers for intestinal cells and liver hepatocytes (D'Amour et al.,2005). Human embryoid bodies that were differentiated using activin and RA to enrich the FoxA2/Sox17/Pdx1 positive populations of cells were transplanted into STZ-treated mice (Shim et al.,2007). While insulin production was minimal in these cells before transplantation, after transplantation, the Pdx1-positive cells were more differentiated and expressed insulin (C-peptide and pro-insulin) and glucagon. In some cases, insulin and glucagon were expressed in the same cell, reminiscent of the insulin/glucagon double-positive cells observed in the developing pancreas, which are not progenitors of β-cells (Herrera,2000). While these grafts rescued the hyperglycemia induced by STZ injections, it is not clear if any of the grafted cells were bona fide β-cell progenitors capable of giving rise to mature, self-renewing β-cells. A study by Jiang et al., resulted in a population of cells expressing C-peptide, insulin, glucagon, and glut2 before transplantation under the kidney capsule (Jiang et al.,2007b). After transplantation, the grafts contained multiple pancreatic cell types, including endocrine, ductal, and exocrine cells. Moreover, the endocrine population of cells appeared more mature, having secretory vesicles and being glucose responsive. These cells also expressed additional β-cell markers Pdx1, Nkx6.1, and Pax6, among others and the grafted cells were able to rescue the STZ-induced hyperglycemia. Importantly, mice were followed for 3 months after transplantation and no teratoma formation was observed (Jiang et al.,2007b). Again, these studies highlight the importance of using methods of differentiation that will guide cells down a developmental path to becoming fully committed, mature cells that are not capable of uncontrolled growth. While the Jiang et al. study is promising, more thorough transplantation studies are needed to determine the risks (teratoma formation) of ESC-DE transplantations.


There has been great progress made over the past few years in directing ESCs into the pancreatic endocrine lineage. The successful approaches in this field have all been based on an embryonic blueprint where ESCs are differentiated in a step-wise manner that mimics pancreas development in vivo. Experimental approaches that deviate from embryonic principles have proven less effective at generating pancreatic cell types. For example, attempts to shortcut development by expressing pancreatic transcription factors including Pdx1, Ngn3, and Pax4 in ESCs are not an efficient method to generate pancreatic cells (Blyszczuk et al.,2003; Vincent et al.,2006) for the likely reason that the ESCs are not yet competent to respond to these factors. Now that efficient protocols have been established for generating foregut endoderm cells, it would be interesting to re-evaluate the ability of these pancreatic transcription factors to efficiently direct differentiation. Differentiation protocols that were not based on pancreas development, such as the Nestin-selection approach discussed above that generated insulin-expressing neurons, have frustrated subsequent efforts at generating bona fide β-cells.

Studies of pancreas development continue to help researchers design better protocols and overcome obstacles toward generating mature β-cells and eliminating residual hESCs that are capable of forming teratomas. For example, most published ESC studies describe a population of insulin-expressing cells that arises early in these cultures that often times coexpresses other hormones such as glucagon. Lineage tracing studies have shown that these cells are probably not progenitors of the β-cell lineage but are possibly the transient population of hormone-positive cells that arise early during pancreas development, which do not give rise to the endocrine cells found in adult animals (Herrera,2000,2002). Researchers are focusing on populations of cells that coexpress β-cell progenitor markers (Nkx2.2, Nkx6.1, Pax4, Pax6) but are insulin−. These may be progenitor cells that can become bona fide β-cells that are capable of self-renewal.

Whether or not early hormone-expressing cells in ESC cultures have an in vivo equivalent, or are aborted attempts at endocrine differentiation, it is clear that we still lack information on the full complement of factors that are required for generating functional endocrine cells. There is still a considerable amount of information in the current literature that could be used to improve existing ESC differentiation protocols. For example, little attention has been placed on developing conditions that promote endocrine vs. exocrine fate by controlling the levels of notch, TGFβ, and Wnt signaling. There is every reason to believe that further studies of normal pancreatic development and β-cell maturation will continue to increase both the efficiency and functionality of HESC-derived insulin-expressing cells.


J.M.W. and J.R.S. are supported by a Career Development Award from the Juvenile Diabetes Research Foundation and the NIH. J.R.S. is supported by an NIH training grant in Developmental and Perinatal Endocrinology.