The molecular basis and prospects in pancreatic development


*Author to whom all correspondence should be addressed.


Studies on the signaling mechanism that control the specification of endoderm-derived organs have only recently begun. While many studies revealed genes involved in the differentiation, growth and morphogenesis of the pancreas through studies of mutant mice, still little is known about how endoderm give rise to specific domains. Although many genes are known to have a role in pancreatic differentiation, growth and morphogenesis, few genes are known to take part in the specification of the pancreas so far. Hallmarks as well as gene markers for early development of the pancreas, which are however still very limited, will be useful for dissecting early events in pancreatic specification. Here, I give a summary on the origin of the dorsal and ventral pancreatic progenitors, signals for inductions, and genes so far known to function in pancreatic differentiation. I also give a future prospect in the use of ES cells and other experimental models, towards a comprehensive understanding of gene networks in the progenitor cells or intermediate cell types which arise during various stages of differentiation.

When and where do the pancreatic progenitors arise?

The adult mammalian pancreas is a heterogeneous organ composed of three major cell types: (i) exocrine cells produce digestive enzymes, which are secreted into the gut via duct cells; (ii) duct cells; and (iii) endocrine cells. There are four endocrine cell types: (i) alpha cells, which makes glucagon; (ii) β cells produce insulin; (iii) delta cells make somatostatin; (iv) and PP cells make pancreatic polypeptide. Endocrine cells are organized into islets of Langerhans, which disperse throughout the exocrine tissue. Exocrine tissues organized into dense epithelial acini, make up over 95% of the pancreas. The islets make up only about 1–2% of the pancreas.

At the 18-somite stage in chick embryos, radio ablation studies show that the dorsal pancreatic endoderm is at a level of somites 7–15 (LeDouarin 1964). However, recent DiI (1,1′-dioctadecyl-3,3,3′-tetramethyl–indocarbocyanine perchlorate) injection studies at the 10-somite stage in the chick embryo, show that the dorsal prepancreatic endoderm locates at the levels of 4–7 somites in the medial region within the lateral borders of the somites (Matsushita 1996).

Studies by CM-DiI solution injection showed that the ventral prepancreatic endoderm at 10-somite stage embryos is located at a position lateral to somites 7–9 in the chicken embryo (Kumar et al. 2003). In mice, the ventral bud of the pancreas is derived from Pdx1-expressing cells at the lip of the anterior intestinal portal (AIP), as the flat endoderm sheet folds and fuses ventrally to form the gut tube (Offield et al. 1996; Deutsch et al. 2001). The liver is also derived from the endoderm of the AIP from the same general domain of cells as the ventral pancreas (Gualdi et al. 1996; Deutsch et al. 2001).

Extrinsic signals

Induction signals for the dorsal pancreas

Development of the pancreas has been studied in mice and chickens. The pancreas forms from the fusion of dorsal and ventral buds. Prior to morphogenesis, the endoderm fated to become the dorsal pancreas bud is a layer of epithelial cells of endoderm, which is formed by the end of gastrulation. Derivatives of the respiratory and digestive system rise from specific regions of the endoderm, as a result of interaction of the endoderm with the adjacent ectoderm or mesoderm. Studies by Wells & Melton (2000) showed that factors produced in the adjacent mesoderm and ectoderm (e.g. one of the candidates is FGF4 in mice) subdivide the endoderm into anterior and posterior domains along the anterior–posterior axis at the late gastrulation stage.

Kim et al. (1997) reported that commitment of the endoderm to a pancreatic fate may occur as early as the 10–12-somite stage in mice and the 13-somite stage in chickens. Prior to this stage, the prepancreatic endoderm, at the level of 7–16 somites, isolated and cultured alone, did not turn on Pdx1 (or Isl1 and Pax6 expression in the dorsal pancreas). But when cultured in combination with notochord, the prepancreatic endoderm turned on Pdx1, Isl1 and Pax6 expression. This suggests that the dorsal prepancreatic endoderm prior to stage 13 in chick embryos has not acquired the ability to differentiate pancreatic fate autonomously, but requires signals from the notochord for the induction and maintenance of pancreatic commitment. The ventral prepancreatic endoderm, which was not in contact with the notochord, does not require notochord to establish its pancreatic fate (Kim et al. 1997). The candidate mediators secreted from the notochord include activin-βB and fibroblast growth factor-2 (FGF2) (Hebrok et al. 1998). When applied to chick embryo cultures of the foregut, these growth factors mimic the notochord in repression of the signaling factor Sonic hedgehog (Shh) in the dorsal pancreatic endoderm and permit expression of pancreatic genes such as glucagon and insulin. Expression of Shh, a potent intercellular patterning signal, is excluded in the dorsal and ventral prepancreatic endoderm (Ahlgren et al. 1997; Apelqvist et al. 1997).

Later, the dorsal aorta take the place of the notochord and come in close proximity to both the dorsal and ventral endoderm; this has also been shown to be sending permissive signals required for the induction of the pancreas (Lammert et al. 2001). Co-culture of prepatterned pancreas endoderm and aorta endothelium result in endocrine cell differentiation. Transgenic expression of VEGF-A under the control of the Pdx1-promoter, attract endothelial cells to the stomach and duodenum resulting in ectopic insulin-expressing cells (Lammert et al. 2001).

Taken together, the role of the signals from the dorsal mesoderm is to ensure the absence of Hh molecules in the pancreatic endoderm. The ventral endoderm is devoid of Hh even in the absence of the mesoderm, and this is biased toward the pancreas.

Inducing signals for the ventral pancreas

The identity of the ventral pancreatic endoderm established seemed to depend on the inducing signals from the splanchnic mesoderm that lies directly beneath the ventral prepancreatic endoderm (Kumar et al. 2003). The ventral prepancreatic endoderm when isolated and cultured alone did not initiate pancreatic gene expression, but when co-cultured with splanchnic mesoderm, it adopted a pancreatic fate. Interestingly, posterior splanchnic mesoderm is able to induce the more anterior endoderm to switch to a posterior fate. When the ventral prepancreatic endoderm at the 10-somite stage is transplanted to a posterior position, it turns off Pdx1 expression and switches on CdxA expression. But when transplanted to an anterior position the ventral prepancreatic endoderm maintained Pdx1 expression (Kumar et al. 2003). This commitment of the Pdx1 domain is observed as early as the 6-somite stage, at which stage the prepancreatic domain is transplanted into a 14-somite stage embryo. These signals are instructive as they are able to induce a broad spectrum of pancreatic differentiation in endoderm anterior to the endogenous pancreas domain. This induction can be mimicked by bone morphogenetic protein (BMP) and activin families of signaling molecules. The lateral plate mesoderm appears to pattern the endoderm in a posterior–dominant fashion analogous to the patterning of the neural tube. The molecular identity of the lateral plate mesoderm signals have yet to be defined.

Recent mouse studies have shown that the ventral endoderm can give rise to liver, pancreas or intestine depending on extrinsic or intrinsic signals they receive. The liver and ventral pancreas are derived from same general domain of cells at the AIP (Gualdi et al. 1996; Deutsch et al. 2001). The ventral mesoderm is devoid of Hh signaling, and the default fate of the ventral foregut endoderm is to activate the pancreas gene program. Fibroblast growth factor (FGF) signaling from the cardiac mesoderm diverts this endoderm to express genes for the liver instead of those for the pancreas (Jung et al. 1999). Therefore, interaction with the cardiac mesoderm is required for proper hepatic differentiation (Gualdi et al. 1996). At the 6–7-somite stage, the cardiogenic mesoderm signals; these candidate signals include FGF and BMP (Jung et al. 1999; Rossi et al. 2001), induce the liver program in explants of adjacent endoderm and simultaneously inhibit the pancreas program (Jung et al. 1999; Deutsch et al. 2001). Ventral endoderm cells that do not receive signals from the cardiac mesoderm can execute the pancreas program. Recently, Kawaguchi et al. (2002) showed that ventral endoderm cells lacking the Ptf1a gene expression are converted to an intestinal fate, thus suggesting the multiple lineage regulation in progenitor cells in the ventral endoderm fate.

Intrinsic signals

Hlxb9 gene

Hlxb9 gene (encoding the Hb9 homeoprotein) is expressed throughout both early pancreatic buds, and in the dorsal bud it precedes Pdx1. In mice lacking Hlxb9, the dorsal prepancreatic endoderm fails to express Pdx1 and formation of the dorsal pancreatic bud never initiates (Harrison et al. 1999; Li et al. 1999; Li & Edlund 2001). The ventral pancreas, remarkably, develops normally until later stages at which point Hlxb9 is required for β-cell maturation. This demonstrates that Hlxb9 functions upstream of Pdx1 in the dorsal but not the ventral pancreatic bud.

Hex gene

Hex gene functions in cell proliferation and correct positioning of the definitive ventral endoderm (Bort et al. 2004).

At embryonic day (E)8.5 in the mouse, proliferation of cells at the leading edge of the ventral–lateral endoderm, where the liver and ventral pancreas emerge, help close off the foregut. During this time, the endoderm grows adjacent to and beyond the cardiogenic mesoderm, executing the pancreas program. In Hex null embryos, a defect in definitive endoderm growth beyond the cardiogenic mesoderm allows for hepatic but not pancreatic induction. However, when Hex-null ventral endoderm is isolated prior to its interaction with the cardiogenic mesoderm and is cultured in vitro, it activates early pancreas genes showing that the mutant endoderm maintains its intrinsic competence to initiate the pancreatic program (Bort et al. 2004).


Ptf1a (also known as Ptf1a-p48) is detectable in acinar cells from E15.5 throughout adulthood and consists of a hetero-oligomer comprised of three subunits. One of the three proteins, Ptf1a, is a transcription factor of the basic helix-loop-helix (bHLH) type, and was originally identified as an exocrine pancreatic-specific transcription factor. Mice nullizygous for the Ptf1a gene die within hours of birth and have no detectable exocrine pancreatic tissue. Dorsal endocrine cells develop and migrate from the intestinal mesentery to colonize the spleen (Krapp et al. 1996). However, by a recombination-based lineage tracing in vivo, it was shown that Ptf1a is expressed at these early stages in the progenitors of pancreatic ducts, exocrine and endocrine cells and supports the specification of precursors of all three pancreatic cell types (Kawaguchi et al. 2002). In the absence of Ptf1a, ventral pancreas fails to form and switches to adopt a duodenal–intestinal fate (Kawaguchi et al. 2002), thereby indicating that Ptf1a is responsible for specification of the ventral pancreas and also needed for robust outgrowth of the dorsal bud.

Taken together, the prospective ventral–lateral endoderm cell may be competent to execute at least three different fates. Hepatogenic signaling from the cardiac mesoderm induces the liver fate, an as yet not identified signal induces or permits the pancreatic fate, and in the absence of both cardiac signaling and the expression of Ptf1a, the cell executes an intestinal–duodenal fate.


In mouse embryos, Pdx1 expression is by far the first sign of differentiation detected at E8.5 in the dorsal endoderm of the gut. At E9.5, Pdx1 expression marks the dorsal and ventral pancreatic buds, and the duodenal endoderm. Removal of Pdx1 by gene targeting arrests pancreatic development after initial bud formation, and results in an animal with no pancreas (Johansson & Wiles 1995; Ahlgren et al. 1996; Offield et al. 1996). Therefore, Pdx1 is an essential molecule for the pancreatic development. Pdx1 expression marks a pluripotent population of cells that gives rise to all cell types of the neonatal pancreas (endocrine, exocrine and duct) and epithelium of the duodenum and posterior stomach (Gu et al. 2002).

Ectopic expression of Pdx1 expands the domain of dorsal pancreatic progenitors (Grapin-Botton et al. 2001; Horb et al. 2003). However, Pdx1 does not trigger complete cell differentiation from embryonic endodermal cells (Grapin-Botton et al. 2001). Pdx1 function is continuously required for epithelial proliferation (Holland et al. 2002). FGF signaling is required to maintain β-cell differentiation; acting in a Pdx1 dependent autoregulatory loop, mice with attenuated FGFR1c signaling but not those with reduced FGFR2b signaling, develop diabetes with age and exhibit a decreased number of β-cells (Hart et al. 2000). Cells that have adopted an endocrine cell fate express the bHLH transcription factor Neurogenin 3 (Ngn3) (Gu et al. 2002).

The particular cell functions that Pdx1 regulate are unknown. With the possible exception of Pax4 (Smith et al. 2000; Chakrabarti et al. 2002), no gene targets of Pdx1 in the prepancreatic endoderm or the early pancreatic buds have been identified. Pdx1 was cloned and identified as a β- and δ-cell-specific regulatory factor for transcriptional expression of insulin and somatostatin genes and has been shown to regulate the expression of Glut-2, Islet amyloid polypeptide (IAPP) and glucokinase. In the adult, Pdx1 expression is maintained in the duodenal epithelium and in the insulin-secreting islet β-cells, where it plays a critical role in the regulation of insulin gene transcription (Offield et al. 1996). Pdx1 is a glucose-responsive regulator of insulin gene expression. There is increasing evidence that glucose and a number of hormones such as insulin, glucagon-like peptide-1 (GLP-1), and transforming growth factor β (TGFβ), are able to regulate Pdx1 on the transcriptional, post-transcriptional or post-translational level. For instance, the function of Pdx1 in response to glucose is regulated by both its phosphorylation and nuclear translocation (McKinnon & Docherty 2001).

Pdx1 physically interacts with a variety of other nuclear proteins (Asahara et al. 1999; Ohneda et al. 2000), including another homeodomain factor Pbx1 (Asahara et al. 1999). Pbx family members are widely expressed throughout the developing embryo as well as the pancreas, and dimerize with a variety of Hox and parahox homeodomain transcription factors (Chang et al. 1995), thereby modulating both their DNA-binding specificity and function. Pbx1 is expressed in pancreatic epithelium as well as surrounding mesenchyme (Kim et al. 2002). Pbx1 exerts a growth-promoting effect within the epithelium, by acting through interaction with Pdx1 (Dutta et al. 2001). Pdx1/Pbx1 complexes are necessary for the expansion of the pancreatic buds but not for the specification of the different pancreatic cell types (Kim et al. 2002).

Ngn3 and Notch signaling

Ngn3 is a bHLH protein which has a role as a key regulator of endocrine development, which is expressed exclusively in scattered cells in the pancreatic epithelium and not in differentiated endocrine cells (Apelqvist et al. 1999; Jensen et al. 2000a; Schwitzgebel et al. 2000). Lineage tracing using Cre recombinase in transgenic animals has demonstrated that these Ngn3 expressing cells function as endocrine cell precursors (Gu et al. 2002). Animals lacking Ngn3 fail to develop any endocrine cells, while exocrine tissue and pancreatic ducts are nearly normal (Gradwohl et al. 2000). Ngn3 has been shown to activate the promoter for the gene encoding another member of the family of neural pro-neuronal HLH factors, NeuroD1. NeuroD1 is expressed slightly later than Ngn3 during pancreatic development, but unlike Ngn3 it persists in the mature islet cells and plays a role in the expression of a number of endocrine products including insulin. Misexpression of Ngn3 is sufficient to induce endocrine differentiation throughout the gut epithelium (Grapin-Botton et al. 2001), although the cells produced are almost exclusively α cells.

Studies of the Ngn3 promoters have identified several transcription factors that may play this role. It contains multiple-binding sites for the Hes1 repressor, which potently inhibits the promoter in transient transfections (Lee et al. 2001), as well as binding sites for several other transcriptional activators broadly expressed in the endoderm and the pancreatic buds, including Hnf1, Hnf3β/Foxa2 and Hnf6 (Jacquemin et al. 2000; Lee et al. 2001). Genetic evidence in mice supports a role for Hnf6 as an upstream activator of Ngn3 expression.

Several recent studies suggest the involvement of Notch in the regulation of pancreatic exocrine and endocrine cell fate. Notch signaling restricts Ngn3. to scattered cells within the pancreatic epithelium. While cell expressing Ngn3. are destined to become endocrine cells, cells in which Ngn3 expression is extinguished by Notch signaling can become part of the exocrine pancreas. Recent works have suggested that when an intracellular fragment of Notch1 (Notch1-IC), an activated form of Notch 1 is introduced in the developing mouse pancreas, inhibited both endocrine and exocrine differentiation (Hald et al. 2003; Murtaugh et al. 2003; Esni et al. 2004). When NotchI-IC is misexpressed in adult exocrine pancreatic tissue, mature acinar cells are replaced by a nestin-positive precursor population, through a process of apparent dedifferentiation (Miyamoto et al. 2003). Endogenous Notch pathway activation do not inhibit Ptf1a expression, suggesting Notch signaling may regulate the sequential recruitment of endocrine and exocrine cell types from a common precursor pool in the developing mouse pancreas. These aspects of the Notch functions correlate well with those found with the expression of Hes1, a downstream effector of Notch signaling. Hes1 is absent from both endocrine and exocrine cells. Hes1 knockout mice exhibit a rapid depletion of precursor cells and result in precocious development of endocrine cells (Jensen et al. 2000b).


The paired-homeodomain factor Pax4 is expressed selectively in the developing pancreas and is required for the normal development of β- and δ-cells. Despite its critical role in β and δ cell genesis, Pax4 by itself is insufficient to drive Ngn3-expressing precursor cells to a β- or δ-cell fate (Grapin-Botton et al. 2001). The Pax4 gene appears to be a direct target of Ngn3. Ngn3 can bind to and activate the Pax4 gene promoter (Smith et al. 2000). In the absence of Ngn3, the expression of Pax4 is lost (Gradwohl et al. 2000). Studies of the Pax4 gene promoter implicate the more broadly expressed factors Hnf1 and Hnf4 in cooperative activation of Pax4 expression along with n Ngn3 (Smith et al. 2000).

Nkx2.2 and Nkx6.1

The Nkx2.2 and Nkx6.1 genes are the NK- homeodomain genes likely to act as β-cell competence factors. Nkx2.2 is expressed early in developing pancreatic buds and is later restricted to α-, β- and PP cells of islets. Endocrine cells of Nkx2.2 mutants express endocrine pancreas-specific proteins such as synaptophysin, prohormone convertase 1/3, IAPP, but are arrested just lacking insulin expression (Sussel et al. 1998). Nkx6.1 is expressed primarily in β-cells of adult islets. Nkx6.1 mutants have a selective reduction of β-cells but other endocrine cell types are normal (Sander et al. 2000).

Other genes

The late factors Pax6, Isl1, Brn4 and Pdx1 function in the final steps of islet-cell differentiation, after Ngn3 expression and in conjunction with hormone gene expression, in the maintenance of the final differentiated islet cell phenotypes. The paired-homeodomain factor Pax6 and the LIM-homeodomain factor Isl1 are expressed in all islet cells and their loss causes defects in the generation of all endocrine cell subtypes (Ahlgren et al. 1997; St-Onge et al. 1997; Sander et al. 2000).

Gene expression profiles of transcription factors and signaling molecules in pancreatic progenitors

Recently, 3400 genes expressed in the pancreas were used to generate an endocrine pancreas microarray (PancChip), which is available through the β Cell Biology Consortium (Scearce et al. 2002). The PancChip will be a valuable diagnostic tool for the genetic analysis of pancreatic cell samples. Gu et al. (2004) reported the transcription profiles of the pancreatic and endocrine progenitor cell. The stages they used were: endoderm before pancreas specification, early pancreatic progenitor cells, endocrine progenitor cells and adult islets of Langerhans. These expression profile studies provide a genetic baseline and highlight genes involved in endoderm plasticity, specification of the prepancreatic cells from their gastrointestinal neighbors, specification of the endocrine and exocrine compartments of the embryonic pancreas or functional β cells, respectively (Gu et al. 2004).

Prospects towards a comprehensive understanding of gene networks in developmental stages of the pancreas using simple experimental models

The formation of the three primary germ layers, ectoderm, mesoderm and endoderm, is an early distinction between groups of cells in developing embryos. Within our limited knowledge of genes involved in endodermal differentiation, many of the genes identified in endoderm induction are conserved among species such as the nematodes, Xenopus, zebrafish, chick and mouse. The use of different experimental systems, such as ascidians, chick embryos or ES cells, which have the advantage in manipulating in vitro, will be extremely useful for identifying molecular function at early stages of the endoderm development (Fig. 1).

Figure 1.

Schematic drawings of prospects in studies of pancreatic development using different experimental models.

Chick embryos

The chick is an excellent experimental model for manipulation of early embryos. Lineage tracing of the pancreatic progenitor cells can be readily performed at early stages. As described above, the dorsal prepancreatic region is mapped to the level of 4–7 somites in 10-somite stage embryos (Matsushita 1996). Fate mapping of cells which give rise to the pancreas at stages earlier than the 10-somite stage will be informative for studies of early inductive signals which might possibly be secreted from adjacent cells in mesodermal or ectodermal layers. Using DiI crystals which allow a fine fate mapping, we mapped dorsal pancreatic precursors first appeared at endoderm layer underline the areas caudal to the node at around stage 5, a stage where gastrulation has just finished. We also observed the migratory route of the prepancreatic progenitors from stage 5 and up to the 10-somite stage (Katsumoto et al. 2005, unpubl. data). Identification of the prepancreatic progenitor cells and how they migrate to form pancreatic buds will enable studies of the expression profiles of the progenitor cells and signals for specifying the cells, as well as enable gene manipulations in these cells.

Ciona intestinalis

Another example of an animal model system for studying endodermal differentiation is Ciona intestinalis, an ascidian, one of the animals whose genome has been sequenced. The ascidian tadpole is composed of only 2600 cells which constitute a small set of larval organs including the epidermis, central nervous system and notochord and tail muscle along with the rudiments of the adult gut, mesodermal organs and gonads. Analysis of the Ciona genome demonstrates that ascidians contain the basic complement to ancestral chordate genes. The simplicity of the Ciona genome will aid researchers in revealing complex developmental processes in vertebrates (Satoh et al. 2003). The developmental fate of each blastomere is restricted to one tissue at or before the 110-cell stage. The lineage leading to the formation of each organ, including the endoderm, is well characterized. The presumptive endoderm blastoderm show a high potential for autonomous differentiation when they are isolated from early embryos (Nishida 1987; Nishida 1992). Therefore, isolation and culture in vitro of the blastomere which are restricted for endodermal development will allow investigation into the gene expression profiles during early stages of endodermal developments.

ES cells as a tool for studies of pancreatic development

ES cells are useful for recapitulating early pancreatic specification in vitro (refer to accompanying text for details). We have developed an efficient culture method for deriving endodermal cells from ES cells using co-culture protocols (Shiraki et al. 2005, unpubl. data). Intermediate cells of endodermal lineages at different developmental stages can be derived from ES cells in vitro when cultured under appropriate conditions. Lineage tracing using appropriate promoters driving the expression of fluorescent proteins is a useful strategy for sorting out cells of interests. The ES cell derived Pdx1 expressing cells were isolated by fluorescence activated cell sorting (FACS) and analyzed with DNA microarrays. We found that these Pdx1-positive cells showed similar expression profiles with E7.5 endodermal cells reported by Gu et al. (2004) (Yoshida et al. 2005, unpubl. data). Therefore, the ES cells in the in vitro differentiation system is suggested to recapitulate developmental events in vivo and thus provide a very useful tool for yielding intermediate cells at various developmental stages. Studies of the expression profiles of these intermediate cells at pancreatic development will be informative to our understanding of pancreatic development. Figure 1 shows the schematic drawing of a future vision on studies of development of pancreas and the unspecified endoderm. The molecules or signal pathways revealed by the ES cell system could be compared with the profiles of endodermal or pancreatic cells at normal development in vivo. Because ES cells can proliferate nearly indefinitely in appropriate conditions, they serve as a source for studies of developmental biology, particularly when it is hard to obtain a large number of cells at the early stage of pancreatic development, such as when the endoderm is established and when the pancreas is specified from the endoderm. Results obtained in ES cell in vitro differentiation could be compared with those in vivo in mice or other simple experimental animals where the early endodermal cells are easier to approach. Genes or signal pathways thus identified could be overexpressed, or their functions inhibited in the mouse or other animal model systems to study their function in vivo.