We first examined the expression of the Notch signaling components in the biliary system and other foregut-derived organs during mouse embryogenesis. We carried out semiquantitative RT-PCR using gallbladder, liver and pancreas isolated from normal CD1 mouse embryos (Fig. 1). We detected Jag1 and Jag2, encoding Notch ligands, in the foregut, gallbladder and pancreas. Of the four Notch genes, Notch1 and Notch2 were expressed in the gallbladder and pancreas after embryonic day (E) 12.5. We observed high levels of Hes1 mRNA in the foregut, gallbladder and pancreas but not in the liver throughout development. On the other hand, we did not detect Hes5, encoding another transcriptional repressor of the Hes family14, in any endodermal organ. These findings suggest that the Notch pathway may contribute to biliary and pancreatic development.

Figure 1: Semiquantitative RT-PCR analysis of expression of Notch and Hes genes in normal mouse embryos.
figure 1

Notch signaling components are expressed in the developing gallbladder and the pancreas. AGM, aorta-gonad-mesonephros region; HPRT, hypoxanthine phosphoribosyltransferase.

To further clarify the role of the Notch pathway in biliary development, we next analyzed biliary morphogenesis in Hes1-deficient mice2. At E9.5, Hes1−/− embryos had no apparent abnormalities: both wild-type and Hes1−/− mice had the bile duct budding from the foregut (Fig. 2a,d). At E11.5, normal elongation of the extrahepatic biliary duct and branching of the ventral pancreas from the common bile duct were clearly observed in Hes1+/+ mice (Fig. 2b,c). Hes1−/− mice, however, had a short, truncated remnant of the common bile duct and proliferation of aggregated epithelial cells that looked like budding vesicles with a central cavity, resembling the future pancreatic acini (Fig. 2e). Staining with a bile duct–specific lectin15, Dolichos biflorus agglutinin (DBA), showed that the aggregated epithelial cells had lost DBA binding (Fig. 2f), like the branching pancreas in wild-type mice (Fig. 2c). These results imply a functional role of Hes1 in maintaining the biliary properties, repressing the vesicular formation of epithelial cells and promoting the growth of the ductal structure of the prospective biliary epithelium.

Figure 2: Abnormal development of biliary epithelium in Hes1−/− embryos.
figure 2

(a,b,d,e) Staining of developing extrahepatic bile ducts with hematoxylin and eosin. Arrows in a,d indicate biliary budding from the hepatic diverticulum (HD). In the mutant (e), a remnant of the common bile duct and proliferation of aggregated epithelial cells are observed. (c,f) Adjacent sections stained with biliary-specific DBA. The ventral pancreas (c, red arrow) and aggregated epithelial cells (f, red asterisk) are negative for DBA binding, whereas the normal extrahepatic biliary system (c) and the remnant of the bile duct (f, arrowhead) are positive. CBD, common bile duct; Du, duodenum; GB, gallbladder; Li, liver; Pa, ventral pancreas; PV, main portal vein. Scale bars, 100 μm.

Because most homozygous Hes1 mutant mice die by E18.5, we examined E17.5 mice in detail to see the ultimate form of biliary morphogenesis. All Hes1−/− embryos that we examined had agenesis of the gallbladder and cystic duct (Fig. 3g), whereas wild-type littermates had these organs (Fig. 3a). We examined serial sagittal sections from the duodenal papilla to intrahepatic bile ducts by DBA staining. In wild-type mice, we observed well-formed extrahepatic bile ducts (Fig. 3b–e). In Hes1−/− embryos, the duodenal portion of the common bile duct was small, truncated halfway along and connected to a pancreas-like structure (Fig. 3h,i), and we observed pancreatic tissue at the missing part of the common bile duct. On the other hand, intrahepatic bile ducts or primordia of hepatic ducts, probably originating from the hepatoblast, developed at the porta hepatis in Hes1−/− mice, although these twisted and fragmented lumen structures were never connected with the common bile duct (Fig. 3i–k). The common bile duct, which was observed in horizontal sections at the level of the first part of the duodenum (Fig. 3f), was totally replaced by pancreatic tissue in Hes1−/− embryos (Fig. 3l).

Figure 3: Gallbladder agenesis and hypoplasia of extrahepatic bile ducts in E17.5
figure 3

Hes1-deficient mice. Macroscopic view, showing a normal gallbladder and cystic duct (a) and the absence of these organs in the mutant (g). (be and hk) Serial sagittal sections of the biliary system stained with DBA. In the wild-type mice, the extrahepatic biliary system from the duodenal papilla (b, arrow) to the common bile duct (CBD; d) and cystic duct (CD; e) are completely formed. In the mutant, the small remnant of the common bile duct disappears halfway along (h,i, arrows). The fragmented lumen structures (ik, arrowheads) are never connected with the common bile duct. (f,l) Horizontal sections stained with hematoxylin and eosin and antibody to amylase (dark purple), showing that the common bile duct in Hes1+/+ mouse (f) is totally replaced by pancreatic tissue (Pa) in Hes1−/− littermates (l). Du, duodenum; Li, liver; PV, main portal vein. Scale bars, 100 μm.

To examine whether apoptotic cell death contributes to the failure of biliary development in Hes1−/− mice, we carried out TUNEL staining of the hypoplastic bile ducts. We found no more apoptotic cells in biliary epithelium of mutant mice than in that of wild-type mice (data not shown).

In Hes1-deficient mice, the hypoplastic biliary system seemed to be converted to pancreatic tissue. To test this possibility, we examined the expression of various pancreatic markers during cholangiogenesis. We found four types of endocrine cells in biliary epithelium of Hes1−/− mice: cells expressing glucagon (α cells), insulin (β cells), somatostatin (δ cells) and pancreatic peptide (PP cells; Fig. 4h–m). None of the insulin-positive cells in the mutant bile ducts coexpressed glucagon (Fig. 4o), indicating that these were not early embryonic multihormonal cells, but rather cells along the normal differentiation pathway to mature β cells16,17. These ectopic insulin-positive cells also coexpressed Ipf1, Isl1 and Glut2 (Fig. 4p–r), like mature pancreatic β cells. The hormone-expressing cells formed clusters, leaving the epithelial layer around the biliary lumen (Fig. 4o), like the pancreatic islet cells migrating out of the pancreatic ducts. Mature islets did not appear in either the biliary system or pancreas in Hes1-deficient mice throughout development. Electron microscopic studies showed that the biliary epithelial cells possessed numerous secretory granules (Fig. 4s,t), confirming at the single-cell level that they had differentiated into endocrine cells. Cells positive for postmitotic endocrine marker Isl1 (ref. 18) or exocrine product carboxypeptidase A were abundantly distributed in the mutant bile ducts (Fig. 4j), suggesting that widespread pancreatic differentiation had replaced the biliary epithelium. In contrast to pancreatic development in the mutant5, which showed accelerated differentiation of only glucagon-expressing endocrine cells, biliary epithelium in Hes1-deficient mice showed exocrine differentiation also. Amylase immunostaining clearly showed acinar development from the common bile ducts of Hes1−/− mice (Fig. 4n). These biliary cells of pancreatic nature were never detected in the normal cholangiogenesis of wild-type littermates (Fig. 4a–f), although small numbers of exocrine cells expressing amylase did appear in bile ducts of wild-type mice in the late embryonic stage (Fig. 4g). These findings indicate that inactivation of Hes1 causes the biliary epithelium to adopt the pancreatic fate and to fully differentiate into endocrine and exocrine cells that are components of premature islets and acini.

Figure 4: Lack of Hes1 leads to pancreatic cell-type differentiation and pancreatic morphogenesis in the biliary ducts.
figure 4

Pancreatic cell-type differentiation occurs in Hes1- deficient bile ducts, as indicated by the presence (arrowheads in in) of the endocrine markers glucagon (h,i), insulin (k), pancreatic peptide (PP; l), somatostatin (m) and Isl1 (j, red) and of the exocrine markers carboxypeptidase A (CA; j, green) and amylase (n). Cytokeratin immunostaining (a,b,di and kn) is observed in the epithelial cells of the bile duct, duodenum and pancreas. bg, in and the outlined areas in a and h show the common bile duct. Pancreatic acini (n, arrows) are budding from the mutant bile duct. (o) Insulin (red) and glucagon (green) are detected in separate sets of biliary epithelial cells located in the outlined area in k. (pr) Mature β cells are generated in E16 Hes1-deficient bile duct. Confocal micrographs showing a small cluster of the ectopic insulin-positive β cells coexpressing nuclear Ipf1 (p, green), nuclear Isl1 (q, red) and cell membranous Glut2 (r, green). (s,t) Electron micrographs (EM) of biliary epithelial cells in E12.5 Hes1−/− mice. These cells containing tight junctions (TJ) and microvilli (MV) had numerous secretory granules (t, arrowheads). The outlined area in s is magnified in t. CBD, common bile duct; Du, duodenum; Li, liver; Lu, lumen of the bile duct; N, nucleus; PV, main portal vein. Scale bars: an,100 μm; pr, 10 μm; s, 3 μm.

The phenotypic conversion from biliary to pancreatic development in the Hes1 mutant mice prompted us to examine the expression of Hes1 and Neurog3 (encoding neurogenin 3) in the developing biliary system to see whether the pancreatic differentiation program5,6,7,8 operates ectopically in the mutant. In situ hybridization showed that Hes1 was expressed in the epithelial cells of the extrahepatic biliary system throughout normal cholangiogenesis (Fig. 5a–c). In addition, in Hes1−/− embryos, Neurog3 was highly expressed ectopically in the developing biliary epithelium (Fig. 5f,g), with a pattern similar to that seen in the normal pancreas. In wild-type littermates, the Neurog3 signal was confined to the pancreas and not detected in the biliary epithelium (Fig. 5d,e). Taken together, these findings suggest that the pancreatic endocrine differentiation program of the biliary epithelium in Hes1−/− mice is triggered by the ectopic expression of Neurog3, which acts as a proendocrine gene to specify cell fate12,13.

Figure 5: Hes1 represses Neurog3 expression in the developing biliary epithelium.
figure 5

(ac) Hes1 is continuously expressed from the budding primordium of the common bile duct (a, arrow) to the epithelia of the extrahepatic biliary system. (dg) Ectopic Neurog3 expression of Hes1−/− embryos in the primordium of the bile duct (f, arrowhead), the abnormally aggregated biliary epithelial cells (f, red asterisk; see also Fig. 2e) and the remnant of the bile duct (g, red arrowhead). In Hes1+/+ littermates, Neurog3 is never detected in the biliary system, despite strong positive signals in the budding (d, arrow) and the mature ventral pancreas (Pa; e). CBD, common bile duct; CD, cystic duct; Du, duodenum; GB, gallbladder; HD, hepatic diverticulum; Li, liver; PV, main portal vein. Scale bars, 100 μm.

Our results show that endodermal Hes1 controls not only endodermal endocrine differentiation5,6 but also biliary organogenesis. Furthermore, they also indicate that the prospective biliary epithelium maintains the potential for pancreatic differentiation in the absence of Hes1. The ontogenetic relation between the common bile duct and the pancreas in sea lampreys supports the common evolutionary origin of these two organs. In this primitive vertebrate, adult endocrine pancreas tissue develops from the larval extrahepatic common bile duct through the process of transdifferentiation19. At a more general level, multipotency or plasticity, such as that seen in the biliary cells, may be a common characteristic of endoderm-derived organogenesis, similar to the conversion from pancreatic progenitors to duodenal cells that has recently been reported in Ptf1a-deficient mice20.

The development of endoderm-derived organs, including the pancreas and liver, is regulated by mesenchymal-epithelial cell interactions8,21,22,23. Expression of Ipf1 (ref. 24) in endodermal epithelial cells renders those cells responsive to mesenchyme-derived differentiation signals that induce pancreatic morphogenesis25. During mouse embryogenesis, Ipf1 was strongly expressed in the biliary system and the pancreas (Fig. 6). In addition, Hes1 represses Neurog3 (refs. 5,26), and we observed upregulation of expression of Neurog3 in bile ducts of Hes1-deficient mice (Fig. 5f,g). The facts suggest that Hes1 inhibits the pancreatic endocrine differentiation program in the biliary epithelium by repressing expression of Neurog3, probably by keeping the biliary epithelial cells unresponsive to mesenchyme-derived differentiation signals for the adjacent ventral pancreas. Some differentiation factors that are normally repressed by Hes1 may inhibit the development of pancreatic exocrine differentiation in the biliary system. Our finding that inactivation of Hes1 induces the conversion of biliary epithelium to pancreatic tissue may provide a substantial basis for future research concerning reconstitutional therapy of various endodermal cell types, including pancreatic β cells, using the abundant biliary cells.

Figure 6: Ipf1 protein is expressed in the biliary epithelial cells and the pancreas.
figure 6

Nuclear Ipf1 is detected in the gallbladder (GB), common bile duct (CBD) and the branching ventral pancreas (Pa) at E11.5 (see also Fig. 2b). Du, duodenum; Li, liver; PV, main portal vein. Scale bar, 100 μm.



We used Hes1-deficient mice established on a CD1 genetic background2 and control CD1 mice in this study. We bred the mice on site and analyzed embryos at E9.5 to E17.5. We carried out experiments using protocols approved by the Laboratory Animal Resource Center, University of Tsukuba.


We used primers and reaction conditions as described for amplification27. We isolated the gallbladder, liver, pancreas and brain from CD1 embryos and neonatal mice with forceps under a microscope. We used brain cDNA and water as positive and negative controls, respectively. We carried out semiquantitative PCR for Notch-related molecules using quantitatively normalized samples as described27.

Histological analyses.

We carried out all histological analyses of the embryonic biliary system using serial sagittal sections, except for those shown in Figure 3f,l. To select a central longitudinal section of the biliary system, we completely sectioned whole embryos and stained every 3rd (E9.5) or every 20th (E17.5) section with hematoxylin and eosin. We then stained all adjacent sections with DBA or by immunohistochemistry or in situ hybridization.

We carried out DBA staining using DBA–horseradish peroxidase (EY Laboratories) as described15. For immunostaining, we used rabbit antibody to glucagon (DAKO), guinea pig antibody to insulin (DAKO), mouse antibody to insulin (sigma), rabbit antibody to somatostatin (NICHIREI), rabbit antibody to pancreatic peptide (DAKO), rabbit antibody to cytokeratin (DAKO), rabbit antibody to carboxypeptidase A (ANAWA), rabbit antibody to amylase (Biomeda), rabbit antibody to Ipf1 (CHEMICON), rabbit antibody to Glut2 (Alpha Diagnostic) and mouse antibody to Isl1 (clones 39.4D5 and 40.2D6, Development Studies Hybridoma Bank, University of Iowa). Immunohistochemical signals were detected with peroxidase-labeled Envision+ antibody to rabbit IgG (DAKO), diaminobenzidine and ammonium nickel sulfate. The immunofluorescent secondary antibodies used were Cy3-conjugated antibody to rabbit IgG (Jackson ImmunoResearch Laboratories) and Alexa488-conjugated antibody to mouse IgG (Molecular Probes). For TUNEL assays, we used an In situ Cell Death Detection Kit (Roche Molecular Biochemicals). We carried out in situ hybridization of Hes1 and Neurog3 using digoxigenin-labeled cRNA probes according to the reported protocol28.