Notch2 signaling promotes biliary epithelial cell fate specification and tubulogenesis during bile duct development in mice


  • Potential conflict of interest: Nothing to report.


Intrahepatic bile duct (IHBD) development begins with the differentiation of hepatoblasts into a single continuous biliary epithelial cell (BEC) layer, called the ductal plate. During ductal plate remodeling, tubular structures arise at distinct sites of the ductal plate, forming bile ducts that dilate into the biliary tree. Alagille syndrome patients, who suffer from bile duct paucity, carry Jagged1 and Notch2 mutations, indicating that Notch2 signaling is important for IHBD development. To clarify the role of Notch2 in BEC differentiation, tubulogenesis, and BEC survival, we developed a mouse model for conditional expression of activated Notch2 in the liver. We show that expression of the intracellular domain of Notch2 (Notch2ICD) differentiates hepatoblasts into BECs, which form additional bile ducts in periportal regions and ectopic ducts in lobular regions. Additional ducts in periportal regions are maintained into adulthood and connect to the biliary tight junction network, resulting in an increased number of bile ducts per portal tract. Remarkably, Notch2ICD-expressing ductal plate remnants were not eliminated during postnatal development, implicating Notch2 signaling in BEC survival. Ectopic ducts in lobular regions did not persist into adulthood, indicating that local signals in the portal environment are important for maintaining bile ducts. Conclusion: Notch2 signaling regulates BEC differentiation, the induction of tubulogenesis during IHBD development, and BEC survival. (HEPATOLOGY 2009.)

During liver development both hepatocytes and biliary epithelial cells (BECs) arise from common bipotential progenitors called hepatoblasts.1–3 Hepatoblasts in the liver parenchyma differentiate into hepatocytes, whereas those adjacent to the portal mesenchyme differentiate into BECs. Early intrahepatic bile duct (IHBD) development in both humans and rodents is characterized by the formation of the so-called ductal plate, a single continuous cell layer containing BECs. During ductal plate remodeling tubular structures arise at distinct sites of the ductal plate and are subsequently incorporated into the portal mesenchyme. Postnatal nontubular ductal plate remnants get eliminated, whereas the tubular structures become mature IHBDs that dilate into the biliary tree.4–7 It is generally assumed that the periportal environment is required for BEC differentiation and tubulogenesis, but the exact mechanisms underlying these processes are poorly understood. Several factors contributing to bile duct development have recently been identified, including Notch2.6, 8 The Notch signaling pathway is highly conserved throughout evolution and plays an important role in cell fate determination by way of cell-cell contacts. Mammals express four Notch receptors (Notch1-4) with five ligands (Dll1, Dll3, Dll4, Jagged1, and Jagged2). Ligand-binding to Notch receptors on neighboring cells leads to the proteolytic processing and translocation of the Notch intracellular domain (NotchICD) into the nucleus. NotchICD then forms a complex with RBPjκ, leading to the transcriptional activation of Notch effector genes, such as Hairy and Enhancer of Split homologs (e.g., Hes1).9, 10

Alagille syndrome (AGS) is a rare hereditary multisystem disorder caused by haploinsufficiency of Jagged1 (OMIM #118450)11–13 as well as Notch2 mutations (OMIM #610205).14 AGS patients display a wide range of developmental abnormalities in liver, heart, eye, skeleton, and kidney. The abnormalities in liver development are characterized by neonatal jaundice, impaired differentiation of IHBDs, and chronic cholestasis.11–13 Mice with a haploinsufficiency for Jagged115 or a liver-specific ablation of Jagged116 exhibit no IHBD abnormalities. However, mice heterozygous for the Jagged1 and Notch2 null allele recapitulated most of the developmental abnormalities seen in human Alagille syndrome, including IHBD development defects.17 Similarly, liver-specific deletion of Notch2 using albumin-Cre (AlbCre) mice caused severe defects in IHBD development, characterized by disorganized tubular structures accompanied by portal inflammation, portal fibrosis, and foci of hepatocyte feathery degeneration in adulthood.18, 19 The formation of a ductal plate and the presence of BECs in these mice suggested that Notch2 is dispensable for BEC differentiation and mostly required for IHBD development. However, a role for Notch2 signaling in BEC differentiation cannot be excluded, because AlbCre-mediated Notch2 ablation experiments are not entirely conclusive. It is known that albumin expression begins around embryonic day (E)13.5 and that AlbCre-mediated gene ablation occurs progressively with age.20 Thus, as pointed out by Geisler et al.,18 partial Notch2 ablation or embryonic Notch2 levels may still allow differentiation of hepatoblasts into BECs, which also starts around E13.5. Unfortunately, mice with a complete ablation of Notch2 die before the onset of BEC differentiation,21 which prevented studying possible effects of Notch2 signaling in this process. Supporting that Notch signaling plays a role in BEC differentiation, it was shown that expression of activated Notch1 in cultured hepatoblasts represses hepatocyte differentiation and induces the expression of BEC markers.22 However, expression of activated Notch1 did not produce cells with the morphological characteristics of mature BECs and did not result in the formation of tubular structures. Notably, conditional ablation of Notch1 in the liver revealed that Notch1 signaling is dispensable for IHBD development.18, 23 Thus, the role of Notch2 during BEC fate specification, the induction of tubulogenesis and survival of BECs is still controversial. For this reason we wanted to further address the role of Notch2 during IHBD development by generating mice that express Notch2ICD in hepatoblasts, which mimics ligand-induced activation of Notch2. These mice not only allowed us to study the role of Notch2 signaling in its cognate portal environment, but also at ectopic lobular sites. Lobular Notch2ICD expression shed light on the inherent potential of Notch2 signaling in hepatoblasts, because it dissociated Notch2 signaling from local cues in the portal environment.


AGS, Alagille syndrome; BEC, biliary epithelial cell; CAGS, chicken β-actin; DBA, dolichos biflorus agglutinin; ES, embryonic stem; HNF, hepatocyte nuclear factor; IHBD, intrahepatic bile duct; Notch2ICD, Notch2 intracellular domain; RMCE, recombinase-mediated cassette exchange; ZO1, zona occludens 1.

Materials and Methods

Generation of N2ICD and N2ICD/AlbCre Mice.

Recombinase-mediated cassette exchange (RMCE) was used to introduce the complementary DNA (cDNA) encoding the intracellular domain of Notch2 (Notch2ICD; Swiss-Prot #O35516, aa1699-aa2470) with a C-terminal human 5xMyc-tag into the modified Rosa26 locus of Balb/c mouse embryonic stem (ES) cells (Fig. 1A). The Myc-tag allows immunohistochemical detection of transgenic Notch2ICD protein expression. This anti-Myc antibody (A-14, Santa Cruz, Santa Cruz, CA) is specific for the human Myc epitope and exhibits no crossreactivity with the endogenous mouse c-Myc in immunohistochemical experiments (Supporting Fig. 1). The RMCE plasmid contained a pBS-SK(+) backbone, with flipase recognition target (FRT) sites flanking the chicken β-actin (CAGS) promoter,24 a floxed transcriptional STOP cassette,25 the Notch2ICD-5xMyc cDNA, and an HSVtkNeo cassette. ES cells with cassette exchange events were identified by polymerase chain reaction (PCR). ES cells with preserved karyotypes were used for blastocyte injection as described.26 Chimeric mice were mated with Balb/c mice. Mice with the conditional Notch2ICD-Myc transgene (N2ICD mice) were identified by genotyping. N2ICD mice were born at a Mendelian ratio and exhibited no overt phenotype. N2ICD mice were crossed with AlbCre mice expressing Cre-recombinase under the liver-specific albumin promoter.20 Single-transgenic AlbCre and N2ICD littermates and wildtype mice were used as controls in our experiments. Genotyping of the mice was performed by TaqMan analysis. In brief, small tail biopsies were digested in proteinase-K containing lysis buffer overnight at 55°C. The 1:10 diluted digested samples were then genotyped by TaqMan-PCR using the following 5′-3′ primers and probes: Cre-forward (gccgcgcgagatatgg), Cre-reverse (gccaccagcttgcatgatc), Cre-probe (Fam-ccgcgctggagtttcaataccgg-Tamra), Rosa26-N2ICD-fw (atatccgcggtggagatcaa), Rosa26-N2ICD-rv (tagaccaggctgggctaaa), Rosa26-N2ICD-probe (MGB-cggtaccagatctc-Tamra). Southern blots with genomic liver DNA were done using a 1.2-kb hybridization probe derived from the neomycin resistance gene using a standard protocol (Fig. 1A,B). All animal experiments were performed in accordance with governmental guidelines and approved by the veterinary office of Basel.

Figure 1.

Liver-specific expression of Notch2ICD. (A) Scheme depicting the generation of N2ICD mice, which allow for Cre-mediated expression of Notch2ICD. The RMCE plasmid contains the ubiquitously active CAGS promoter. The CAGS promoter is silenced by a STOP sequence flanked by loxP sites. The Notch2ICD is C-terminally tagged with a human 5xMyc epitope. The CAGS promoter, STOP sequence, the Notch2ICD, and a neomycin resistance cassette (neoR) are flanked as a group by FRT sites. Flipase-mediated recombination in modified mouse ES cells was used to exchange a hygromycin resistance cassette (hygR) flanked by FRT sites with the Notch2ICD targeting construct in the ROSA26 locus. Correctly recombined ES cells were used to generate N2ICD mice. N2ICD mice were crossed with AlbCre mice to induce Cre-mediated excision of the STOP cassette flanked by loxP sites, resulting in liver-specific Notch2ICD expression. (B) Site-specific integration of the Notch2ICD targeting construct into the ROSA26 locus and Cre-mediated excision of the STOP cassette was analyzed on Southern blot using a neoR-specific hybridization probe and genomic liver DNA. Recombination of the Notch2ICD targeting construct into the Rosa26 locus yields a 5.6kb EcoRV fragment in N2ICD mice, which is absent in wildtype mice. N2ICD/AlbCre mice display an additional 4.1 kb band, indicating excision of the STOP cassette. (C) Western blot analysis of liver homogenates shows that Notch2ICD levels are highly increased in N2ICD/AlbCre mice compared to wildtype and single transgenic N2ICD mice. Myc-tagged Notch2ICD was only present in N2ICD/AlbCre mice, showing that the STOP cassette in N2ICD mice is not leaky. Actin controls for sample loading. (D) Light cycler quantitative real-time PCR for the Notch target gene Hes1 shows a 2-fold increase in Hes1 mRNA levels in N2ICD/AlbCre mice (n = 3), confirming that the Myc-tagged Notch2ICD protein is functional. Mean ± standard error of the mean (SEM); **P ≤ 0.01.

Immunohistochemistry, Western Blots, and Quantitative Real-Time PCR (RT-PCR).

Mice were sacrificed by cervical dislocation and livers were removed. Samples for protein, DNA, and RNA analysis were immediately frozen at −80°C. Samples for histological examination were fixed overnight with 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) at 4°C. Tissue used for zona occludens 1 (ZO1) staining was freshly frozen and postfixed in methanol at −20°C for 10 minutes. Ten-μm-thick cryostat sections were mounted, washed in PBS, and incubated for 1 hour in blocking solution (2% bovine serum albumin [BSA] and 0.2% Triton X-100 in PBS) followed by overnight incubation at 4°C in blocking solution containing primary antibodies or dolichos biflorus agglutinin (DBA, biotinylated or fluorescein-coupled, Vector Laboratories, Burlingame, CA). The primary antibodies used were rabbit anti-hepatocyte nuclear factor (HNF)4α (H-171, Santa Cruz), rabbit anti-HNF1β (H-85, Santa Cruz), goat anti-Myc (A-14, Santa Cruz), rabbit anti-ZO1 (Zymed, San Francisco, CA), rabbit anti-Ki67 (Novocastra, UK) and rat anti-Notch2 (C6551.6DbHN, Developmental Studies Hybridoma Bank, Iowa City, IA). After rinsing in PBS, sections were incubated in blocking solution containing secondary antibodies for 1 hour at room temperature. The secondary antibodies used were Cy3-conjugated donkey anti-rabbit and Cy5-conjugated donkey anti-sheep (Jackson Immunoresearch, UK). Four-μm-thick paraffin-embedded tissue sections were rehydrated and antigen retrieval was performed in citrate buffer (10 mM, 0.01% Tween20, pH 6.0) for 20 minutes. Incubation with the primary antibodies was performed as described above. Immunostaining was completed using the Vectastain ABC Kit (Vector Laboratories). Cellular colocalization of immunohistochemical markers was quantified (≥3 mice per genotype and age) using Volocity Software (Improvision, UK). Immunoblotting and light cycler quantitative RT-PCR was performed as described.27


Liver-Specific Transgenic Expression of Notch2ICD.

To study the role of Notch2 signaling during IHBD development, we developed a mouse model for conditional expression of Notch2ICD in hepatoblasts (Fig. 1A). We generated N2ICD mice harboring a Notch2ICD-Myc fusion construct downstream of a floxed STOP cassette and crossed them with AlbCre mice,21 which yielded N2ICD/AlbCre mice. The expression of Cre recombinase from the albumin promoter leads to the excision of the STOP cassette and the expression of the Myc-tagged Notch2ICD protein. In the embryonic liver the onset of albumin expression in hepatoblasts occurs around E13.5, when bipotential hepatoblasts begin to differentiate into either hepatocytes or BECs.5 Successful excision of the STOP cassette in genomic N2ICD/AlbCre liver DNA was confirmed by Southern blot analysis (Fig. 1B). Western blot analysis of liver homogenates confirmed the expression of the Myc-tagged Notch2ICD protein in N2ICD/AlbCre mice (Fig. 1C). N2ICD/AlbCre livers exhibited a 2-fold increase in the mRNA level of the Notch2 target gene Hes1, thus confirming that the Myc-tagged Notch2ICD protein is signaling-competent (Fig. 1D).

Notch2ICD Regulates Differentiation of Hepatoblasts into BECs.

To study the fate of Notch2ICD-expressing cells, we performed coimmunostaining studies at different timepoints during embryonic and postnatal development. At E16.5 transgenic Notch2ICD expression, detected by Myc-staining, was observed in a mosaic pattern in both periportal and lobular regions (Fig. 2A). 96% of Notch2ICD-expressing cells were positive for the hepatoblast marker HNF4α and 70% expressed HNF1β, a transcription factor mediating BEC fate (Fig. 2B).28 At this developmental stage, BECs in both N2ICD/AlbCre and control mice showed negligible DBA staining, a lectin that specifically binds to mature BECs.29 This is consistent with previous reports showing that at E16.5 BECs are still at an immature stage.30 At E18.5 the number of HNF1β+/Myc+ cells increased to more than 80%, whereas the number of HNF4α+/Myc+ cells decreased to ≈30%, indicative of a differentiation of HNF4α+/HNF1β+ cells to HNF4α−/HNF1β+ cells (Fig. 2B). HNF1β+ cells close to the periportal environment were strongly DBA+, whereas those in lobular regions were weakly DBA+ (Fig. 2A). Notably, both Myc+ and Myc− BECs within the portal region were DBA+ (data not shown). At birth (postnatal day [P]0), HNF1β+/Myc+ cells in the lobules became strongly DBA+, suggesting maturation of HNF1β+/DBA− to HNF1β+/DBA+ cells in a portal-to-lobular manner (Fig. 2A). During early postnatal development the number of HNF1β+/Myc+ cells further increased and peaked at P4, whereas the number of HNF4α+/Myc+ cells decreased to less than 5% at P4. This suggests a progressive differentiation from HNF4α+/HNF1β+ into HNF4α−/HNF1β+ cells. Interestingly, a small subset of Myc+ cells remained HNF4α+/HNF1β− and their number increased progressively from P4 to P90 (Fig. 2A, left panels). Due to the misspecification of lobular Myc+ hepatoblasts into HNF4α−/HNF1β+/DBA+ BECs cells, the number of HNF4α+ cells was extremely low at P4, but increased progressively with age until a normal level of about 60% was reached at P90 (Fig. 2C). We therefore assessed proliferation in N2ICD/AlbCre mice by Ki67 staining (Supporting Fig. 2A,B). Increased proliferation was observed at P4 and P90 but not at E18.5. Proliferation was restricted to morphologically identified hepatocytes, whereas proliferation in DBA+ BECs was negligible (Supporting Fig. 2A). In addition, costaining for Myc and Ki67 showed that Notch2ICD expression did not alter proliferation (data not shown). The postnatal increase in hepatocyte proliferation in N2ICD/AlbCre mice resulted in increased liver weight (Supporting Fig. 2C). We assume that the increased postnatal proliferation of hepatocytes in N2ICD/AlbCre mice compensates for the extremely low number of hepatocytes seen at P4 (Fig. 2A, 2C). In summary, our data support that Notch2ICD expression in bipotential hepatoblasts promotes their progressing differentiation from HNFα+/HNF1β+/DBA− to HNF4α−/HNF1β+/DBA− cells and eventually to HNF4α−/HNF1β+/DBA+ BECs (Fig. 5).

Figure 2.

Notch2ICD differentiates hepatoblasts into BECs. (A) Liver sections were prepared from mouse embryos (E16.5, E18.5), postnatal (P0, P4, P10), and adult (P90) N2ICD/AlbCre and control mice (n = 3-5). Markers for hepatoblasts/hepatocytes (HNF4α) and BECs (HNF1β and DBA) were used to follow differentiation of hepatoblasts into BECs. HNF1β is expressed at early stages of BEC differentiation, whereas DBA binds to mature BECs. The portal vein is outlined with a dashed line. Immunostaining using an anti-Myc antibody reveals a mosaic expression pattern of the Myc-tagged Notch2ICD in both portal and lobular regions. Coimmunostaining for Myc and HNF4α (left panels) or HNF1β (middle left panels) shows progressing differentiation of Notch2ICD-expressing hepatoblasts toward the biliary lineage. HNF1β/DBA coimmunostaining shows that the maturation of BECs occurs in a portal (E18.5) to lobular (P0-P4) manner in N2ICD/AlbCre mice (middle right panels). BEC differentiation through a HNF1β+/DBA− intermediate stage is also seen in control sections (right panels). Ductal plate BECs (arrows) persist to P90 in N2ICD/AlbCre mice, whereas they are gradually eliminated during postnatal development in control mice. Multiple ectopic tubular structures (arrow heads) are present throughout the liver in P4 N2ICD/AlbCre mice. Ectopic tubular structures and DBA markers gradually disappeared within lobular regions after P4, whereas periportal extra ducts survive into adulthood (asterisk). (B) The percentage of Myc+ cells expressing HNF4α and HNF1β was determined by coimmunostaining (E16.5-P90) and quantified. Both HNF4α and HNF1β are coexpressed in Myc+ cells at E16.5–E18.5. From P0 onwards Myc+ cells mostly express the BEC marker HNF1β+. Notably, a small percentage of Myc+ cells remain HNF4α+ and their percentage increases progressively from P4 to P90. (C) The number of HNF4α and HNF1β-expressing cells in N2ICD/AlbCre mice normalized to 4′,6-diamidino-2-phenylindole (DAPI) shows an increase in HNF1β+ cells, accompanied by a dramatic decrease of HNF4α+ cells, which peaks at P4. After P4 the number HNF4α+ cells increases progressively until P90. Bar diagrams (B,C) show mean ± SEM. Scale bar = 50 μm.

Figure 5.

Model of IHBD development in N2ICD/AlbCre mice. Upon Notch2 activation, bile duct development begins with the differentiation of HNF4a+ hepatoblasts into HNF1β+/DBA+ BECs, passing HNF4α+/HNF1β+, and HNF4α−/HNF1β+ BEC differentiation stages. Although this is the predominant differentiation pathway of Notch2ICD-expressing hepatoblasts, we also observed some Myc+/HNF4α+ cells with mature hepatocyte morphology. These cells could either derive from lobular HNF4α−/HNF1β+/DBA+ BECs that transdifferentiate into hepatocytes or from hepatoblasts that already were committed to become hepatocytes, possibly due to delayed expression of Notch2ICD (dashed arrows). Ectopic Notch2 signaling promotes BEC differentiation and the formation of tubular structures in the lobules, showing that Notch2 signaling can promote tubulogenesis remote from the periportal environment. This is consistent with a conditional lack of Notch2 signaling (Notch2 cKO) disrupting tubulogenesis (*Geisler et al.,18 Lozier et al.,19 highlighted in gray). Notch2ICD-expressing tubular structures are only maintained in the periportal environment. Tubular structures in the lobules are eliminated during postnatal development, presumably leaving behind dedifferentiated HNF1β+/DBA− bile duct cells. This suggests that cues from the periportal environment are required for maintaining bile ducts in adult livers.

Notch2ICD-Expressing Biliary Epithelial Cells in the Lobules form Ectopic Tubular Structures.

During ductal plate remodeling, tubular structures forming bile ducts arise at distinct sites of the ductal plate. Findings from AGS patients11–14 and mouse studies15–19, 30 both implied a role for Notch2 in ductal plate remodeling and tubulogenesis. We therefore examined whether Notch2ICD expression in Alb-Cre/N2ICD mice is able to induce tubulogenesis at ectopic lobular sites. We analyzed tubulogenesis at P4 by costaining for Myc/HNF1β/DBA (Fig. 3A-D) and HNF1β/hematoxylin (Fig. 3E,F). We found that Notch2ICD-expressing BECs in P4 N2ICD/AlbCre mice indeed formed ectopic tubular structures in the lobules (Figs. 2A, 3A-D). Most of these ectopic tubular structures had a lumen (Fig. 3E) and were HNF1β+/DBA+ (Fig. 3C,D). However, some ectopic tubular structures were disorganized and some HNF1β+/DBA+ cells did not participate in tubular structures. Ectopic tubular structures first appeared at P0 and their number increased until P4. At P10 most ectopic tubular structures had disappeared (Fig. 2A). These data suggest that Notch2 signaling promotes tubulogenesis, but that it is not sufficient to maintain tubular structures into adulthood.

Figure 3.

Notch2ICD induces ectopic tubular structures remote from the periportal environment. Liver cryosections from P4 N2ICD/AlbCre mice were stained for Myc, HNF1β, and DBA (A-D). Paraffin sections from P4 N2ICD/AlbCre and control mice were stained for the BEC marker HNF1β and hematoxylin (E,F). (A-D) Ectopic tubular structures containing BECs are present throughout the lobules of N2ICD/AlbCre mice, as demonstrated by HNF1β staining (A). The outlined area was magnified showing HNF1β (B), Myc/HNF1β (C), and HNF1β/DBA (D) stained ectopic tubular structures. (E,F) HNF1β/hematoxylin-stained N2ICD/AlbCre (E) lobular sections show ectopic tubular structures, mostly containing a lumen. Control mice do not express HNF1β in the lobules (F). Cell clusters with strong hematoxylin staining represent hepatic hematopoietic stem cells. Scale bar (A) = 200 μm, (B-F) = 50 μm.

Transgenic Notch2ICD Expression Results in an Increased Number of Bile Ducts.

We next investigated whether the expression of Notch2ICD results in an increased number of bile ducts in the periportal environment. We stained liver sections from P90 N2ICD/AlbCre (Fig. 4A) and littermate control (Fig. 4B) mice for DBA and hematoxylin and quantified the number of bile ducts per portal tract. We observed up to seven bile ducts per portal tract in N2ICD/AlbCre mice, whereas control mice never showed more than four ducts per portal tract. Portal tracts with five or more bile ducts were seen in 8.3% of all N2ICD/AlbCre portal tracts analyzed (Fig. 4G). N2ICD/AlbCre mice also showed a significant increase in the mean number of DBA+ ducts per portal tract compared to control littermates (2.20 ± 0.06 versus 1.52 ± 0.04; Mann-Whitney test P = 0.0022, n = 6 mice per genotype). We next addressed whether Myc+/DBA+ ducts are connected to the biliary canaliculi network, a tight junction network transporting bile from hepatocytes to the bile ducts.31 Myc+/DBA+ ducts stained for the tight junction marker ZO1 at P10 and P90 (Fig. 4E,F). This indicates that Myc+/DBA+ ducts are connected to the biliary network, which suggests that these bile ducts are functional. In support of this, bilirubin, alanine aminotransferase (ALAT), alkaline phosphatase (AP) (Supporting Fig. 3), and gamma glutamyl transferase (GGT) levels (<5 U/L for all animals tested) were normal in N2ICD/AlbCre mice.

Figure 4.

Adult (P90) N2ICD/AlbCre mice have an increased number of bile ducts per portal tract. (A,B) Hematoxylin- and DBA-stained paraffin-embedded sections of N2ICD/AlbCre (A) mice and control (B) livers showing six and two ducts (arrowheads) per tract, respectively. Five and more bile ducts per tract were only seen in N2ICD/AlbCre mice, but were relatively rare (see Fig. 4G for quantification). Some DBA+ ductal plate cells (arrows) were still present in N2ICD/AlbCre mice at P90 (A) but not in control mice (B). (C,D) Ectopic lobular BECs in N2ICD/AlbCre mice remain HNF1β+ (C). No lobular HNF1β+ cells are found in control mice (D). (E,F) Coimmunostaining for Myc, DBA, and the tight junction marker ZO1 shows that Myc+ periportal bile ducts are connected to the biliary canaliculi network in both P10 (E) and P90 (F) N2ICD/AlbCre mice. Areas outlined in (C,E,F) are enlarged in the corresponding insets. (G) Number of bile ducts per portal tract in N2ICD/AlbCre and control littermate mice (n = 6 per genotype). N2ICD/AlbCre mice had up to seven bile ducts per portal tract, controls never had more than four. pv, portal vein. Bar diagram (G) shows mean ± SD. Scale bar (A-D) = 50 μm, (E-F) = 100 μm.

Myc+/DBA+ ducts in the periportal region of N2ICD/AlbCre mice persisted into adulthood. In contrast, ectopic tubular structures were no longer observed in the lobules of adult mice. However, some loosely associated Myc+/HNF1β+ cells were still present at P90 in the lobules of N2ICD/AlbCre mice but not of control mice (Fig. 4C,D). It appears that the numbers of these cells decreased after P4 and that they lost their DBA marker (Fig. 2A). We further observed that ductal plate cells, which normally get eliminated within the first 2 weeks after birth, are still present at P90 in N2ICD/AlbCre mice and that these cells express Notch2ICD (Figs. 2A, 4A). This suggests that Notch2ICD promotes BEC survival. However, Notch2ICD expression is clearly not sufficient to maintain bile ducts in the absence of additional portal signals.


Our understanding of the mechanisms regulating IHBD development has improved since the discovery of Jagged1 and Notch2 mutations in AGS patients.11–14 There is experimental evidence showing that Notch2 signaling is required for tubulogenesis of bile ducts during ductal plate remodeling.15–19, 30 Whether Notch2 signaling is required for the differentiation of hepatoblasts into BECs is less clear. Notch2 knockout studies using AlbCre mice for conditional gene ablation tend to support that Notch2 is dispensable for BEC differentiation.18, 19 However, AlbCre-driven gene ablation in embryos is incomplete and occurs progressively with age.20 Therefore, BEC development could still be supported by residual Notch2 signaling. Moreover, Notch1ICD expression in cultured hepatoblasts caused the induction of BEC markers, whereas hepatocyte markers were repressed, pointing to a possible role for Notch signaling during BEC differentiation.22 It is known that periportal hepatoblasts differentiate into BECs, whereas lobular hepatoblasts differentiate into hepatocytes.6 One possibility is that cell-cell contacts of Notch2 receptor bearing hepatoblasts and ligand-expressing cells in the portal environment are responsible for the spatially confined formation of BECs. To test this possibility, we developed a novel mouse model that allows for conditional expression of Notch2ICD in bipotential hepatoblasts. We show that Notch2ICD expression not only induces BEC differentiation in periportal but also in lobular regions, clearly demonstrating that Notch2 signaling can induce BEC differentiation remote from periportal cues. Notch2ICD expression in HNF4α+ hepatoblasts triggers progressing differentiation to HNF4α−/HNF1β+/DBA+ BECs through HNFα+/HNF1β+/DBA− and HNF4α−/HNF1β+/DBA− intermediate stages. Although this is clearly the predominant differentiation pathway of Notch2ICD-expressing hepatoblasts, we also observed Myc+/HNF4α+ cells with mature hepatocyte morphology in P90 N2ICD/AlbCre mice. These cells could derive from lobular HNF4α−/HNF1β+/DBA+ cells that transdifferentiate into hepatocytes due to the lack of signals required for the maintenance of the BEC phenotype in the lobular environment. Alternatively, these cells might represent a subset of hepatoblasts that were already committed to the hepatocyte fate at the time when Notch2ICD expression occurred. It is therefore possible that progressing transdifferentiation of HNF4α−/HNF1β+/DBA+ cells into hepatocytes as well as a compensatory proliferation of Notch2ICD-expressing hepatocytes contribute to the increase in Myc+/HNF4α+ cells observed from P4 to P90 (Fig. 2B).

At P4, N2ICD/AlbCre livers featured multiple ectopic tubular structures, mostly containing a lumen. Remarkably, tubulogenesis of BECs not only occurred in periportal but also in lobular regions, suggesting that Notch2 signaling can promote tubulogenesis in the absence of periportal cues. However, diffusible periportal cues may still be necessary for ectopic BEC differentiation and the formation of ectopic tubular structures. At P10 most ectopic tubular structures in the lobule were lost. In contrast, the additional periportal tubular structures survived into adulthood, which resulted in an increased number of ducts per portal tract. The periportal ducts were connected to the biliary canaliculi network, because Myc+/DBA+ ducts stained for ZO1, thus indicating functional bile ducts. N2ICD/AlbCre mice displayed preserved liver chemistries, supporting that their biliary system functions normally. In summary, our data indicate that signals from the periportal environment are crucial for preserving functional bile ducts in adult livers.

It was proposed that Notch2 signaling is absent in ductal plate cells that do not contribute to bile duct formation and that these cells get eliminated.30 In agreement with this, ductal plate cells were progressively eliminated within the first 2 weeks after birth in control mice. However, in N2ICD/AlbCre mice Notch2ICD-expressing ductal plate cells persisted into adulthood, implicating Notch2 signaling in BEC survival. Notch1ICD was shown to increase survival of neuronal precursor cells by up-regulating the antiapoptotic proteins Mcl1 and Bcl2.32 It is possible that Notch2 promotes survival by way of the same mechanism.

The present study shows that transgenic expression of Notch2ICD in hepatoblasts triggers differentiation into BECs, induces tubulogenesis, and promotes BEC survival. However, the spatial and temporal profile of Notch2 signaling during normal IHBD development remains to be addressed in more detail. Kodama et al.30 showed that Notch2 and Hes1 expression is restricted to the parts of the ductal plate where tubulogenesis occurs. Hes1 null mice lacked proper ductal plate remodeling despite the presence of a ductal plate,30 similar to mice with a liver-specific Notch2 ablation.18, 19 As reviewed by Frederic Lemaigre,8 these findings suggest that the Jagged1-Notch2-Hes1 cascade induces ductal plate remodeling. However, it is still unclear how Notch2 contributes to the formation of mature bile ducts emerging from the single-layered ductal plate. Our data show that hepatoblasts differentiate into BECs upon activation of the Notch2 pathway. Jagged1 expression was found within ductal plate BECs.33 Thus, ductal plate BECs expressing Jagged1 could activate Notch2 in adjacent hepatoblasts, differentiating them into BECs. Our data also show that activated Notch2 induces BECs to form tubular structures. It is therefore possible that Notch2 regulates ductal plate remodeling by differentiating neighboring hepatoblasts into BECs, which in turn participate in tubulogenesis.

The main findings of our study are summarized in the model presented in Fig. 5. We demonstrate that transgenic Notch2ICD expression in bipotential hepatoblasts leads to their differentiation into BECs and to the formation of additional tubular structures in portal regions and ectopic tubular structures in lobular regions. Additional periportal ducts are connected to the biliary tight junction network and maintained in adult mice. In contrast, ectopic lobular ducts are lost during postnatal development, suggesting a crucial role for the portal environment in mediating bile duct maintenance. Remarkably, ductal plate cells that express transgenic Notch2ICD persist into adulthood, suggesting that Notch2 also plays a role in BEC survival.


We thank Lukas Landmann, Beat Erne, Thomas Zeis, Luigi Terracciano, and Vincenza Carafa-Tornillo for help with the histological analysis, Freddy Radtke and Andrea Durham for the mNotch2ICD plasmid, and Mira Susa for functional testing of the RMCE plasmid in cultured cells. We thank Martin Gassmann, David Semela, Franziska Schatzmann, Brian Hemmings, Adrian Merlo, Hans-Rudolph Brenner, Audree Pinard, and Nicole Schaeren-Wiemers for helpful discussions and critical reading of the article. We also thank Renato Zedi and his team for excellent animal caretaking.