Canonical Notch2 signaling determines biliary cell fates of embryonic hepatoblasts and adult hepatocytes independent of Hes1


  • Potential conflict of interest: Nothing to report.


Notch signaling through the Notch2 receptor is essential for normal biliary tubulogenesis during liver development. However, the signaling events downstream of Notch2 critical for this process are less well defined. Furthermore, whether Notch signaling also underlies adult hepatic cell fate decisions is largely unknown. By implementing different genetic mouse models, we provide a comprehensive analysis that defines the role of Notch in cell fate control in the developing and adult liver. We show that cell-specific activation of Notch2 signaling by a Notch2IC (N2IC) transgene leads to rapid biliary specification of embryonic hepatoblasts, but also—when expressed in up to 6-month-old adult livers—rapidly reprograms adult hepatocytes to biliary cells with formation of tubular-cystic structures. When directed specifically to the adult biliary and facultative liver progenitor cell compartment, Notch2 is capable of inducing a ductular reaction. Furthermore, we characterized the significance of key effectors of canonical Notch signaling during normal development and in N2IC-expressing models. We demonstrate that tubule formation of intrahepatic bile ducts during embryonic development as well as N2IC-induced specification and morphogenesis of embryonic hepatoblasts and biliary conversion of adult hepatocytes all critically rely on canonical Notch signaling via recombination signal binding protein (RBP)-Jκ but do not require Hes1. Conclusion: Notch2 appears to be the main determinant not only of biliary commitment of embryonic hepatoblasts during development but also of biliary reprogramming of adult hepatocytes. Notch2-dictated cell fates and morphogenesis in both embryonic hepatoblasts and adult hepatocytes rely on canonical Notch signaling but do not require Hes1. Adult liver cells possess a remarkable plasticity to assume new cell fates when embryonic signaling pathways are active. (HEPATOLOGY 2013)

Hepatocytes and biliary cells both arise from embryonic bipotential hepatoblasts during liver development. In the adult liver, in the setting of acute or chronic liver injury, hepatocytes are believed to arise also from a facultative liver stem cell compartment when proliferation of mature hepatocytes is impaired. Then, proliferating oval-shaped cells can be observed, a process called oval cell reaction or ductular reaction.1, 2 Oval cells are proposed to derive from the biliary compartment and to reside in a quiescent state in the Canals of Hering. Recent evidence from lineage-tracing studies suggests that they are, in fact, a progeny of the embryonic ductal plate,3 which is a periportal layer of embryonic biliary precursors occurring during liver development. Although controversially debated, there is the hypothesis that not only the biliary compartment (via oval cells) can give rise to hepatocytes, but also that mature hepatocytes can function as facultative stem cells and replenish the biliary compartment by transdifferentiation when proliferation of both mature bile ducts and progenitor cells is impaired.1, 4 The cellular mechanisms that potentially drive transdifferentiation of adult hepatocytes in vivo are unknown. However, emerging concepts suggest that signaling pathways essential for embryonic liver development may also navigate cell fates in liver repair during adulthood.5 In this context, the Notch signaling pathway, for which a fundamental role in embryonic bile duct development is well established,6-10 has recently come into focus.11

Bile duct development starts with embryonic hepatoblasts forming a single-layered ring of condensed epithelial cells expressing biliary-specific proteins around the larger portal veins called the ductal plate. In a second step, adjacent hepatoblasts at distinct sites of the ductal plate also become committed to the biliary lineage, thus forming a second ductal plate layer. Lumina arise at these sites that after further remodeling give rise to intrahepatic bile ducts (IHBDs).12 Notch signaling is reported to be critical for both sequential steps, biliary specification of hepatoblasts as well as tubule formation.6 While the cellular and molecular events leading to the formation of the first ductal plate layer are less well characterized, recent evidence supports the concept that Notch is especially important for the specification of the second ductal plate layer and consecutive tubule formation. Summing up the results from genetic mouse models, this Notch signaling axis comprises the Notch ligand Jagged1 in the portal mesenchyme as well as the Notch receptor Notch2, the DNA-binding recombination signal binding protein (RBP)-Jκ, and the Notch target gene Hes1 in hepatoblasts. Mice with genetic deletion of the Jagged1 gene in the portal mesenchyme13 or mice with hepatoblast-specific deletion of Notch2 or RBP-Jκ, the key effector protein of canonical Notch signaling,6, 10 all display similar defects in perinatal biliary tubulogenesis.7, 9, 10 Hes1 is believed to be the critical Notch-effector for biliary tubulogenesis, because Hes1 null animals were reported to also lack biliary tubule formation at birth.14

Here, by implementing distinct genetic Notch gain-of-function and loss-of-function mouse models specific for embryonic and adult liver cell compartments, we provide a comprehensive analysis to define the role of Notch in cell fate control in the developing and adult liver. We show that RBP-Jκ-dependent Notch2-signaling directs biliary cell fate and subsequent morphogenesis of hepatoblasts to give rise to bile ducts, processes that surprisingly we find not to require Hes1. Furthermore, we provide direct in vivo evidence that adult hepatocytes possess the plasticity to transdifferentiate to tubule forming biliary cells and that this event can be induced by Notch2. Beyond that, we identify Notch2 as a potential regulator of the adult liver progenitor cell niche.


IHBD, intrahepatic bile ducts; HNF1β, hepatocyte nuclear factor 1β; N2IC, Notch2 intracellular domain.

Materials and Methods

Mouse Strains.

For activation of Notch2 signaling in hepatoblasts, mice carrying one conditional Notch2IC allele in the murine rosa26-locus (R26loxP-Stop-loxP-Notch2IC, henceforth R26N2IC animals)15 were crossed with transgenic mice carrying a Cre gene under control of the albumin enhancer promoter (AlbCre mice).16 In these mice (R26N2ICAlbCre) Cre-mediated excision of a STOP-cassette leads to the expression of the Notch2 intracellular domain (N2IC) under the transcriptional control of the CAGGS promoter. For expression of N2IC in adult livers, R26N2IC animals were crossed with MxCre animals to obtain R26N2ICMxCre animals, where Cre expression is effectively induced in liver cells after intraperitoneal poly(I)-poly(C) injection17 (polyIC, 10 μg/g or 2.5 μg/g body weight [BW] intraperitoneal, Amersham Biosciences). HNF1βCreERT2 animals18 were used for conditional expression of N2IC specifically in the liver biliary and progenitor cell compartment (R26N2ICHNF1βCreERT2). Cre expression was induced by a single intraperitoneal injection of tamoxifen (Sigma, dissolved in corn oil at 100 μg/g BW). Cre expression in HNF1βCreERT2 or MxCre animals was analyzed in R26loxP-Stop-loxP-tdTomato or R26loxP-Stop-loxP-lacZ (henceforth R26Tom and R26lacZ) reporter mice.19 For further genetic characterization of the Notch signaling pathway conditional mutants for RBP-Jκ (RbpjF/F)20 or Hes1 (Hes1F/F)21 were interbred with AlbCre, MxCre and/or R26N2IC animals to obtain RbpjF/FAlbCre, Hes1F/FAlbCre, R26N2ICRbpjF/FAlbCre, R26N2ICHes1F/FAlbCre, R26N2ICRbpjF/FMxCre, or R26N2ICHes1F/FMxCre animals. A schematic overview of all mouse lines is given in Supporting Fig. 1 and Supporting Table 1. Genotyping was done by polymerase chain reaction (PCR) of tails (Supporting Table 2).

All strains were maintained on a mixed background. Cre-negative littermates served as control unless stated otherwise. Mice were handled according to protocols that follow the national guidelines for ethical animal treatment and all experiments were performed according to the protocols approved by our Institutional Animal Care and veterinarian office. Additional details and methods are provided in the Supporting Materials and Methods.


Activation of Notch2 Signaling in Embryonic Hepatoblasts Results in Perinatal Lethality with Loss of Normal Liver Architecture Due to Aberrant Bile Duct Formation.

To characterize the role of Notch2 and its downstream signaling in biliary cell fate control and morphogenesis we generated a mouse line that specifically expresses N2IC in hepatoblasts (R26N2ICAlbCre). At birth, livers of R26N2ICAlbCre animals showed an induction of Notch target genes (Fig. 1A). Newborn mutant animals were smaller and frequently displayed large cysts filled with bile or blood (Fig. 1A). Histological analysis of N2IC-expressing livers revealed a complete loss of the normal liver architecture reminiscent of a “Swiss cheese pattern” (Fig. 1B; Supporting Fig. 2A). N2IC-expressing cells lined tubular, tubular-cystic, and microcystic structures (Supporting Fig. 2A) which stained positive for biliary markers such as hepatocyte nuclear factor 1β (HNF1β), Sox9, DBA, or E-cadherin (Fig. 1C) but lost the hepatocyte marker HNF4α (Supporting Fig. 2B/C). Just like regular bile ducts in controls, these cells displayed apical ciliogenesis identified by acetylated-tubulin staining (Fig. 1D). Concomitantly, albumin expression was strongly reduced in livers of R26N2ICAlbCre animals (Fig. 1E). These results show that activation of Notch2 signaling rapidly converts hepatoblasts toward the biliary lineage, which is in line with previous reports.22 However, the phenotype observed in our R26N2ICAlbCre mice was more dramatic and almost all mutant animals died within 24 hours after birth. Macroscopic examination of surviving adult R26N2ICAlbCre animals revealed hepatomegaly with widespread yellowish lesions (Supporting Fig. 3A). In histological analysis these lesions consisted of N2IC-expressing multicystic structures positive for biliary markers (Supporting Fig. 3A,B, upper row). These tumors were reminiscent of extensive biliary hamartoma; however, detection of focal epithelial dysplasia and the presence of inflammatory stroma development suggest their malignant potential (Supporting Fig. 3B, lower panel). We also noted large areas with apparently normal hepatocytes staining negative for N2IC and Sox9 (Supporting Fig. 3C), indicating that these hepatocytes have escaped Cre-driven recombination and failed to be committed to the biliary lineage.

Figure 1.

Activation of Notch2 signaling in embryonic hepatoblasts results in perinatal lethality with loss of normal liver architecture due to aberrant formation of biliary structures. (A) Hepatoblast-specific expression of N2IC in livers of newborn R26N2ICAlbCre animals results in increased expression of Notch target genes as assessed by real-time RT-PCR. (Bottom) Newborn R26N2ICAlbCre are smaller than controls and frequently display bile-filled or hemorrhagic cysts (asterisks). (B) Hematoxylin and eosin (H&E) staining of control and R26N2ICAlbCre livers reveals loss of normal liver architecture with panlobular appearance of tubular-cystic structures in mutant animals. (C) Immunostaining for N2IC, HNF1β, Sox9, E-cadherin, and DBA demonstrate the biliary phenotype of N2IC-expressing ectopic tubules in livers of R26N2ICAlbCre mice at P0. (D) Immunofluorescence staining with the cilia marker anti-acetylated tubulin shows the preserved apicobasal polarity of these structures (arrows mark a WT bile duct with stained cilia). (E) Real-time RT-PCR analysis shows reduced albumin mRNA expression in R26N2ICAlbCre animals at P0. Data shown in (A,E) are mean values ± standard deviation (SD) (n = 3-6 for each genotype). PV, portal vein. BD, bile duct. Scale bars in (B) top = 100 μm, bottom 25 μm; (C) = 25 μm; (D) = 15 μm.

Conditional Activation of Notch2 Signaling in Adult Livers Results in Rapid Conversion of Mature Hepatocytes to Biliary Cells.

Results from R26N2IC AlbCre animals show that Notch2 activation directs cell fates of hepatoblasts to form biliary-like structures that persist into adulthood. It is unknown to what extent adult hepatocytes retain this cellular plasticity in response to Notch signals. We therefore sought to characterize the potential of Notch2 to mediate biliary conversion of adult hepatocytes. For this, we analyzed R26N2ICMxCre animals in which N2IC expression can be induced in liver cells upon pIC injection. Seven days after pIC injection 4 to 6-week-old R26N2ICMxCre mice displayed enlarged icteric livers (Fig. 2A, left). Histologically, tubular-cystic and microcystic structures replaced the entire liver (Fig. 2A, right) and a loss of regular hepatocytes with horizontal polarity was noted. N2IC-expressing cells stained positive for HNF1β and CK19 and a significant portion of these structures exhibited typical cholangiocytic vertical polarity. They were delineated by a continuous basal layer of laminin and showed apical expression of mucin-1 as well as apical formation of primary cilia (Fig. 2B), arguing for their biliary phenotype. These morphological changes were accompanied by Sox9 expression and decline of the hepatocyte markers HNF4α and albumin (Fig. 2C). Formation of tubular-cystic structures first appeared at day 5-6 after pIC injection, most pronounced around portal tracts, and from day 7 onwards progressively involved the liver lobule accompanied by a portal-to-central loss of HNF4α. By day 10 nearly all N2IC-expressing cells were HNF4α-negative (Supporting Fig. 4A,B). N2IC expression in all hepatocytes (pIC 10 μg/g BW) led to the formation of predominantly cystic architectural changes involving the entire liver. In contrast, “low dose pIC” injection (pIC 2.5 μg/g BW), resulting in Cre expression in less that 50% of hepatocytes, led to the formation of singular ectopic biliary-like ductules in otherwise normal liver parenchyma (Fig. 2D; Supporting Fig. 4A/B).

Figure 2.

Conditional activation of Notch2 signaling in adult livers results in rapid conversion of mature hepatocytes to biliary cells. For postnatal inducible expression of N2IC 4 to 6-week-old (A-C) or 6-month-old (E) R26N2ICMxCre or R26N2IC animals were treated with a single injection of pIC (1× pIC intraperitoneal, 10 μg/g) and analyzed at the indicated times. (A) Seven days post pIC injection R26N2ICMxCre animals display enlarged and icteric livers (left) with loss of normal liver architecture replaced by tubular-cystic structures (right). (B) N2IC-expressing tubules stain positive for HNF1β and CK19 and exhibit apicobasal polarity as assessed by staining of mucin-1, laminin, and acetylated tubulin. (C) pIC-induced hepatic expression of N2IC results in increased Sox9 protein levels and in loss of HNF4α or albumin expression as assessed by immunoblot (left) and real-time RT-PCR (right) analysis. (D) Cre expression profile after “normal dose pIC” (10 μg/g BW) and “low dose pIC” (2.5 μg/g BW) was assessed by X-gal staining in R26LacZMxCre reporter animals (upper panel). “Low dose pIC” treatment of R26N2ICMxCre animals resulted in the appearance of periportal, lobular, and pericentral ectopic biliary structures in otherwise normally structured livers (lower panels). (E) Livers of 6-month-old mice retain sensitivity to N2IC-induced transdifferentiation of hepatocytes as assessed by H&E, HNF4α, and N2IC stainings of livers of R26N2ICMxCre animals 10 days after pIC injection. Real-time RT-PCR data shown in (C) are mean values ± SD (n = 3-6 for each genotype). PV, portal vein. BD, bile duct. Scale bars in (A) = 200 μm; (B) = H&E stainings 50 μm, immunofluorescence 20 μm, all other 25 μm; (D) = 50 μm (X-gal) and 100 μm/25 μm (HNF1β); (E) = 100 μm and 50 μm, respectively.

Finally, when 6-month-old R26N2ICMxCre animals were examined 10 days after pIC injection, their livers displayed the same morphological changes as young animals (Fig. 2E). Significant persistence of HNF4α-positive hepatocytes was due to less effective Cre-induced N2IC expression (Fig. 2E) in older mice rather than due to reduced susceptibility of aged hepatocytes to N2IC-induced biliary conversion.

Conditional Activation of Notch2 Signaling Specifically in the Adult Biliary and Progenitor Compartment Results in a Ductular Reaction.

Cre expression in MxCre animals is not exclusively restricted to hepatocytes. To investigate if the dramatic phenotype observed in R26N2ICMxCre animals in fact results from hepatocyte transdifferentiation and not from expansion of preexistent biliary cells or progenitors, we generated R26N2ICHNF1βCreERT2 animals using the HNF1βCreERT2 mouse strain.18 To assess the cell-specific Cre expression profile, we analyzed R26TomHNF1βCreERT2 reporter mice that express the fluorescent protein tdTomato after Cre expression. tdTomato expression in 7-week-old animals 7 days after tamoxifen injection was observed in biliary ducts and in small periportal ductules (Fig. 3A). Labeling efficacy of HNF1β-positive bile ducts and periportal cells was 83.6% (range 75%-100%, n = 3 animals) which exclusively costained for Sox9 (Supporting Fig. 5A). Adult Sox9-positive cells are known to harbor cells of the adult progenitor cell compartment and give rise to oval cells after various liver damage protocols.23, 24 No HNF4α-positive hepatocytes expressing tdTomato were observed (Supporting Fig. 5B). We therefore concluded that the HNF1βCreERT2 mouse strain effectively induces Cre expression in the adult biliary and hepatic progenitor cell compartment. Seven days after tamoxifen treatment livers of R26N2ICHNF1βCreERT2 animals displayed an increase in panCK-positive cells that were strictly confined to the periportal area while the lobular liver parenchyma did not show any abnormalities (Fig. 3B). The periportal ductular structures stained positive for HNF1β and N2IC and were proliferative as assessed by Ki67 staining, suggesting that HNF1β-positive hepatic progenitors give rise to these cells (Fig. 3C). The morphological changes observed in livers of R26N2ICHNF1βCreERT2 mice resembled histological features of a ductular reaction and were obviously different from the panhepatic phenotype observed in R26N2ICMxCre animals.

Figure 3.

Conditional expression of N2IC specifically in the adult biliary and liver progenitor cell compartment results in a ductular reaction. (A) Cre expression in the HNF1βCreERT2 mouse strain is restricted to HNF1β-positive bile ducts (arrow) and periportal ductules (arrowheads) as assessed by analysis of HNF1β and tdTomato expression in 6-week-old R26TomHNF1βCreERT2 reporter mice 7 days after tamoxifen injection (intraperitoneal, 100 μg/g BW). (B) Directing N2IC expression to the HNF1β-expressing cellular compartment of R26N2ICHNF1βCreERT2 animals 7 days after tamoxifen treatment results in the periportal appearance of panCK-positive ductules reminiscent of an oval cell response. The outlined area is magnified in the right panel. (C) HNF1β, N2IC, and Ki67 immunostainings demonstrate proliferative periportal ductules (left images). The majority of Ki67-expressing cells colabel for HNF1β as assessed by immunofluorescence (right images). PV, portal vein. Scale bars in (A) = 25 μm; (B) = left 200 μm, right 25 μm; (C) = 25 μm.

Our observations show that the lobular biliary structures in livers of R26N2ICMxCre mice arise from mature hepatocytes rather than from activation of the hepatic adult progenitor cell compartment. Moreover, our results suggest that Notch2 signaling is capable of promoting the expansion of HNF1β-positive cells resulting in a ductular reaction.

Instructive Notch Signaling Inducing Biliary Cell Fate and Morphogenesis Critically Depends on RBP-Jκ, but Not Hes1.

Next, we intended to characterize the role of the Notch key effectors RBP-Jκ and Hes1 in normal development and in our N2IC-expressing models. For this, we analyzed mice carrying conditional knockout alleles for Rbpj and Hes1 (RbpjF/F and Hes1F/F animals).20, 21 After hepatoblast-specific deletion of RBP-Jκ (RbpjF/FAlbCre) portal tracts lacked mature bile ducts as assessed by panCK staining at postnatal day (P)10 (Fig. 4), confirming the central role of the canonical Notch pathway for perinatal bile duct maturation.6, 10 Surprisingly, biliary morphology was normal in Hes1F/FAlbCre mice. Bile ducts were well incorporated within the portal mesenchyme and indistinguishable from controls (Fig. 4), suggesting that Hes1 is dispensable for perinatal tubulogenesis. Effective deletion of Hes1 was confirmed and expression of additional Notch targets was analyzed by real-time reverse-transcription polymerase chain reaction (RT-PCR) analysis at birth, when also no biliary abnormalities were observed (Supporting Fig. 6A,B).

Figure 4.

Conditional hepatoblast-specific deletion of Rbpj, but not Hes1, leads to impaired IHBD development. IHBD status in control, RbpjF/FAlbCre and Hes1F/FAlbCre mice at P10 was assessed by panCK and Sox9 immunostaining. In control and Hes1F/FAlbCre animals panCK- and Sox9-expressing biliary cells form mature ducts well incorporated into the portal mesenchyme (asterisks mark portal tracts). RbpjF/FAlbCre livers lack normal bile ducts and display irregular periportal panCK-expressing cells of intermediate phenotype. In contrast to control and Hes1F/FAlbCre animals, in RbpjF/FAlbCre livers Sox9-expressing hepatocytes are detected in the periportal area and along the interlobular septs (arrowheads). Outlined areas are magnified below each panel. Scale bar = 100 μm and 25 μm in the magnified panels.

Of note, many periportal hepatocytes showed enhanced panCK staining at P10 in RbpjF/FAlbCre animals. Moreover, while Sox9 expression was restricted to mature bile ducts in control and Hes1F/FAlbCre animals, Sox9-positive cells with hepatocyte morphology were detected in RbpjF/FAlbCre livers along the interlobular septs connecting the portal tracts (Fig. 4). When livers were analyzed later at P20, these intermediate cells formed irregular ductules spreading from portal tracts along the interlobular septs (Supporting Fig. 6C). We interpret the appearance of these intermediate cells as a compensatory transdifferentiation response to the lack of normal bile ducts. We suggest that due to the loss of RBP-Jκ, biliary transdifferentiation of hepatocytes to mature biliary epithelial cells is impaired or severely delayed.

To assess the contribution of RBP-Jκ and Hes1 in N2IC-induced biliary specification and morphogenesis of embryonic and adult liver cells, we generated R26N2ICRbpjF/FAlbCre, R26N2ICHes1F/FAlbCre, R26N2IC RbpjF/FMxCre, and R26N2ICHes1F/FMxCre animals, respectively. The additional genetic inactivation of Rbpj fully rescued the perinatal lethal phenotype observed in R26N2ICAlbCre animals now displaying a normal liver architecture at birth (Fig. 5A). N2IC-expressing hepatocytes in R26N2ICRbpjF/FAlbCre livers had normal hepatocyte morphology lacking expression of biliary markers such as HNF1β (Fig. 5A). In contrast, R26N2ICHes1F/FAlbCre animals all died within 24 hours after birth. Their livers displayed the same structural pathology as R26N2ICAlbCre animals where the loss of Hes1 did not prevent N2IC-positive hepatoblasts to acquire a biliary phenotype and form tubular-cystic structures (Fig. 5A). In analogy, analysis of 5 to 6-week-old R26N2ICRbpjF/FMxCre mice 7 days after pIC injection demonstrated that N2IC-induced transdifferentiation of mature hepatocytes can be prevented by concomitant inactivation of Rbpj. N2IC-expressing cells in R26N2ICRbpjF/FMxCre animals were HNF1β-negative and maintained typical hepatocyte morphology (Fig. 5B). As with embryonic inactivation of Hes1, the additional inactivation of Hes1 in N2IC-expressing hepatocytes in R26N2ICHes1F/FMxCre mice had no visible impact on the N2IC-induced formation of biliary tubular-cystic structures (Fig. 5B). Congruent with the histological results, the additional deletion of Rbpj, but not Hes1, in embryonic and adult N2IC-expressing mice reversed the rapid decline in albumin expression (Fig. 5C,D). Of note, induction of Hes1 mRNA expression was prevented in embryonic and adult N2IC-expressing models, both after deletion of Rbpj and Hes1 (Fig. 5C,D). However, only deletion of RBP-Jκ resulted in phenotypic rescues in these models, indicating that canonical Notch targets other than Hes1 are decisive to determine N2IC-induced biliary cell fates and morphogenesis. Of note, in our model deletion of Hes1 clearly preceded the formation of biliary microcysts as demonstrated by analyzing R26N2ICHes1F/FMxCre mice 4 days after pIC injection (Supporting Fig. 7A,B).

Figure 5.

Instructive Notch signaling inducing biliary fate and morphogenesis critically depends on RBP-Jκ but not on Hes1. (A) Representative H&E and immunostaining for HNF1β and N2IC of transgenic hepatoblast-specific N2IC-expressing livers at birth without (R26N2ICAlbCre) and with concomitant genetic inactivation of Rbpj (R26N2ICRbpjF/FAlbCre) or Hes1 (R26N2ICHes1F/FAlbCre) reveal dependence of N2IC-induced biliary differentiation and morphogenesis on RBPJ-κ but not Hes1. (B) Equal histological analysis of livers from 6-week-old control, R26N2ICRbpjF/FMxCre, and R26N2ICHes1F/FMxCre animals 7 days after pIC-injection demonstrates that N2IC-induced transdifferentiation of adult hepatocytes is prevented after genetic inactivation of Rbpj but remains unchanged after deletion of Hes1. (C,D) Albumin and Hes1 mRNA expression levels were analyzed by real-time RT-PCR in (C) control, R26N2ICAlbCre, R26N2ICRbpjF/FAlbCre, and R26N2ICHes1F/FAlbCre animals at P0 as well as in (D) control, R26N2ICMxCre, R26N2ICRbpjF/FMxCre, and R26N2ICHes1F/FMxCre animals 7 days after pIC injection. Real-time RT-PCR data shown are mean values ± SD (n = 3-6 for each genotype). Scale bar in (A) = 100 μm for all H&E stains and 25 μm for N2IC and HNF1β immunohistochemistry.


In our study, embryonic expression of N2IC in hepatoblasts of R26N2ICAlbCre mice resulted in rapid replacement of the entire liver by biliary tubular-cystic structures, confirming that Notch2 signals convert hepatoblasts to the biliary lineage and promote tubulogenesis. This observation is in line with a previous study using an equivalent transgenic approach, where a similar phenotype with ectopic periportal and lobular tubule formation in newborns was observed.22 Tchorz et al.22 described postnatal gradual “regression” of the lobular tubules in their mouse model by P10 and concluded that additional signals besides N2IC may be required for maintenance of lobular ducts. In our study, almost all R26N2ICAlbCre mice died shortly after birth, which is understandable, considering that virtually no hepatocytes remained to preserve liver function. However, those animals reaching adulthood displayed both lobular areas with ectopic bile ducts and hamartoma-like biliary tumors but also areas with normal hepatocytes lacking N2IC expression (Supporting Fig. 3C). From these results we argue that Notch2 signaling is capable of forming lobular biliary structures that do not require additional periportal signals for survival. However, the compromised metabolic function of R26N2ICAlbCre livers necessitates wildtype hepatocytes, having escaped recombination, to gradually repopulate the liver, a well-known phenomenon termed therapeutic liver repopulation.25 Subtle differences in timing of Cre expression as well as different transgene levels may explain the different capacity of wildtype hepatocytes to repopulate the liver as well as tumor formation in the two N2IC-expressing mouse lines in our and Tchorz et al.'s study.22

While AlbCre-mediated deletion of Rbpj resulted in severe postnatal IHBD morphogenesis defects in RbpjF/FAlbCre mice, biliary tubulogenesis was normal in Hes1F/FAlbCre animals. Of importance, hepatoblast Cre expression occurs rather late in AlbCre animals starting at around E14.5 during embryogenesis.26 Therefore, when using AlbCre mice, early ductal plate phenotypes may be missed because recombination events may be incomplete by the time formation of the first ductal plate layer occurs.6 Nevertheless, the AlbCre mouse strain is highly suitable for studying tubulogenesis, a process that involves specification of the second ductal plate layer and intense remodeling well beyond birth.7, 9, 10 The hitherto existing concept that Hes1 is the critical Notch target for IHBD remodeling is derived from immunohistochemical analysis6, 14 and a single study analyzing Hes1 null animals, where ductal plates but no tubule formation was observed at P0.14 However, most Hes1 null animals die by E18.5 from severe neural tube defects and have gallbladder agenesis and hypoplasia of extrahepatic bile ducts never connecting with the IHBD system, possibly interfering with proper ductal plate remodeling.27, 28 Moreover, those Hes1 null animals reaching birth all die within the first 24 hours and therefore are of limited informative value to study the impact of Hes1 on IHBD tubulogenesis as this process extends several days beyond birth. In addition to our observations in RbpjF/FAlbCre and Hes1F/FAlbCre animals we found that N2IC-induced morphogenetic effects in R26N2ICAlbCre animals could be reverted by the additional genetic deletion of Rbpj, but not Hes1. Although Hes1 has been clearly demonstrated to be expressed in developing bile ducts,6, 14 it is increasingly accepted that Hes1 may not be a perfect readout for Notch activity because Hes1 expression is also regulated independently of Notch.29 In support of this evidence, embryonic deletion of Jagged1 in the portal mesenchyme resulted in severe IHBD morphogenesis defects without altered expression of Hes1.13 Furthermore, Hes1 may even function as a Notch suppressor30; in this context, we observed enhanced expression of Hey1 and Hes5 after genetic deletion of Hes1 in N2IC-expressing livers of R26N2ICHes1F/FAlbCre animals (Supporting Fig. 8), which might also argue for redundancy of these Notch targets as proposed in brain development.31 Inactivation of the Notch target gene Sox9 results in IHBD maturation defects32 and, therefore, Sox9 is a likely candidate to contribute to N2IC-expressing tubulogenesis in our model. However, because IHBD malformations are much more pronounced after genetic deletion of Jagged1, RBP-Jκ or Notch2 than after deletion of Sox9,6, 7, 10, 13, 32 additional Notch targets yet to be identified are likely involved to drive Notch-induced biliary tubulogenesis. Taken together, our results underline the vital role of canonical Notch signaling but clearly argue against a pivotal role of Hes1 as the key Notch target in IHBD formation. It should be also kept in mind that besides Notch other signaling pathways such as the TGFβ or Wnt/β-Catenin pathway act in concert to induce biliary lineage defining proteins such as Sox9, HNF1β, CK19, or osteopontin.12 The observation that Sox9 expression is induced in periportal and interlobular hepatocytes of P10 RbpjF/FAlbCre livers that later acquire an intermediate phenotype (Fig. 4) underscores that induction of biliary proteins can take place in the absence of canonical Notch signaling.

Remarkably, adult hepatocytes fully retained their susceptibility to N2IC-induced biliary reprogramming. Similar to embryonic hepatoblasts, N2IC-induced formation of biliary-like structures strongly depended on canonical Notch signaling through RBP-Jκ but occurred in the absence of Hes1. We are well aware that our N2IC-expressing mouse models by no means represent a physiological condition when normally exact timing of fine-tuned Notch dosages navigates developmental cell fates. However, our results provide a proof of principle that adult mouse hepatocytes are capable to undergo rapid biliary transdifferentiation in vivo when embryonic signaling pathways are reactivated. It has been a debate on principles whether biliary transdifferentiation of hepatocytes happens in vivo and whether this represents a general regeneration mechanism in response to injury.1, 33 Numerous studies have demonstrated the capability of isolated hepatocytes to undergo biliary transformation in vitro (for review, see Ref.1). However, data that unambiguously show hepatocyte transdifferentiation in vivo are scarce. Only by generating chimeric rats by hepatocyte transplantation and combining bile duct ligation with the application of a biliary toxin one group was able to demonstrate that transplanted hepatocytes can give rise to bile ductules.4 Zong et al.6 demonstrated that inducible transgenic Notch1IC (N1IC) expression using the AlbCreERT2 promoter resulted in the appearance of some ectopic tubular structures of biliary phenotype when 6-day-old mice were subjected to repetitive tamoxifen injections for 3 weeks. Their findings were suggestive that the sensitivity of embryonic hepatoblasts to Notch signals extends to young hepatocytes shortly after birth; however, in that study it could not be ruled out that progenitor cells gave rise to the ectopic biliary structures observed. In another recent study by Fan et al.,34 the adenoviral delivery of N1IC together with constitutively active AKT1 led to the lobular appearance of singular hepatobiliary hybrid cells that, however, rapidly clonally expanded to give rise to invasive cholangiocytic tumors. After combining this model with hepatocyte lineage tracing using adenoviral transfer of transthyretin-Cre into R26EYFP reporter animals the authors concluded that the biliary tumors were of hepatocyte origin. This conclusion supports the concept that adult hepatocytes may change cell fates upon stimulation with N1IC. Nevertheless, some concerns with this adenoviral hepatocyte fate-tracing model remain in terms of possible Cre expression in the biliary compartment during malignant transformation. In our study, we used the HNF1βCreERT2 mouse line to specifically direct N2IC expression to the biliary and facultative progenitor compartment. Using this approach, we show that the lobular biliary structures in R26N2ICMxCre animals were not the progeny of N2IC-expressing biliary cells or progenitors, thereby circumventing potential confounding variables that may arise from hepatocyte transplantation or adenoviral models.

In comparison to the above-mentioned studies by Zong et al. and Fan et al.6, 34 using N1IC, it is interesting that using N2IC we observed strongly accelerated biliary conversion of hepatoblasts and hepatocytes and eventually premalignant biliary tumor development in surviving R26N2ICAlbCre animals without the additional implementation of oncogenes. This may be due to context-dependent differences between Notch1 and Notch2, different transgene expression levels, or targeting of different cellular compartments including microenvironmental interaction issues. Regarding different roles of Notch1 and Notch2, we previously established a pivotal role of Notch2, but not Notch1, during development.7 In line with these data it is also reasonable to propose Notch2 as the primary Notch receptor mediating biliary transdifferentiation of adult hepatocytes.

We did not detect any lobular biliary structures in R26N2ICHNF1βCreERT2 livers; however, we identified Notch2 signaling as a potential regulator of the adult liver progenitor compartment, as a significant ductular reaction was observed. This attractive presumption is supported by recent data derived from rodent oval cell models by means of pharmacological Notch inhibition that suggested Notch to be important for an effective oval cell response.11, 35

In summary, not only embryonic hepatoblasts but also mature hepatocytes exhibit a marked plasticity to canonical Notch2 signaling by biliary lineage commitment and morphogenesis. The particular transcriptional target(s) of Notch2 mediating lineage commitment and morphogenesis of both embryonic and adult liver cells remain to be identified. Nevertheless, our results support the emerging concept that liver repair mechanisms in adulthood such as ductular reactions or transdifferentiation of hepatocytes recapitulate developmental processes using signaling pathways that are normally active during embryogenesis. However, the precise role of Notch in hepatocyte transdifferentiation and progenitor cell-mediated liver regeneration awaits to be characterized in the setting of specific liver injury models with genetic compartment-specific Notch loss-of-function and gain-of-function models.


We thank Stephanie Dürl and Silvia Krutsch for excellent laboratory assistance. We thank Ryoichiro Kageyama, (Kyoto University, Japan) for providing Hes1F/F mice.