Inducible inactivation of Notch1 causes nodular regenerative hyperplasia in mice

Authors

  • Adrien Croquelois,

    1. Ludwig Institute for Cancer Research, Lausanne Branch, University of Lausanne, Epalinges, Switzerland
    2. Division of Hypertension, University of Lausanne Medical School, Lausanne, Switzerland
    Search for more papers by this author
    • A.C., A.B., and L.T. contributed equally to this work.

  • Alex Blindenbacher,

    1. Department of Research, University Hospital Basel, Basel, Switzerland
    Search for more papers by this author
    • A.C., A.B., and L.T. contributed equally to this work.

  • Luigi Terracciano,

    1. Institute of Pathology, University Hospital Basel, Basel, Switzerland
    Search for more papers by this author
    • A.C., A.B., and L.T. contributed equally to this work.

  • Xueya Wang,

    1. Department of Research, University Hospital Basel, Basel, Switzerland
    Search for more papers by this author
  • Igor Langer,

    1. Department of Surgery, University Hospital Basel, Basel, Switzerland
    Search for more papers by this author
  • Freddy Radtke,

    1. Ludwig Institute for Cancer Research, Lausanne Branch, University of Lausanne, Epalinges, Switzerland
    Search for more papers by this author
  • Markus H. Heim

    Corresponding author
    1. Department of Research, University Hospital Basel, Basel, Switzerland
    2. Division of Gastroenterology and Hepatology, University Hospital Basel, Basel, Switzerland
    • Department of Research, University Hospital Basel, Hebelstrasse 20, CH-4031 Basel, Switzerland
    Search for more papers by this author
    • fax: (41) 61 265 53 52.


  • Conflict of interest: Nothing to report.

  • See Editorial on page 439

Abstract

The discovery that the human Jagged1 gene (JAG1) is the Alagille syndrome disease gene indicated that Notch signaling has an important role in bile duct homeostasis. The functional study of this signaling pathway has been difficult because mice with targeted mutations in Jagged1, Notch1, or Notch2 have an embryonic lethal phenotype. We have previously generated mice with inducible Notch1 disruption using an interferon-inducible Cre-recombinase transgene in combination with the loxP flanked Notch1 gene. We used this conditional Notch1 knockout mouse model to investigate the role of Notch1 signaling in liver cell proliferation and differentiation. Deletion of Notch1 did not result in bile duct paucity, but, surprisingly, resulted in a continuous proliferation of hepatocytes. In conclusion, within weeks after Notch1 inactivation, the mice developed nodular regenerative hyperplasia without vascular changes in the liver. (HEPATOLOGY 2005;41:487–496.)

During organogenesis of the liver, both hepatocytes and biliary epithelial cells are derived from common precursors, the hepatoblasts.1, 2 The molecular signals determining this cell fate decision are poorly understood. Important insights came from the discovery that the human Jagged1 gene (JAG1) is the Alagille syndrome (AGS) disease gene.3, 4 AGS is a dominantly inherited disorder characterized by developmental abnormalities such as paucity of interlobular bile ducts, characteristic facies, peripheral pulmonary artery stenosis, butterfly-like vertebral anomalies and posterior embryotoxon.5, 6 The AGS disease gene JAG1 encodes a ligand in the Notch signaling pathway. This pathway is an evolutionary conserved mechanism to control cell fates through local cell interactions.7

In mice and humans, four Notch family receptors (NOTCH1 through NOTCH4) have been described. They interact with membrane-bound ligands that are encoded by the Jagged (JAG1 and JAG2) and Delta-like (DLL1, DLL3, and DLL4) gene families. The specific ligand-receptor interactions have not been fully characterized, but they appear to be variable depending on the tissue studied. In the liver, JAG1 is expressed during embryogenesis in the ductal plates and after birth in bile ducts, endothelial cells, and hepatocytes.8–10 All four Notch family members are expressed as well. In mice, Notch1 expression is fairly stable during gestation as well as after birth, whereas Notch2 expression peaks at embryonic day 15, at the time when bile duct differentiation starts.10Notch3 and Notch4 expression are low during embryogenesis and peaks by 12 weeks of age postnatally.10 Mice homozygous for a targeted null mutation of Jag1 die in utero because of vascular defects, and heterozygous Jag1+/− mice exhibit eye defects but no paucity of bile ducts or other AGS associated phenotypes.11Notch1 and Notch2 are both essential for normal embryonic development, because mice with disruptions of either gene die in mid-gestation.12–14 On the contrary, mice with targeted disruptions of Notch3 or Notch4 develop normally.15, 16

Recently, a mouse model of AGS was found by crossing Jag1 mutant mice (Jag1dDSL) with mice harboring a hypomorphic Notch2 mutant allele (Notch2del1).17 Doubly heterozygous Jag1dDSL/+Notch2del1/+ mice suffer from jaundice, growth retardation, impaired differentiation of intrahepatic bile ducts, and defects in heart, eye, and kidney development. The development of AGS in the double but not in the single heterozygous mice strongly indicates that Jagged1 cannot be the only ligand for the Notch2 receptor, or that Notch2 is not the only receptor for Jagged1. Jagged1 must therefore activate an additional partially redundant Notch receptor. Because conventional gene-targeted mice for the Notch3 and Notch4 receptor do not exhibit any obvious phenotypes, we concentrated our efforts on the consequences of a loss of Notch1 function in the liver.

To circumvent the embryonic lethality of Notch1 deficiency and to explore the consequences of postnatal inactivation of Notch1, we have previously successfully used the Cre-loxP system. For example, inactivation of Notch1 in bone marrow progenitors has shown an essential role of Notch1 signaling in T- versus B-cell fate specification.18, 19 More recently, we showed that an inducible inactivation of Notch1 in the mouse epidermis results first in epidermal hyperplasia and subsequent development of skin tumors.20 Here we have used this conditional Notch1 knockout model to study the role of Notch1 signaling for proliferation and differentiation of hepatocytes and biliary epithelial cells (BECs). Because conventional gene targeting of the Notch1 gene results in an early embryonic lethal phenotype, developmental abnormalities caused by Notch1 inactivation cannot be studied directly. We therefore used partial hepatectomy and bile duct ligation experiments to analyze potential proliferation or survival defects of hepatocytes and BECs after inducible Notch1 inactivation. Surprisingly, we found no defects in the proliferation or survival of BECs in the absence of Notch1, but we found a striking nodular regenerative hyperplasia of the liver.

Abbreviations:

JAG1, Jagged1 gene; AGS, Alagille syndrome; BECs, biliary epithelial cells; IFN, interferon; PIPC, Poly(I)-poly(C); DBA, dolichos bioflorus agglutinin; BSA, bovine serum albumin; BrdU, 5-bromo-2′-deoxyuridine; PH, partial hepatectomy; NRH, nodular regenerative hyperplasia; IL, interleukin; HCC, hepatocellular carcinoma.

Materials and Methods

Animals, Induction of Notch1 Deletion, Genotyping, Southern Blots.

The animals were maintained in the animal facility of the Department of Research, University Hospital Basel, in a specific pathogen-free environment on a 12-hour light and 12-hour dark schedule. The animals had free access to food and drinking water. For breeding, mice homozygous for the loxed Notch1 and heterozygous for the MxCre recombinase (Notch1lox/loxMxCre+/−) were used. Newborn mice were genotyped as previously described.18 Except when indicated otherwise, control mice were Notch1lox/loxMxCre−/− mice obtained during breeding with Notch1lox/loxMxCre+/− mice. For some experiments, Notch1wt/wtMxCre+/− mice were used as additional controls.

Notch1wt/wtMxCre+/− mice have wild-type Notch1 alleles and were heterozygous for MxCre. They were obtained by backbreeding Notch1lox/loxMxCre+/− to C57BL/6 mice. The Mx1 protein is part of the murine defense mechanism against viral infections. It is not expressed in healthy mice but is transcriptionally induced on viral infection as a consequence of interferon alpha (IFNα) secretion. The double-stranded RNA polyiosinic-polycytidylic acid (PIPC) mimics a viral infection and thereby induces the cells of the immune system to secrete IFNα, which leads to the transcriptional activation of the Mx promoter. For induction of the Cre recombinase (and inactivation of the Notch1 gene), 8- to 12-week-old mice were injected intraperitoneally 3 times with 300 μg PIPC (Fluka, Buchs, Switzerland) at 3-day intervals. Because the immune system of newborn mice is not fully developed at birth and therefore cannot be induced efficiently enough by PIPC to secrete IFNα, 5000 IU murine IFNα (R&D Systems, Minneapolis, MN) was injected at postnatal days 2, 4, 6, and 8 (4 times) to inactivate the Notch1 gene shortly after birth. The efficiency of Notch1 deletion was regularly checked by DNA extraction from livers and subsequent Southern blots analysis with a Notch1-specific probe as previously described.18 This system can only be used to inactivate floxed target genes after birth. It is not suitable for inducible inactivation of genes in utero, because IFNα does not efficiently pass the placenta (F. Radtke, unpublished observations, September 1999), which is why we only focused on postnatal inactivation of Notch1. The study was approved by the animal care committee of the Kanton Basel.

Macroscopic and Microscopic Assessment of Livers After Notch1 Deletion.

Mice were killed at different times after birth and after inactivation of Notch1 as indicated in the text. After measuring the body weight, the livers were removed, inspected by eye, and weighed, and representative examples were photographed. From all livers, parts were fixed in formaldehyde and prepared for microscopic assessment using standard methods (hematoxylin-eosin staining, Masson staining). For dibenzanthracene Dolichos bioflorus agglutinin (DBA, from Vector Laboratories, Burlingame, CA) staining of biliary epithelial cells, slides were warmed at 60°C, rehydrated, and boiled in 0.01 mol/L sodium citrate in a microwave oven twice for 5 minutes. After blocking in 1% bovine serum albumin (BSA), slides were covered with 200 μL 20 μg/mL DBA in 1% BSA (Fluka) for 45 minutes in a humid chamber at room temperature. After 4 times rinsing in phosphate-buffered saline, slides were covered with 200 μL 1:2000 streptavidine alkaline phosphatase in 1% BSA for 30 minutes in a humid chamber at room temperature. Signals were detected with the Fast Red dye from Sigma (Fluka) according to the manufacturer's instructions. Slides were then stained with hematoxylin.

Formalin-fixed, paraffin-embedded liver tissue was used for immunohistochemical staining of Notch1 (1:400, H-131, Santa Cruz, Santa Cruz, CA). Primary antibody to Notch was incubated overnight after antigen retrieval using microwaves at 98°C, 60 minutes in citrate buffer 10 mmol/L pH 6.0. After washing in Tris-buffered saline and incubation with biotinylated antibodies, visualization was performed with diaminobutyric acid peroxidase kit and Elite Vectastain ABC Kit according to the manufacturers' instructions (Vectorstain Laboratories Inc., Burlingame, CA).

For 5-bromo-2′-deoxyuridine (BrdU) labeling of cells in the S phase of the cell cycle, mice were injected intraperitoneally with 50 μg/g body weight BrdU (Sigma, Fluka) 1 hour before they were killed. Alternatively, BrdU was provided in the drinking water at a concentration of 0.8 mg/mL. Liver tissue was immediately fixed in 4% formaldehyde. BrdU incorporation in liver cells was detected by immune histochemistry by using an anti-BrdU antibody (Roche Diagnostics, Rotkreuz, Switzerland) and avidin biotin complex (ABC-Elite, Vectra Laboratories, Geneva, Switzerland).

Partial Hepatectomy.

Ten- to twelve-week-old Notch1lox/loxMxCre−/− (controls) and Notch1lox/loxMxCre+/− (knockout) were injected intraperitoneally 3 to 4 times in 2- to 3-day intervals with 300 μg PIPC. Seven to ten days later, partial hepatectomy (PH) was performed as previously described,21 with slight modifications. Animals were anesthetized with isoflurane. After a midline laparotomy, the left and the middle lobes of the liver were ligated at the base and resected. The abdominal wall and the skin were sutured separately. The entire procedure was performed under sterile conditions. Surgery was performed between 9:00 AM and 1:00 PM. The resected liver lobes were frozen in liquid nitrogen and kept at −70°C for further processing. Analysis of recovery of the liver weight after PH was done as follows: Before surgery, each mouse was weighed, and the total liver weight was calculated by multiplying the total body weight by 0.05 for controls and 0.07 for knockout mice. Mice were then killed at different times after PH. The liver was removed and weighed. Proliferation of hepatocytes after PH was analyzed with BrdU staining as described previously. For these experiments, mice were injected intraperitoneally with 50 μg/g body weight BrdU 1 hour before they were killed. For analysis of bile duct regeneration after PH, slides were stained with DBA and hematoxylin as described previously.

Bile Duct Ligation.

Animals were anesthetized with isoflurane. Surgery was done with magnifying glasses. After a midline laparotomy, the liver was mobilized and the dorsal surface was displayed. The hepatoduodenal ligament was identified, and the common bile duct was separated from the portal vein and the hepatic artery. The common bile duct was ligated with a 3–0 Dexon suture. The abdominal wall and the skin were sutured separately.

Surgery was performed between 9:00 AM and 1:00 PM. Mice were killed 2 days after the operation. Morphology and number of bile ducts were analyzed with DBA-hematoxylin and hematoxylin-eosin–stained liver sections. Proliferation of biliary epithelial cells was measured with BrdU incorporation. For this purpose, the animals were injected intraperitoneally with 50 μg/g body weight BrdU 1 hour before they were killed.

Results

Inducible Inactivation of Notch1.

To inducibly inactive the Notch1 gene in the liver, we made use of the previously described Notch1lox/loxMxCre+/− mice.18Notch1lox/loxMxCre+/− mice are homozygous for a Notch1 allele that has two loxP sites flanking the leader peptide of Notch1, and heterozygous for an Mx-Cre transgene. The expression of the Cre recombinase can be induced through the activation of the Mx promoter by treatment with IFN or PIPC.22 IFN-inducible, Cre-mediated inactivation of the floxed Notch1 allele (Fig. 1A) has previously been shown to be close to 100% in liver and bone marrow.18 Here we further tested the efficiency of Notch1 inactivation in the liver over a prolonged period. After 3 injections of 300 μg PIPC at 3-day intervals, Southern blot analysis of DNA isolated from liver indicated a complete inactivation already at the day of the third injection (day 0). The deletion efficiency remained stable in animals tested after different periods for over 1 year, suggesting that the Notch1-deficient hepatocytes were not replaced over time by a small population of cells that may have escaped Notch1 inactivation (Fig. 1B shows representative examples after 5 days, 200 days, and 360 days).

Figure 1.

Inducible deletion of the leader peptide of Notch1. (A) Schematic representation of the targeting strategy. Notch1lox/lox mice are homozygous for a genetically modified Notch1 containing two loxP sites (triangles) flanking the leader peptide of Notch1. In the presence of Cre recombinase this leader peptide is excised, and no functional Notch1 is expressed at the cell surface. The Notch1lox/lox mice were crossed to Mx-Cre transgenic mice. For breeding, mice homozygous for the floxed Notch1 allele and heterozygous for the MxCre recombinase transgene were used. The offspring was homozygous for the floxed Notch1 allele, and either heterozygous for the MxCre transgene (Notch1lox/loxMxCre+/−), or lacking the MxCre transgene (Notch1lox/loxMxCre−/−). The expression of Cre recombinase can be induced through the activation of the Mx promoter by intraperitoneal injection of IFN or PIPC. (B) Efficiency of Notch1 inactivation in the liver. Mice were injected 3 times with 300 μg PIPC intraperitoneally at 3-day intervals and then killed at indicated times after the last injection. DNA was extracted from livers and tested by Southern blots using a Notch1 probe. Shown are examples of Notch1lox/loxMxCre+/− (knockout) and Notch1lox/loxMxCre−/− (wildtype, control) at days 0, 5, 200, and 360. Deletion efficiency was over 95% already at day 0, and remained so up to 1 year. (C) Immunohistochemical staining of a liver section from a control (Notch1lox/loxMxCre−/−) mouse shows Notch1 signal in hepatocytes (double arrow) and biliary epithelial cells (single arrow; representative example, immunostaining with Santa Cruz antibody H-131). (D) Deletion of Notch1 by MxCre induction in Notch1lox/loxMxCre+/−mice results in Notch1 signal loss in both hepatocytes (double arrow) and biliary epithelial cells (single arrow; representative example, immunostaining with Santa Cruz antibody H-131).

Whereas Notch1 could be detected in hepatocytes and BECs in control (Notch1lox/loxMxCre−/−) mice, no such signals were present in Notch1lox/loxMxCre+/− after induction of MxCre (Fig. 1C-D).

Inactivation of Notch1 Causes Nodular Regenerative Hyperplasia of the Liver.

Inactivation of Notch1 consistently resulted in a change of the surface and structure of the liver within the first week (compare Fig. 2A and B). The nodular deformation of the liver surface was observed at all times after Notch1 inactivation up to 400 days. The liver appeared to be enlarged, and a careful analysis of the liver weight confirmed an increase of the liver mass relative to the body weight in Notch1 knockout animals (Fig. 2C). The histological analysis of the liver sections from control and Notch1 knockout animals showed no signs of steatosis, inflammatory infiltrates, excess storage of glycogen, or blood congestion in the latter. There was, however, a striking ill-defined nodulation of the liver parenchyma without fibrotic septae (Fig. 3B,D,F). Nodules were regular and small in all cases. Areas of hepatocellular atrophy surrounding areas of large hepatocytes with evident compression of hepatocytes in the periphery of the nodules were frequently observed. Nodular regenerative hyperplasia (NRH) has been associated with disturbed blood flow in the portal vein and obliteration of small portal veins.23 We therefore looked carefully at the livers of all Notch1-deficient mice for obliterations of small portal veins or alterations in arterial vessels or central veins, but could not detect any such changes (data not shown). To confirm our observations, we sectioned an entire liver of a Notch1-deficient mouse with NRH into 991 consecutive sections and analyzed all of them for the presence of thrombi or abnormalities of the intrahepatic portal venous system or the intrahepatic arteries but could not detect any (data not shown). Conversely, focal sinusoidal dilatation as well large vascular space abutting directly on hepatic parenchyma were frequent findings. The increase in liver weight together with the nodular deformation could be the result of increased proliferation in the absence of Notch1. We therefore tested whether Notch1 inactivation induces proliferation of hepatocytes. Ten- to twelve-week-old Notch1lox/loxMxCre+/− (knockout) and Notch1lox/loxMxCre−/− (control) mice were given BrdU in the drinking water for 10 consecutive days. The mice were injected with PIPC to induce the expression of Cre recombinase on days 1, 4, and 7 of BrdU administration. To exclude nonspecific effects of Cre recombinase expression, Notch1wt/wtMxCre+/− mice were used as additional controls. These mice express Cre recombinase on PIPC treatment but are wild-type for Notch1 and therefore have normal Notch1 signaling. The control mice showed very little BrdU-positive hepatocytes compared with Notch1-deficient animals, which exhibited a dramatic increase in BrdU-labeled hepatocytes (Fig. 4A-B). The quantitative analysis shows a mean of 3.2% BrdU+ hepatocytes for control animals and a mean of 25.8% for the Notch1-deficient animals (Fig. 4C), suggesting that the absence of Notch1 in the liver causes spontaneous proliferation of hepatocytes. The increase in proliferating hepatocytes in Notch1-deficient animals apparently accounts for the increased liver mass in Notch1-deficient animals (Fig. 2C).

Figure 2.

Inactivation of Notch1 causes nodular deformation of the liver surface. (A-B) 55-day-old mice injected with IFNα as newborns were killed and their livers removed. (A) Normal liver of a control mouse. (B) The macroscopic pictures show that Notch1-deficient mice have a nodular liver surface (and enlarged livers). (C) Ten- to 12-week-old mice were injected intraperitoneally 3 to 4 times in 2- to 3-day intervals with 300 μg PIPC and killed 7 to 10 days later. Bar diagram of the liver weight in percent of the body weight of Notch1lox/loxMxCre+/− (knockout), of Notch1lox/loxMxCre−/− (control A) and of Notch1wt/wtMxCre+/− (control B).

Figure 3.

Deletion of Notch1 causes nodular regenerative hyperplasia. Newborn Notch1lox/loxMxCre−/− (controls) (A,E) and Notch1lox/loxMxCre+/− (knockout) (B,F) were injected intraperitoneally at day 2, 4, 6, and 8 with murine IFNα and killed 55 days after birth. Ten- to twelve-week-old Notch1lox/loxMxCre−/− (controls) (C) and Notch1lox/loxMxCre+/− (knockout) (D) were injected intraperitoneally 3 to 4 times in 2- to 3-day intervals with 300 μg PIPC and killed 200 days after birth. (A) Hematoxylin staining of a liver section from a 55-day-old control mouse, final magnification 25×. Normal liver parenchyma. (B) Hematoxylin staining of a liver section from a 55-day-old Notch1 knockout mouse, final magnification 25×. Nodular transformation of the parenchyma without inflammation. (C) Hematoxylin staining of a liver section from a 200-day-old control mouse, final magnification 100×. Normal liver parenchyma. (D) Hematoxylin staining of a liver section from a 200-day-old Notch1 knockout mouse, final magnification 100×. Nodular transformation of the parenchyma without inflammation. (E) Masson (collagen) staining of a liver section from a 55-day-old control mouse, final magnification 25×. (F) Masson staining of a liver section from a 55-day-old Notch1 knockout mouse, final magnification 25×. Nodular transformation of the parenchyma without fibrous septae.

Figure 4.

Inactivation of Notch1 causes spontaneous proliferation of hepatocytes. Four-month-old mice were given BrdU in the drinking water (0.8 mg/mL) for 10 days. On days 1, 4, and 7, they were injected intraperitoneally with 300 μg PIPC. At day 10, mice were killed, and liver sections were stained for BrdU incorporation. (A) Representative example of a control mouse (Notch1lox/loxMxCre−/−), final magnification 100×. (B) Representative example of a Notch1 knockout mouse (Notch1lox/loxMxCre+/−), final magnification 100×. (C) Bar diagram showing the percentage of BrdU-positive hepatocytes counted in 4 high-power fields of 4 control (Notch1lox/loxMxCre−/−) and 6 knockout mice (Notch1lox/loxMxCre+/−). Shown are the mean and its 95% confidence interval.

Liver Regeneration After Partial Hepatectomy in Notch1-Deficient Mice.

Spontaneous proliferation of Notch1-deficient hepatocytes under steady-state conditions leads to the question of how these hepatocytes react under a challenge situation such as liver regeneration. Surgical removal of the median and left liver lobes in mice and rats is a well-studied model for liver regeneration. This PH induces a compensatory hyperplasia of the remaining right lobes that restores the original liver mass within 7 to 10 days. During this time, previously quiescent cells such as hepatocytes, biliary epithelial cells, Kupffer cells, stellate cells, and sinusoidal endothelial cells reenter the cell cycle and undergo one or more rounds of cell division.24 The regenerative response is proportional to the amount of liver removed.25 To study the role of Notch1 signaling during liver regeneration, Notch1lox/loxMxCre+/− knockout mice and both Notch1lox/loxMxCre−/− and Notch1wt/wtMxCre+/− control mice were subject to PH and then killed at different times after PH. PH was performed 7 to 10 days after the last PIPC injection, at a time when the liver of Notch1 knockout mice is already enlarged to 7% relative to their body weight. Despite the observed spontaneous proliferation of hepatocytes after Notch1 inactivation (Fig. 4), the restoration of the liver mass was clearly slower in the Notch1 knockout mice compared with both groups of control mice (Fig. 5A). Nine days (216 hours) after PH, the liver mass in the controls was restored to 100% of the original liver mass. In Notch1 knockout mice, only 70% of the original mass was restored after 9 days. Of note, 4 weeks after PH, all Notch1 knockout mice had livers restored to the same weight as before surgery (100% restoration). In agreement with the slower restoration of liver mass in Notch1 knockout mice, a quantitative assessment of liver cell proliferation using the BrdU incorporation method indicated a reduced proliferative response after PH in Notch1 knockout mice (Fig. 5B). Forty hours after PH, when hepatocyte DNA replication is at its maximum in normal mice, the control mice showed a mean of 28.9% of hepatocytes labeled with BrdU, whereas only 8.2% labeled cells were found in the average in Notch1 knockout mice (Fig. 5B). As outlined above, the regenerative response after PH is proportional to the amount of liver removed.25 The experimental data published so far on this subject could also be interpreted differently: The regenerative response after PH is inversely proportional to the amount of liver that remains intact after PH relative to the total body weight. As shown in Fig. 5C, this second interpretation seems to be correct at least in the case of Notch1 knockout mice. In this figure, the average liver mass at different times after a two-thirds hepatectomy is expressed as liver mass relative to body weight, and not as liver mass relative to original liver as in Fig. 5A. In control mice, a two-thirds hepatectomy reduces the mass of the remaining liver to 1.66% of the body weight. In Notch1 knockout mice with their enlarged liver (7% of body weight), a two-thirds hepatectomy reduces the mass of the liver to 2.33% of the body weight. This value is close to the reduction obtained by a hemihepatectomy. Interestingly, the restoration of the liver mass relative to the body weight is not significantly different in Notch1 knockout mice compared with controls (Fig. 5C), and, indeed, 9 days after PH, both groups have livers regenerated to 5% of the body weight. We conclude that the regenerative response in Notch1 knockout mice is not impaired, but on the contrary seems to be as precisely correlated to the body weight as in normal control mice. As mentioned above, analysis of Notch1 knockout mice 4 weeks after PH indicated that the continuous proliferation of hepatocytes beyond the restoration of the liver mass to 5% of the body weight resulted again in hepatomegaly and NRH.

Figure 5.

Liver regeneration after partial hepatectomy (PH). Ten- to 12-week-old mice were injected intraperitoneally 3 to 4 times in 2- to 3-day intervals with 300 μg PIPC. Seven to ten days after the last injection, PH was performed. (A) Recovery of liver mass after two-thirds hepatectomies relative to original liver mass. Notch1-deficient animals (gray diamonds, Notch1lox/loxMxCre pos) have an increased liver mass before PH (Fig. 4), but after removal of two thirds of this increased liver, the remaining liver shows a delayed restoration of the mass relative to the preoperative liver size. When analyzed 4 weeks after PH, the preoperative, increased liver weight was restored. Controls are mice without the MxCre transgene (black squares, Notch1lox/lox & MxCre neg) and MxCre transgenic mice without the floxed Notch1 gene (open circles, Notch1wt/wt & MxCre pos). (B) DNA synthesis after PH is decreased in Notch1-deficient mice. DNA synthesis was measured by BrdU incorporation in 8 control and 7 Notch1-deficient mice. Thirty-nine hours after PH, mice were injected intraperitoneally with BrdU at a dose of 50 μg/g body weight. One hour later (40 hours after PH), mice were killed. BrdU-positive cells were stained by using an anti-BrdU antibody. For each animal, 700 to 1000 hepatocytes were analyzed. The average percentage of BrdU-positive hepatocytes was 29% in controls and 8% in Notch1-deficient mice. The mean and the 95% confidence interval are shown in the bar diagram. (C) Recovery of liver mass after PH relative to the body weight of the mice. Notch1-deficient animals (gray diamonds, Notch1lox/lox and MxCre pos) have an increased liver mass before PH (7% of body weight). After two-thirds hepatectomies, their liver mass is restored relative to the body weight with similar kinetics as in the control animals without the MxCre transgene (black squares, Notch1lox/loxMxCre neg).

Notch1 Signaling Is Not Required for Development, Survival, Regeneration, or Proliferation of Bile Ducts.

Because Notch1 is a potential receptor for Jagged1, and because mutations in the human Jagged1 gene (JAG1) have been shown to be the cause of AGS, which is characterized amongst other things by a paucity of intrahepatic bile ducts, we made a systematic analysis of bile ducts in our conditional Notch1 knockout mouse model. First, liver sections obtained from Notch1 knockout and from control mice at different times after inactivating Notch1 in 10- to 12-week-old animals were stained with DBA (a lectin that binds specifically to biliary epithelial cells) and hematoxylin. Whereas we consistently observed NRH in Notch1 knockout mice, no paucity of bile ducts was seen in animals up to 400 days after Notch1 inactivation. Representative samples are shown in Fig. 6A-B. We conclude that for the normal survival and presumably slow turnover of BECs, Notch1 signaling is not required. Next, we analyzed the regenerative capacity of bile ducts after PH. Nine days after PH, both Notch1 knockout and control mice showed the same regeneration of bile ducts (Table 1 and Fig. 6C-D), suggesting that Notch1 signaling is dispensable for proliferation and differentiation of BECs even in the situation of a PH-induced compensatory hyperplasia. Finally, we used the well-described experimental model of bile duct ligation to induce ductular proliferation. Histological analysis of liver section 2 days after bile duct ligation showed the same pattern of ductular proliferation in both Notch1 knockout and control mice (Table 1 and Fig. 6E-F). The proliferation of BECs after bile duct ligation was quantified using the BrdU labeling method. As shown in Fig. 7, no difference was observed between Notch1 knockout and control mice. We conclude that for proliferation of BECs during a ductular reaction, Notch1 signaling is not required.

Figure 6.

Bile ducts are normal in Notch1-deficient mice. (A-B) Newborn Notch1lox/loxMxCre−/− (controls) (A) and Notch1lox/loxMxCre+/− (knockout) (B) were injected intraperitoneally at days 2, 4, 6, and 8 with murine IFNα and killed 55 days after birth. (A) DBA staining of a liver section from a 55-day-old control mouse, final magnification 250×. Shown is a portal field with a normal bile duct. (B) DBA staining of a liver section from a 55-day-old Notch1 knockout mouse, final magnification 250×. Shown is a portal field with a normal bile duct. (C-D) 10- to 12-week-old mice were injected intraperitoneally 3 to 4 times in 2- to 3-day intervals with 300 μg PIPC. Seven to ten days after the last injection, partial hepatectomy was performed. 9 days after partial hepatectomy, mice were killed, and liver sections were stained with DBA. (C) Representative example from a control mouse, final magnification 250×. (D) Liver section from a Notch1 knockout mouse, final magnification 250×. (E-F) Ten- to 12-week-old Notch1lox/loxMxCre−/− (controls) (C,E) and Notch1lox/loxMxCre+/− (knockout) (D,F) were injected intraperitoneally 3 to 4 times in 2- to 3-day intervals with 300 μg PIPC. Seven to ten days after (E) the last injection, the common bile duct (F) was ligated. Two days later, the mice were killed. (E) DBA staining of a liver section from a control mouse, final magnification 250×. Normal ductular proliferation. (F) DBA staining of a liver section from a Notch1 knockout mouse, final magnification 250×. Normal ductular proliferation.

Table 1. Number of Bile Portal Tracts and Bile Ducts (Stained With DBA) in Representative Liver Sections of Notch1lox/loxMxCre−/− (Controls) and Notch1lox/loxMxCre+/− (Knockout) Mice 9 Days After Partial Hepatectomy (PH) or 2 Days After Bile Duct Ligation (BDL)
TreatmentMouse No.StrainPortal TractsBile Ducts
PH30.10Knockout911
PH30.9Knockout1214
PH30.12Knockout1113
PH30.11Control1011
PH13.2Knockout1414
PH13.3Control1315
BDL4.6Control932
BDL4.5Knockout1229
Figure 7.

Quantification of bile duct proliferation after bile duct ligation (BDL). Proliferation of biliary epithelial cells (BECs) was quantified by dividing the number of BrdU-positive BECs by the total number of BECs at different times after BDL. For each time point and mouse strain, at least 4 individual mice were analyzed. Shown are the mean and the 95% confidence interval. No difference between Notch1-deficient and wild-type mice was found.

Discussion

Our understanding of the pathogenesis of AGS has considerably improved after the seminal discovery that mutations in JAG1 are responsible for the disease.3, 26 Jagged1 is a member of a family of membrane-bound ligands that can interact with Notch family receptors. The specific ligand–receptor interactions between the 5 different ligands and the 4 Notch receptors have not been fully characterized but appear to be variable, depending on the tissue studied. Likewise, the Notch family member involved in the pathogenesis of paucity of bile ducts in patients with AGS has not been identified. Based on our analysis of bile duct number and morphology in Notch1 knockout mice, we conclude that Notch1 is not a critical receptor for BEC survival signals originating from Jagged1. During the time of our study, T. Gridley and colleagues reported that mice doubly heterozygous for the Jag1 null allele and a Notch2 hypomorphic allele showed most of the phenotypes observed in AGS patients.17 The evidence from this work as well as from our present work points to a critical role of Notch2, and not Notch1, for the maintenance of bile duct integrity in the liver. However, we cannot exclude the possibility that Notch1 inactivation in the context of Jagged1 haploinsufficiency (mice generated by crossing Jag−/+ mice to Notch1lox/loxMxCre+/− mice) would also result in paucity of bile ducts. In this regard, it is noteworthy that Notch1, Notch2, as well as Notch3 have all been found to be expressed in bile duct cells.27 It is possible that inactivation of one of them alone does not decrease intracellular Notch signaling enough to cause paucity of bile ducts, and that only the combination of Jagged1 haploinsufficiency and inactivation of one of the Notch receptors lowers intracellular signaling below a critical threshold. In this context, we also would like to point out another important difference between our work and the work by T. Gridley and colleagues. In our model, Notch1 was inactivated after birth, whereas the doubly heterozygous Jag1 null/Notch2 hypomorphic mice suffered from reduced Notch signaling already during embryonal development. Whether postnatal inactivation or reduction of Jagged1 and Notch2 mediated Notch signaling would also result in the development of AGS is not yet clear. Inducible inactivation of both genes after birth would help to clarify this question.

The striking liver phenotype caused by inactivation of Notch1 after birth was the spontaneous proliferation of hepatocytes, resulting in a picture identical to that of human NRH. We could not find the vascular changes such as obliteration of small portal veins or central veins that have been postulated to be the cause of NRH. Interestingly, hyperproliferation of hepatocytes by continuous overstimulation with interleukin (IL)-6 in a double transgenic mouse model expressing high levels of both human IL-6 and soluble IL-6 receptor under the control of liver-specific promoters also resulted in NRH.28 Obviously, NRH is not a disease with a unique cause, but rather a pathological change of the liver architecture that can be the result of different underlying pathogenetic mechanisms. In the case of Notch1 inactivation, the most likely mechanism is a continuous proliferative activity of hepatocytes. We conclude that Notch1 signaling is an important inhibitor of hepatocyte proliferation.

Interestingly, Jagged1 has been found to be expressed in hepatocytes, bile duct cells, portal veins, hepatic arteries, and hepatic veins.9 Therefore, mutual inhibition of cell proliferation between adjacent hepatocytes through reciprocal Jagged1–Notch1 interactions could be the cause of most hepatocytes in a normal liver being in a quiescent G0 stage of the cell cycle for extended periods. Seemingly contradictory to this interpretation, Köhler and colleagues recently reported an activation of Notch signaling during liver regeneration after PH and concluded that Notch/Jagged signaling potentially induces cell growth and differentiation after PH.29 They observed upregulation of Jagged1 and Notch1 protein in plasma membrane fractions of rat liver extracts 2 to 4 days after PH. However, Notch1 Western blots were difficult (because of low expression levels of Notch1), and semiquantitative real-time polymerase chain reaction showed a decrease of Notch1 mRNA during the first 2 to 4 days after PH.29 The same authors also reported that pretreatment of rats with silencing RNA for Notch1 and Jagged1 resulted in a decreased BrdU incorporation into hepatocytes at days 2, 3, and 4 after PH.29 This latter result is consistent with our results shown in Fig. 5B. However, because we observed a normal increase in liver weight relative to body weight in our Notch1 knockout mice compared with the controls, we conclude that liver regeneration is not impaired in Notch1-deficient mice, but on the contrary is as precisely regulated as in the wild-type control mice. Interestingly, Köhler and colleagues also report that they found no significant difference in the recovery of liver weight after PH.

As outlined in Results, we believe that the decreased proliferative response as measured by BrdU incorporation after PH is a result of the increased liver mass of Notch1-deficient mice before PH. After removal of the middle and the left liver lobes, the remnant liver in control mice is 1.66% of the body weight, but in the Notch1-deficient mice it is 2.33% of the body weight. Because the regenerative response is inversely correlated with the amount of remaining liver after PH (and not because the remaining hepatocytes cannot signal through Notch1), the Notch1 mice have a less vigorous proliferative response after a two-thirds hepatectomy. We have recently found that tissue-specific deletion on Notch1 in the epidermis results in skin hyperplasia and predisposes to spontaneous basal cell-carcinoma–like tumors over time.20 In this model, Notch1 deletion results in increased and sustained expression of Gli2, a downstream component of the Sonic-hedgehog (Shh) signaling pathway, and increased β-catenin signaling. The results indicate that Notch1 functions as a tumor-suppressor gene in mammalian skin.20 Interestingly, mutations in the β-catenin gene have been found in up to 40% of primary hepatocellular carcinomas (HCCs),30, 31 and even more frequently in hepatoblastomas.32, 33 Furthermore, overexpression of Notch1 in HCC cell lines was recently shown to inhibit cell growth in vitro and in vivo.34 Of note, a possible link between NRH and HCC that had arisen in livers without cirrhosis was found in a study on histological slides of 342 patients without cirrhosis with an HCC. Among the 342 samples, the authors found 23 with an NRH, and 74% of them showed liver cell dysplasia.35

In conclusion, these observations together with the function of Notch1 to inhibit hepatocyte proliferation under nonpathological conditions raise the intriguing possibility that Notch1 signaling may be part of a tumor suppressor–like program in the liver. Although Notch1 deficiency in the liver did not result in the development of liver carcinomas, the spontaneous proliferation of Notch1−/− hepatocytes may represent the first genetic hit in a cascade of multiple mutations that have to be acquired before hepotocyte growth control is completely deregulated. Future work is needed to test this hypothesis, for example, by investigating whether Notch1 deficiency facilitates chemical-induced liver carcinogenesis.

Acknowledgements

The authors thank Vincenza Carafa for technical assistance.

Ancillary