A mouse model of Alagille syndrome: Notch2 as a genetic modifier of Jag1 haploinsufficiency

The Notch signaling pathway is an evolutionarily conserved intercellular signaling mechanism, and mutations in its components disrupt cell fate specification and embryonic development in organisms as diverse as insects, nematodes and mammals (for reviews, see Artavanis-Tsakonas et al., 1999; Kadesch, 2000; Mumm and Kopan, 2000; Weinmaster, 2000). Recent work has established that several human diseases are caused by mutations in components of the Notch signaling pathway. Mutations in the JAG1 gene, which encodes a ligand for Notch-family receptors, cause Alagille syndrome (Li et al., 1997; Oda et al., 1997). Alagille syndrome (OMIM #118450) is a pleiotropic developmental disorder characterized by neonatal jaundice and impaired differentiation of intrahepatic bile ducts (Alagille et al., 1987; Krantz et al., 1997; Emerick et al., 1999). The bile duct paucity causes chronic cholestasis, the disruption of bile flow from the liver into the gut. Accompanying features of this syndrome include congenital heart defects, skeletal defects, eye abnormalities, a characteristic facial appearance, and kidney abnormalities (reviewed by Krantz et al., 1997; Emerick et al., 1999). Alagille syndrome exhibits autosomal dominant inheritance, and the first mutations identified in the JAG1 gene in Alagille syndrome patients were inactivating mutations, generally leading to premature truncation of the JAG1 protein (Li et al., 1997; Oda et al., 1997). An extensive survey of the types and frequency of JAG1 mutations in Alagille syndrome patients revealed that patients with large deletions encompassing the entire JAG1 gene had the same phenotype as patients with intragenic JAG1 mutations, suggesting that haploinsufficiency for the JAG1 gene was a primary cause of Alagille syndrome (Spinner et al., 2001). Missense mutations in the JAG1 gene have also been observed in Alagille syndrome patients (Krantz et al., 1998; Morrissette et al., 2001; Spinner et al., 2001). Some missense mutations that have been studied in detail result in defective intracellular transport and processing of the mutant JAG1 protein, further supporting JAG1 haploinsufficiency as the cause of the majority of Alagille syndrome cases (Morrissette et al., 2001).

We have described previously the phenotype of mice homozygous and heterozygous for a targeted null mutation of the Jag1 gene, Jag1^dDSL (Xue et al., 1999). Mouse embryos homozygous for the Jag1^dDSL null mutation died from vascular defects by embryonic day 10 (E10) of gestation. Mice heterozygous for the Jag1^dDSL mutation, whose genotype mimics that of human Alagille syndrome patients, proved to be a disappointing animal model for this disease. The Jag1^dDSL/+ heterozygous mice exhibited anterior chamber eye defects, but did not exhibit other phenotypes associated with Alagille syndrome in humans (Xue et al., 1999).

We have also recently characterized a targeted mutation of the Notch2 gene, Notch2^del1 (McCright et al., 2001). Alternative splicing of the Notch2^del1 mutant allele leads to the production of two different in-frame transcripts that delete either one or two EGF repeats of the Notch2 protein, suggesting that this allele is a hypomorphic Notch2 mutant allele. Approximately 50% of Notch2^del1/Notch2^del1 homozygotes died perinatally from defects in glomerular development in the kidney (McCright et al., 2001). Analysis of the kidney defects in Notch2^del1/Notch2^del1 homozygotes suggested that the Jag1 gene encoded the ligand that was signaling to the Notch2 receptor during glomerular development. This model was supported by the analysis of mice that were heterozygous for both the Jag1^dDSL and Notch2^del1 alleles (McCright et al., 2001). These double heterozygotes exhibited kidney defects that were similar to, although less severe than, the kidney defects observed in Notch2^del1/Notch2^del1 homozygotes.

In this paper, we demonstrate that Jag1^dDSL/+ Notch2^del1/+ double heterozygous mice exhibit other phenotypes characteristic of human Alagille syndrome patients. In addition to the kidney glomerular defects, Jag1^dDSL/+ Notch2^del1/+ double heterozygous mice exhibit jaundice, growth retardation, paucity of intrahepatic bile ducts, heart defects, and eye defects. This demonstrates that the Notch2 and Jag1 mutations interact to create a more representative mouse model of Alagille syndrome, and provides a possible explanation of the variable phenotypic expression observed in Alagille syndrome patients. The defects in bile duct epithelial cell differentiation and morphogenesis in the Jag1^dDSL/+ Notch2^del1/+ double heterozygotes are similar to defects in epithelial morphogenesis observed in Notch pathway mutants in Drosophila, suggesting that a role for the Notch signaling pathway in regulating epithelial morphogenesis is evolutionarily conserved.

Official nomenclature for the mutant alleles used in these studies are Jag1^dDSL (Xue et al., 1999): Jag1^tm1Grid, and Notch2^del1 (McCright et al., 2001): Notch2^tm1Grid. Mice were analyzed on a mixed C57BL/6J × 129S1/SvImJ background (C57BL/6J backcross generation of N3 or greater).

For detection of Notch2 protein, a rabbit polyclonal antibody generated to amino acids 1-255 of the human NOTCH2 protein (sc-5545, Santa Cruz) was used. Paraffin sections were dewaxed in xylene, rehydrated through graded ethanol washes and treated at 95°C with 10 mM sodium citrate twice for 5 minutes to release the antigen. Slides were treated with 0.3% H₂O₂ in phosphate-buffered saline (PBS) for 20 minutes to block endogenous peroxidase activity, washed twice in PBS, and blocked with 1% non-fat milk in PBS, 0.05% Tween 20 (PBST) for 30 minutes at room temperature. Slides were washed with PBS and incubated with a 1:25 dilution of the rabbit anti-Notch2 antibody. They were then washed in 50 mM Tris, 150 mM NaCl, 0.05% Tween 20 (TBST) three times and incubated with horseradish peroxidase-conjugated donkey anti-rabbit IgG secondary antibody (Jackson ImmunoResearch) at a final concentration of 16 μg/ml in TBST. After incubation at room temperature for 2 hours the slides were washed three times in TBST and once in PBS. Secondary antibody was visualized using DiaminoBenzidine (Sigma) substrate including a final concentration of 0.2% NiCl. Slides were counterstained with Eosin before mounting.

For detection of Jag1 protein, a goat polyclonal antibody generated to the extracellular domain of the rat Jag1 protein (AF599, R & D Systems) was used. For antigen release, sections were treated with 0.1% trypsin at 37°C for 10 minutes. The secondary antibody was horseradish peroxidase-conjugated donkey anti-rabbit IgG (Jackson ImmunoResearch) at a final concentration of 16 μg/ml in TBST.

For lectin binding, biotinylated wheat germ agglutinin (B-1025, Vector Laboratories) and biotinylated Dolichos biflorus agglutinin (B-1035, Vector Laboratories) were used at 20 μg/ml. For detection of lectin binding, horseradish peroxidase-conjugated avidin (A-2004, Vector Laboratories) or fluorescein-conjugated avidin (A2011, Vector Laboratories) were used at 10 μg/ml.

Mouse embryos and neonates were partially dissected to expose the heart, and a 1:1 dilution of India ink/PBS was injected into the left ventricle. Torsos were fixed overnight in 4% paraformaldehyde. Hearts then were dissected from surrounding tissues, dehydrated through a methanol series and cleared in 2:1 benzyl alcohol:benzyl benzoate.

Our previous characterization of kidney defects in Notch2^del1/Notch2^del1mutants indicated that the likely ligand signaling to the Notch2 receptor was the Jag1 protein (McCright et al., 2001). Therefore, we intercrossed Jag1^dDSL/+ mice (Xue et al., 1999) with Notch2^del1/+ mice (McCright et al., 2001) to assess dosage-sensitive genetic interactions in this double mutant combination. The Jag1^dDSL/+ Notch2^del1/+ (henceforth abbreviated J1N2+/–) double heterozygous mice exhibited jaundice (Fig. 1) and postnatal growth retardation (Table 1). J1N2+/– double heterozygous neonates were born at the expected Mendelian ratio, but approximately 50% of the double heterozygotes died within the first week after birth (Table 1).

The neonatal jaundice and cholestasis observed in Alagille syndrome patients is a direct consequence of the paucity of intrahepatic bile ducts (Krantz et al., 1997; Emerick et al., 1999). The J1N2+/– double heterozygous mice exhibited defects in intrahepatic bile duct differentiation. Intrahepatic bile ducts develop as part of the portal triad, which includes the portal vein, the hepatic artery and the bile duct. During late embryonic and early postnatal development, bile duct epithelial cells differentiate from hepatoblasts adjacent to the portal veins (Gall and Bhathal, 1989; Shiojiri and Nagai, 1992; Vijayan and Tan, 1997; Kanno et al., 2000; Shiojiri et al., 2001). Examination of histological sections of the livers of J1N2+/– mice revealed that few morphologically identifiable bile ducts were present (Fig. 2B). Analysis of Dolichos biflorus agglutinin (DBA) lectin expression, a marker for bile duct epithelial cells (Shiojiri and Nagai, 1992), revealed that DBA-positive cells formed patent bile ducts adjacent to the portal veins in wild-type mice (Fig. 2C). In J1N2+/– mice, DBA-positive cells were present in small numbers adjacent to the portal veins, but these cells were not arranged into patent epithelial ducts (Fig. 2D). Wheat germ agglutinin (WGA) is a more widely expressed marker for a subset of hepatoblasts that includes precursors to bile duct epithelial cells (Shiojiri and Nagai, 1992). No obvious differences in the numbers or distribution of WGA-positive cells were observed between J1N2+/– double heterozygotes and their littermates (Fig. 2E,F).

Some J1N2+/– double heterozygotes survived to adulthood. However, these animals had elevated levels of alanine aminotransferase and alkaline phosphatase, which are indicative of liver and biliary dysfunction (Table 2). Elevated levels of these enzymes are consistent with chronic cholestasis. Livers of adult J1N2+/– double heterozygotes exhibited an abnormal proliferation of cells adjacent to the portal veins (Fig. 2H) and bile pigment accumulation in the hepatic parenchyma, which are defects commonly associated with chronic cholestasis in humans (Cotran et al., 1999). The J1N2+/– double heterozygotes also exhibited elevated blood urea nitrogen levels (Table 2). This finding is consistent with the kidney glomerular defects we reported previously in J1N2+/– double heterozygotes (McCright et al., 2001), and with the renal defects observed in some Alagille syndrome patients (Alagille et al., 1987; Habib et al., 1987; Krantz et al., 1997; Emerick et al., 1999).

We also examined whether defects in differentiation of bile duct epithelial cells could be observed in mice homozygous for the Notch2^del1 mutation. These mice, and their littermate controls, had to be analyzed at P0 because of the perinatal lethality accompanying the defects in kidney glomerular development observed in Notch2^del1/Notch2^del1 mice (McCright et al., 2001). Analysis of DBA lectin expression at P0 in liver sections from J1N2+/– and Notch2^del1/Notch2^del1 mice (Fig. 3) revealed that almost no DBA-positive cells could be observed adjacent to the portal veins in either the J1N2+/– (Fig. 3B) or the Notch2^del1/Notch2^del1 (Fig. 3D) mice. This demonstrates that similar bile duct epithelial cell differentiation defects occur in both the J1N2+/– and Notch2^del1/Notch2^del1 mice. We were unable to assess whether bile duct epithelial cell differentiation defects occur in mice homozygous for the Jag1^dDSL mutation because these embryos die from vascular defects at E10 (Xue et al., 1999), prior to the differentiation of bile duct epithelial cells.

More than 95% of Alagille syndrome patients exhibit congenital heart defects, including peripheral pulmonic stenosis, pulmonic valve stenosis, atrial and ventricular septal defects, coarctation of the aorta, and tetralogy of Fallot (Alagille et al., 1987; Krantz et al., 1997; Emerick et al., 1999; Krantz et al., 1999; Eldadah et al., 2001). The J1N2+/– mice exhibited multiple cardiac defects. Analysis of whole-mount preparations of India ink-injected hearts revealed right ventricular hypoplasia and narrowing of the pulmonary artery (Fig. 4). Of nine hearts from J1N2+/– mice examined in whole mount, all displayed right ventricular hypoplasia while 6/9 displayed narrowing of the pulmonary artery. Histological analysis of the J1N2+/– double heterozygotes revealed additional heart defects, including atrial (Fig. 5A,B) and ventricular (Fig. 5C,D) septal defects and dextropositioning of the aorta (Fig. 5E,F). Dextropositioning of the aorta was observed most commonly (14/14 J1N2+/– mice examined), followed by atrial (12/14 J1N2+/– mice) and ventricular (6/14 J1N2+/– mice) septation defects.

Alagille syndrome patients also exhibit, with high penetrance, anterior chamber eye defects, vertebral anomalies such as butterfly vertebrae, a characteristic facial appearance and kidney abnormalities (Alagille et al., 1987; Krantz et al., 1997; Emerick et al., 1999). Mice heterozygous for the Jag1^dDSL targeted null mutation (J1/+), as well as mice heterozygous for the Coloboma deletion (Cm/+), which deletes the Jag1 gene, exhibit anterior chamber eye defects (Xue et al., 1999). The eye defects observed in the J1N2+/– mice appeared similar to those we observed previously in J1/+ and Cm/+ mice (data not shown). We also prepared Alizarin Red- and Alcian Blue-stained skeletal preparations of the J1N2+/– mice. We examined these for craniofacial, vertebral or other skeletal anomalies, but did not detect any obvious skeletal defects in the J1N2+/– mice (data not shown). We have reported previously the defects in kidney glomerular development observed in J1N2+/– mice (McCright et al., 2001).

To determine whether the phenotypes observed in the J1N2+/– mice correlated with domains of gene expression, we analyzed expression of the Jag1 and Notch2 proteins during bile duct and heart development. During bile duct formation, Jag1 protein was expressed in the portal veins and hepatic arteries (Fig. 6A,C). Jag1 immunoreactivity was observed in both endothelial cells and non-endothelial supporting cells. However, Jag1 protein was not expressed in bile ducts (Fig. 6C,D). Notch2 protein was expressed in cells surrounding the portal vein, the hepatic artery and the bile ducts, but was not expressed in the bile ducts themselves (Fig. 6B, and data not shown).

We also tested whether Notch2 immunoreactivity could be detected in sections of livers from Notch2^del1/Notch2^del1 homozygous mutants. In both wild-type littermates (Fig. 6E) and Notch2^del1/Notch2^del1 mutants (Fig. 6F) at E18, Notch2 immunoreactivity was highest in groups of cells adjacent to the developing portal triads. No differences between the Notch2^del1/Notch2^del1 mutants and littermate controls in the numbers or distribution of Notch2-positive cells were observed, demonstrating that the protein product of the alternatively spliced Notch2^del1 mutant allele (McCright et al., 2001) is expressed in Notch2^del1/Notch2^del1 homozygous mutant embryos.

During heart development, Jag1 protein was expressed at embryonic day 13 (E13) in the aorta, the pulmonary trunk and in coronary arteries (Fig. 7A,C). This pattern of expression is similar to that previously described for Jag1 RNA expression in mouse (Loomes et al., 1999) and human (Loomes et al., 1999; Crosnier et al., 2000; Jones et al., 2000) embryos. Notch2 immunoreactivity was detected in cells surrounding the aorta, the pulmonary trunk and coronary arteries at E13 (Fig. 7B,D). In addition, we detected Notch2 immunoreactivity in the myocardium and the walls of the atria (Fig. 7B,D). By E16 Jag1 immunoreactivity was detected primarily in coronary blood vessels, and Notch2 immunoreactivity was detected in cells surrounding the coronary vessels and was greatly reduced in the myocardium (data not shown). These studies demonstrate that during both bile duct and cardiac development, Jag1-expressing cells are found near or adjacent to Notch2-expressing cells, suggesting that the Jag1 protein is a physiological ligand for the Notch2 receptor during development of these tissues. This hypothesis is supported by the dosage-sensitive genetic interaction we observe in the J1N2+/– double heterozygotes, and with our previous studies on the roles of these genes during kidney development (McCright et al., 2001).

We demonstrate here that, while mice heterozygous for the Jag1^dDSL mutant allele are a disappointing Alagille syndrome model (Xue et al., 1999), J1N2+/– double heterozygotes reproduce most of the clinically relevant phenotypes observed in human Alagille syndrome patients. J1N2+/– mice exhibit jaundice, growth retardation, and bile duct, heart, eye and kidney abnormalities that are similar or identical to the abnormalities observed in Alagille syndrome patients. However, J1N2+/– mice do not appear to exhibit the vertebral or craniofacial abnormalities seen in Alagille syndrome patients. The vertebral and craniofacial anomalies seen in Alagille syndrome patients are relatively minor: butterfly (clefted) vertebrae and a characteristic facial appearance. Butterfly vertebrae result from clefting abnormalities of the vertebral bodies, and the characteristic facial features of Alagille patients include a prominent forehead, deep set eyes, pointed chin, and a saddle or straight nose with a bulbous tip (Krantz et al., 1997).

Examination of Alizarin Red-Alcian Blue stained skeletal preparations of J1N2+/– mice did not reveal any obvious skeletal abnormalities. During embryogenesis, the Jag1 gene is expressed in the most recently formed somite (Zhang and Gridley, 1998). Analysis, at E10, of mouse embryos homozygous for the Jag1^dDSL mutation did not reveal any defects in somite formation, although the vascular defects present in the homozygous mutant embryos precluded the analysis of somite or vertebral defects at later embryonic stages (Xue et al., 1999). We do not have an explanation for why the J1N2+/– mice do not exhibit obvious vertebral or craniofacial abnormalities. Perhaps the reduction in dosage of the Jag1 gene can be compensated for by expression of the Notch ligands encoded by the Dll1 and/or Dll3 genes. Homozygous mutants for either of these genes exhibit substantial defects in somite development (Hrabé de Angelis et al., 1997; Kusumi et al., 1998; Bulman et al., 2000).

Alagille syndrome exhibits high penetrance but extremely variable expressivity (Alagille et al., 1987; Krantz et al., 1997; Emerick et al., 1999). Frequently, one of the parents of a child identified with Alagille syndrome has a subclinical presentation of the disease that has not been diagnosed previously. In addition, no genotype/phenotype correlations have been identified between the different types of JAG1 mutations that give rise to Alagille syndrome, including nonsense mutations, missense mutations and total gene deletions (Krantz et al., 1998; Spinner et al., 2001). One possible explanation for this variable expressivity is the existence of genetic modifiers of the disease phenotype in the human population.

Previous work has shown that the Notch signaling pathway is exquisitely sensitive to the gene dosage of various pathway components (reviewed by Artavanis-Tsakonas et al., 1991). This dosage sensitivity has formed the basis for several genetic screens in Drosophila that have identified both new alleles of previously known pathway components, and novel genes whose expression impacts Notch pathway function (Brand and Campos-Ortega, 1990; Verheyen et al., 1996; Go and Artavanis-Tsakonas, 1998; Cornell et al., 1999; Purcell and Artavanis-Tsakonas, 1999). Gene dosage sensitivity and genetic modifiers of Notch pathway components have also been demonstrated in mammals. Notch1/Notch 4 double mutant embryos and mice exhibit synergistic genetic interactions (Krebs et al., 2000), and a mutation in the Lunatic fringe (Lfng) gene partially suppresses the inner ear phenotype of Jag2 homozygous mutant mice (Zhang et al., 2000). Genetic modifiers of the Jag2^sm (syndactylism) mutation, a hypomorphic Jag2 mutant allele, also have been described (Sidow et al., 1997).

We have shown here that the Notch2 gene acts as a genetic modifier to interact with a Jag1 mutation to create a more representative mouse model for Alagille syndrome. We hypothesize that similar genetic interactions may occur in human Alagille syndrome patients, and that particular NOTCH2 alleles may influence the severity of Alagille syndrome phenotypes. This hypothesis can be tested by determining whether segregation of different NOTCH2 alleles in families with Alagille syndrome correlates with phenotypic severity. Further studies of J1N2+/– mice may lead to additional insights into the pathogenesis of Alagille syndrome in humans.

The analysis of the J1N2+/– mice demonstrates that Notch signaling is required for bile duct formation in mice. We are not aware of any other mouse mutant that disrupts bile duct development in a similar manner. Analysis of DBA lectin expression (a marker for bile duct epithelial cells) in the livers of J1N2+/– mice indicates that a few DBA-expressing cells differentiate in their normal position adjacent to the portal veins. However, these cells do not form morphologically normal epithelial bile ducts. Neither Jag1 nor Notch2 protein is expressed in bile duct epithelial cells themselves. Instead, Jag1 protein is expressed in the hepatic vasculature in both endothelial cells and periendothelial supporting cells. Notch2 protein is expressed in a subset of hepatoblasts adjacent to the Jag1-expressing cells. We propose that at least some of the Notch2-expressing cells are bile duct epithelial cell precursors, and that the decreased Notch2 signal in the J1N2+/– mice results in the differentiation of fewer bile duct epithelial cell precursors. This model is supported by the finding that differentiation of DBA-positive bile duct epithelial cell precursors is inhibited in Notch2^del1/Notch2^del1 mutant homozygotes.

Notch signaling is essential for the development and proper morphogenesis of multiple epithelial tissues in Drosophila (Hartenstein et al., 1992). A requirement for Notch signaling has been demonstrated during development of the endoderm (Tepass and Hartenstein, 1995), trachea (Ikeya and Hayashi, 1999; Llimargas, 1999; Steneberg et al., 1999) and ovarian follicular epithelium (Goode et al., 1996; Zhao et al., 2000) (for a review, see Dobens and Raftery, 2000). The defects in bile duct epithelial cell differentiation and morphogenesis in J1N2+/– mice are similar to the defects in epithelial morphogenesis of Notch pathway mutants in Drosophila. We have previously described the defects in development of the kidney glomeruli in Notch2^del1/Notch2^del1 homozygous mutant mice. In the majority of glomeruli of these mice, glomerular podocyte precursors differentiate but do not epithelialize, remaining instead as a dysmorphic aggregate of cells (McCright et al., 2001). The phenotypes of the J1N2+/– and Notch2^del1/Notch2^del1 mice suggest that a role for the Notch signaling pathway in regulating epithelial morphogenesis has been conserved between insects and mammals.

The phenotype exhibited by J1N2+/– mice is an example of nonallelic noncomplementation, in which recessive mutant alleles in two distinct genes fail to complement one another (Yook et al., 2001). Nonallelic noncomplementation is often interpreted as evidence of a physical interaction between the products of the two noncomplementing genes. Two models have been proposed to explain nonallelic noncomplementation; the gene dosage model and the poison model (Stearns and Botstein, 1988; Fuller et al., 1989; Regan and Fuller, 1990) (reviewed by Yook et al., 2001). In the gene dosage model, a mutant phenotype results from the simultaneous reduction in gene dosage at the two interacting loci. In the poison model, at least one of the interacting loci must make an altered protein product that binds to and impairs the function of the protein encoded by the other interacting locus.

Here we demonstrate nonallelic noncomplementation between a null Jag1 allele and a hypomorphic Notch2 allele. At present we cannot distinguish between the gene dosage and poison models. We show here by immunohistochemistry that the protein product of the alternatively spliced Notch2^del1 mutant allele (McCright et al., 2001) is expressed in Notch2^del1/Notch2^del1 homozygous mutant embryos. Production of an altered Notch2 protein would be consistent with the poison model, but we cannot exclude a gene dosage model. Two missense alleles of the Jag1 gene that are likely to be hypomorphic mutant alleles have been isolated recently from large scale mouse mutagenesis screens (Kiernan et al., 2001; Tsai et al., 2001). We are also constructing additional targeted Notch2 mutant alleles, including a definitive null mutation. Analysis of these mutant alleles in different combinations and gene dosages should permit us to distinguish between the gene dosage and poison models in this model for Alagille syndrome.

We thank S. Ackerman, W. Frankel and T. O’Brien for comments on the manuscript. This work was supported by grants from the NIH (NS36437) and the March of Dimes Foundation to T. G., by a National Research Service Award to B. M., and by a Core grant (CA34196) from the National Cancer Institute to the Jackson Laboratory.