Hepatoblasts give rise to both mature hepatocytes and cholangiocytes. While Notch signaling has been implicated in the formation of bile ducts composed of cholangiocytes, little is known about the mechanism of lineage commitment of hepatoblasts. Here we describe the role of the Notch pathway in hepatoblast differentiation. Immunohistochemical analysis showed that Jagged1 was expressed in the cells surrounding the portal veins and Notch2 was expressed in most hepatic cells at mid gestation when ductal plates are formed surrounding the portal veins. Interestingly, the Jagged1+ cells were adjacent to ductal plates, suggesting that the Notch signaling is activated in hepatoblasts that undergo differentiation into cholangiocytes. In fact, expression of the Notch intracellular domain in Dlk+ hepatoblasts inhibited hepatic differentiation and significantly reduced the expression of albumin, a marker of both hepatoblasts and hepatocytes. Furthermore, the addition of Matrigel to the hepatoblast culture upregulated the expression of cytokeratin 7 and 19, integrin β4, and HNF1β, which are known to be expressed in cholangiocytes. By contrast, downregulation of the Notch signaling by siRNA specific for Notch2 mRNA as well as by the γ-secretase inhibitor L-685,458 promoted the hepatic differentiation. Consistent with the previous finding that mature cholangiocytes strongly express HNF1β, but barely express HNF1α, HNF4, and C/EBPα, activation of the Notch signaling upregulated HNF1β expression, whereas it downregulated the expression of HNF1α, HNF4, and C/EBPα. These results suggest that the Notch signaling contributes to form a network of these transcription factors suitable for cholangiocyte differentiation.
The liver consists of endodermal components, hepatocytes and cholangiocytes, and various types of nonparenchymal cells such as sinusoidal endothelial cells, stellate cells, Kupffer cells and blood cells. Hepatocytes and cholangiocytes constitute, respectively, the liver parenchyma and the lumen of intrahepatic bile ducts, extra-hepatic bile ducts and gallbladder. They originate from a common progenitor, the hepatoblasts, which are derived from the ventral foregut endoderm and the main component of the liver primordium (Duncan and Watt, 2001; Shiojiri, 1997; Zaret, 2000). Hepatoblasts give rise to mature hepatocytes in the liver parenchyma, whereas they differentiate into cholangiocytes in the periportal area. It is thus reasonable to consider that a specific environment of periportal area is needed for hepatoblasts in order to become cholangiocytes. However, it was difficult to characterize hepatoblasts because there was no means of isolating them. Recently, several monoclonal antibodies against cell surface molecules were used to isolate hepatoblasts from mouse and rat fetal livers and the isolated hepatoblasts were shown to proliferate clonally and differentiate into two lineages (Kubota and Reid, 2000; Suzuki et al., 2000, 2002; Tanimizu et al., 2003). Thus, it became possible to characterize the growth and differentiation potential of hepatoblasts and to investigate the mechanism by which they give rise to hepatocytes and cholangiocytes.
Studies on human genetic diseases and mutant mice implicate the Notch signaling pathway in the formation of bile ducts. Alagille syndrome is a human pleiotropic genetic disease involving developmental disorders of the liver, heart, eye, skeleton, craniofacial appearance and kidney and is known to be caused by an allelic null mutation of the jagged1 gene (JAG1) (Krantz et al., 1997; Li et al., 1997). The liver abnormality is characterized by jaundice and an impaired differentiation of intra-hepatic bile ducts. Although mice heterozygous for the jagged 1 gene do not exhibit abnormalities of intra-hepatic bile ducts (Xue et al., 1999), those heterozygous for the jagged 1 null allele and Notch2 hypomorphic allele exhibit developmental abnormalities that are characteristics of human Alagille syndrome including intra-hepatic bile duct malformation (McCright et al., 2002). Thus, Notch signaling is essential for bile duct formation in mice as well as humans.
The Notch signaling pathway is an evolutionarily conserved mechanism that controls a wide variety of developmental processes (Artavanis-Tsakonas et al., 1999; Greenwald, 1998; Milner and Bigas, 1999). Notch proteins are large membrane-bound receptors that interact with membrane-bound ligands belonging to the Delta-Serrate-Lag-2 (DSL) family. Four Notch genes have been identified in mammals and their mutations cause severe abnormalities in organ development and adult homeostasis (Hamada et al., 1999; Joutel et al., 1996; Krebs et al., 2000; Shutter et al., 2000; Spinner, 2000; Swiatek et al., 1994). In mammals, three Delta (Dll1, Dll3 and Dll4) and two Jagged proteins (Jag 1 and Jag 2) have been identified as Notch ligands and are indispensable for the development and function of organs (de Angelis et al., 1997; De Bellard et al., 2002; Dunwoodie et al., 2002; Jiang et al., 1998; Xue et al., 1999). The Notch signaling is initiated by the interaction between the receptors and ligands, which triggers the proteolytic cleavage of the Notch receptor at the membrane proximal region by the γ-secretase presenilin (Kidd et al., 1998; Ray et al., 1999; Struhl and Greenwald, 1999). The Notch intracellular domain (NICD) then translocates to the nucleus where it associates with CBF1/RBPJκ and Deltex via the RAM domain (Kurooka et al., 1998; Tamura et al., 1995; Zhou et al., 2000) and cdc10-ankyrin repeat domain (Matsuno et al., 1995), respectively. The NICD/RBPJκ complex transactivates protein targets such as Hes1, Hes5, Hes7, Hes-related protein (HERP) 1 and HERP2 (Bessho et al., 2001; Iso et al., 2001; Ohtsuka et al., 1999).
The Notch signaling pathway controls various biological events, e.g. segregation of neural and epidermal cells from proneural cluster in the ventral ectoderm of Drosophila embryos, gliogenesis from multipotent neural progenitor cells (Morrison et al., 2000; Tanigaki et al., 2001), T cell differentiation from lymphoid precursors (Milner and Bigas, 1999; Pui et al., 1999; Radtke et al., 1999), and maintenance of the immature state of hematopoietic stem cells (Carlesso et al., 1999; Jones et al., 1998; Varnum-Finney et al., 1998). During the processes of hematopoiesis, Notch receptors and ligands are expressed in distinct cells, e.g. a Notch is expressed in hematopoietic cells, and its ligand is expressed in stroma cells (Jones et al., 1998; Varnum-Finney et al., 1998).
Several transcription factors known as liver-enriched transcription factors play key roles in liver organogenesis and metabolic functions of the liver (Akiyama et al., 2000; Clotman et al., 2002; Coffinier et al., 2002; Parviz et al., 2003; Pontoglio et al., 1996). Among them, hepatic nuclear factor (HNF) HNF1β (TCF2) and HNF6 (Onecut1) are highly expressed in cholangiocytes and have been implicated in the formation of bile ducts. HNF6 null mice exhibit liver abnormalities, e.g. no gallbladder, abnormal extra-hepatic bile ducts, and an impaired development of the intra-hepatic bile ducts at the perinatal stage (Clotman et al., 2002). Conditional inactivation of the HNF1β gene in the liver results in severe jaundice due to abnormalities of gallbladder and intra-hepatic bile ducts (Coffinier et al., 2002). Furthermore, HNF6 binding sites are present in the promoter of the HNF1β gene and HNF6 was shown to activate the HNF1β promoter, indicating that HNF6 is upstream of HNF1β in the bile duct development (Clotman et al., 2002). By contrast, HNF1α, HNF4, and C/EBPα (Cebpa) are strongly expressed in mature hepatocytes and play essential roles in the differentiation and metabolic functions of hepatocytes. Inactivation of the HNF1α gene does not affect embryonic development but results in death around weaning because of liver dysfunction and kidney abnormalities (Akiyama et al., 2000; Pontoglio et al., 1996). Targeted disruption of the HNF4 gene in embryonic liver results in less glycogen being accumulated and in an abnormal liver architecture due to a reduced expression of glycogen synthetase and cell adhesion molecules, respectively (Parviz et al., 2003). C/EBPα null mice die as neonates because of hypoglycemia associated with the impaired expression of gluconeogenic enzymes such as glucose-6-phosphatase (G6Pase) and phosphoenolpyruvate carboxykinase (PEPCK) (Wang et al., 1995).
While studies using mutant mice have shown that the liver-enriched transcription factors are required for liver development, it still remains unknown how extracellular signals control these transcription factors to determine the fate of hepatoblasts. In this paper, we describe the possibility that Notch signaling controls the network of liver-enriched transcription factors. The activation of Notch signaling in hepatoblasts resulted in inhibition of hepatic differentiation and induction of several cholangiocytic characteristics. We also found that ectopic expression of the Notch intracellular domain (NICD) altered the expression of several liver-enriched transcription factors. Taken together, we suggest that Notch signaling plays a key role in the differentiation of hepatoblasts.
Materials and Methods
Mice and plasmids
C57BL6 mice (Nippon SLC, Japan) were used for all the experiments. The retrovirus vector pMX-GFP encodes green fluorescence protein (GFP) controlled by the SV40 promoter. The vectors pMX-NICD-GFP and pMX-mutant NICD-GFP encode, as well as GFP, the Notch1 intracellular domain (NICD) and mutant NICD possessing two amino acid substitutions in the fourth ankyrin repeat domain (Kopan et al., 1994; Zhou et al., 2000), respectively. Rat Hes1 cDNA, originally cloned in the pUC vector, was excised by EcoRI digestion and inserted into the EcoRI site of the retroviral vector to make pMX-Hes1-GFP.
Mouse embryos and the liver of a postnatal (P) 7 days (P7) mouse were embedded in OCT compound (Sakura Finechemical, Tokyo, Japan). Frozen thin sections were prepared using a Microtome cryostat HM 500 (Microm, Walldorf, Germany) and mounted on glass slides (Matsunami Glass, Japan). The sections were fixed in PBS containing 4% paraformaldehyde (PFA) at 4°C for 10 minutes. After washing and blocking procedures, sections were incubated with antibody (Ab) solutions. Rabbit anti-mouse cytokeratin 19 Ab (CK19) (Tanimizu et al., 2003), and mouse monoclonal anti-α-smooth muscle actin Ab (DAKO, Carpinteria, CA) antibodies were used at 1:500. Goat anti-rat Jagged 1 Ab (R&D Systems, Minneapolis, MN), rat monoclonal anti-human Notch1 Ab [Developmental Studies Hybridoma Bank (DSHB), Univ. of Iowa, IA], and rat monoclonal anti-Notch2 Ab (DSHB) were used at 1:50. Sections were further incubated with biotin-conjugated anti-rabbit IgG Ab, anti-mouse IgG Ab or anti-rat IgG Ab (Vector, Burlingame, CA). Signals were visualized with the Vectastain ABC kit (Vector) and 3,3-diaminobenzidine tetrahydrochloride (DAB) (Roche). A section of E (embryonic day) 15.5 liver was incubated with anti-Jagged 1 and rabbit anti-bovine keratin (DAKO) antibodies. The signals of Jagged1 and keratin were visualized with Cy3-conjugated anti-goat IgG Ab (Rockland, Gilbertsville, PA) and FITC-conjugated anti-rabbit IgG Ab (Rockland), respectively. The samples were examined under a microscope, Nikon Eclipse E800 (Nikon, Tokyo, Japan).
For immunocytochemistry, Dlk+ cells were plated onto gelatin-coated chamber slides (Nunc, Roskilde, Denmark). After 4 or 7 days of incubation, they were fixed in 4% PFA at 4°C for 10 minutes and permeabilized in methanol (MeOH) at room temperature for 5 minutes. After two washes in PBS and blocking with 3% goat serum, they were incubated with 2 μg/ml of rabbit anti-albumin Ab (Nordic, Sweden) or rabbit anti-HNF4 Ab (Santa Cruz Biotechnology, Santa Cruz, CA). The signals were visualized with rhodamine-conjugated anti-rabbit IgG Ab (Chemicon, Temecula, CA).
Culture and retroviral infection
Liver cells were prepared from mouse E14.5 liver according to the method of Kamiya et al. (Kamiya et al., 1999). The method to isolate Dlk+ hepatoblasts from E14.5 livers was described in a previous report (Tanimizu et al., 2003). Dlk+ cells resuspended in Dulbecco's modified Eagle's medium (DMEM) (Nissui, Tokyo, Japan) were cultured in the same conditions used for the primary culture of fetal hepatic cells, in which hepatic differentiation is induced by oncostatin M (OSM) (Kamiya et al., 1999). Alternatively, after 5 days of incubation, cells were overlaid with Matrigel (Becton Dickinson) and incubated for an additional 5 days according to the previous report (Kamiya et al., 2002).
For the production of retrovirus, BOSC23 cells were used as packaging cells. BOSC23 cells (6×106 cells) were grown for 24 hours in a 10 cm dish in DMEM containing 10% fetal bovine serum (FBS). After replacing the medium with serum-free medium, OPTI-MEM (Gibco, BRL, Gaithersburg, MD), cells were transfected with 5 μg of plasmid DNA of pMX-GFP, pMX-NICD-GFP, pMX-mutantNICD-GFP or pMX-HES1-GFP using Lipofectamine™ and Plus™Reagent (both Invitrogen Corp., Carlsbad, CA) according to the manufacturer's instructions. Three hours later, to each plate 1 volume of fresh DMEM containing 20% FBS was added. Twenty-one hours later, the medium was replaced with fresh medium and after another 24 hours of incubation the virus-containing medium was harvested and centrifuged at 9000 rpm for 18 hours. Before transfection, the virus pellet was resuspended in 2 ml of fresh DMEM containing 10% FBS. Eight hours after plating Dlk+ cells on gelatin-coated dishes, the medium was replaced with 1 ml of retrovirus solution. GFP expression was examined by FACScallibur™ (Becton Dickinson, San José, CA) to evaluate the infection efficiency 4 days after transfection.
Cell culture with small interfering RNA (siRNA)
A 19-bp long sequence of mouse Notch2 cDNA and its complementary sequence were synthesized by Japan Bioservice (Saitama, Japan). A two nucleotide 3′ overhang was added to each RNA strand. The sequences we used are as follows: 5′-GAUGUGGACAGUGUCUGUTT-3′ and 5′-ACAGACACUCGUCCACAUCTT-3′. Eighteen hours before lipofection, E14.5 Dlk+ cells were plated on gelatin-coated dishes. Both oligonucleotides were diluted to a concentration of 20 μM in annealing buffer (100 mM potassium acetate, 30 mM HEPES-KOH pH 7.4, and 10 mM magnesium acetate). After incubation at 90°C for 1 min, the solution was kept at 37°C for 60 minutes to generate duplex short-interfering RNA (siRNA). The siRNA solution (10 μl) was mixed with 170 μl of OPTI-MEM and the mixture was then added to 20 μl of Oligofectamine™ (Invitrogen) solution which was prepared by mixing 4 μl of Oligofectamine and 16 μl of OPTI-MEM. The mixed solution was kept at room temperature for 20 min and then added to each well of a 6-well plate filled with 800 μl of OPTI-MEM. After 4 hours, one volume of fresh DMEM containing 20% FBS was added. As the control, the annealing buffer without siRNA was mixed with Oligofectamine solution and added to the medium. Four days after lipofection, mRNAs of Notch2 and GAPDH were detected by northern blot analysis. The signals were quantified using the NIH image program. The Notch2 levels were normalized with the corresponding GAPDH level.
Cell culture with L-685,458
L-685,458 (Bachem Bioscience, King of Prussia, PA), an aspartyl protease transition state mimic, inhibits γ-secretase, presenilin, by cleaving Notch in the membrane proximal region (Figueroa et al., 2002; Martys-Zage et al., 2000; Shearman et al., 2000). L-685,458 was dissolved to 1 and 5 mM in DMSO and 2 μl of the solution was added to 2 ml of the medium at 8 hours after the plating of Dlk+ cells. The same volume of DMSO was added to the control culture.
Northern blotting analysis
Total RNA was extracted from cultured cells. Ten micrograms of each RNA sample was loaded onto a 1.2% agarose gel containing 5.5% formaldehyde. After electrophoresis, RNA was transferred to a nylon membrane (Roche, Basel, Switzerland) and hybridized with a dioxygenin (DIG)-labeled DNA probe at 48°C overnight. After washing and blocking procedures, the membrane was incubated with alkaline phosphatase (AP)-labeled anti-DIG Ab (Roche). The signal was developed in CDP-star solution (Roche).
Reverse-transcription polymerase chain reaction
cDNA was prepared from 1 μg of total RNA using random hexamer primers and the First strand synthesis kit (Amersham Pharmacia Biotech., Piscataway, NJ). One-fifteenths of the cDNA solution was used for each reaction. The thermal cycle (denaturation at 94°C for 30 seconds, annealing at the temperature for each pair of primers for 30 seconds and extension at 72°C for 2 minutes) was repeated 30 times.
Quantitative analysis of mRNA levels was performed with a LightCycler (Roche). Each cDNA sample (1 μl) was mixed with a pair of primers and the reaction mixture of the LightCycler FastStart DNA SYBR I kit (Roche). The thermal cycle (denaturation at 95°C for 5 seconds, annealing at 55°C for 5 seconds, and extension at 72°C for 30 seconds) was repeated 35 times. To make a standard curve, the cDNA solution of the control culture was serially diluted and used as a template for RT-PCR. The analysis was repeated using three sets of samples prepared from three independent cultures. The quantitative analysis was performed according to the manufacturer's instructions. Statistical analysis to obtain P-values was performed using the JMP program. The following primers were used:
Notch1 (5′-GTGTGGATGAGGGAGATAAAC-3′ and 5′-GCATAGACAGCGGTAGAAAG-3′),
Notch2 (5′-TTGCTGTCGGAAGAT-3′ and 5′-CATGTGGTCAGTGAT-3′),
Notch3 (5′-CTGTGAAGTCAACGTGGATGA-3′ and 5′-CACAAGAATGAGCCCTGTGTA-3′),
Notch4 (5′-AAGAGGGAGAGGCTGAAGAAA-3′ and 5′-AGCCAGCATCAAAGGTGTAGT-3′),
Jagged1 (5′-TGCAGCTGTCAATCACTTCG-3′ and 5′-CAGAATGACGCTTCCTGTCG-3′),
Jagged2 (5′-AAGGACATACTCTACCAGTGC-3′ and 5′-ACGTCCTTGGTACTTCTGACG-3′),
Delta1 (5′-CTGAGGTGTAAGATGGAAGCG-3′ and 5′-CAACTGTCCATAGTGCAATGG-3′),
HNF1α (5′-ACGTCCGCAAGCAGCGAG-3′ and 5′-TACACTCTTCCACCAAGGTC-3′),
HNF1β (5′-GCAAACCGCCGGAAGGAAG-3′ and 5′-GGTTCTGAGATTGCTGGGG-3′),
HNF4 (5′-GGTCTGCCAGTGATGCAC-3′ and 5′-CAGGAGCTTGTAGGATTCAG-3′),
HNF6 (5′-GAAGTAATTCAGGGCAGATG-3′ and 5′-CAGCCACTTCCACATCCTCCG-3′),
CK7 (5′-GATGACCTCCGCAACACC-3′ and 5′-TCCAGCAGCTTGCGGTAG-3′),
Integrin β4 (5′-GACCTATGAAGAAGGTGCTC-3′ and 5′-GCTCAGATGCGTGCCATAG-3′),
GAPDH (5′-ACCACAGTCCATGCCATCAC-3′ and 5′-TCCACCACCCTGTTGCTGTA-3′).
Expression of Notch2 and Jagged1 in developing liver
First, we examined the RNA expression of Notch and their ligands in developing liver by RT-PCR (Fig. 1A). Specific bands of Notch1 and Notch2 RNA were clearly detected throughout the development, whereas the RNA expression of Notch3 and Notch4 was barely detectable. Jagged1 RNA expression was strongly expressed in fetal liver and weakly in postnatal liver, whereas the expression of Jagged2 and Dll1 was barely detectable throughout development (data not shown). Based on these results, we further examined the expression of Notch1, Notch2 and Jagged1. Immunohistochemical analysis showed that Notch2 was widely expressed in both E14.5 and E17.5 livers (Fig. 1C and E), whereas Notch1 protein was not detected in fetal liver. Furthermore, although Notch1 mRNA was detected by RT-PCR, it was barely detectable by northern analysis when compared with Notch2 mRNA (Fig. 1A lower panel). These results suggest that Notch2 is expressed most abundantly in fetal liver among four Notch receptors.
Seven days after birth, Jagged1 expression was detected around portal veins and in hepatic arteries (Fig. 1G) as described in a previous report (McCright et al., 2002). However, at this stage, Jagged1+ cells were not in contact with CK19+ cholangiocytes that formed bile ducts (Fig. 1G and H). Because it is believed that hepatoblasts start to differentiate into cholangiocytes during mid-gestation, we examined the expression of Jagged1 protein in fetal livers to investigate the role of Jagged1 in cholangiocyte differentiation. In E15.5 liver, cholangiocytes forming ductal plates that eventually become bile ducts, were stained with anti-keratin Ab (Fig. 1K). Consistent with the results of the in situ hybridization described in a previous report (Loomes et al., 2002), Jagged1 protein was strongly expressed in the cells surrounding the portal veins (Fig. 1J). Interestingly, the ductal plates were in close contact with Jagged1+ cells (Fig. 1L). Although it has not been elucidated whether or not Jagged1 is expressed in ductal plates (Crosnier et al., 2000; Loomes et al., 2002; Louis et al., 1999; McCright et al., 2002), our result clearly indicates that the Jagged1+ cells are adjacent but not identical to ductal plates.
The expression of Jagged1 protein around portal veins and the expression of Notch2 protein in most hepatic cells, including hepatoblasts, suggest that Notch signaling is activated in hepatoblasts upon interaction with Jagged1+ cells, which might be the first step in cholangiocyte differentiation.
Expression of the Notch intracellular domain inhibits hepatic differentiation and induces cholangiocytic characteristics
The expression pattern of Jagged1 and Notch2 in fetal liver suggested that the Notch pathway is activated where cholangiocytic differentiation begins. To investigate the function of Notch in the lineage commitment of hepatoblasts, Dlk+ hepatoblasts were infected with pMX-GFP encoding GFP, and pMX-NICD-GFP encoding both the Notch intracellular domain (NICD) (a constitutively active form of Notch) and GFP. In primary culture, fetal hepatocytes are induced by oncostatin M (OSM) to express hepatocyte differentiation markers such as G6Pase, tyrosine amino transferase (TAT) and carbamoyl phosphate synthetase (CPS) and exhibit a morphology similar to mature hepatocytes (Kamiya et al., 1999). Under the same conditions, Dlk+ hepatoblasts isolated from E14.5 liver are also induced to differentiate (N.T., unpublished data). Four days after the retroviral infection, fluorescence-activated cell sorter (FACS) analysis showed that over 70% of cells expressed GFP (data not shown); also albumin expression was significantly downregulated in GFP+ cells expressing NICD (arrows in Fig. 2B) compared with GFP+ cells in the control culture (Fig. 2A). Furthermore, 7 days after infection with pMX-NICD-GFP, albumin was expressed in GFP– cells but completely disappeared from GFP+ cells (Fig. 2D), whereas albumin was detected throughout the control culture (Fig. 2C). Seven days after infection, albumin mRNA was also significantly downregulated by NICD (Fig. 2E). In addition, NICD completely blocked the expression of TAT and CPS, whereas they were induced in the control culture (lane 1 in Fig. 2E). Thus, the Notch signal prevented the hepatic differentiation of hepatoblasts. However, except for the reduced expression of albumin, we did not find any characteristic features in the cultured cholangiocytes.
To dissect the Notch signaling, we infected Dlk+ cells with pMX-mutantNICD-GFP encoding the mutant NICD. This mutant contains two amino acid substitutions in the fourth ankyrin repeat and thereby fails to transmit through this domain the signal, e.g. the interaction with SKIP (Zhou et al., 2000), which is indispensable for the biological function of the NICD/RBPJκ complex (Kopan et al., 1994). Expression of the mutant NICD affected neither albumin expression nor hepatic differentiation (Fig. 2E, lane 3), indicating that the signal transmitted through the ankyrin repeat domain is essential for the inhibition of hepatic differentiation and reduction of albumin expression. Furthermore, we also infected Dlk+ cells with a virus encoding Hes1, one of the target transcription factors induced by Notch through RBPJκ. Ectopic expression of Hes1 partially downregulated albumin expression and significantly inhibited the expression of TAT and CPS (lane 2 in Fig. 2E), suggesting the possibility that NICD blocks hepatic differentiation and alters the characteristics of hepatoblasts, at least in part, through RBPJκ via Hes1.
In the primary culture system, hepatic maturation is further promoted by Matrigel, leading to the expression of tryptophan oxygenase and serine dehydrogenase (Kamiya et al., 2002). In this process, the interaction between the cell and extracellular matrix (ECM) suggested to play a key role in the induction of hepatic maturation. Because the interaction might also be important for the formation of bile ducts, we considered that Matrigel treatment confers cholangiocyte characteristics to NICD-expressing cells. After incubation of the cells with OSM for 5 days, Matrigel was overlaid with the cultured Dlk+ cell layer for an additional 5 days. Under such condition, some characteristic features of cholangiocytes appeared in the cells expressing NICD, i.e. upregulation of cholangiocyte marker genes such as CK7, CK19, HNF1β, and integrin β4 (Fig. 2F). However, even under this condition, we did not observe a morphological differentiation of cholangiocytes, such as the formation of tubule structures.
Downregulation of Notch signaling promotes hepatic differentiation
Because activation of the Notch pathway significantly inhibited hepatic differentiation, downregulation to the basal level of Notch signaling should promote hepatic differentiation. To test this possibility, we interfered with the Notch signaling by altering the Notch2 mRNA level and also the proteolysis of Notch.
To alter the Notch2 mRNA level, siRNA against Notch2 was introduced into Dlk+ cells. Four days after lipofection, the level of Notch2 mRNA was reduced to about 60% compared with the mRNA level in the control culture (Fig. 3A) and restored in the following 2 days (data not shown). By transiently downregulating Notch2 mRNA, expression of CPS was induced 6 days after lipofection, which suggests that Notch signaling suppresses hepatic differentiation (Fig. 3A).
Notch is cleaved in the membrane proximal region by presenilin to produce NICD, which then enters the nucleus and activates target genes. L-685,458, an aspartyl protease transition-state mimic, inhibits NICD production by blocking the γ-secretase activity of presenilin (Martys-Zage et al., 2000). Northern blotting showed that addition of L-685,458 slightly promoted the expression of TAT and CPS (Fig. 3B). Furthermore, L-685,458 induced the formation of cell clusters, in which cells exhibited the characteristics of differentiated hepatocytes: highly condensed cytosol and clear round-shape nuclei (Fig. 3C).
Notch controls the expression of liver-enriched transcription factors
Liver-enriched transcription factors such as the hepatocyte nuclear factor (HNF) and the CCAAT/enhancer binding protein (C/EBP) play essential roles in liver development and function. In adult liver, HNF1α, HNF4 and C/EBPα are strongly expressed in hepatocytes, whereas HNF1β and HNF6 are mainly expressed in cholangiocytes. To address the mechanism by which Notch signaling inhibits hepatic differentiation and controls the lineage commitment of hepatoblasts, we examined the expression of these five transcription factors. Because albumin expression was already downregulated 4 days after the viral infection, we investigated the effect of NICD on these transcription factors at this time point. Real-time quantitative RT-PCR showed that mRNA levels of HNF1α and HNF4 were reduced to about 70 and 40% of the control value, respectively (Fig. 4A and C), while the HNF1β mRNA level was elevated about two-fold (Fig. 4B). The reduction of HNF4 by NICD was also shown by immunocytochemistry; HNF4 expression disappeared from GFP+ cells 4 days after the infection with NICD-expressing virus (Fig. 4E, image 2), whereas it was detected throughout the control culture (Fig. 4E, image 1). Furthermore, northern blot analyses showed that the C/EBPα level in cells infected with NICD-expressing virus was about 60% of the control value (Fig. 4F). None of these effects were observed when expressing the mutant NICD. Thus, activation of the Notch pathway alters the expression of the above transcription factors via the signal through the ankyrin repeat domain of NICD, leading to cholangiocyte differentiation. Although NICD was also expected to increase the level of HNF6, an essential factor for bile duct development, it did so only slightly (Fig. 4D).
As described above, ectopic expression of Hes1 partially downregulates albumin expression and significantly inhibits hepatic differentiation. However, the effect of Hes1 on nuclear factors was quite different from that of NICD. Although Hes1 reduced the HNF4 level (Fig. 4C), it had no effect on HNF1β and C/EBPα mRNA levels (Fig. 4B and F). Furthermore, HNF1α expression was slightly upregulated by Hes1, whereas it rather was downregulated by NICD (Fig. 4A). Thus, both Hes1-dependent and -independent pathways might be necessary to change the characteristics of hepatoblasts.
The Notch signaling pathway controls many developmental processes through the direct contact between receptor-expressing cells and ligand-expressing cells, because both Notch receptors and their ligands are transmembrane proteins. In fetal liver, Jagged1 protein is strongly expressed in cells surrounding the portal veins, while Notch2 protein is expressed in most hepatic cells including hepatoblasts. Thus, hepatoblasts and Jagged1+ cells are in periportal zones in close contact with each other. Furthermore, ductal plates are formed at the boundary of the two types of cells. These results strongly suggest that Notch signaling is activated in hepatoblasts in the periportal area, leading to the differentiation of hepatoblasts into cholangiocytes. Consistent with this idea, NICD downregulates the expression of albumin, and the combination of NICD expression and Matrigel treatment results in primary cultures of Dlk+ hepatoblasts with increased expression of genes specific for cholangiocytes, such as the CK7, CK19, HNF1β and integrin β4. Furthermore, NICD strongly inhibits hepatic differentiation. Thus, the ectopic expression of NICD in hepatoblasts inhibits the differentiation leading to hepatocytes and induces characteristics of cholangiocytes. Consistently, hepatic differentiation was promoted by interfering with the Notch signaling pathway, i.e. downregulating the Notch2 gene expression with siRNA and using the presenilin inhibitor L-685,458. These results indicate that the basal level of Notch signaling activated in primary Dlk+ culture negatively regulates hepatic differentiation. Although strong Jagged1 expression was observed in cells surrounding the portal veins, it was also detected in isolated Dlk+ cells by RT-PCR analysis (data not shown). Because of such low-level expression of Jagged1, the Notch pathway might be weakly activated in hepatoblasts by the direct contact of the cells in primary culture.
Liver development as well as homeostasis is regulated by liver-enriched transcription factors. HNF4 (Parviz et al., 2003) and C/EBPα (Wang et al., 1995) are essential for the development of embryonic hepatocytes and HNF1α plays important roles in the functional maturation of hepatocytes in the postnatal period (Akiyama et al., 2000; Pontoglio et al., 1996), whereas HNF1β and HNF6 are required for the development of bile ducts (Clotman et al., 2002; Coffinier et al., 2002). Most of these transcription factors are expressed in hepatoblasts of fetal liver, but they show different expression patterns in adult liver; HNF1α, HNF4 and C/EBPα are strongly expressed in mature hepatocytes but barely detectable in cholangiocytes. By contrast, HNF1β and HNF6 are strongly expressed in cholangiocytes and HNF1β is weakly expressed in mature periportal hepatocytes (Coffinier et al., 2002). These expression profiles of HNFs and C/EBPα suggest that the expression of these factors is dramatically altered during the lineage determination of hepatoblasts. In fact, 4 days after infecting Dlk+ hepatoblasts with NICD-expressing virus, the albumin level was reduced and the expression of these nuclear factors was altered: HNF1α, HNF4 and C/EBPα were down- and HNF1β was upregulated. These changes coincide with the expression of those factors in adult cholangiocytes as described above. However, morphogenesis of the bile ducts was not induced by activation of the Notch pathway alone. Notch signaling might allow hepatoblasts to differentiate into cholangiocytes by regulating the level of the liver-enriched transcription factors, but it is not enough to induce the further differentiation of cholangiocytes to form tubule structures.
It was reported that intra-hepatic bile duct malformation occurs in HNF1β null mice, HNF6 null mice, or Jagged1 and Notch2 doubly heterozygous mice, liver-enriched transcription factors are important for bile duct development. A study on the regulation of the HNF1β promoter indicated that HNF6 controls the level of HNF1β (Clotman et al., 2002). Consistent with this finding, the HNF1β level was significantly reduced in the liver of HNF6 null mice at E12.5 and 14.5 (Clotman et al., 2002). However, the level was restored in the liver of HNF6 null mice at P3 and, moreover, not all livers of HNF6 null mice exhibited weak HNF1β expression (Clotman et al., 2002). Thus, it remains unclear whether or not HNF6 solely controls HNF1β expression. Whereas the precise linkage between Notch and HNF6 and/or HNF1β is still unknown, Notch signaling probably occurs upstream HNF6 because Jagged1 and Hes1 were expressed normally in the liver of HNF6 null mice (Clotman et al., 2002). In our experiments, NICD upregulated HNF1β transcription but not HNF6 RNA expression in cultured Dlk+ hepatoblasts. Taken together, Notch signaling might control HNF1β gene expression via another pathway, distinct from that regulated by HNF6.
Bile duct development can be divided into two sequential steps; the lineage commitment of hepatoblasts and morphological differentiation of cholangiocytes. Because ductal plate-like structures are present at E17.5 in livers of both HNF6–/– and HNF1β–/– mice, HNF6 and HNF1β are essential for the formation of intact bile duct structures but not for lineage commitment. By contrast, it is not known whether Notch2 and Jagged1 doubly heterozygous mice exhibit ductal plate-like structures during mid gestations. Based on the finding that NICD induced cholangiocytic characteristics in cultured Dlk+ hepatoblasts, we postulate that the Notch signaling is essential for the lineage commitment of hepatoblasts. Our results that, in hepatoblasts, NICD downregulates the expression of the genes HNF1α, HNF4 and C/EBPα, and upregulates the expression of the HNF1β gene suggest that an alteration in the expression of these transcription factors by the Notch signaling pathway is a necessary step for the initiation of cholangiocyte differentiation.
Neither hepatic differentiation nor the expression of transcription factors were affected by the NICD mutant, indicating that the signal via the ankyrin repeat domain of NICD is essential for controlling of hepatoblast differentiation. The Hes1 gene is a direct target of Notch signaling through RBPJκ, that associates with the ankyrin domain of NICD via SKIP (Zhou et al., 2000). However, the effect of Hes1 on the expression of HNF1α, HNF1β and C/EBPα was quite different from that of NICD, though its overexpression resulted in a significant inhibition of hepatic differentiation and reduction of HNF4. Furthermore, Hes1 knockout mice show the normal development of intra-hepatic bile ducts, whereas they exhibit abnormalities in the formation of extra-hepatic bile ducts (Sumazaki et al., 2004). These results indicate that Hes1 is not an essential mediator of the Notch pathway in the development of intra-hepatic bile ducts. In addition to Hes1, we detected the expression of HERP1, HERP2, and Deltex1 in Dlk+ hepatoblasts infected with pMX-NICD-GFP (data not shown). Thus, they might involve in the regulation of the expression of liver-enriched transcription factors by NICD.
In this study, we demonstrated that Notch signaling alters the network of liver-enriched transcription factors into a state, suitable for hepatoblasts to give rise to cholangiocytes. In vivo, this pathway might be activated only in the periportal area by cells strongly expressing Jagged1 in the portal veins. Furthermore, our results suggest that the lineage determination of hepatoblasts can be induced in vitro by activating or inhibiting the Notch signaling.
We are grateful to all members of the KAST laboratory of Stem Cell Regulation for technical assistance and helpful discussion. We thank Shigeru Chiba for valuable suggestions. We also thank Masato Nakafuku, Ryuichiro Kageyama, and Yukiko Gotoh for providing us with retroviral vectors. This work was supported in part by Grants-in-Aid for Scientific Research and the Special Coordination Fund for Promoting Science and Technology from the Ministry of Education, Culture, Sports, Science and Technology, a research grant from the Ministry of Health, Labour and Welfare, the Japan Government and CREST of Japan Science and Technology.
- Accepted February 19, 2004.
- © The Company of Biologists Limited 2004