Potential conflict of interest: Nothing to report.
Vascular endothelial cells (ECs) play a critical role in angiogenesis and organogenesis, especially in embryonic liver development. Hypoxia-inducible transcription factors (Hifs) are a key trigger of hypoxic signals, a primary stimulus of angiogenesis. The aryl hydrocarbon receptor nuclear translocator (Arnt), also called Hif-1β, serves as an obligate heterodimerization partner of Hif-1α and Hif-2α. Using Cre-Lox technology, the mouse Arnt gene was specifically disrupted in endothelial cells. The resulting mice, designated ArntΔEC, developed impaired hepatic vasculature, liver necrosis, and degenerative lesions in cardiac myocytes at the late embryonic stage (E16.5-E18.5), leading to approximately 90% neonatal lethality. Low serum glucose, downregulation of glucose transporter-1 and glucose-6-phosphatase mRNA, and hepatocyte proliferation were observed in ArntΔEC embryos. Magnetic resonance imaging on E16.5 embryonic livers revealed that ArntΔEC mice had a significant volume of avascular region. ArntΔEC mice that survived to the adult stage were fertile, showed normal behavioral activity, but had smaller livers with mild portal fibrosis, dilated blood vessels, abnormal collagen accumulation, and remarkable iron deposition. ArntΔEC mice had reduced adiposity, impaired serum lipid homeostasis, and a higher respiratory exchange ratio, indicating they utilized relatively more carbohydrates than their ArntF/F counterparts. In conclusion, endothelial Arnt plays a pivotal role in embryonic liver development. Adult ArntΔEC mice carrying embryonic hepatic defects developed what was possibly an early stage of cirrhosis with consequences of limited oxygen availability and altered lipid metabolism. (HEPATOLOGY 2006.)
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Angiogenesis, remodeling, and expansion of new blood vessels from preexisting vessels are responsible for vascular networking during later embryogenesis.1 Angiogenic factors secreted from surrounding tissues stimulate endothelial cells (ECs) to build new blood vessels to accommodate the oxygen, nutrient, and cellular waste exchange capacity. ECs communicate with neighboring parenchymal cells by promoting secretion of growth and survival factors. Vascular ECs differentiate to meet organ-specific homeostasis and thus are crucial for proper organ development, especially during embryonic liver development.2 The hepatic microvasculature consists of larger vessels such as portal and central venules and hepatic arterioles and sinusoids lined with sinusoidal endothelial cells (SECs). Failure of SEC function can result in hepatocellular carcinomas, cirrhosis, or hemorrhage.
Hypoxia is the main stimulus for angiogenesis, mediated by signaling through hypoxia-inducible transcription factors (Hifs). The Hifs are composed of α/β heterodimers, with the α and β subunits having at least two isoforms. The α subunits are hypoxia-inducible transcription factors, whereas the β subunits are constitutively expressed nuclear proteins. Although the α subunit heterodimerizes with only the β subunit, the Hif-1β, also called Arnt (aryl hydrocarbon receptor nuclear translocator), heterodimerizes with the AhR (aryl hydrocarbon receptor), Sim, and Per and can also form a homodimer.3, 4 Hif-1α-null embryos exhibit morphological defects from E8.0 to E8.5 and die by E11.5–7 Targeted disruption of the Arnt gene causes embryonic lethality between E9.5 and E10.5, mainly from abnormal vasculature development.8, 9 The embryonic expression profiles of Hif-1α, Hif-2α, Arnt/Hif-1β, Arnt2/Hif-2β, and AhR suggest they play unique and distinct roles in embryogenesis with limited redundancy.10, 11
As ECs play a key role in hypoxia-induced angiogenesis, embryonic lethality of Hif-1α or Arnt-null mice might be primarily caused by deficiency in EC function.9, 12–14 To investigate the role of endothelial Arnt during embryonic angiogenesis and pathophysiological alterations in adult mice, the Arnt gene was disrupted in ECs. Disruption of this subunit would be expected to inactivate both Hif-1α and Hif-2α because heterodimerization with Arnt is required for their transactivation activity.15
Mice with a targeted disruption of the Arnt (exon 6) gene were previously described,16 except the neomycin resistance gene cassette was excised using EIIA-Cre mice.17, 18 The Arntflox/flox mice, hereafter called ArntF/F, were further crossed with transgenic mice carrying the Tie2/Tek promoter-driven Cre recombinases, which is selectively expressed in the endothelial cell lineage (generously provided by Masashi Yanagisawa).19–21 Male mice homozygous for the Arnt floxed allele and hemizygous for Tie Cre (hereafter called ArntΔEC) were bred with ArntF/F female mice and the offspring genotyped at P21 by PCR for the Cre transgene.16 Embryos were collected from pregnant mice (counting noon of the day a vaginal plug was detected as 0.5 d.c.p.), and their genotypes were determined by PCR analysis of embryonic tail or yolk sac DNA. All mice were housed in a specific pathogen-free environment on a 12-hour light-dark cycle and fed NIH standard rodent chow (NIH-07) ad libitum. The National Cancer Institute Animal Care and Use Committee approved all protocols for mouse use.
Quantitative real-time PCR (qPCR) was carried out as described previously.22 Primers were designed for qPCR using Primer Express software (Applied Biosystems). Primer sequences are available on request. Values were quantified using the comparative CT method, and samples were normalized to 36B4 mRNA.
Tissues were fixed in 10% neutral phosphate-buffered formalin, embedded in paraffin, sectioned 4–6 μM, and stained with hematoxylin and eosin. Sections prepared from adult livers were stained with Masson's trichrome and Picro Sirius Red staining for collagen deposition and Prussian Blue to detect iron load. Pregnant females at the E17.5 and E19.5 gestational stages were injected intraperitoneally with 500 μL of 2.5 mg/mL BrdU (Sigma-Aldrich, St. Louis, MO) in phosphate-buffered saline (PBS); 2 hours later, the embryos were necropsied. Incorporated BrdU was detected using an ABC Mouse Vectastain Elite Kit (Vector Laboratories, Burlingame, CA) with a mouse monoclonal anti-BrdU antibody (Dako Cytomation, Carpinteria, CA). The BrdU labeling index was determined by counting the total number of positive hepatocytes per field. At least six fields per slide (at random high-power fields; magnification, 200×), from three to five embryonic livers were counted on each gestational day.
Vascular Imaging Using Magnetic Resonance.
Embryos at various gestational stages were harvested and their umbilical cords perfused with PBS followed by gelatin containing 25 mg/mL contrast media (Magnevist; Berlex Laboratory, Wayne, NJ). Embryos were then fixed in 10% neutral phosphate-buffered formalin overnight at 4°C. Three ArntF/F and six ArntΔEC embryonic livers were imaged. MR images were acquired with a Bruker Avance 7T/89 mm NMR spectrometer (Bruker Instruments, Inc., Billerica, MA) with a microimaging apparatus equipped with Micro 2.5 gradients and a 15-mm Birdcage RF coil. The T1-weighted images shown were acquired using a gradient-recalled echo, a spoiled FLASH sequence with the following parameters: TR/TE = 200/4.6 milliseconds, RF flip-angle = 30°, and BW = 100 kHz.23 The specimens were immersed in Fomblin, a perfluronated polyether oil (Solvay Solexis, Thorofare, NJ) prior to imaging to eliminate any magnetic susceptibility–induced distortions associated with the air–tissue interface. Image data matrix size was dependent on the orientation and size of a specific sample; an acquired image resolution of 50 μm isotropic was maintained.
Serum Glucose and Lipid Analysis.
Serum glucose levels were measured with a Glucometer Elite (Bayer, Mishawaka, IN) from blood withdrawn from the tail vein of adult mice, or from decapitated embryos. For lipid analysis, blood was withdrawn at 10:00 am from the retro-orbital sinuses of mice in a nonfasting state. Plasma free fatty acids (Wako Chemicals, Richmond, VA), total cholesterol, cholesterol esters, triglycerides, and phospholipids (Sigma Diagnostics, St. Louis, MO) were analyzed as previously described.2, 24 Serum insulin levels were measured by radioimmunoassay (Linco Research, St. Charles, MO).
Food Intake, Indirect Calorimetry, and Body Composition.
Daily food intake was measured in individually caged 12-month-old females. O2 consumption and CO2 production were measured at 30°C using the Oxymax system (Columbus Instruments, Columbus, OH) as previously described.25 Mice had free access to food and water during the measurement. Motor activities were measured by an infrared beam interruption (Opto-Varimex mini, Columbus Instruments, Columbus, OH) for 24 hours. Resting energy expenditure was calculated as the average of points with fewer than six ambulating beam brakes per minute. Body composition was measured in unanesthetized mice using an NMR analyzer Bruker minispec q10 (Bruker Optics Inc., Billerica, MA).
Isolation and Culture of Endothelial Cells.
Lungs from ArntF/F and ArntΔEC were isolated aseptically and rinsed in sterile PBS (KD Medical, Columbia, MD). The tissue was minced finely into 1 × 2 mm squares and digested in 20 mL of collagenase A (1 mg/mg; Sigma) at 37°C for 1 hour. Following digestion the cells were filtered through sterile 40-μm nylon mesh (BD Biosciences, San Jose, CA) centrifuged at 400g for 10 minutes. The cell pellet was resuspended in 4 mL of DMEM (Invitrogen) supplemented with 10% fetal bovine serum (FBS) and incubated with Dynabeads (Invitrogen) coated with anti-rat CD-31 antibody (BD Biosciences) for 30 minutes at 4°C with constant agitation. The cell-bound beads were washed five times with DMEM supplemented with 10% FBS and once with DMEM alone and cultured in bovine endothelial cell growth medium (Cell Application Inc.) supplemented with 10% FBS on collagen I (Sigma) coated 60-mm tissue culture dishes. For CD-31 staining, the cells were plated directly on chamber slides (BD Biosciences) coated with collagen I. CD-31 was detected using an ABC Mouse Vectastain Elite Kit (Vector Laboratories) with anti-rat CD-31 antibody.
Generation of Endothelial Cell–Specific Disruption of Arnt.
The Arnt gene was disrupted in endothelial cells by mating ArntF/F mice, in which exon 6 of the mouse Arnt gene was flanked with loxP sites18 with mice carrying the Cre recombinase gene driven by the endothelial cell-specific Tie2/Tek promoter.19–21 To assess endothelial-specific recombination of the Arnt allele, lung endothelial cells were immunopurified using CD-31 antibody. The endothelial cells were determined to be 90%–95% pure as assessed by CD-31 immunostaining, whereas no staining was detected in nonendothelial cells (Fig. 1A). PCR analysis was used to estimate the extent of recombination of the floxed Arnt allele. The recombined allele, amplified as a 340-base-pair (bp) product, was detected in whole lung tissue from ArntΔEC mice and was absent from lung DNA isolated from ArntF/F mice (Fig. 1B). The intact floxed Arnt allele, amplified as a 300-bp product, was detected in lung DNA from ArntΔEC mice and was the only band evident in lung DNA from ArntF/F mice (Fig. 1B). The null allele was the only band detected in endothelial cells isolated from ArntΔEC mice; in contrast no recombination was detected in endothelial cells isolated from ArntF/F mice and from lung cells that did not bind CD-31 antibody (nonendothelial cells; Fig. 1C). RT-PCR analysis demonstrated that the level of Arnt mRNA was significantly lower in endothelial cells isolated from ArntΔEC than in those isolated from ArntF/F mice (Fig. 1D).
Disruption of Arnt in Endothelial Cells Caused Partial Embryonic Lethality.
Only 13% of weanlings at P21 carried the Cre transgene and thus had a disrupted Arnt gene in ECs (Table 1). More than 1,300 mice were genotyped, and only 203 were identified as ArntΔEC. There was no gender preference in survival of ArntΔEC and ArntF/F mice; among the ArntΔEC mice that survived at P21, 46% were female and 54% were male. To determine the critical developmental stage of embryonic lethality in the ArntΔEC mice, neonates and live embryos at various developmental stages were genotyped. The ratios of ArntΔEC mice to total live animals at E16.5 and E18.5 were 57% and 54%, respectively. The ratio of ArntΔEC neonates at P2 was approximately 10%, similar to that obtained from P21 animals, indicating that most of the ArntΔEC embryos were dying during the prenatal stage.
Live ArntΔEC embryos and their yolk sacs at E16.5 and E18.5 were visually indistinguishable from littermate ArntF/F animals. Dead ArntΔEC embryos were easily identified by no blood in the umbilical cords or placentas at E16.5 or later developmental stages with significant edema (Fig. 2A-B). At E16.5, ArntΔEC embryos were smaller than littermate controls, although the difference was not significant (Fig. 2C). The difference in body weight gradually increased, and at P15, ArntΔEC mice were 12% lower in body weight than ArntF/F mice (P < .001; Fig. 2D).
E16.5 ArntΔEC Embryos Showed Impaired Energy Homeostasis.
The E16.5 ArntΔEC embryos exhibited severe hypoglycemia, with blood glucose levels as low as 31 ± 2 mg/dL compared to 115 ± 20 mg/dL in controls (P < .01; Fig. 2E). Consistent with low blood glucose levels, downregulation of genes involved in gluconeogenesis was observed in embryonic liver, including glucose-6-phosphase (G-6-P) and GLUT-1, but no significant differences between genotypes in expression of GLUT-2 or in mRNA of hepatocyte-enriched transcription factors HNF4α and C/ebpα was observed (Fig. 2F).
Arnt in Endothelial Cells Is Essential for Normal Embryonic Liver Development.
To determine the cause of embryonic lethality of the ArntΔEC mice, the major organs were examined at various embryonic stages. Embryonic livers lacking Arnt in the ECs showed focal areas of necrosis, especially at the edges of each liver lobe (Fig. 3A). Lesions with hemorrhage were observed as early as stage E16.5 (Fig. 3B). Frequently, lesions found at E16.5 were small and red to darker colored, whereas the E18.5 livers included white lesions, which were further identified as necrosis (inset). ArntΔEC mice also developed focal areas of heart hemorrhage and myocardial degeneration at stage E18.5 (Fig. 3C). In vivo labeling with 5-bromodeoxyuridine (BrdU) revealed that cell proliferation in ArntΔEC mice was significantly lower than in ArntF/F mice in both E17.5 and E19.5 embryonic livers (Fig. 3D). Hepatic BrdU incorporation at E17.5 was 344 ± 88 (per field, n = 9) in ArntΔEC and 739 ± 90 (per field, n = 12) in ArntF/F (P < .0001). Similar results were observed in E19.5 embryonic livers. In contrast, TUNEL assays showed no significantly increased apoptotic cell death in the embryonic livers from ArntΔEC mice (data not shown).
MR Imaging Showed Abnormal Hepatic Vascular Development in ArntΔEC Embryos.
To obtain a more global view of embryonic vascularization, magnetic resonance imaging (MRI) was used. E16.5 embryos were perfused with the contrast media, Magnevist, via the umbilical cord, and whole embryos of both genotypes were imaged. Excised livers were imaged at a higher resolution to identify the hepatic defect. In contrast to ArntF/F mice, which developed an organized and evenly distributed vascular network throughout the entire liver, littermate ArntΔEC embryos had extremely poor vasculature. Several central veins with dilated and serpiginous channels and areas without microvessels were identified in the ArntΔEC mice. Overall, the ArntΔEC embryonic livers did not develop a normal vascular network (Fig. 4A-B). MRI analysis of six ArntΔEC embryos revealed that the degree of abnormality varied from moderate to severe, which is a possible explanation of why about 13% of ArntΔEC mice survived to the adult stage. Each slice of the MR image obtained from embryonic livers was recalculated to a volumetric unit to determine the vascularized volume to total volume of livers. The average avascular volume was 9.8% of the ArntF/F livers and 29.6% of the ArntΔEC livers (Fig. 4C; movies can be viewed at http://ccr.cancer.gov/staff/links.asp?profileid=5801).
Adult ArntΔEC Mice Had Reduced Body Weight.
Approximately 13% of litters carried the Cre transgene and successfully reached weaning age. ArntΔEC mice and their gender- and body-weight-matched littermates were housed together for further study. ArntΔEC mice were fertile and showed normal behavioral activity and survival rates until 12 months of age. Especially noteworthy was that female ArntΔEC mice were gaining less weight and, starting from age 16 weeks, were significantly smaller than ArntF/F littermates (Fig. 5A–B). At the age of 12 months, the ArntΔEC mice had livers that were 32% smaller and fat mass 48% lower than in controls, whereas lean mass was reduced only by 7% (Fig. 5B–C). ArntΔEC mice also showed decreased free fluid content (Fig. 5C). When organ weight was expressed as a percentage of body weight, only reduction in fat mass remained significantly different between genotypes. These results indicated that ArntΔEC mice were lower in body weight mainly because of a reduction in fat mass.
ArntΔEC Mice Developed Liver Lesions.
Adult ArntΔEC mice developed small livers starting from the age of 5 weeks. At the age of 12 months, ArntΔEC mice exhibited hepatic portal arteriolar hyperplasia, hypertrophy, and minimal to mild portal fibrosis (Fig. 6A). H&E staining of liver sections from ArntΔEC mice showed an increased numbers of bile ducts as well as arterioles in the portal areas. Some portal areas had mild oval cell and bile duct hyperplasia (Fig. 6B). ArntΔEC mice showed significant amounts of portal collagen deposition, whereas ArntF/F mice did not show any obvious collagen staining (Fig. 6C–D). ArntΔEC livers also demonstrated significant amounts of iron deposition as determined with Prussian blue staining (Fig. 6E). Interestingly, ArntΔEC mice did not show elevated serum AST and ALT levels, which indicates no significant liver toxicity (data not shown).
ArntΔEC Mice Had Increased Metabolic Rate and Favored Carbohydrate over Fatty Acid Oxidation.
To study the mechanism of reduced adiposity in the ArntΔEC mice, food intake, metabolic rate, and activity were measured (Fig. 7). Daily food intake and motor activity were similar in both the ArntΔEC and the ArntF/F groups (Fig. 7A–B). In contrast, ArntΔEC mice showed significantly higher resting and total oxygen consumption at a thermoneutral temperature (30°C), suggesting an increased basal metabolic rate (Fig. 7C). ArntΔEC mice also exhibited higher resting and total respiratory exchange ratios (RERs). The RER is the ratio of CO2 produced to O2 consumed (Fig. 7D), with oxidation of carbohydrates producing an RER of 1.00, and fatty acid oxidation resulting in an RER of approximately 0.700. These results demonstrate that the ArntΔEC mice were hypermetabolic and used relatively more carbohydrates as an energy source than did control mice.
ArntΔEC Mice Developed Impaired Serum Lipid Homeostasis.
To investigate the influence of liver vasculature and the lower body and fat mass of ArntΔEC mice on lipid homeostasis, serum chemistry was examined. Overall, the serum lipid level of ArntΔEC mice was significantly lower than that of ArntF/F mice (Table 2). The serum cholesterol level of ArntΔEC mice was also significantly lower than that of ArntF/F mice. The most significant difference between ArntF/F and ArntΔEC mice was in serum triglyceride levels, which were 178 ± 17 and 80 ± 5 mg/dL, respectively (P = 3.79 × 10−5). Serum phospholipids levels was also lower in ArntΔEC mice than in ArntF/F mice. Glucose and insulin levels after 12 hours of fasting were not statistically different between groups. There was no observable difference in bile acid, direct bilirubin, and total bilirubin levels between ArntF/F and ArntΔEC mice (data not shown).
P values were calculated by the two-tailed Student t test.
Body weight (g)
40.7 ± 1.7
31.9 ± 1.5
Fat weight (g)
15.7 ± 1.5
8.2 ± 1.0
Total cholesterol (mg/dL)
114 ± 11
71 ± 5
178 ± 17
80 ± 5
3.79 × 10−5
182 ± 15
127 ± 6
Free cholesterol (mg/dL)
15 ± 3
7 ± 1
Cholesterol esters (mg/dL)
101 ± 11
64 ± 4
Disruption of Arnt in ECs would be expected to inactivate Hif-1α and Hif-2α as well as AhR. Onset of developmental expression of Tie2/Tek starts at E8.5, reaching a maximum at E9.5, at which time Arnt expression in ECs should be extinguished and any resultant phenotypes would emerge.19 Indeed, EC-specific disruption of Arnt caused partial embryonic lethality; approximately 90% of the embryos died before the postnatal stage with evidence of liver necrosis and heart hemorrhage. Interestingly, expression of Arnt in ECs does not appear to be required for blood vessel formation or for the development and normal functioning of most tissues except perhaps the liver and heart. Thus, Arnt may have a principle role in angiogenesis, primarily during embryonic liver development. MR imaging revealed that ArntΔEC embryos developed extremely poor vasculature with dilated and serpiginous channels. The avascular volume of ArntΔEC livers was 29.6% compared to the avascular volume of ArntF/F livers of 9.8%. Because about 20%–25% of oxygen exchange occurs in hepatocytes via serpiginous channels, the avascular regions observed in ArntΔEC embryonic livers probably resulted in poor oxygen, nutrient, and waste exchange. The hypoglycemia observed in ArntΔEC embryos was likely a consequence of hepatic defects caused by irreversible damage to the dynamics of embryonic development and/or through direct effects mediated by Hif-1α target genes such as Glut-1.26, 27 Alternatively, profound hypoglycemia could lead to perfusion defects in various organ systems that could trigger chronic hypoxia within the organ, thus leading to necrosis. BrdU staining and TUNEL assays revealed that the abnormal hepatic development and smaller livers were a consequence of impaired hepatocyte proliferation rather than accelerated apoptosis. The hypoglycemia from liver failure, possibly as a result of a defect in the hepatic vasculature in the ArntΔEC embryo, is probably one of the major causes of the defect in systemic embryonic development that results in small embryos. Although the ArntΔEC mice also exhibited severe heart hemorrhage, it is unclear whether this was a result of liver failure or a direct cause of embryonic mortality.
Partial postnatal lethality and abnormal vasculature of ArntΔEC mice were distinct from the phenotypes observed in the Ahr-null mice.28, 29 The Ahr-null mice developed portosystemic shunting, an abnormal circulatory transition from fetus to newborn, and thus manifested abnormal vasculature after birth.30, 31 However, a recent study revealed that the Ahr-null mice exhibited fetal hepatic necrosis due in part to abnormal perfusion of the developing liver.32 The necrosis was near veins/arteries, whereas in the ArntΔEC mice, necrosis was in distant areas of the liver associated with microvessels. Several other phenotypic differences between the two models suggests that Arnt disruption in ECs is not a direct result of inactivation of Ahr signaling. Partial embryonic lethality in the ArntΔEC mouse line was also distinct from the embryonic lethality of Hif-1α5–7 and Arnt-null8, 9 mice. Most importantly, ArntΔEC embryos did not demonstrate any obvious placental defect commonly observed in both Hif-1α and Arnt-null embryos. Embryonic defects from abnormal placental development frequently cause embryonic lethality in relatively earlier developmental stages, evidence in support of the contention that abnormal development of ArntΔEC mice may not be a direct result of placental defects. Interestingly, embryonic lethality was not reported in a mouse line lacking the Hif-1α gene in ECs,33 suggesting that the ArntΔEC embryonic defect is not solely a result of Hif-1α signal transduction. The embryonic liver defect may be the result of inactivation of both Hif-1α and Hif-2α and perhaps even AhR.
The few mice that survived to the adult stage gained less body weight mainly because of a reduction in fat mass. Adult ArntΔEC mice developed hepatic portal arteriolar hyperplasia, hypertrophy, and portal fibrosis with a significant amount of collagen and iron deposition. Collagenization of the space of Disse in liver, a hallmark of cirrhosis, also limited hepatic function. The small and fibrotic livers with collagen deposition were somewhat similar to those found in the Ahr-null mice.28, 29 However, the iron deposition found in the ArntΔEC livers was not detected in Ahr-null mice. Although ArntΔEC mice exhibited hepatic vascular defects from the late embryonic stage to adulthood, the livers of the Ahr-null mice at 4 weeks did not show any histological abnormalities. Thus, even with the ArntΔEC and Ahr-null mice having similar smaller livers with collagen deposition, the causation of the hepatic alteration of these two mouse models may differ.
Compared with the ArntF/F mice, the ArntΔEC mice tended to prefer carbohydrates as an energy source. This might have been caused by abnormal lipid metabolism, possible because of the low oxygen availability as a result of the fibrotic liver and its altered vascular network. Lower serum cholesterol levels, including free cholesterol and cholesterol esters, are also indications of abnormal hepatic function. The ArntΔEC livers either did not synthesize cholesterols to the same extent as the ArntF/F livers or had an increased capacity for bile acid synthesis.
In summary, studies using ArntΔEC mice revealed that Arnt in ECs is critical for embryonic vascular development of the liver. The precise mechanism of Arnt function during embryogenesis remains to be determined.