Endothelial Notch signaling is essential to prevent hepatic vascular malformations in mice

Abstract

Liver vasculature is crucial for adequate hepatic functions. Global deletion of Notch signaling in mice results in liver vascular pathologies. However, whether Notch in endothelium is essential for hepatic vascular structure and function remains unknown. To uncover the function of endothelial Notch in the liver, we deleted Rbpj, a transcription factor mediating all canonical Notch signaling, or Notch1, specifically from the endothelium of postnatal mice. We investigated the hepatic vascular defects in these mutants. The liver was severely affected within two weeks following endothelial deletion of Rbpj from birth. Two-week old mutant mice had enlarged vessels on the liver surface, abnormal vascular architecture, and dilated sinusoids. Vascular casting and fluorosphere passage experiments indicated the presence of porto-systemic shunts. These mutant mice presented severely necrotic liver parenchyma and significantly larger hypoxic areas, likely resulting from vascular shunts. We also found elevated levels of VEGF receptor 3 together with reduced levels of ephrin-B2, suggesting a possible contribution of these factors to the generation of hepatic vascular abnormalities. Deletion of Rbpj from the adult endothelium also led to dilated sinusoids, vascular shunts, and necrosis albeit milder than that in mice with deletion from birth. Similar to deletion of Rbpj, loss of endothelial Notch1 from birth led to similar hepatic vascular malformations within two weeks.

Conclusions

Endothelial Notch signaling is essential for the development and maintenance of proper hepatic vascular architecture and function. Our findings may help understand the molecular pathogenesis of hepatic vascular malformation and the safety of therapeutics inhibiting Notch.

The liver has a wide range of functions key for organismal homeostasis, including blood detoxification and lipid metabolism, and the presence of a proper vasculature is critical for its performance. Most vascular diseases of the liver are rare; however, without timely diagnosis, they can lead to significant morbidity and mortality(1). Hepatic vascular malformations (HVMs), characterized by the abnormal shunting of liver blood vessels, are such lesions. In patients, while HVMs can be idiopathic or acquired as a consequence of underlying cirrhosis, liver injury or hepatocellular carcinoma, most cases are congenital, as in hereditary hemorrhagic telangiectasia (HHT)(2). The incidence of HVM detection is rising, in part due to increased use of ultrasonography or other liver imaging assays in patients.

The liver receives blood from two sources (the hepatic artery and the portal vein), increasing the classes of vascular shunts that can develop. Shunts may be: 1) arteriovenous shunts linking the hepatic artery and the hepatic vein, 2) porto-systemic shunts between the portal vein and the hepatic vein, and 3) arterioportal shunts connecting the hepatic artery and the portal vein. Any or all of these shunt classes can form in a given liver(1). Clinical manifestation of HVMs is affected by the type and degree of blood shunting, and symptoms can include high-output heart failure, portal hypertension, hepatic encephalopathy, and biliary ischemia(1). A better understanding of the molecular pathways that govern the function of liver vasculature will aid in the diagnosis and treatment of these liver vascular pathologies.

The highly conserved Notch signaling pathway is involved in many developmental and pathological processes, including the establishment of arterial EC identity and in the regulation of the angiogenic response(3),(4). Upon ligand binding, the Notch receptor undergoes a cascade of proteolytic cleavages that results in the release of its intracellular domain (ICD). The Notch ICD translocates to the nucleus, where it binds to the transcription factor Rbpj (recombination signal binding protein for immunoglobulin kappa J region), and activates the transcription of target genes. Though mammals have four Notch receptors (Notch1-4), endothelial cells (ECs) predominantly express Notch1 and Notch4(4). Mice lacking Notch4 do not develop an overt phenotype, while mice deficient for Notch1 present arrested growth and die at embryonic stages with severe vascular anomalies(5). Hence, Notch1 is considered to be the main Notch receptor in the endothelium; however, deletion of both Notch1 and Notch4 results in a more pronounced phenotype than that of Notch1 alone(5), suggesting synergistic functions of the receptors.

The Notch signaling pathway has been identified as a regulator of liver vasculature maintenance. Mice with deletion of Notch1 develop hepatic vascular tumors(6, 7), while deletion of Rbpj in mice results in venooclusive disease(8). These loss of function studies were performed via non cell-specific ablation of Notch signaling. To elucidate the role of Notch signaling specifically in the endothelium, we have used mice with endothelial deletion of Rbpj (to delete signaling via both endothelial Notch receptors) or Notch1 at different growth stages. Our results indicate a critical role for endothelial Notch signaling in the maintenance of liver sinusoid architecture and functionality and in the prevention of HVMs.

MATERIALS & METHODS

Mouse Experiments

Animal experiments were performed in compliance with the University of California, San Francisco (UCSF) Institutional Animal Care and Use Committee (IACUC) guidelines under animal protocols AN085404 and AN102764. All animals received humane care according to the criteria outlined in the “Guide for the Care and Use of Laboratory Animals” prepared by the National Academy of Sciences and published by the National Institutes of Health (NIH publication 86-23 revised 1985).

Rbpj^flox/flox(9), Cdh5(PAC)-CreERT2(10), Rosa26^mT/mG (mT/mG)(11), Notch1^flox/flox(12), and ephrin-B2^tau-lacZ(13) mice were obtained from T. Honjo (Kyoto University, Japan), R. Adams (Max Planck Institute for Molecular Biomedicine, Münster, Germany), L. Luo (Stanford University, California, USA), F. Radtke (École Polytechnique Fédéral de Lausanne, Switzerland), and D. Anderson (California Institute of Technology, USA), respectively.

Detailed methods for Cre activation, histology, immunostaining, βgal detection, two-photon microscopy, vascular shunting assay, portal pressure measurement, liver panel markers, LSEC enrichment, real-time PCR, time-lapse imaging, and statistical analysis are included in the Supporting Information.

RESULTS

Vascular shunts develop in postnatal Rbpj^iΔEC mice

To interrogate the role of Notch signaling in the liver endothelium, we deleted Rbpj, which encodes the Rbpj transcription factor downstream of Notch and essential for canonical Notch signaling(3). Endothelial deletion of Rbpj was mediated by tamoxifen-inducible Cre-recombinase activity driven by the vascular endothelial (VE)-Cadherin (Cadherin5) promoter in Rbpj^flox/flox; Cdh5(PAC)-CreERT2 mice (Rbpj^iΔEC). Inclusion of the mT/mG reporter(11) allowed us to monitor Cre-recombinase activity and to visualize ECs. Tamoxifen injection of Rbpj^iΔEC mice at postnatal day (P)1 and P2 resulted in absence of Rbpj specifically in the ECs of the liver (Fig.S1a,b,c). mGFP expression in Rbpj^iΔEC mice bearing mT/mG was observed throughout the liver endothelium at P14 (Fig. S1d) and throughout the vessels of the retina, where we observed a dramatic increase in retinal angiogenesis at P7 (Fig.S1e,f), consistent with previous reports(14, 15). Taken together, these results indicate specific, efficient endothelial deletion of Rbpj in our mouse model, including the liver vessels.

Analysis of P14 Rbpj^iΔEC pups revealed a dramatically affected liver. Gross examination revealed enlarged vessels on the liver surface, particularly along lobe edges. Importantly, large swathes/regions of the hepatic parenchyma were discolored (Fig. 1a,b). Liver sections revealed the presence of dilated sinusoids in Rbpj^iΔEC pups (Fig.1c,d and Fig.S2). Sinusoid dilation, most prominent around the central venules, was significantly increased in Rbpj^iΔEC mice, when compared with heterozygous and wild type control littermates (Fig.1e). Perturbations to normal sinusoid morphology were severe enough to interrupt normal hepatic plate architecture. Moreover, enlarged sinusoids were observed connecting the portal and the central venules directly (Fig.1c,d), suggesting the formation of vascular shunts. To gain a deeper understanding of these vessel irregularities, we performed casting of the liver vasculature. When casting resin was injected in the left ventricle of controls, the arterial and portal vascular branches were filled (Fig.2a). The density of the resin prevented penetration of the sinusoids of healthy livers. However, when Rbpj^iΔEC mice were injected with the resin, both the portal branches and the central venules were filled, indicating abnormal shunting between these two vascular systems (Fig.2b).

Figure 1. Endothelial specific deletion of Rbpj at birth led to dilated sinusoids.

a,b, Liver from heterozygous control (a) and Rbpj^iΔEC mutants (b) at postnatal (P) day 14 (n=47). Scale=1mm. c,d, Endothelial cell (EC)-mGFP expression of the mT/mG reporter in liver sections from control (c) and Rbpj^iΔEC (d) mice. DAPI stains cell nuclei. Arrowheads indicate enlarged connections between the portal and the central venules. Scale=100μm (n=6) e, Sinusoid width was significantly increased in the central vein area of Rbpj^iΔEC. Data show mean ± s.e.m. (n=3-4 mice); ***p<0.001. p.v. portal venule, cv: central venule, EC: Endothelial Cell.

Figure 2. Endothelial specific deletion of Rbpj at birth resulted in liver vascular shunts.

a,b, Vascular casting of control mice (a). Direct shunts were visible between the portal and central venules (arrowhead) in Rbpj^iΔEC mutants (b) (n=11). c,d, Top, FITC-fluorescent microspheres in the liver and lungs of control (c) and Rbpj^iΔEC (d) mice at P14. Below, bright field (BF) images indicate tissue orientation (n=4).

To further demonstrate the presence of vascular shunts, we analyzed the circulation pattern of size-limiting (15μm) microspheres. FITC-conjugated beads, injected into the portal vein of P14 control mice, were arrested in the liver (Fig.2c). In Rbpj^iΔEC mice, the beads bypassed liver sinusoids and lodged in the lungs (Fig.2d), indicating the presence of porto-systemic vascular shunts. Taken together, these data suggest that Rbpj^iΔEC mice develop vascular abnormalities that lead to the formation of vascular shunts.

Endothelial loss of Rbpj at birth results in liver necrosis, malfunction and hypoxia by P14

Histological analysis of liver, hematoxylin and eosin (H&E) sections, revealed dilated sinusoids and expanded subcapsular sinuses, consistent with the enlarged vessels seen near the liver surface. Additionally, a variable degree of hepatocyte damage, ranking from atrophy to severe panacinar necrosis, was observed in severe cases, forming an arc following the zone typical of ischemic injury in the milder injuries (Fig.3a,b and Fig.S3a,b).

Figure 3. Immature Rbpj^iΔEC mice showed increased hypoxia, poor perfusion and liver necrosis, but unaltered EC proliferation.

a,b, Representative H&E images of liver sections from control (a) and Rbpj^iΔEC (b) mice at P14. Arrowheads indicate enlarged subcapsular vessels connected to the central venules. (n=6). Scale=50μm. c,d, In contrast to controls (c), liver sections of Rbpj^iΔEC mice (d) presented numerous hypoxic and poorly lectin-perfused areas. Scale=200μm. e, Hypoxyprobe⁺ area was significantly increased in Rbpj^iΔEC mice. Data presented are mean ± s.e.m (n=3) **p<0.01. f, The number of pHH3⁺ ECs per HPF (high-power field) was not significantly changed in Rbpj^iΔEC. Data presented are mean ± s.e.m (n=7).

The porto-systemic shunting and abnormal vessel structure in Rbpj^iΔEC mice led us to hypothesize that liver necrosis might result from poor tissue oxygenation. To test this, P14 Rbpj^iΔEC pups were injected with pimonidazole hydrochloride (Hypoxyprobe-1), which binds to hypoxic tissue. To visualize the vascular perfusion of the liver, mice were co-injected with FITC conjugated Lycopersicon esculentum lectin. Immunostaining against Hypoxyprobe-1 showed significantly increased hypoxic area in Rbpj^iΔEC mice when compared to the heterozygous control (Fig.3c,d,e). Of note, hypoxic areas in Rbpj^iΔEC mice corresponded to regions with reduced lectin perfusion.

Because of the high levels of necrosis and hypoxia observed, we hypothesized that liver function was impaired in these mice. Rbpj^iΔEC mice presented higher serum levels of bilirubin, alanine transaminase (ALT), aspartate transaminase (AST), and alkaline phosphatase, as well as lower levels of total protein and albumin (Fig.S4), suggesting that Rbpj^iΔEC mice have a poorly perfused liver vasculature that results in hypoxia, liver damage and malfunction.

Because previous work using global Notch1 or Rbpj deletion reported increased EC proliferation, we investigated the livers of Rbpj^iΔEC mice at P7, before the onset of severe necrosis. Phospho-histone-H3 immunostaining appeared increased in Rbpj^iΔEC liver ECs; however, this difference was not statistically significant (Fig.3f, Fig.S5), suggesting that increased endothelial proliferation does not contribute to the vascular phenotypes observed in P14 Rbpj^iΔEC mice.

Endothelial deletion of Rbpj in adult mice also induces vascular shunts

To determine whether the vascular defects following loss of endothelial Rbpj were age dependent, we deleted Rbpj by injecting tamoxifen in 6 week-old mice and analyzed the liver vasculature 6 weeks later (adult Rbpj^iΔEC). We found Rbpj efficiently and specifically deleted in the adult liver endothelium at this time point (Fig.S6a,b). Similarly, mGFP expression derived from the mT/mG reporter was observed throughout the adult liver vasculature (Fig.S6d,e).

Adult Rbpj^iΔEC mice presented enlarged vessels on the liver surface (Fig.4a); however, this phenotype was not as severe as that observed following endothelial Rbpj deletion from birth. Similarly, adult Rbpj^iΔEC showed a modest discoloration of the liver parenchyma, while the discoloration following endothelial Rbpj deletion from birth was more pronounced. Consistently, histological analysis of the liver from adult Rbpj^iΔEC mice revealed less robust sinusoid dilation (Fig.4b). Hepatocyte atrophy was observed following the same arc pattern present in the necrotic hepatocytes of mice with loss of endothelial Rbpj form birth, indicative of a milder ischemic insult in the adult mice. Also, wide plates of hepatocytes with a nodular regenerative hyperplasia (NRH)-like appearance were noticed (Fig.4b and Fig.S7).

Figure 4. Adult Rbpj^iΔEC mice presented dilated sinusoids.

a, Liver from adult control (left panel) and Rbpj^iΔEC (right panel) mice, showing enlarged vessels in mutants. (n=24). Scale=1mm. b, H&E staining of adult control (left panel) and Rbpj^iΔEC (right panel) mice (n=5). Scale=50μm. c, EC-mGFP expression from the mT/mG reporter in liver sections. In contrast to controls (left panel), there were enlarged sinusoids between the portal and the central venules in adult Rbpj^iΔEC (right panel) mice. Scale=100μm. DAPI stains cell nuclei. pv: portal venule, cv: central venule, EC: Endothelial Cell. d, Sinusoid width was significantly increased in the area surrounding the central vein of adult Rbpj^iΔEC when compared to control littermates. Data show mean ± s.e.m. (n=3 mice); ***p<0.001.

Further analysis also showed an abnormal vascular architecture (Fig.4c). Quantification of sinusoid width around central venules demonstrated that sinusoids significantly wider than heterozygous controls (Fig.4d), consistent with that observed following Rbpj deletion at birth.

To determine whether deletion at 6 weeks also led to vascular shunting, we injected fluorescent microspheres in the portal vein. Microspheres were lodged in the liver of control mice, while no beads were observed in the lungs (Fig.5a). In adult Rbpj^iΔEC mice, most of the microspheres passed through the liver vasculature and were detected in the lungs, demonstrating direct porto-systemic shunting (Fig.5b). Altogether, these results indicate that while the liver phenotype is less severe than deletion at birth, endothelial deletion of Rbpj at 6 weeks also results in direct porto-systemic shunting.

Figure 5. Adult Rbpj^iΔEC mice developed liver vascular shunts and portal hypertension.

a,b, Top, fluorescent microspheres in control mice (a) were retained in the liver, while in adult Rbpj^iΔEC mice (b), arrested in the lung. Below, bright field (BF) images indicate tissue orientation. (n=3). c, Portal pressure was significantly increased in Rbpj^iΔEC mice. Data show mean ± s.e.m. (n=4); *p<0.05. d, Spleen/body weight ratio was significantly increased in adult Rbpj^iΔEC mice compared to wild-type and heterozygous controls. Data show mean ± s.e.m. (n=4-6); ** p<0.01.

Endothelial deletion of Rbpj in adult mice leads to increased portal pressure

The observation that mice with adult deletion of Rbpj develop a NRH-like phenotype led us to consider the possibility that these mice also present increased portal pressure. Direct measurement of portal pressure showed significantly increased values in Rbpj^iΔEC mice when compared to control (Fig.5c). Similarly, a significantly increased spleen/body weight ratio, an indirect surrogate of elevated portal pressure was observed in Rbpj^iΔEC mutants (Fig.5d). In the absence of observed liver fibrosis on histology, these results are suggestive of presinusoidal portal hypertension, as commonly encountered in nodular regenerative hyperplasia, in Rbpj^iΔEC mice. We cannot rule out, however, that hemodynamic changes resulting from other abnormalities in Rbpj^iΔEC mice(16) contribute to the nodular phenotype, and the elevated portal pressure observed in these mice.

Endothelial deletion of Notch1 at birth resembles that of Rbpj

Since Rbpj mediates canonical Notch signaling, we have used it as a proxy to interrogate the role of Notch signaling in the liver vasculature; however, to confirm that the phenotypes we observed following Rbpj deletion reflect loss of canonical Notch signaling, we investigated the role of the Notch1 in the liver vasculature. Notch1^flox/flox; Cdh5(PAC)-CreERT2 (Notch1^iΔEC) mice were injected with tamoxifen at birth, and the liver vasculature was analyzed at P14. Successful and efficient Notch1 deletion was verified (Fig.S8).

Similar to that observed in P14 Rbpj^iΔEC, Notch1^iΔEC mice had enlarged vessels on the liver surface as well as hepatic discoloration, though these phenotypes were milder than those observed in Rbpj^iΔEC mice (Fig.6a,b). When the vascular organization was analyzed, liver sinusoids appeared dilated around the central vein area. As observed in Rbpj^iΔEC mice, the architecture and disposition of these vessels was severely altered, suggesting direct shunting between the portal and central veins (Fig.6c,d).

Figure 6. Endothelial deletion of Notch1 in immature mice resulted in liver vascular shunts.

a,b, Liver from control (a) and Notch1^iΔEC mutant mice (a) at P14. Scale=1mm. (n=15) c,d, Compared to controls (c), EC-mGFP expression from the mT/mG reporter in Notch1^iΔEC mutant (d) liver sections showed enlarged and abnormal vessels. DAPI stains cell nuclei. Scale=100μm e,f Top, fluorescent microspheres injected in controls (e), and Notch1^iΔEC mutants (f). Below, bright field (BF) images indicate tissue orientation. (n=3). pv: portal venule, cv: central venule. EC: Endothelial Cell.

To determine whether Notch1^iΔEC mice also develop porto-systemic shunts, we injected fluorescent microspheres into the portal vein. Lungs from control mice lacked microspheres, which were retained in the liver sinusoids (Fig.6e). In contrast, the lungs of Notch1^iΔEC mice contained microspheres (Fig.6f), indicating the presence of vascular shunts. Taken together, these data show that loss of Notch1 receptor function at birth in ECs recapitulates the phenotypes observed in Rbpj^iΔEC mice.

Given that previous publications have reported the development of NRH and angiosarcoma in the liver, following global deletion of Notch1, we explored the longer-term effects of endothelial Notch1 deletion. Survival of Notch1^iΔEC and control mice was monitored up to 80 days after birth, with no significant differences observed (Fig.S9c). Additionally, we evaluated H&E sections from P80 Notch1^iΔEC mice and found dilated sinusoids and slight differences in hepatocyte size between the portal and central regions (Fig.S9a,b). However, we did not observe the NRH-like phenotype detected in adult Rbpj^iΔEC, and we found no evidence of vascular tumors. These results indicate that long-term endothelial deletion of Notch1 from birth does not result in similar liver parenchymal abnormalities to those of Rbpj^iΔEC, or liver vascular tumors, within this timeframe.

Endothelial deletion of Rbpj leads to abnormalities in isolated liver sinusoid endothelial cells

To gain cell biological insight into the role of endothelial Notch signaling in the liver vasculature, we isolated liver sinusoid endothelial cells (LSECs) from Rbpj^iΔEC and control mice at P7, to avoid secondary effects that might result from the hypoxia and hepatocyte necrosis observed in P14 Rbpj^iΔEC mice. We imaged isolated cells every 15min for 48h under culture conditions. We evaluated the rate of cell proliferation and cell death of LSECs from Rbpj^iΔEC and found no significant differences when compared to controls (FigS10a,b). However, Rbpj^iΔEC LSECs exhibited an abnormal cell:cell interface that acquired the appearance of membranous “holes” (Movies 1,2). Consistent with time-lapse imaging, isolated and fixed LSECs showed an increased number of membranous “holes” per cell (Fig.7a,b,c). Interestingly, a similar aberrant cell-cell interface was observed in liver sections in the sinusoids of adult Rbpj^iΔEC mice (Fig.7a,b). These results suggest that endothelial Rbpj may be involved in regulating endothelial cellular properties that affect EC:EC interactions in the liver.

Figure 7. Primary liver ECs from Rbpj^iΔEC mice exhibited aberrant cell-cell interface and decreased ephrin-B2 expression.

a,b, As compared to controls (a), liver ECs from Rbpj^iΔEC mice (b), either cultured at P7 (top) or in fixed adult tissue (bottom), displayed aberrant cell-cell interface, manifesting in membranous “holes” (arrowheads). mGFP from the mT/mG reporter indicates ECs. DAPI stains cell nuclei. Scale=25μm. c, The number of membranous “holes” per P7 liver EC. Data show mean ± s.e.m. (n=3 mice); * p<0.05. d,e, ephrin-B2^tau-lacZ expression was decreased in P14 Rbpj^iΔEC liver (e), as compared to controls (d). ephrin-B2^tau-lacZ expression is shown in whole liver (top panels in d,e) and in liver sections (bottom panels in d,e). Nuclear fast red stains nuclei. Scale=1mm (top panels in d,e); 25μm (bottom panels in d,e). f, Quantitative RT-PCR analysis of P7 liver ECs. Data are presented as mean ± s.e.m. (n=3). * p <0.05.

Endothelial deletion of Rbpj leads to downregulation of ephrin-B2

We investigated molecular changes to liver endothelium, following endothelial deletion of Rbpj using freshly isolated LSECs from P7 Rbpj^iΔEC and control mice. We first evaluated Notch downstream genes, including the arterial marker ephrin-B2 and the venous marker Ephb4, previously described to be modulated by Notch in the liver endothelium liver(6, 17) and involved in AVM formation in the brain(18). While Ephb4 was not altered significantly, ephrin-B2 was significantly reduced in LSECs from P7 Rbpj^iΔEC mice, when compared to control (Fig.7f).

We also evaluated expression of candidate genes previously shown to be regulated by Notch and involved in vascular shunting, Matrix Gla protein(19), Smad4(20), and Pten(21). We did not observe significant changes in the levels of these genes (Fig.7f), indicating that modulation of these pathways is not likely involved in the formation of porto-systemic shunts, following endothelial deletion of Rbpj.

We further investigated expression of ephrin-B2 in vivo by crossing Rbpj^iΔEC mice with a mouse line carrying lacZ under the control of the ephrin-B2 promoter (ephrin-B2^tau-lacZ). In control mice, β-galactosidase (produced by ephrin-B2^tau-lacZ) was detected in the arteries, portal venules, and some periportal sinusoids (Fig.7d); however, Rbpj^iΔEC-ephrin-B2^tau-lacZ mice showed drastically reduced levels of the reporter (Fig.7e). These results suggest a possible role for ephrin-B2 in the development of porto-systemic shunts, following endothelial loss of Notch signaling.

Loss of endothelial Notch signaling results in upregulation of VEGFR3 in dilated sinusoids

Recent studies have shown that the hyper-angiogenic response observed in postnatal retina in Notch loss-of-function mice is mainly mediated by VEGFR3(22). We therefore hypothesized that Rbpj^iΔEC livers would have altered VEGFR3 expression. mRNAs from the livers of P14 mice were analyzed by RT-PCR, and Vegfr3 (Flt4), but not its ligands Vegfc and Vegfd, was found significantly increased in Rbpj^iΔEC mice (Fig.8a). Interestingly, Flt4 levels in P7 LSECs did not show significant differences between Rbpj^iΔEC and control mice (Fig.S11a).

Figure 8. VEGFR3, but not vegfc or vegfd, was upregulated in Rbpj^iΔEC mice.

a, Quantitative RT-PCR analysis of P14 whole liver tissue. Data are presented as mean ± s.e.m. (n=3). * p <0.05. b,c,d,e VEGFR3 immunostaining in liver sections control (b,d) and Rbpj^iΔEC (c,e) mice indicated an increase in VEGFR3 in P14 (c) and P7 (e) mutants. Scale=50μm. (n=4) pv: portal venule, cv: central venule.

Immunostaining demonstrated that VEGFR3 was highly upregulated in P14 and P7 Rbpj^iΔEC liver vessels. This upregulation was most pronounced in the sinusoids nearest the central venule, where sinusoid dilation is maximal (Fig.8b,c,d,e). Consistent with these data, VEGFR3 upregulation was observed in the sinusoids of P14 Notch1^iΔEC mice (Fig.S12). Immunostaining against VEGFR2 revealed no differences in Rbpj^iΔEC vs. control mice (Fig.S13). Similarly, Kdr (Vegfr2) levels in P7 LSECs of Rbpj^iΔEC mice were not significantly different than Controls (Fig.S11b). Together, these results suggest a post-transcriptional regulation of VEGFR3 by Notch signaling in liver sinusoids and that VEGFR3 may be involved in the development of vascular anomalies, following endothelial Notch loss of function.

DISCUSSION

Here, we show for the first time that endothelial deletion of Notch signaling results in HVMs

While previous studies have reported similar vascular and parenchymal phenotypes in the liver, following global deletion of Notch during adulthood(23), our results delineate that impaired Notch signaling in the endothelial lineage is responsible for the observed hepatic abnormalities. Disruptions to liver parenchyma and sinusoid architecture and impairments to liver function were very severe following loss of endothelial Notch function in immature mice, but were milder when endothelial Notch signaling was deleted during adulthood, suggesting differential temporal regulation of Notch. Additionally, VEGFR3 was increased in the abnormal sinusoids of Notch mutant mice, and ephrin-B2 was reduced, suggesting that these factors are involved in the formation of liver vascular malformations. Together, these results support a role for Notch in postnatal liver endothelium to establish adequate density and interface of the hepatic vasculature (Fig.S14).

Endothelial Notch signaling regulates sinusoid diameter and prevents porto-systemic shunting

Loss of Notch signaling in ECs leads to dilated sinusoids, originating around the central vein area, that ultimately form direct shunts between the portal and central veins and result in the development of HVMs. Increased liver sinusoid diameter has also been reported in mice with non-specific loss of Notch1(6) and in mice treated with antibodies against the Notch ligand Dll4(24). Observations in other vascular beds, such as the developing retina, support a role for endothelial Notch in regulating vessel diameter. Increased diameter of venule branches was observed in Rbpj^iΔEC mice(15, 16). Additionally, endothelial Notch was reported to regulate vessel diameter in the yolk sac(25).

In the retina(15) and in liver sinusoids(6), increased vessel diameter is consistent with increased angiogenesis and hypervascularity, triggered by loss of Notch signaling. It is likely that, in the presence of a hypervascular network generated by loss of endothelial Notch signaling, the enlarged sinusoids form as a consequence of increased angiogenesis. In these settings, blood flow through the liver will opt for the path of least resistance within the wider, non-remodeled, sinusoids, selecting for their growth at the expense of neighboring branches. As sinusoid dilation and porto-systemic shunting progress, tissue perfusion and oxygenation is impaired, as evidenced by increased hypoxia in mutant livers. Consistent with this, similar hemodynamic changes, caused by enlarged vessel diameter, contribute to the development of vascular malformations in two different models of brain AVMs(26, 27). Progressive, steal-mediated enlargement of liver sinusoids in mice with loss of endothelial Notch may similarly lead to the development of vascular malformations. Together, our data suggest that endothelial Notch regulates sinusoid diameter and prevents porto-systemic shunting in the postnatal liver.

The phenotypes resulting from the ablation of Rbpj reveal temporal differences in Notch signaling

Our results suggest a critical role for endothelial Notch in the early postnatal liver. Endothelial ablation of Notch signaling at this time results in a disruption of the developing vascular network that rapidly leads to severe HVMs and liver malfunction. Meanwhile, deletion of Notch in mature mice requires more time for the appearance of HVMs. These results suggest that endothelial Notch regulates liver vasculature in a temporally controlled manner. This temporal discrepancy is likely due to the quiescent state of the adult vasculature, wherein ECs are not exposed to the myriad of angiogenic stimuli present in the early postnatal liver. This age-dependent phenotype is consistent with the previous proposition that an active angiogenic environment is key for the development of vascular malformations(28, 29).

Notch4, in addition to Notch1, may be required in ECs to prevent HVMs

Endothelial Rbpj deletion results in more severe liver vascular malformations than endothelial deletion of Notch1. Because ablation of Rbpj impairs all canonical Notch signaling, it is possible that other Notch receptors are important. Since Notch4 is expressed in ECs, it is possible that Notch4 deficiency, in concert with Notch1 deficiency, is required for liver vascular malformations. Indeed, deletion of both Notch1 and Notch4 in embryos results in a more pronounced phenotype than that of Notch1 deletion alone(5). Importantly, mice lacking Rbpj also die at embryonic stages and display similar defects to those of the Notch1/Notch4 double knockouts(30). Therefore, it is likely that the increased severity of hepatic vascular malformations in Rbpj^iΔEC livers, as opposed to Notch1^iΔEC livers, results from loss of Notch signaling via both Notch1 and Notch4 receptors. However, Rbpj has been reported to act independently of Notch signaling(31). Thus it is also possible that non-Notch mediated effects of Rbpj are important in preventing porto-systemic shunting.

VEGFR3 upregulation in endothelium of Rbpj mutants supports a role in HVM formation

Our results show an upregulation of VEGFR3 but not its ligands in mice with endothelial deletion of Notch signaling. VEGFR3 staining was more robust in the sinusoids associated with the central vein area, where the vascular abnormalities are first detected and more severe. Surprisingly, we did not see a difference in the VEGFR2 levels, consistent with the work of Dill et al.(6) who detected no changes in the Vegfr2 levels from Notch1^null sinusoid ECs, and also with the lack of change in VEGFR2 albeit upregulation of VEGFR3 in the lung and developing retina of Rbpj^iΔEC mice(6). Interestingly, we observed no differences in levels of Vegfr3 mRNA in P7 isolated LSECs. Similar uncoupled regulation of mRNA and protein levels of VEGFR3 was also observed by Benedito et al.(22) who concluded that Notch likely regulates VEGFR3 by additional post transcriptional mechanisms. These studies also conclude that VEGFR3 kinase activity (as opposed to the ligand binding domain) mediates deregulated angiogenesis in settings where Notch activity is low or absent(22), consistent with the vascular abnormalities seen in Rbpj^iΔEC and Notch1^iΔEC livers.

Taken together, these findings lead us to propose a role for VEGFR3 dysregulation, likely through its ligand independent kinase activity, in the development of HVMs. While previous reports have shown that inhibition of VEGFR2 ameliorates HVMs in patients(32), recent findings indicate that endothelial deletion of VEGFR2 results in decreased levels of VEGFR3(22), suggesting that targeting both VEGFR2 and the VEGFR3 activity might be the optimal strategy for the treatment of HVMs. Certainly, further studies to determine the specific role(s) of VEGFR3 and 2 signaling in the pathogenesis of HVMs are warranted.

Reduced ephrin-B2 in the Rbpj^iΔEC liver suggests it may be a molecular mediator downstream of Notch signaling in HVM formation

Our results show a downregulation of ephrin-B2 in the endothelium of Rbpj^iΔEC mice, which is in agreement with previous works(6, 16, 30). Ephrin-B2 is regulated by Notch and selectively expressed in arteries(13, 18, 27). Interactions between ephrin-B2 and its receptor EphB4 are thought to segregate ECs between arteries or veins, therefore establishing arteriovenous boundaries needed for a functional vasculature(33, 34). In the liver, ephrin-B2 is expressed not only in the hepatic arteries, but also in the portal venules and the periportal sinusoids. Ephrin-B2 is likely to be involved in the establishment of porto-systemic hierarchy. In a setting where loss of function of Notch leads to increased vascular density, additional loss of ephrin-B2 would hamper an adequate establishment of vessel structure and promote the development of porto-systemic shunts.

Notch deletion may lead to HVMs in human liver

Previous studies inhibiting the Notch signaling pathway in different species have shown a highly conserved function for the Notch pathway in the liver(24), raising the possibility that inhibition of Notch signaling also results in HVMs in humans. Consistent with this, mutations in RBPJ and NOTCH1 recently have been reported as the most common cause of Adams-Oliver syndrome(35, 36), a disease that presents a high frequency of vascular malformations, suggesting that perturbation of Notch signaling in the human liver also results in HVMs.

Our findings reveal an important role for Notch signaling in the development of vascular malformations in the early postnatal and adult mouse liver. Given the conservation of this signaling pathway, it is possible that disruption of Notch signaling in humans results in a similar vascular phenotype in the liver. As there are currently more than 20 ongoing clinical trials testing Notch pathway inhibitors (http://www.cancer.gov/clinicaltrials), we propose that the livers of the individuals in these trials should be diligently monitored for detrimental effects on their vasculature and function.

List of Abbreviations

ALT

alanine transaminase

AST

aspartate transaminase

AVM

arteriovenous malformation

cv

central venule

EC

Endothelial cell

GFP

green fluorescent protein

H&E

hematoxylin and eosin

HHT

Hereditary Hemorrhagic Telangiectasia

HPF

High Power Field

HVMs

Hepatic vascular malformations

IACUC

Institutional Animal Care and Use Committee

LSEC

Liver Sinusoid Endothelial Cells

NICD

Notch intracellular domain

NRH

nodular regenerative hyperplasia

O/N

over night

P

Postnatal day

PBS

phosphate-buffered saline

pHH3

Phospho Histone H3

RT-PCR

Real time PCR

pv

portal venule

Rbpj

recombination signal binding protein for immunoglobulin kappa J region

s.e.m.

standard error of the mean

UCSF

University of California, San Francisco

VE

vascular endothelial

VEGFR

vascular endothelial growth factor receptor