Liver Endothelial Heg Regulates Vascular/Biliary Network Patterning and Metabolic Zonation Via Wnt Signaling.

BACKGROUND & AIMS: The liver has complex interconnecting blood vessel and biliary networks; however, how the vascular and biliary network form and regulate each other and liver function are not well-understood. We aimed to examine the role of Heg in mammalian liver development and functional maintenance.
METHODS: Global (Heg-/-) or liver endothelial cell (EC)-specific deletion of Heg (Lyve1-Cre;Hegfl/fl ) mice were used to study the in vivo function of Heg in the liver. Carbon-ink anterograde and retrograde injection were used to visualize the 3-dimensional patterning of liver portal and biliary networks, respectively. RNA sequencing, histology, and molecular and biochemical assays were used to assess liver gene expression, protein distribution, liver injury response, and function.
RESULTS: Heg deficiency in liver ECs led to a sparse liver vascular and biliary network. This network paucity does not compromise liver function under baseline conditions but did alter liver zonation. Molecular analysis revealed that endothelial Heg deficiency decreased expression of Wnt ligands/agonists including Wnt2, Wnt9b, and Rspo3 in ECs, which limits Axin2 mediated canonical Wnt signaling and the expression of cytochrome P450 enzymes in hepatocytes. Under chemical-induced stressed conditions, Heg-deficiency in liver ECs protected mice from drug-induced liver injuries.
CONCLUSION: Our study found that endothelial Heg is essential for the 3-D patterning of the liver vascular and indirectly regulates biliary networks and proper liver zonation via its regulation of Wnt ligand production in liver endothelial cells. The endothelial Heg-initiated changes of the liver metabolic zonation and metabolic enzyme expression in hepatocytes was functionally relevant to xenobiotic metabolism and drug induced liver toxicity.

The Heg gene in endothelial cells regulates the formation of blood vessel and bile duct networks. Deletion of Heg downregulates the expression of drug metabolism enzymes and protects the liver from toxin-induced liver injury.

The liver interwoven vascular and biliary networks are organized as numerous lobules, the functional and structural unit of liver, to perform its diverse functions including nutrient metabolism, detoxification, protein synthesis, and bile secretion. The formation of the 3-dimensional (3-D) network of blood vessels and the biliary system has been closely explored in recent years.2, 3, 4, 5 Endothelial signaling has been shown to mediate liver development and regeneration,, but the molecular regulation of the liver vascular network development is largely unknown. In each lobule, blood flows from the portal tract through sinusoidal vessels to a draining central vein. From the portal vein (PV) to the central vein (CV), hepatocytes can be divided into 3 zones according to the molecular signatures and functional preferences. A range of factors, such as oxygen tension, metabolic hormone gradient, and angiocrine morphogens, have been implicated to contribute to hepatic zonation., Wnt ligands/agonists including Wnt2, Wnt9b, and Rspo3 produced from pericentral liver endothelial cells (ECs) have been shown to play a dominant role in the establishment and maintenance of hepatic zonation.,11, 12, 13 However, the upstream signaling component that drives the expression of Wnt ligands in liver ECs is not known.

Biotransformation of xenobiotics is an important step of liver detoxification. The liver contains 2 classes of enzymes essential for the metabolism of xenobiotics. Group I enzymes, including cytochrome P450, flavin-containing monooxygenase, and epoxide hydrolases, catalyze the oxidative and reductive reaction of xenobiotics; and Group II enzymes, including glutathione-S transferase, UDP-glucuronosyltransferase, N-acetyltransferases, and sulfotransferase, catalyze conjugation reactions of metabolites from the group I reaction., These enzymes are also regulated by zonation morphogens and differentially expressed in different hepatic zones. For example, Cyp2e1 is predominantly expressed in pericentral zone 3 and intermediate zone 2 cells, and its expression is positively regulated by Wnt signaling. Down-regulation of Wnt/β-Catenin signaling decreases cytochrome P450 expression in zone 3 hepatocytes, and the liver gains resistance from toxic compound-induced injury.,

Heart-of-glass (Heg) is a type I transmembrane protein with a large extracellular domain and a relative short, but highly conserved intracellular domain that interact with Krit1 to engage Heg with the cerebral cavernous malformation (CCM) signaling complex. The interaction between Heg and CCM proteins is required for normal heart and vessel development. Heg and ccm2 have been shown to regulate hepatocyte polarity in zebrafish whereby the loss of Heg or ccm2 expression causes mis-patterning of the vessel and biliary networks. Here, we demonstrate that deletion of Heg in mouse liver ECs decreased the density of vascular and biliary networks and altered liver metabolic zonation, consequent to the downregulation of Wnt ligands from endothelial cells. This metabolic zonation change altered the expression level of biotransformation enzymes and protected the liver from hepatotoxin-induced injury. Our data also demonstrated that Heg is an upstream regulator of Wnt ligands that mediated the crosstalk between liver EC and hepatocyte.

Results

Loss Of Heg in Liver Endothelial Cells Decreases the Density of the Liver Vascular Network

Heg has been shown to regulate cardiovascular development., In zebrafish, heg deficiency impaired hepatocyte polarity and canaliculi development. To investigate the role of Heg in the mammalian liver, we generated the global Heg-deficient (Heg) mice. The liver size and weight were comparable between Heg and littermate control mice at 3 months of age (Figure 1A and B). The level of liver enzymes and bilirubin in plasma were also similar between control and Heg mice (Figure 1C). Histology analysis revealed no differences except a slight increase of inflammatory cell infiltration in the periportal area of Heg liver in (Figure 1D). Gene expression analysis revealed a small increase of Acta2 and F4/80, but not other markers of fibrosis and inflammation, in the Heg livers compared with those of control littermates (Figure 1E and F).

Figure 1

A-B, Micrograph of livers (A) and quantitation of liver/body weight ratio (B) of Heg and littermate control mice at 3 months of age (n = 4 per genotype). C, Serum level of liver enzymes (ALP/ALT/AST), protein (TP/ALB), and bilirubin (TBIL) show no significant difference between control and Heg mice at 12 weeks of age. D, H&E staining, sirius red staining, immunostaining of CD45 and Desmin. E, mRNA level of liver injury markers (Col1a1, Acta2, Desmin, Ddr2 and Tgfβ). F, mRNA level of macrophage markers (F4/80, Il1β, and Il6). Data are presented as mean ± SD using unpaired Student t test. ∗P < .05. Scale bars represent 5 mm in panel A and 50 μm in panel B.

To investigate the potential role of Heg in the development of liver vascular network, we performed the anterograde injection of Indian ink into the common PV followed by tissue clarification to visualize the vascular network of the portal tract. Vascular branching and density were reduced in Heg livers compare to the littermate controls (Figure 2A and B). Pecam immunostaining of the liver sections confirmed the decreased vascular density and revealed mis-patterning of the vascular and sinusoid network in the livers of the Heg mice (Figure 2C and D). Glutamine synthase (GS) staining to label hepatocytes surrounding the CVs further confirmed the reduction in the density of CVs (Figure 2E). To quantify the vessel density, we marked the vessels surrounded by GS+ cells as CVs and the vessels without GS+ cells surrounding as PVs, and found that the densities of the central and portal vessels were reduced by 42% and 57% in the Heg livers compared with littermate controls, respectively (Figure 2E and F).

Figure 2

A-B, 3-D visualization of portal vein system with anterograde injection of Indian ink (A) and quantitation of peripheral vessel branches (B) to indicate the decreased vessel branching and density in the livers of Heg mice at 6 weeks of age. C-D, Pecam staining (C) and quantitation of vessel coverage area (D) of liver sections of to indicate the decreased vessel density in the livers of Heg mice at 3 months of age. E, GS staining to mark pericentral hepatocytes and the decrease of vessel density in the liver of Heg mice at 3 months of age. F, Quantitation plot indicating the decreases of CV and PV density in the liver of Heg mice at 3 months of age (n = 4 per genotype). Data are presented as mean ± SD using unpaired t test. ∗P < .05; ∗∗∗P < .001. Scale bars represent 5 mm in panel A, 50 μm in panel C, and 400 μm in panel E.

Heg is highly expressed in liver ECs, as indicated by gene expression plot from single cell sequencing dataset (Figure 3A) and gene expression analysis of isolated liver cell groups (Figure 3B). RNAscope analysis indicate Heg is highly expressed in ECs of the PVs and sinusoid and CVs, with some expression detected in hepatocytes and epithelial cells of the bile duct (Figure 3C and D). To test whether it is indeed the liver endothelial Heg that initiate signaling to regulate the patterning of the vascular network, we generated mice that specifically lack Heg in liver ECs by crossing the Heg mice with the Lyve1-cre mice, which express Cre recombinase specifically in the CV and PV and sinusoidal ECs in livers (Figure 3E and F). The exon 1 of the Heg gene were flanked with loxP sites in the Heg mice (Figure 3E). The expression level of Heg in liver ECs of the Lyve1-Cre;Heg (hereafter denoted as Heg) mice was almost absent as assayed with quantitative polymerase chain reaction (qPCR) analysis of samples from livers of embryos at E12.5 and mice at the age of 3 days or 6 weeks (Figure 3G). Similar to the Heg mice, no growth defects were observed in the livers of the Heg mice at 2 months of age. 3-D casting of the PV system revealed a dynamic 3-D expansion of the portal vascular network during postnatal development in both the Heg and the Heg mice, as indicated by the increased density of ink dye in the central areas of liver lobes from ages 4 to 12 weeks. However, it was noted that the Heg livers showed a decreased density of the portal vascular network at all maturation stages compared with Heg mice (Figure 3H and I). GS immunostaining confirmed the significantly reduced CV and PV density, but the CV/PV ratio was significantly increased in the Heg mice compared with littermate controls (Figure 3J and K). CV lumen coverage area was not affected, whereas the percentage of PV lumen coverage area was significantly increased in the Heg mice (Figure 3K). These data suggest that an expansion of the PV lumen compensated for the reduced PV density in the Heg liver. In addition, the percentage of GS stained area was significantly reduced in the Heg liver (Figure 3K), suggesting that the reduction of pericentral metabolic enzyme capacity was not compensated (Figure 3J and K). Collectively, these results all indicated that endothelial-specific loss of Heg significantly reduced the vascular branching in the liver.

Figure 3

Decrease of liver vascular network density in the mice.A, Analysis of single-cell RNA-seq data from Tabula Muris show Heg is predominantly expressed in ECs in the liver. B, qPCR analysis of Heg expression in isolated liver cells of 6-week-old mice show Heg is selectively expressed in ECs. C, RNAscope analysis show Heg is expressed (red dots) in ECs, hepatocytes, and epithelial cells of bile duct in livers of mice at age of 6 weeks. D, Quantification of Heg RNAscope signal in different liver cell types. E, Schematic of the generation of the Lyve1-Cre;Heg mice. F, Co-immunostaining of Pecam and GFP on liver section of 3-week-old Lyve1Cre; Rosa mice show the Lyve1-Cre driven recombination in liver ECs including ECs of sinusoid, CVs, and PVs. G, qPCR analysis of Heg deletion efficiency in isolated liver ECs from embryo (E12.5), neonatal (P3) and adult (6 week) control and Lyve1-Cre;Heg mice. H-I, 3-D visualization of PV system with Indian ink injection (H) and quantitation of peripheral vessel branches (I) in the liver of Lyve1-Cre;Heg and control mice from weaning to adult (4, 8, and 12 weeks of age). J, GS immunostainings indicate the decreased vessel density in the liver of Lyve1-Cre;Heg mice at 3 months of age. K, Quantitative plots indicate the density and coverage area of CV and PV, and the GS+ areas in the liver of Lyve1-Cre;Heg and control mice at 3 months of age (n = 4 per genotype). Scale bars represent 5 μm. Data are presented as mean ± SD using unpaired t test. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001. Scale bars represent 50 μm in panel C and F, 5 mm in panel G, and 400 μm in panel I.

Loss of Heg in Liver Endothelial Cells Decreases Biliary Network Density

The liver triad consists of PV, bile duct, and hepatic artery. During liver development in mammals, the bile ducts form around the PV following the inductive signaling from PV, whereas the bile ducts guide hepatic artery development. We performed the retrograde ink injection into the common bile ducts to visualize the 3-D network of the biliary network and found the density of the biliary network was also decreased in the Heg livers (Figure 4A and B). Although less branching of the biliary system in the Heg livers was observed, the diameter of the branches of the bile ducts appear to larger in the Heg livers compared with controls (Figure 4A), although we were unable to precisely measure the diameters of comparable grade of bile ducts. The reduction of biliary network density was also confirmed with immunohistochemical staining for cytokeratin 19 (CK19). Interestingly, despite this decrease in overall biliary density, the number of bile ducts for each of the portal tracts was not affected (Figure 4C and D). In addition, no obvious defect of the periportal lymphatic vessels were found in the Heg livers as indicated by stainings of Podoplanin, which is expressed in lymphatic vessels and well-lumenized periportal bile ducts (Figure 4E and F).

Figure 4

Decrease of liver biliary network density in the mice.A-B, 3-D visualization of biliary network (A) and quantitation of peripheral biliary branches (B) in the liver of Lyve1-Cre;Heg and control mice at 4, 8 and 24 wk of age. C-D, Immunostaining of CK19 (C) and quantitative plots of total and portal CK19 counts (D) indicating the bile duct density and bile duct/PV ratio in Lyve1-Cre;Heg and control mice at 3 weeks of age. E-F, Immunostainings of Podoplanin (Pdpn) and CK19 (E) and quantitative plots of the number of Pdpn+ lymphatic vessels (F) show no obvious defect in periportal lymphatic vessels in the liver of Lyve1-Cre;Heg mice at 8 weeks. ∗∗P < .01; ∗∗∗P < .001; Scale bars represent 5 mm in panel A, 100 μm in panel C, and 50 μm in panel E.

Loss of Heg in Liver Endothelial Cells Impairs Liver Zonation

Liver lobules are spatially zonated between the portal and central veins. Based on gene expression and metabolic profiles, liver lobule can be divided into three zones: periportal zone (zone 1), intermediate zone (zone 2), and pericentral zone (zone 3). Each zone can be identified with specific biomarkers such as E-Cadherin (ECad) for zone 1, Cyp2e1 for zone 2 and 3, and GS for zone 3. Due to the decreased vascular and biliary density in the livers of the Heg deficient mice (Figure 2, Figure 3, Figure 4), we hypothesized that Heg deficiency may affect liver lobule structure and metabolic zonation. We performed immunostaining of zonation marker proteins to assess zonal distribution. Although GS+ and Cyp2e1+ cells were comparable between the Heg and control livers, we identified an expanded zone 1 in the Heg liver based on ECad staining (Figure 5A and B). Additionally, the number of cell layers in zone 1 was significantly increased in the Heg livers compared with that in the control livers (from average of 14 layers in the control to 20 layers in the Heg livers). The numbers of cell layers in zone 2 and zone 3 remained the same between the control and Heg livers (Figure 5C–E). The percentage of GS+, Cyp2e1+, or ECad+ cells in each cell layers were comparable between control and knockout mice, with the notable expansion of ECad+ in the Heg liver (Figure 5D). Accordingly, the total cell layers of the liver lobules increased to an average 34 layers in the Heg mice compared with an average of 28 layers in control mice (Figure 5E). In each liver lobule, the GS and Cyp2e1 marked zone 3 and zone 2 area remained the same, whereas the area of ECad marked zone 1 increased by ∼3 fold in area (Figure 5F). At the whole liver level, there was a significant decrease in the percent of GS+ area and no changes to the Cyp2e1+ area, and we noted a modest but significant increase in the ECad+ area in the Heg mice compared with controls (Figure 5G). The decreased expression of GS was also observed at the protein levels, and a trend of increasing Ecad protein expression was also observed, but quantification did not reach statistical significance (Figure 5H and I). Gene expression analysis of purified hepatocytes showed decreased expression of zone 3 marker genes (Axin2, Glul, Oat) in the Heg-deficient liver compared with controls, whereas the expression of zone 1 markers (Cyp2f2, Hsd17, and Arg1) remained the same despite an expansion in cell layers in each lobule (Figure 5J). Gene expression analysis of purified liver ECs showed that Heg deficiency in ECs altered the expression of certain endothelial zonation genes such that expression of periportal gene Esm1 was increased and levels of pericentral genes such as Lhx6 and Fgfr2 were decreased in the Heg mice (Figure 6A). There was an expansion in Ace2 expression, a gene predominantly expressed in zone 1 endothelium, coinciding with the expansion of zone 1 in Heg liver (Figure 6B). Immunostaining with Ki67 revealed an increased proliferative activity in zone 1 that was not seen in zones 2 and 3 of maturing Heg liver (3 weeks). Ki67 positive cells were significantly reduced in mature mouse livers (6 weeks), and differences in cellular proliferation between the 2 groups were no longer observed (Figure 6C and F).

Figure 5

Deletion of A-B, Co-immunostainings of zonal markers, GS (zone 3) and E-Cadherin (ECad, zone 1), Cyp2e1 (zone 2/3), and ECad (zone 1) to indicate the zonal distribution within the lobules of Lyve1-Cre;Heg and control livers at 8 weeks of age. The dash lines mark the edge of ECad+ zones. C, Liver sections of control and Lyve1-Cre;Heg mice at 8 weeks were stained with Cyp2e1 and ECad+, example of onion layers of Cyp2e1+ hepatocytes and ECad+ hepatocytes were marked surrounding CVs and PVs, respectively, Cyp2e1+ layers as zone 2, ECad+ layers as zone1. D, Quantitative plots of the average number of GS+, Cyp2e1+, and ECad+ layers in each liver lobules (n = 3 per genotype). E, A continuous plot of the percent of GS+, Cyp2e1+, and ECad+ in each cell layer from pericentral to periportal indicating the increase zone 1 cell layers (from 14 to 20) and total cell layers (from 28 to 34) in each liver lobule in control and Lyve1-Cre;Heg mice. F, Quantification of GS+, Cyp2e1+, and ECad+ area in each liver lobule. G, Quantification of the percentage of GS+, Cyp2e1+, and ECad+ area in total liver area (n = 5–7 per genotype). H-I, Western blots (H) and quantification (I) of zonation maker, GS, Cyp2e1, and ECad in the whole lysates of livers from 8-week control and Lyve1-Cre;Heg mice. J, qPCR analysis of the expression level of marker genes of pericentral (Axin2, Glul, Oat, Cyp7a1, Rdh9) and periportal (Cyp2f2, Hsd17, Arg1, Ecad) zones in livers of 8-week control and Lyve1-Cre;Heg mice (n = 4 per genotype). Data are presented as mean ± SD using unpaired Student t test. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001. Scale bars represent 200 μm in panel A and B and 150 μm in panel C.

Figure 6

Deletion of A, qPCR analysis of the expression level of marker genes of periportal (Sdc1, Esm1, Ace2, Cxcl9) and pericentral (Lhx6, Fgfr2, Cdk1) zone ECs in livers of 8wk control and Lyve1-Cre;Heg mice (n = 3 per genotype). B, Immunostainings of zonal liver EC marker Ace2 (zone 1) to indicate LEC zonal distribution within the liver lobules of 8 week control and Lyve1-Cre;Heg mice. C-F, Co-immunostainings of Ki67/ECadherin and quantitative plots show liver proliferation at different zone in the liver of Lyve1-Cre;Heg and control mice at 3-week (C and D) and 6-week (E and F) time points. Data are presented as mean ± SD using unpaired Student t test. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001. Scale bars represent 200 μm for top images and 50 μm for bottom images in panels B, C, and E.

Heg Regulates the Expression of Wnt Ligands in Endothelial Cells and Wnt Signaling in Hepatic Cells

To determine how Heg, as an endothelial transmembrane receptor, can affect liver lobule patterning and metabolic zonation, we profiled gene expression levels in ECs isolated from control and Heg livers. RNA-seq analysis revealed that Wnt ligands (Wnt2, Wnt9b, and Rspo3) were among the list of top downregulated genes in ECs from the Heg liver (Figure 7A). Gene Ontology (GO) term analysis further confirmed cell-cell signaling by Wnt as one of the top pathways altered in ECs of Heg-deficient livers (Fig. 7B and C). Wnt2, Wnt9b, and Rspo3 were significantly downregulated in ECs of Heg liver (Figure 7D), and these reductions in Wnt genes were further confirmed in purified liver ECs isolated from the Heg mice (Figure 7E). Wnt ligands are predominantly expressed by ECs of the CV, and RNAscope analysis demonstrated downregulation of Wnt2 in ECs of the CV in the Heg mice (Figure 7F and G). Wnt ligands are major angiocrine factors that regulate zonation via canonical signaling in liver parenchymal cells. In the isolated hepatocytes, the expression of known Wnt targets such as Axin2, Glul, and Oat, were significantly downregulated, whereas the expression of Wnt receptors and co-receptors such as Fzds, Lgr4 and Znf3 were not altered (Figure 7H and I). The expression of β-Catenin mRNA was also not significantly altered in the Heg hepatocytes (Figure 7I). Immunostaining revealed that β-Catenin protein was predominantly localized to the cytoplasmic membrane of hepatocytes, and its expression levels were similar in pericentral zone 3 and periportal zone 1 hepatocytes of Heg and control livers (Figure 7J). On the other hand, phosphorylated β-Catenin was predominantly found in the cytoplasm, with increased expression observed in the pericentral hepatocytes of the Heg livers (Figure 7K). Expression of phosphorylated β-Catenin was minimum in both the control and Heg peri-portal hepatocytes (Figure 7K). This finding is consistent with previous reports that pericentral ECs are the predominant source of Wnt ligands. Diminished Wnt ligands production promote the increase of phosphorylated β-Catenin and result in the degradation of β-Catenin in pericentral hepatocytes. To further confirm whether Heg regulate the expression of Wnt ligands/agonists, we knocked down HEG expression with shRNAs in human umbilical vein endothelial cells (HUVECs) that express detectable levels of RSPO3, but not WNT2 and WNT9b. Similar to our observations in the Heg-deficient ECs, loss of HEG levels downregulated the expression of RSPO3 in HUVECs in vitro (Figure 7L). These results suggest Heg is an upstream signal that can positively regulate the expression of Wnt ligands in liver ECs and indirectly maintain canonical Wnt signaling in hepatocytes.

Figure 7

Endothelial A-C, Transcriptomic analysis of liver endothelial cells of Lyve1-Cre;Heg and littermate control mice at 6 weeks of age. Volcano plot (A) shows Wnt2, Wnt9b, and Rspo3 are among the top downregulated genes in Heg-deficient liver ECs, GO term analysis (B) and GSEA analysis (C) identified Wnt signaling as a top function that is altered in Heg-deficient liver ECs. D, Transcripts per million (TPM) counts of all detectable Wnt ligands and Wls shows Wnt2, Wls, Rspo3, and Wnt9b are Wnt signaling genes expressed at high level in liver ECs. E, qPCR analysis confirms Wnt2, Wnt9b, and Rspo3 are significantly downregulated in the ECs of Lyve1-Cre;Heg liver (n = 4 per genotype). F-G, RNAscope analysis and quantification confirm Wnt2 expression is decreased in the ECs of central veins in livers of 6wk Lyve1-Cre;Heg mice. H-I, qPCR analysis of the expression of Wnt target genes and Wnt receptor and co-receptor show the downregulation Wnt target genes in purified hepatocytes from 6-week control and Lyve1-Cre;Heg mice (n = 4 per genotype). J-K, Immunostainings of β-catenin (J) and phosphorylated β-catenin (K) show the constant expression of membrane localizing β-catenin and upregulated phosphorylated β-catenin in livers of 8wk Heg-deficient mice. L, qPCR analysis shows the decreased expression of RSPO3 in HUVECs after HEG gene knockdown with shRNA (n = 3 independent repeat experiments). Data are presented as mean ± SD using unpaired Student t test. ∗P < .05; ∗∗∗P < .001. NS, Not significant.

Heg is Required to Maintain Wnt Ligand Expression in Liver Endothelial Cells and Indirectly Regulate Liver Metabolic Zonation

Lyve1-Cre constitutively drives gene deletion from the developmental stage. To test whether it is the developmental defects and vessel density changes indirectly affect Wnt expression or Heg regulates Wnt expression in ECs in a cell-autonomous manner, we crossed the Cdh5-CreERT2 mice with the Heg mice to generate the Cdh5-CreERT2;Heg (Heg) and induced Heg gene deletion in endothelium after the liver is matured. Heg deletion in adult mice for 2 weeks does not change liver vessel density (Figure 8A and B), but can increase the area of zone 1 (ECad+) and decrease the area of zone 2/3 (Cyp2e1+) (Figure 8C-F). Similar to that of Heg mice, induced deletion of Heg also down-regulated the expression of Wnt2, Wnt9b, and Rspo3 in liver ECs, and the expression of Wnt target genes, Axin2, and Oat in hepatocytes (Figure 8G and H). Together with the phenotype observed in Heg mice, these data suggest Heg in liver ECs regulate the development/patterning of liver vasculature and promote the expression of Wnt ligands, thus, indirectly maintaining appropriate metabolic zonation in the liver.

Figure 8

Induced deletion of A, GS immunostainings indicate no difference of vessel density between livers of 6-week-old control and Cdh5-CreERT2;Heg mice 2 weeks after tamoxifen induction. B, Quantitative plots indicate the density and coverage area of CV and PV in the liver of control and Cdh5-CreERT2;Heg mice at 6 weeks (n = 3 per genotype). C-D, Co-immunostainings of zonal markers, GS (zone 3) and ECad (zone 1), Cyp2e1 (zone 2/3), and ECad (zone 1) to indicate the zonal distribution within the lobules of 6-week control and Cdh5-CreERT2;Heg mice. The dash lines mark the edge of ECad+ zones; scale bars represent 200 μm. E, Quantification of GS+, Cyp2e1+, and ECad+ area in each liver lobule (n = 3 per genotype). F, Quantification of the percentage of GS+, Cyp2e1+, and ECad+ area in total liver area (n = 3 per genotype). G-H, qPCR analysis confirms Wnt2, Wnt9b, Rspo3, and Heg are significantly downregulated in the liver ECs (G) and Wnt targets, Axin2, and Oat are significantly downregulated in the hepatocytes (H) of 6-week Cdh5-CreERT2;Heg mice (n = 3 per genotype). Scale bars represent 50 μm.

Loss of Heg in Liver Endothelial Cells Reduces the Expression of Xenobiotic Biotransformation Enzymes in Hepatocytes

To determine whether the decreased vascular and biliary network density can alter liver zonation and impair liver function, we performed liver chemistry analysis and histology in control and Heg livers. Serum biochemistry analysis revealed no obvious changes of liver enzymes (alkaline phosphatase [ALP], alanine aminotransferase [ALT], and aspartate aminotransferase [AST]), proteins involved in protein synthesis (total protein [TP] and albumin [ALB]), and bile secretion (total bilirubin [TBIL]) between the Heg and control mice (Figure 9A). No notable histological differences, such as lymphocytes infiltration and fibrosis, were detected between control and Heg mice by hematoxylin and eosin (H&E), CD45, Sirius red, and desmin staining (Figure 9B). Gene expression analysis also did not reveal any significant differences in the expression of inflammatory and fibrosis marker genes between the control and Heg livers (Figure 9C and D).

Figure 9

A, Serum level of liver enzymes (ALP/ALT/AST), ALB, and TBIL show no significant difference between control and Lyve1-Cre;Heg mice at 12 weeks. B, H&E staining, sirius red staining, and immunostaining of Desmin in control and Lyve1-Cre;Heg liver sections. C, mRNA levels of liver injury markers (Col1a1, Acta2, Desmin, Ddr2, and Tgfβ), and (D) mRNA level of macrophage markers (F4/80, Il1β, and Il6). Data are presented as mean ± SD using unpaired Student t test. Scale bars represent 50 μm.

Even though we did not detect an obvious defect in the livers of Heg-deficient mice at baseline, we reasoned that Heg regulated Wnt signaling might alter the metabolic processes at the molecular level. RNA-seq analysis revealed that differentially expressed genes between control and Heg liver tissues were enriched in drug and chemical metabolism pathways (Figure 10A–C). For example, genes of cytochrome p450 mediated drug metabolism, fatty acid metabolism, and glutathione metabolism were downregulated in Heg-deficient liver (Figure 10D). Among the top genes in the list of differentially expressed genes, Ces1d, Cyp4a14, and Cyp2b9 were significantly decreased in the Heg livers (Figure 10E and F).

Figure 10

Endothelial A, Volcano plot show the DEGs (red, up-regulated genes; green, down-regulated genes) in livers of Heg-deficient vs control mice at 12 weeks of age. B, GO term analysis and gene expression heatmap (C) revealed that genes with functions relevant to drug and chemical metabolism pathways are among the top DEGs. D, GSEA analysis shows the gene sets representing metabolism pathways are downregulated in Lyve1-Cre;Heg liver. E-F, qPCR analysis show the change of mRNA expression levels of xenobiotic transform enzymes in livers of 12-week Heg-deficient mice (n = 4–6 per genotype). Data are presented as mean ± SD using unpaired Student t test. ∗P < .05; ∗∗∗P < .001. NS, Not significant.

Loss of Heg in Liver Endothelial Cells Prevents Liver From Hepatotoxin-induced Injury

Detoxification is a critical function of the liver to protect the body from chemically induced injuries. Although the Heg-deficient livers at baseline were functionally similar to control livers, the observed changes in liver zonation (Figures 5 and 8) and the downregulation of drug metabolizing enzymes (Figure 10) in the Heg-deficient mice suggest Heg may have a role in regulating liver detoxification. When toxins exceed the liver detoxification capacity, they can induce liver injuries and cause structural and functional damages to the liver., Carbon tetrachloride (CCl4) is a hepatotoxin widely used to induce liver injury. We treated mice with CCl4 (1 mL/kg, 3 times a week for 3 weeks) to determine whether Heg-deficiency in liver ECs will compromise the toxin handling capability of the liver. CCl4 treatment led to a significant increase in liver weight in control mice, but this increase was not observed in the Heg-deficient mice (Figure 11A). To our surprise, CCl4 treatment caused less damage to the Heg livers than that of the control livers, as evidenced by reduced sirius red and desmin staining and increased retention of GS+ cells in CCl4-treated Heg-deficient liver compared with the CCl4-treated control livers (Figure 11B–E). These data suggested that the loss of Heg in liver ECs protected the liver from CCl4-induced damage. Further liver function analysis also revealed Heg mice tend to have reduced elevation of ALT and AST levels in response to CCl4-induced damage (Figure 11F). Similarly, the increases in serum levels of different forms of bilirubin were also less profound in the Heg mice than that of the control mice (Figure 11F). These results together suggest that Heg-deficiency in liver ECs was protective against CCl4-induced liver damage.

Figure 11

A, Ratio of liver weight and body weight show less profound increase of liver weight of 11-week Lyve1-Cre;Heg mice after 3 weeks CCl4 treated. B-E, Sirius red staining (B), Desmin staining (C), GS staining (D), and quantification (E) show less severe fibrosis and better pericentral protection in in Lyve1-Cre;Heg mice. F, Serum liver enzyme (ALP/ALT/AST) and bilirubin (TBIL/DBIL/IBIL) levels indicate less damage to liver function in CCl4 treated Lyve1-Cre;Heg mice (n = 5 pair for sham group and n = 6 pair for CCl4 group). Data are presented as mean ± SD using unpaired Student t test. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001. NS, not significant. Scale bars represent 500 μm in panel B, 100 μm in panels C and D.

Thioacetamide (TAA) is another toxic chemical used to model chronic liver damage where it is converted to toxic metabolites by Cytochrome P450 enzymes or monooxygenases., Treatment with TAA causes compensatory growth of liver and increase of liver weight in control mice (Figure 12A). TAA-induced liver weight gain was significantly diminished in the Heg mice compared with that of littermate controls (Figure 12A). Chronic TAA treatment led to massive infiltration of inflammatory cells to the liver parenchyma in control mice, whereas the number of infiltrated inflammatory cells was significantly less in the TAA-treated livers of the Heg mice (Figure 12B). Similarly, sirius-red staining revealed less fibrosis in the Heg liver. This was further confirmed by reduced expression of collagen and desmin immunostaining in the Heg livers (Figure 12C–D). It is known that TAA causes ductular reaction and immunostaining of CK19 demonstrated severe ductular reaction in the TAA-treated control mice, but this was again decreased in the TAA-treated Heg livers (Figure 12E). qPCR analysis also showed reduced levels of markers for macrophage (F4/80), fibrosis (Col1a1, Acta2, Desmin, and Timp), and ductal reaction (CK19) in the TAA-treated Heg livers (Figure 12F). Consistent with the reduced liver damage as shown by histological analysis, Heg-deficient mice also had significantly lower serum levels of ALP and bilirubin in comparison to the control mice after TAA treatment (Figure 12G). These results suggest that Heg deficiency in liver ECs protected the liver from TAA-induced liver injury.

Figure 12

A, Ratio of liver weight over body weight show less profound compensatory liver weigh growth of 16wk Lyve1-Cre;Heg mice after 8 weeks TAA treated. B-F, H&E staining (B), Sirius red, Col1a1, and Desmin staining (C), quantification (D), CK19 staining (E), and qPCR analysis (F) of marker genes show less lymphocyte infiltration, less fibrosis, and milder ductal reaction in TAA treated Lyve1-Cre;Heg mice. G, Serum liver enzyme analysis (ALP/ALT/AST) and bilirubin (TBIL/DBIL/IBIL) indicate less damage to liver function in CCl4 treated Lyve1-Cre;Heg mice. Data are presented as mean ± SD using unpaired Student t test (n = 6 pair for sham group and n = 6 pair for TAA group). ∗P < .05; ∗∗P < .01; ∗∗∗P < .001. NS, not significant. Scale bars represent 100 μm in panels B, C, and E.

To test whether the protective effects to toxin-induced liver injury in Heg liver were specifically due to the down-regulation of xenobiotic biotransforming enzymes, we applied a bile duct ligation model (BDL) that cause liver injury by bile accumulation in liver parenchyma. The liver damage in this model is not mediated by cytochrome P450 enzymes or monooxygenase. BDL caused a severe inflammation response and fibrosis from a week after the procedure. Histology and gene expression changes in 1 to 3 weeks after BDL were similar between Heg mice and its littermate controls (ie, in inflammatory cell infiltration, the expression of fibrosis makers and duct reaction marker were comparable between Heg mice and its littermate controls) (Figure 13). The serum level of liver enzyme and bilirubin also did not differ between Heg mice and its littermate controls (Figure 13I). These results further support our data that a loss of Heg in liver ECs was protective against drug-induced injury due to its primary effect on Wnt expression in endothelial cells, indirectly the Wnt-mediated metabolic zonation and the expression of cytochrome P450 enzymes in hepatocytes.

Figure 13

A-G, H&E (A), Sirius red (B), desmin (C), and CK19 (D) staining and quantification (E-G) of sham group and BDL model at 1-, 2-, and 3-week time points show no difference in the injury severity between control and Lyve1-Cre;Heg livers. H, mRNA level of macrophage marker (F4/80), fibrosis markers (Col1a1, Acta2, Desmin, and Timp), and marker of ductal reaction (CK19) of liver tissues in sham group and BDL model mice. I, Serum level of liver enzymes (ALP/ALT/AST) and bilirubin (TBIL/DBIL/IBIL) of sham group and BDL model mice shows BDL causes sever liver injury, but there is no significant difference between control and Lyve1-Cre;Heg mice. Data are presented as mean ± SD using unpaired Student t test. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001. NS, not significant. Scale bars represent 100 μm in panels A, B, and D and 50 μm in panel C.

Heg-deficient Mice are Protected From Drug-induced Liver Injury

Acetaminophen overdose is a leading cause of drug-induced liver failure., The expression levels of several reported metabolic enzymes involved in xenobiotic transformation for CCl4, TAA, and acetaminophen, including (Ces1d, Cyp4a14, Cyp2b9, Fmo1, Fmo5, and Ugt1a1) was significantly down regulated in the Heg livers (Figure 14A and B). To test the effect of Heg-regulated xenobiotic transformation pathway in handling injury response to clinically relevant drugs, we administrated Heg-deficient and control mice with high doses (300 mg/kg body weight) of acetaminophen. Acetaminophen caused acute liver damage in control mice, 265- and 49-fold increase in plasma ALT and AST, respectively, 24 hours after acetaminophen administration in control mice. Strikingly, only a mild increase in plasma ALT and AST levels were observed in the Heg mice following acetaminophen administration (Figure 14C). It coincided with a significantly reduced number of dead cells in the Heg livers as reflected by H&E and terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assays (Figure 14D–G). These data further suggest Heg-deficiency in ECs can protect the liver from drug-induced liver injury via endothelial cell-hepatocyte communication.

Figure 14

A, Metabolism process of acetaminophen and related enzymes. B, The expression levels of a list of reported enzymes involved in acetaminophen metabolism in the livers of control and Lyve1-Cre;Heg mice as measured by TPM values of RNA-seq. C, Serum liver enzyme analysis (ALT, AST, and ALP) indicate less acute damage to liver function in APAP treated 8-week Lyve1-Cre;Heg mice. D-E, H&E staining (D) and quantification (E) show less damaged area of pericentral hepatocytes in Lyve1-Cre;Heg mice 24 hours post APAP treatment. F-G, TUNEL staining (F) and quantification (G) show less cell death of pericentral hepatocytes in Lyve1-Cre;Heg mice 24 hours post APAP treatment. Data are presented as mean ± SD using unpaired Student t test (n = 4 pair for sham group and n = 4 pair for APAP group), ∗P < .05; ∗∗∗P < .001. Scale bars represent 100 μm in panels B and C.

Discussion

In the present study, we have generated a liver EC-specific knockout of the Heg that has allowed us to: (1) show that liver endothelial Heg signaling regulates vascular and biliary 3-D patterning; (2) identify Heg as an upstream regulator of Wnt ligand expression in liver ECs that regulate liver metabolic zonation; and (3) demonstrate that loss of liver endothelial Heg resulted in a decrease in key metabolic enzymes in hepatic zone 3 that protect the liver from drug-induced liver injury.

Although the molecular mechanisms of bile duct morphogenesis in relation to PVs have been extensively studied, few studies to date have addressed the regulation of liver vascular and bile duct branch patterning or their interdependence during development and adulthood.,,26, 27, 28 In this study we identified endothelial Heg as a key factor in regulating 3-D patterning of the liver vascular and biliary network. Our 3-D visualization of the vascular and biliary networks showed that the loss of Heg limited the branching capacity of the vascular network and indirectly reduced the density of the biliary network. The correlated density changes between vascular and biliary networks indicate that vascular patterning guides biliary patterning during network development and maturation. In mammalian liver development, PVs form before bile ducts and induce cholangiocyte specification., The decreased biliary branch density in Heg-deficient mice is likely to be a secondary effect caused by a decrease in PV branching, leading to a decrease in actual portal tracks rather than a direct effect of Heg on the biliary epithelium. Our results showed that, despite a decrease in portal tracks, the interlobular bile duct per portal track was unaffected and normal. In mice, ECs are essential for liver morphogenesis. In contrast, the patterning of intrahepatic bile duct systems in zebrafish has been demonstrated as not directly dependent on the vascular network. Liver budding and hepatocyte differentiation progress normally in cloche mutant, which lacks endothelial cells. However, endothelial heg and ccm2 have been shown to regulate hepatocyte polarity and canaliculi formation, and a recent study demonstrated that CDK5/Pak/LimK/cofilin in biliary epithelial cell regulated bile duct branching capacity but did not link this to signals from the liver vasculature. Thus, this study showed for the first time that endothelial Heg signaling directly modulates the 3-D patterning of vascular networks and then indirectly affects the 3-D patterning of biliary networks in the mammalian liver. The downstream molecular signals that mediate the role of Heg in vascular patterning remain to be elucidated.

The loss of Heg in liver ECs resulting in the decreased expression levels of Wnt ligands is intriguing as the Wnt/β-catenin signaling is known to be the major regulator of liver zonation.33, 34, 35, 36 Wnt2, Wnt9b, and Rspo3, are the 3 predominant ligands produced from pericentral ECs to shape liver zonation.,, Deletion of Rspo3 or blocking Wnt ligands secretion from liver ECs led to the diminished expression of β-catenin target genes such as Axin2, GS, and Cyp2e1 in the pericentral zone, increased expression of periportal zone genes (Cyp2f2, Hsd17), and the expansion of the periportal E-cad+ zone., Yet, it is not clear what regulates Wnt ligand expression in liver ECs. The molecular and histology phenotypes in our Heg-deficient mice closely resemble mice with disrupted Wnt signaling, such as diminished expression of Wnt target genes in pericentral hepatocytes and expansion of periportal zone. The downregulation of Wnt2, Wnt9b, and Rspo3 expression in Heg-deficient liver ECs strongly support Heg as an upstream positive regulator of Wnt ligand production in liver ECs, providing an important clue to the regulation and potential environmental stimulation of Wnt expression in liver ECs.

One of our findings related to the structure of the normal liver was the expansion of the number of cell layers for each lobule. Based on zonation marker staining, we observed there was an average of 28 cell layers per lobule in wild-type mice, but that increased to 34 layers in the Heg-deficient mice. This was mainly due to the increased layers of E-Cad expressing zone 1 cells, whereas the layers in zones 2 and 3 remained unchanged. A study of the overall numbers of cells in a human liver lobule suggested that a lobule consists of 25 to 30 layers of cells, including 7 to 8 layers of periportal cells and 10 layers of mid-lobular cells. The reason for the increase in the layers of ECad+ zone1 cells is possibly due to increased hepatocyte proliferation, because we detected increased Ki67+ cells in zone 1 of the Heg-deficient liver. It may also be due to a lack of Wnt repression signals or be the result of decreased vascular branching activity itself through a yet undefined mechanism. Therefore, how Heg regulate liver vascular branching should be a focus of future studies.

Another major finding from this study was the discovery that Heg regulated zone 2/3 xenobiotic metabolism. Gene expression data of whole liver tissues revealed significant downregulation of a number of zone 3 metabolic enzymes in the Heg-deficient mice, among them Ces1d, Cyp2b9, and Cyp4a14. The effect on the cytochrome P450 system, in association with fatty acid and glutathione metabolism, suggest that Heg expression in adjacent ECs indirectly regulate hepatocyte detoxification pathways. This observation was not associated with any changes in liver enzymes, liver immune cells, or extracellular matrix deposition at baseline. It is tempting to link our findings with other studies that have identified the pivotal role of Wnt/β-catenin signaling in maintaining GS and Cyp450 enzyme expression in pericentral hepatocytes, which are responsible for xenobiotic metabolism.,, Thus, given our findings of a decrease in endothelial cell Wnt ligands, it is not surprising to find that lack of liver endothelial Heg expression led to a reduction in hepatocyte xenobiotic metabolism pathways, which reduced the severity of acute liver injuries triggered by CCl4 and TAA. Our results are also consistent with the finding that liver-specific deletion of β-catenin promoted the loss of GS expression in the pericentral hepatocyte, and that ectopic expression of RSPO1 caused an expansion of GS+ cells., Therefore, the deletion of Heg in liver ECs decreased Wnt ligand production that had the potential to alter zonation, but confer protection for drug-induced liver injuries. However, we are not able to rule out whether the increase in lobule size can alter the oxygen tension and, therefore, its metabolic function. The pathogenic mechanism of CCl4 and TAA as liver toxins are caused by their conversion to reactive radicals by Cyp450 enzymes to react with nucleic acid, protein, and lipids and cause cell damage., Both CCl4 and TAA cause damage predominantly in the pericentral zone, although TAA induced a severe ductular reaction, whereas CCl4 did not. Both CCl4 and TAA treatment increased the level of ALP, ALT, and AST in the control mice. Heg deficiency significantly normalized the elevation of ALT and AST after CCl4 treatment and prevented ALP upregulation after TAA treatment. In contrast, Heg deficiency did not alter liver injury in a bile duct ligation model which does not rely on hepatocyte detoxification mechanisms. This does imply that the protective function of a loss of Heg in liver endothelial cells on liver injury is specifically metabolic-related by its indirect regulation of the expression of xenobiotic metabolism enzymes in hepatocytes, but not via the effects of liver endothelial signaling to inflammatory, hepatic stellate, or progenitor cells. The clinical significance of Heg regulation of xenobiotic metabolism should be further tested with pharmacological interventions. In this study, we tested the effect of endothelial Heg through overdosing with acetaminophen, which causes liver injury via zone 3 metabolic enzymes., Similar to that of CCl4 and TAA, loss of Heg in liver ECs protected the liver from acetaminophen-induced injuries indirectly via altered gene expression in hepatocytes.

Our current study did not show how Heg regulates vascular patterning in detail. However, RNA-seq data provided some initial clues for future studies. First, Heg was initially identified as an upstream receptor in CCM signaling. It can directly interact with Ccm1 and Rasip1, both of which regulate angiogenesis.,43, 44, 45 Thus, Heg may function through these proteins and their downstream small GTPase or Notch signaling to regulate vessel patterning in the liver vascular bed, but our RNA-sequencing (RNA seq) analysis of isolated liver ECs and whole liver tissue did not enrich genes in the Notch signaling. Second, the GO analysis of RNA-seq data identified about a dozen of genes implicated in blood vessel morphogenesis as the most down-regulated genes in Heg liver ECs such as Rspo3, Cav1, Dll1, Edn1, Bmp4, and Bmper. Among these genes, deletion of Rspo3 expression with the pan-endothelial cre, Cdh5-CreErt2, resulted in reduced vascular density in retina via its regulation of non-canonical Wnt signaling. Downregulation of Ca2+/Nfat signaling was found to be downstream of Rspo3 deficiency to cause the decrease of vascular density by regulating vessel pruning. Whether this Rspo3/Ca2+/Nfat signaling is conserved in the liver and lies downstream of Heg to regulate vessel remodeling and density should be investigated. Other proteins, such as BMP4, BMPER, Dll1, Edn1, and Tspan18, have all been shown to promote angiogenesis and vascular patterning., More investigations are required to identify what are the critical downstream signals that mediate the function of Heg in regulating liver vascular patterning.

Our finding that Heg initiated signaling that regulate vascular/biliary patterning and toxin-induced liver toxicity suggest the possibility that Heg may play a role in humans. Although the predominant Heg-expressing cell in the liver at physiological condition is the ECs, a recent study reported Heg expression to be increased in liver cancer cell lines and primary liver cancer tissues. Relevant to our finding that Heg is required for Wnt ligand production, Heg is found to stabilize β-catenin in these cancer cells and promote cell survival and migration. Further exploration of Heg mutations or variants in vascular- or biliary-related human liver diseases may point to the clinical significance of Heg mediated signaling.

Methods

All authors had access to the study data and had reviewed and approved the final manuscript.

Mice

The Heg and Heg mice have been described previously. Lyve1-Cre, Cdh5-CreERT2, and Rosa26 mice were purchased from Jax Laboratories. Cdh5-CreERT2;Heg mice were given 2 mg tamoxifen per mouse 3 times from 4 weeks of age intraperitoneally. The age-and gender-matched animals were randomized into control and treated groups. The Institutional Animal Care and Use Committee of Tianjin Medical University and The Sydney Local Health District Animal Welfare Committee approved all animal ethics and protocols. All experiments were conducted under the guidelines/regulations of Tianjin Medical University, and Centenary Institute, the University of Sydney, the guideline of National Research Council of the National Academies, and the Animal Research: Reporting of In Vivo Experiments guidelines.

Visualization of Liver Biliary and Vascular Network

The visualization of the biliary tree was performed according to the protocol developed by Kaneko et al with minor modifications. Briefly, mice were anesthetized and the abdomen was opened to expose the extrahepatic common bile duct. Indian ink (Phygene, PH1714) was slowly infused to the intrahepatic biliary tract by retrograde injection from the extrahepatic bile duct using a 30G needle (Terumo syringe, Japan). The injection was stopped when the ink had reached the surface of the liver. After filling the biliary tree with ink, the liver was harvested and subjected to gradual dehydration with 10 vol%, 40 vol%, and 80 vol% ethanol in phosphate buffered saline (PBS), with each incubation time being 1 hour at RT. The liver was finally incubated in 100% ethanol overnight and then soaked in a 2:1 benzyl benzoate:benzyl alcohol solution. In benzyl benzoate:benzyl alcohol solution, the dehydrated liver becomes optically transparent within a day or two and was then imaged (Nikon, SMZ18 or Leica, M165FC).

For the visualization of the liver vascular network, similar procedures were performed for the biliary tree, except the ink was slowly infused to the liver vessels by injection from the portal vein using a 30G needle. The injection was stopped just before the ink reached the liver lobe edge. Tissues were processed and imaged as above.

The vascular and biliary densities were represented by branches of peripheral ink-filled blood or bile duct per mm of lobe edges of the livers.

Immunofluorescence and Immunohistochemistry

After dewaxing and hydration, tissue sections were boiled in 10 mM citrate buffer solution (pH 6.0) for antigen retrieval, then blocked with 10% donkey serum at room temperature, incubated with appropriate primary antibodies overnight at 4°C. For immunohistochemistry, sections were incubated with horseradish peroxidase-conjugated secondary antibodies, then developed using 3,3′-Diaminobenzidine (DAB) tablet (Sigma), counterstained with hematoxylin. For immunofluorescence staining, sections were incubated with fluorescence-conjugated secondary antibodies for 1 hour at room temperature, then counterstained with 4, 6-diamidino-2-phenylindole (DAPI, Solarbio, China) and images were acquired under a fluorescence microscope. The following primary antibodies were used: anti-desmin (Abcam, ab15200, 1:300 dilution), anti-Col1a1 (ABclonal, A16891, 1:300 dilution), anti-Ck19 (Abcam, ab133496, 1:300 dilution), anti-GS (Abcam, ab49873, 1:5000 dilution), anti-Cyp2e1 (Atlas, HPA009128, 1:200 dilution), anti-Ecad (BD, BD610181, 1:300 dilution), anti-CD31 (Dianova, DIA-310, 1:300 dilution), anti-GFP (CST, 2555s, 1:300 dilution), anti-β-catenin (Santa, SC-7963, 1:300 dilution), anti-p-β-catenin (SC-57535, 1:200 dilution), Ace2 (Santa, SC -390851, 1:100 dilution), PDPN(Abcam, ab256559, 1:500 dilution), Ki67 (Abcam, ab15580, 1:300 dilution).

Immunoblot Analysis

Whole liver tissue homogenates were prepared in RIPA buffer (150 mM NaCl, 1% NP-40, 50 mM Tris [pH 8.0]) containing protease inhibitors. Protein concentrations of the homogenates were determined by BCA (Thermo Fisher Scientific, 23252). Protein was separated by electrophoresis using SDS-polyacrylamide gel and transferred to polyvinylidene difluoride membranes. The membranes were blocked with 5% skim milk and probed with primary antibodies anti-GS (Abcam, ab49873, 1:5000 dilution), anti-Cyp2e1 (Atlas, HPA009128, 1:1000 dilution), anti-Ecad (BD, BD610181, 1:1000 dilution) overnight at 4°C and followed with the incubation of respective peroxidase-conjugated secondary antibodies for one hour at room temperature. Blots were developed with ECL Western Blotting Substrate (Thermo Fisher Scientific), and the signals were acquired with MiniChemi610 (Beijing Sage Creation Science Co Ltd) in dark field mode and marker merged mode.

Quantification of Liver Central and Portal Vein Counts and Area

Slides were stained with peri-CV hepatocytes marker (GS), scanned with Pannoramic MIDI (3D HISTECH, Hungary). Images were exported from Pannoramic Scanner, then import to Image J (version 1.48). According to the image ruler, set “Distance in pixels” to 100, “known distance” to 100 μm. Central and portal veins were marked with “wand tool”; liver areas were calculated.

Quantification of Liver Lobule Cell Layers

Liver sections were immunostained with GS, Cyp2e1 or Ecad to mark zone 3, zone 3/2, and zone 1 regions in the liver, respectively. Cell layer were manually marked according to the shape of PV or CV and nucleus of hepatocytes (DAPI labeled). The percentage of GS+, Cyp2E1+, or Ecad+ cell in each layer were calculated (Figure 5). The average number of cell layers for each zone were calculated from liver sections of 3 mice, and 10 sample image fields with similar CV or PV size were assessed for each liver.

Isolation of Liver Endothelial Cells and Hepatocytes

Liver ECs and hepatocytes were isolated as described with minor modifications. Briefly, mice were anesthetized with avertin and perfused with D-Hanks through the portal vein, followed by perfusion with Hanks buffer containing 0.2 mg/mL of collagenase intravenously (Sigma-Aldrich, C5138). The liver was removed and placed in Hank’s buffer with gently agitation, and the cell suspensions were filtered through a 70 μm nylon mesh, centrifuged at 68 g for 5 minutes at 4 °C. The precipitations contain mostly hepatocytes. The supernatant was further centrifuged at 350 g for 5 minutes. The cell pellets were resuspended with Macks Buffer (D-Hanks + 0.5% BSA+2 mM EDTA). Liver EC were then isolated by magnetic-activated cell sorting using anti-CD31 MicroBeads (Miltenyi Biotech, Germany) according to the manufacturers’ instructions. The purity of liver EC was confirmed by qPCR compared with whole liver, hepatocytes, supernatants, and flow-through using different cell markers.

RNA Extraction, Reverse Transcription, and qPCR Analysis

Total RNA was extracted from tissue or cell homogenates using Trizol reagent (Invitrogen) and cDNA was synthesized from 1 μg of total RNA by reverse transcription (GeneStar, China). qPCRs were performed using SYBR QPCR Master Mix (Vazyme), and the primer pairs used are listed at the end of this section. The level of target gene expression was normalized to GAPDH expression in each sample.

RNA Sequencing

Total RNA was isolated from total liver tissue or liver ECs of control and Lyve1-Cre;Heg mice. Only high-quality RNA sample with OD260/280=1.8∼2.2, OD260/230≥2.0, RIN≥7, 28S:18S≥1.0 were used for sequencing library construction. Paired-end RNA-seq sequencing library was sequenced with the Illumina HiSeq xten/NovaSeq 6000 sequencer (2 × 150 bp read length) (majorbio, China). All clean reads were separately aligned to reference genome with orientation mode using HISAT2 (http://ccb.jhu.edu/software/hisat2/index.shtml) software. The mapped reads of each sample were assembled by StringTie (https://ccb.jhu.edu/software/stringtie/index.shtml?t=example) in a reference-based approach. To identify DEGs (differential expression genes) between 2 different samples, the expression level of each transcript was calculated according to the transcripts per million reads method. Differential expression analysis was performed using the DESeq2. Genes with Q value ≤0.05, DEGs with |log2FC| >1 and Q value ≤ 0.05(DESeq2) were considered to be significantly different expressed genes. In addition, GO functional-enrichment analysis was performed to identify which DEGs were significantly enriched in GO terms and metabolic pathways at Bonferroni-corrected P-value ≤ .05 compared with the whole-transcriptome as background. Gene Set Enrichment Analysis (GSEA) was implemented on the R (version 3.6.1). Gene sets with nominal P values less than .05 were considered statistically significant.

RNAscope in situ Hybridization

After deparaffinized in xylene, rehydrated in ethanol, and tissue sections were processed for RNA in situ hybridization using the RNAscope LS Multiplex Fluorescent Reagent Kit following the manufacturer’s instructions (Advanced Cell Diagnostics). The Wnt2 (Mm-Wnt2, Cat. 313601, NM_023653.5, region 857–2086) and Heg (Mm-Heg1-C2, Cat. 510581-C2, NM_175256.5, region 1927–2856) RNAscope probe were used. Sections were counterstained with DAPI and mounted. Positive RNAscope signals were quantified in Image J software, the expression levels of Wnt2 or Heg were presented as number of signals per cell of the specified liver cell type.

In vitro Knockdown of HEG

For in vitro knockdown of HEG in HUVEC, shRNA of human HEG and control nonsense shRNA were purchased from Vigene Bioscience (China). psPAX2, pMD2, and shRNA plasmids were transfected into HEK293T using PEI, 8 hours post-transfection, the medium was removed from the transfected cells and replaced with fresh medium. Virus was harvested from the HEK 293T cells 48 hours and 72 hours post-transfection, and concentrated with 40% PEG8000, store at −80 °C. To knock down of HEG in vitro, the lentiviral vectors were added to the culture media of HUVEC. The sequences of the shRNA are HEG-shRNA-1, 5’- CCGAGCATGTGAAGATGGATATAG-3’;

HEG-shRNA-2, 5’-ACCTTCGTGACAGAGTTTAAA-3’.

Liver Injury Models

For the CCl4 induced liver injury model, 8-week old male littermate mice were given with 1 mL/kg CCl4 (Rhawn, China, mixed with corn oil at a ratio of 1:4) via intraperitonial injection 3 times a week for 3 weeks. Control mice were given same volume of corn oil. The liver and serum samples were collected after the last injection of CCl4.

For the TAA-induced liver injury model, 8-week old female littermate mice were administered 300 mg/L thioacetamide (Solarbio, China) via the drinking water for 8 weeks; control mice were given sterilized water, after which the liver and serum samples were collected for analysis.

For the acetaminophen (APAP)-induced acute liver injury, 8-week old male littermate mice were administered 300 mg/kg APAP (Aladdin, China) via oral gavage after 12 hours of starvation. Control mice were given same volume of PBS. Liver and serum samples were taken at 24 hours or 48 hours post-APAP treatment.

For the BDL model, 8-week old male mice were anesthetized with avertin and subjected to a mid-abdominal incision about 3 cm long, the common bile duct was ligated with 4-0 silk in 2 adjacent positions approximately 1 cm apart. The bile duct was then severed by incision between the two sites of ligation. Sham group was operated similarly, except that the bile duct was not ligated. Liver tissues and serum samples were collected at 1, 2, and 3 weeks after BDL.

For all the injury models, the body and liver weight of the mice were measured. After perfusion with PBS from the portal vein, the livers were fixed in 4% PFA overnight, dehydrated in 100% ethanol, and embedded in paraffin. Sections (6-μm thick) were stained with H&E. Hepatic collagen contents were evaluated by Sirius red staining of paraffin-embedded sections. Liver function-related enzymes (ALT, AST, and ALP) and bilirubin (TBIL, direct bilirubin [DBIL], and indirect bilirubin [IBIL]) levels were measured by an automatic biochemical analyzer (AU5800, Beckman Colter Chemistry System analyzer).

TUNEL Staining

After dewaxing and hydration, tissue sections were boiled in 10-mM citrate buffer solution (pH 6.0) for antigen retrieval, then blocked with 10% donkey serum at room temperature, and incubated with appropriate primary antibodies overnight at 4°C. TUNEL was performed to detect apoptotic nuclei by using terminal deoxynucleotidyl transferase–mediated biotin-conjugated dUTP nick end labeling technique according to the manufacturer's protocol (Servicebio, China), nuclei were counterstained with hematoxylin.

Statistical Analysis

The data in this study are expressed as the mean ± standard deviation (SD) or mean ± standard error of the mean (SEM) as noted in individual figure legends. Statistical analyses were performed using GraphPad PRISM software, version 9.0. The unpaired Student t test were used to assess the differences between groups. Differences were considered statistically significant when P < .05, P < .01, or P < .001.

List of qPCR Primers

Acta2, Fwd 5’-TAGAACACGGCATCATCACCAA-3’, Rev, 5-CAGAGTCCAGCACAATACCAGT-3’;

Apc, Fwd 5’-CGAAAATGGGGTCCAAGGGTA-3’, Rev, 5-CGTAGTTTCACTCCGGGGAAA-3’;

Arg1, Fwd 5’-CGAAGCAAGCCAAGGTTAAAGCCAC-3’, Rev, 5-GGAGGCCTATCTTACAGAGAAGGTC-3’;

Axin2, Fwd 5’-CGGAAACAGCTGAAAACGGATT-3’, Rev, 5-AGGTAGAGACACTTGGCCATTG-3’;

β-Catenin , Fwd 5’-AAGGAAGCTTCCAGACATGC-3’, Rev, 5-AGCTTGCTCTCTTGATTGCC-3’;

Ccnd1, Fwd 5’-CGGCCCGAGGAGCTG-3’, Rev, 5-GGCCAGGTTCCACTTGAGC-3’;

Ces1b, Fwd 5’-AACGGCCAGTCTTCTCTTGAG-3’, Rev, 5-CTGGGGTCTCTTGTCAGACAC-3’;

Ces1d, Fwd 5’-CCCTCTTGGCTCCTTGAGATTT-3’, Rev, 5-CCATCACTGGTAGTCTGCTGTT-3’;

Ces1g, Fwd 5’-GCACACCACTTCCTACCCTC-3’, Rev, 5-AGGGGCACTCCATCATAGGT-3’;

Cyp1a2, Fwd 5’-ATCCTTTGTCCCCTTCACCAT-3’, Rev, 5-GGGAATGTGGAAGCCATTCA-3’;

Cyp4a14, Fwd 5’-GTGCCTGCTTACAGTGTCTCT-3’, Rev, 5-TGGAAGGCTGGAGTCAACATC-3’;

Cyp4a32, Fwd 5’-TGCTAGACAAATGGGAAAGGCT-3’, Rev, 5-ATCTGTGTGATCATGGGCAAGT-3’;

Cyp7a1, Fwd 5’-CACCATTCCTGCAACCTTCT-3’, Rev, 5-TTGGCCAGCATCTGTAATG-3’;

Cyp2e1, Fwd 5’-CACCGTGTTCCGAGGATATGTCATC-3’, Rev, 5-ACACACGCGCTTTCCTGCAGAAAAC-3’;

Cyp2b9, Fwd 5’-CTCTGGCCACCATGAAAGAGT-3’, Rev, 5-AGCAGATGATGTTGGCTGTGA-3’;

Cyp2b10, Fwd 5’-CACGGAGTTCCATCACCAGAA-3’, Rev, 5-CTGTGTGGCACTCCAATAGGT-3’;

Cyp2f2, Fwd 5’-TCGCTTCGACTATGACGATGAG-3’, Rev, 5-TATTGAAGTGGCTCAGTGGGTC-3’;

Col1a1, Fwd 5’-GCTCCTCTTAGGGGCCACT-3’, Rev, 5-ATTGGGGACCCTTAGGCCAT-3’;

Cps1, Fwd 5’-TCTGTTGCTGGTGAAGTGGTT-3’, Rev, 5-AGCCATTGACCCAGACTCTTG-3’;

Ck19, Fwd 5’-GATCGTCTCGCCTCCTACTTG-3’, Rev, 5-CAGGCGAGCATTGTCAATCTG-3’;

Ddr2, Fwd 5’-GGACATCACAGCCTCAAGTCA-3’, Rev, 5-AATTCAATGCCATGACCCCCT-3’;

Desmin, Fwd 5’-GATAGACGACCTGCAGAGGC-3’, Rev, 5-CATACTGAGCCCGGATGTCC-3’;

Ecad, Fwd 5’-CAGTGAAGCGGCATCTAAAGC-3’, Rev, 5-TTTTCATTTTCGGGGCAGCTG-3’;

F4/80, Fwd 5’-TGACTCACCTTGTGGTCCTAA-3’, Rev, 5-CTTCCCAGAATCCAGTCTTTCC-3’;

Fzd4, Fwd 5’-TCCAGCCAGCTGCAGTTC-3’, Rev, 5-TTGTGGTCGTTCTGGGGC-3’;

Fzd5, Fwd 5’-CACAGGTACCTAGCTTGTCGTT-3’, Rev, 5-GCACTCAGTTCCACACCAGATA-3’;

Fzd6, Fwd 5’-GCCGGAACCAAGAGAAGCT-3’, Rev, 5-TTCCAACCCAGAAGACCGC-3’;

Gls2, Fwd 5’-GGAGCGTATCCCTATCCACAAG-3’, Rev, 5-CGTGAACTCCTCAAAGTCAGGA-3’;

Glul, Fwd 5’-GCCAGGAGAAGAAGGGCTACTTTGA-3’, Rev,5-GAGAGGGATCACTGGAAGTCTAGTC-3’;

Hsd17, Fwd 5’-TAGTGCCAAGGACGAGGAGA-3’, Rev, 5-GCAGAGACACGAGGTTTGGA-3’;

IL1b, Fwd 5’-TGTAATGAAAGACGGCACACC-3’, Rev, 5-TCTTCTTTGGGTATTGCTTGG-3’;

IL6, Fwd 5’-CAAAGCCAGAGTCCTTCAGA-3’, Rev, 5-GATGGTCTTGGTCCTTAGCC-3’;

Lgr4, Fwd 5’-CGCCTTCACCCAAGCACTA-3’, Rev, 5-AGATGCCGCAACTGAACGA-3’;

Oat, Fwd 5’-CAATTACCATCCTTTGCCTGTA-3’, Rev, 5-GTACTGCCTGCCTTCCACAT-3’;

Rdh9, Fwd 5’-AGAGGCATGAGAGTGTTGGC-3’, Rev, 5-CTCGATCAAGCCCAACAGGT-3’;

Rnf43, Fwd 5’-AAGCTAATGCAGTCCCACCC-3’, Rev, 5-CAGCACCACTGGCTTAGTCA-3’;

Rspo3, Fwd 5’-GGGGGACTGAAACACGGG-3’, Rev, 5-GCTGCTGCTGCCTCTCTT-3’;

Tgfβ, Fwd 5’-CTCCCGTGGCTTCTAGTGC-3’, Rev, 5-GCCTTAGTTTGGACAGGATCTG-3’;

Timp1, Fwd 5’-CTTGGTTCCCTGGCGTACTC-3’, Rev, 5-ACCTGATCCGTCCACAAACAG-3’;

Wnt2, Fwd 5’-CCTGATGAACCTTCACAACAAC-3’, Rev, 5-TCTTGTTTCAAGAAGCGCTTTAC-3’;

Wnt9b, Fwd 5’-CTGGTGCTCACCTGAAGCAG-3’, Rev, 5-CCGTCTCCTTAAAGCCTCTCTG-3’;

Znf3, Fwd 5’-TCCAGCACCTCCCTCCTC-3’, Rev, 5-CGCTGCTCAGTGTGTCCA-3’.