Liver Endothelial Heg1 Signal Limits Hepatic Steatosis via Wnt-Mediated Fatty Acid Oxidation in Hepatocytes

Abstract

Background & Aims

Fatty liver diseases are highly prevalent worldwide, driven by hepatocyte metabolic dysfunction and alcohol consumption. Both cell-autonomous and non–cell-autonomous mechanisms contribute to hepatic lipid accumulation. Here, we investigated the role of liver endothelial cell–hepatocyte communication in regulating lipid metabolism.

Methods

The in vivo function of Heg1 in liver endothelial cells was investigated using 2 endothelial-specific knockout models, conditional (Lyve1-Cre;Heg^fl/fl) and inducible (Cdh5-CreERT2;Heg^fl/fl), in the context of 3 liver disease models: metabolic dysfunction-associated steatotic liver disease via a high-fat diet, metabolic dysfunction-associated steatohepatitis via a methionine–choline-deficient diet, and alcoholic liver disease via the National Institute on Alcohol Abuse and Alcoholism protocol. Subsequent transcriptomic and lipidomic analyses and tissue immunostainings were used to determine gene and protein expression levels and metabolic changes.

Results

EC-specific Heg1 deletion exacerbated hepatic steatosis under metabolic stress. Mechanistically, Heg1 deletion in liver endothelial cells downregulated bone morphogenetic protein signaling, leading to reduced expression of Wnt2, Wnt9b and Rspo3. This reduction attenuated Wnt signaling activation in hepatocytes, resulting in decreased expression of peroxisome proliferator–activated receptor α and fatty acid oxidation enzymes, ultimately promoting steatosis. Restoration of RSPO3 expression in endothelial cells via a conditional knock-in allele, reversed the steatotic phenotype. Treatment with bone morphogenetic protein proteins and SY-LB-35, a bone morphogenetic protein signaling activator, effectively restores RSPO3 expression.

Conclusions

Heg1 functions as an essential endothelial receptor mediating liver endothelial–hepatocyte crosstalk. By sustaining endothelial Wnt production, Heg1 maintains hepatocyte peroxisome proliferator–activated receptor α activity and fatty acid oxidation, thereby preserving hepatic lipid homeostasis and protecting against steatosis.

Graphical abstract

Summary.

Heg1 in liver endothelial cells promotes the production of secreted factors to maintain fatty acid oxidation in hepatocytes. Deletion of Heg1 in liver endothelial cells downregulates fatty acid oxidation and causes lipid accumulation in hepatocytes.

What You Need to Know.

Background

The contribution of endothelial–hepatocyte crosstalk to liver lipid metabolism is poorly understood. Heg1, a receptor selectively expressed in liver endothelial cells, has an unknown role in hepatic lipid metabolism.

Impact

Endothelial Heg1 regulates the bone morphogenetic protein–WNT–peroxisome proliferator–activated receptor α axis to sustain hepatocyte fatty acid oxidation, thereby preventing hepatic steatosis and revealing a novel endothelial target for therapy.

Future Directions

Future work should focus on identifying Heg1’s endogenous ligands, elucidating its mechanosensory–metabolic crosstalk, and validating the pathway’s function and target efficacy in human fatty liver diseases.

The liver is a vital organ responsible for maintaining systemic homeostasis through its roles in lipid and carbohydrate metabolism and xenobiotic biotransformation and detoxification, as well as the production of hormones, growth factors, and carrier proteins.¹ Fatty liver diseases affect a substantial portion of the global population.² Both metabolic dysregulation and alcohol consumption contribute to hepatic lipid accumulation and the development of fatty liver diseases.³ Despite significant advances in understanding lipid metabolism, effective treatments for fatty liver diseases remain limited.⁴

The liver is primarily composed of hepatocytes, endothelial cells (ECs), Kupffer cells, and hepatic stellate cells (HSCs) that act in concert and interact with other regulatory cell types to maintain liver function and homeostasis.¹ Liver ECs are a heterogeneous population comprising various subtypes with specialized functions, including arterial and central venous ECs, sinusoidal ECs, periportal and pericentral ECs, and lymphatic ECs.^5,6 Similar to the hepatocytes in the hepatic plate, liver ECs are spatially and functionally organized into metabolic zones based on their location relevant to portal and central veins.⁷

Endothelial–hepatocyte crosstalk plays a critical role in liver homeostasis. For example, endothelial cells secrete angiocrine factors to regulate hepatocyte metabolism and proliferation, liver zonation, and regeneration.8, 9, 10, 11, 12, 13, 14, 15 Liver sinusoidal endothelial cells (LSECs) prevent Kuffer cell and stellate cell activation in the homeostatic state.¹⁶ LSECs sense mechanical stresses and secrete angiocrine factors involved in neutrophil recruitment or regulation of bile acid production in hepatocytes.^17,18 LSECs contribute to hepatocyte steatosis through various pathologic mechanisms, including capillarization, uptake of oxidized low-density lipoprotein, and extracellular vesicle–mediated transport of signaling molecules.¹⁶

Our previous work demonstrated that Heg1, a type I transmembrane receptor in liver ECs, is an upstream regulator of Wnt ligands that control liver metabolic zonation.¹² Loss of Heg1 in liver ECs leads to expansion of zone 1 (periportal) and downregulation of xenobiotic metabolism enzymes in hepatocytes, thereby providing protections against drug-induced liver injury. In the present study, we show that Heg1 deficiency in liver ECs results in decreased fatty acid oxidation (FAO) in hepatocytes and increased hepatic lipid accumulation under metabolic stresses. Mechanistically, Heg1 modulates bone morphogenetic protein (BMP) signaling to regulate the expression of Wnt ligands and Wnt enhancer RSPO3 in ECs, which, in turn, maintains Wnt signaling activity in hepatocytes and sustains the expression of peroxisome proliferator–activated receptor α (PPARα) and FAO enzymes, thereby preventing hepatic steatosis.

Results

Loss of Heg1 in Liver Endothelial Cells Confers Hepatic Steatosis in Mice Under High-Fat Diet and Methionine–Choline-Deficient Diet Stresses

In our previous study, we generated mice lacking Heg1 in liver ECs (Heg1^LECKO) by crossing Lyve1-Cre mice with Heg1^fl/fl mice. We performed RNA sequencing (RNA-seq) of liver tissues to investigate the effects of endothelial loss of Heg1 on the molecular basis of liver functions. Gene Ontology (GO) and gene set enrichment analysis of liver transcriptomes from Heg1^LECKO and littermate control mice revealed that the fatty acid metabolism pathway was among the most significantly altered upon Heg1 deletion in ECs.¹² Based on these findings, we reasoned that endothelial Heg1 deficiency may influence hepatic lipid metabolism. However, histology and biochemical analysis of 2-month-old and 1-year-old mice under baseline conditions showed no significant differences in lipid storage between Heg1^LECKO and control livers (Figure 1A). Liver lipid contents were comparable between the 2 groups (Figure 1B). There were no significant differences in inflammatory cell infiltration or fibrosis between Heg1^LECKO and control livers (Figure 1C). Liver and plasma lipid contents were comparable between the 2 groups (Figure 1D). Liver function markers (aspartate aminotransferase [AST] and alanine aminotransferase [ALT]) remained within normal ranges in Heg1^LECKO mice (Figure 1E), and glucose tolerance was also unaffected (Figure 1F). These data suggest that endothelial Heg1 deletion does not significantly impact liver lipid metabolism under homeostatic conditions.

Figure 1.

Heg1-deficiency in liver endothelial cells do not affect liver lipids metabolism under unchallenged condition. (A) H&E and Oil Red O staining in liver sections of 8 weeks and 1 year old mice. Scale bar, 100 μm. (B) Hepatic levels of TG and TC of 1-year-old normal diet-fed mice (n = 5). (C) CD45 and Sirius red staining of liver sections of 1-year-old mice. Scale bar, 100 μm. (D) Serum concentrations of TC and TG in 8-week-old and 1-year-old normal diet-fed mice (n = 5). (E) Serum levels of AST and ALT of the 8-week-old and 1-year-old normal diet-fed mice (n = 6). (F) Glucose tolerance tests of normal diet-fed mice. Results shown were obtained from female mice. Data are presented as mean ± SD using unpaired Student’s t test. H&E, hematoxylin and eosin; SD, standard deviation.

To investigate whether endothelial Heg1 regulates liver metabolism during metabolic stress, we subjected Heg1^LECKO and wild-type littermates (control mice) to a high-fat diet (HFD) for 4 to 12 weeks. Compared with control mice, Heg1^LECKO mice exhibited prominent blood vessels on the liver surface, which is consistent with our previous findings that Heg1 deficiency in liver ECs resulted in a sparse hepatic vascular and biliary network. No other significant gross morphologic abnormalities were noted in the liver (Figure 2A). After fixation in 4% paraformaldehyde (PFA) and immersion in 30% sucrose overnight, control livers sank to the bottom of vials, whereas Heg1^LECKO livers were still floating on the top (Figure 2B), indicating Heg1^LECKO livers may have higher lipid content than controls. Hematoxylin and eosin and Oil red O staining confirmed the increased lipid accumulation in Heg1^LECKO livers (Figure 2C and D). Biochemical assays further demonstrated increased hepatic triglycerides (TGs), and total cholesterol (TC) levels in Heg1^LECKO mice compared with control mice (Figure 2E). Although no significant differences in serum lipid levels were observed between the 2 groups, male mice appeared to have higher serum lipid levels than female mice following HFD feeding (Figure 2F). The serum AST and ALT levels and glucose tolerance also remained normal (Figure 2G and H). CD45 and Sirius Red staining on liver tissue sections of mice fed an HFD for 8 weeks demonstrated increased inflammatory cell infiltration and collagen deposition in Heg1^LECKO livers (Figure 2I and J).

Figure 2.

Deletion of Heg1 in liver ECs aggravates HFD-induced lipid accumulation. (A) Liver images of HFD-fed mice after PBS perfusion. Scale bar, 5 mm. (B) PFA-fixed liver tissues of HFD-fed mice in a 30% sucrose solution. Black arrow, liver of control mice; red arrow, liver of Lyve1-Cre;Heg1^fl/fl mice. (C) H&E and Oil Red O staining of liver sections from the HFD-fed control and Lyve1-Cre;Heg1^fl/fl mice. Scale bar, 100 μm. (D) Lipid droplet quantification of Oil red O staining (n = 5). (E) Hepatic levels of TG and TC of HFD-fed mice (n = 10). (F) Serum TG and TC levels of mice fed with normal diet and HFD (n = 6). (G) Serum levels of AST and ALT fed with normal diet and HFD (n = 6). (H) Glucose tolerance tests of HFD-fed mice. (I) CD45 and Sirius Red staining of liver sections of HFD-fed mice. Scale bar, 100 μm. (J) Quantification of CD45-positive cells and Sirius Red-stained collagen area (n = 6). Results include data from both male and female mice. Data are presented as mean ± SD using unpaired Student’s t test. ∗P < .05, ∗∗P < .01, ∗∗∗P < .001, ∗∗∗∗P < .0001. H&E, hematoxylin and eosin; SD, standard deviation.

Although HFD feeding models metabolic dysfunction-associated steatotic liver disease (MASLD), a methionine–choline-deficient diet (MCD) is widely used to induce metabolic dysfunction-associated steatohepatitis (MASH) mainly by impairing hepatic lipid export.¹⁹ We therefore subjected Heg1^LECKO and control mice to MCD feeding for 2 or 4 weeks. After fixation with 4% PFA and overnight immersion in 30% sucrose, control livers sank, whereas Heg1^LECKO livers remained afloat (Figure 3A). Livers from Heg1^LECKO mice exhibited dramatically increased lipid accumulations and severe steatosis compared with controls (Figure 3B and C). Hepatic TG and TC levels were significantly elevated in Heg1^LECKO mice (Figure 3D). As expected, MCD feeding reduced plasma levels of TGs and TC in mice. However, no significant differences in serum TG and TC levels were observed between Heg1^LECKO and control mice (Figure 3E). MCD feeding led to elevated serum AST and ALT levels, but these were comparable between Heg1^LECKO and control mice (Figure 3F). CD45 staining indicated the Heg1^LECKO liver has aggravated hepatic inflammatory responses, whereas no obvious collagen accumulation was observed (Figure 3G). These findings indicate that deletion of Heg1 in liver ECs promotes diet-induced hepatic steatosis.

Figure 3.

Deletion of Heg1 in liver ECs aggravates MCD-induced lipid accumulation. (A) PFA-fixed liver tissues of MCD-fed mice in a 30% sucrose solution. Black arrow, liver of control mice; red arrow. liver of Lyve1-Cre;Heg1^fl/fl mice. (B) H&E and Oil Red O staining of liver sections from the MCD-fed control and Lyve1-Cre;Heg1^fl/fl mice. Scale bar, 100 μm. (C) Lipid droplet quantification of Oil red O staining (n = 6). (D) Hepatic levels of TG and TC of MCD-fed mice (n = 5). (E) Serum TG and TC levels of normal diet and MCD-fed mice (n = 5). (F) Serum levels of AST and ALT of normal diet and MCD-fed mice (n = 6). (G) CD45 and Sirius Red staining of liver sections of MCD-fed mice and quantification of CD45-positive cells (n = 6). Scale bar, 100 μm. Results include data from both male and female mice. Data are presented as mean ± SD using unpaired Student’s t test. ∗P < .05, ∗∗P < .01, ∗∗∗P < .001, ∗∗∗∗P < .0001. H&E, hematoxylin and eosin; SD, standard deviation.

Loss of Heg1 in Liver Endothelial Cells Downregulates Fatty Acid Oxidation Enzyme Expression Under Metabolic Stresses

To investigate the molecular mechanisms underlying hepatic lipid accumulation in Heg1^LECKO mice, we performed pathway enrichment analysis of differentially expressed genes (DEGs) from the liver RNA-seq dataset. This analysis revealed that the gene sets for lipid metabolism, lipoprotein transportation, PPAR signaling and beta-oxidation were among the top altered pathways (Figure 4A). Among the downregulated genes within these gene sets shown in Figure 4B, many encode key enzymes involved in distinct steps of FAO, including Cact, Acadvl, Acadl, Acadm, Ech1, Hadh, Ehhadh, Acaa1, and Acaa2 (Figure 4C). qPCR analysis confirmed a significant downregulation of FAO-related genes in Heg1^LECKO livers under both HFD and MCD conditions. In contrast, the expression of genes involved in lipid uptake, lipogenesis, and export remained unchanged (Figure 4D and E).

Figure 4.

Heg1 deficiency in liver ECs downregulates genes involved in fatty acid oxidation. (A and B) Transcriptomic analysis of the livers of HFD-fed Lyve1-Cre;Heg1^fl/fl and littermate control mice. (A) GO and Wiki Pathway analysis. (B) Gene expression heatmap (the genes in the FAO pathway are displayed in red fonts). (C) Schematic of FAO steps and enzymes. (D and E) qPCR analysis showing the change of mRNA expression levels of FAO enzymes and mRNA levels of genes for lipid uptake, lipogenesis, and export in livers of mice fed with HFD or MCD for 2 weeks. Results include data from both male and female mice (n = 7). (F) Relative amounts of acylcarnitine species of various carbon chain lengths in livers of male mice feed with HFD for 8 weeks (n = 4). (G) Representative transmission electron microscope images of liver sections from male mice fed with the MCD diet. Scale bar, 500 nm. (H) Liver MDA content of male mice fed with MCD for 2 weeks (n = 5). Data are presented as mean ± SD using unpaired Student’s t test. ∗P < .05, ∗∗P < .01, ∗∗∗P < .001, ∗∗∗∗P < .0001. MDA, malondialdehyde; SD, standard deviation.

In the process of FAO, fatty acids are first conjugated with carnitine to form acylcarnitine, which is then transported to mitochondria for β-oxidation and energy production. Impaired FAO disrupts this metabolic flux, resulting the accumulation of acylcarnitine species. To assess whether reduced FAO enzyme expression in Heg1^LECKO liver leads to acylcarnitine accumulation, we performed gas chromatography-mass spectrometry to profile acylcarnitine in the liver tissues. Acylcarnitines of varying chain lengths were significantly elevated in Heg1^LECKO livers after HFD feeding for 8 weeks (Figure 4F), which consistently demonstrated an impairment in FAO.

Next, we examined the ultrastructure of the hepatic mitochondria using a transmission electron microscope. After 5 days of MCD diet exposure, control mice showed mild mitochondrial swelling, whereas livers of Heg1^LECKO mice displayed severe mitochondrial abnormalities, including membrane rupture, disorganized or fragmented cristae, increased mitochondria fragmentation, and a higher proportion of small-sized mitochondria (diameter <0.5 μm), indicating exacerbated mitochondrial injury (Figure 4G). Consistent with these structural abnormalities, levels of malondialdehyde (MDA), a marker of lipid peroxidation and oxidative stress, were significantly increased in Heg1^LECKO liver (Figure 4H).

These results indicate that the observed lipid accumulation is primarily due to reduced FAO rather than increased lipid uptake or synthesis.

Peroxisome Proliferator–Activated Receptor α Mediates Heg1-Dependent Regulation of Fatty Acid Oxidation

PPARα is the predominant isoform of peroxisome proliferator–activated receptors (PPARs) in the liver to regulate lipid metabolism²⁰ (Figure 5A). RNA-seq analysis of downregulated genes in Heg1^LECKO liver identified the PPAR signaling pathway as one of the most significantly enriched pathways (Figure 4A). Fasting stimulates Ppara expression in the liver, but this induction was impaired in Heg1^LECKO livers. Quantitative polymerase chain reaction (qPCR) analysis revealed a significant downregulation of fasting-induced Ppara messenger RNA (mRNA) in the livers of Heg1^LECKO mice under MCD treatment, but not under HFD treatment (Figure 5B). At the protein level, PPARα was markedly decreased in livers of Heg1^LECKO mice following HFD or MCD treatment, whereas no significant difference was observed between livers of Heg1^LECKO and littermate control mice under normal chow conditions (Figure 5C).

Figure 5.

Heg1 deficiency in liver ECs inhibits FAO by down-regulating PPARα. (A) Schematic representation of PPARα regulation of FAO enzyme expression. (B) mRNA expression of Ppara in the livers in response to fasting (n = 4), HFD (n = 8), and MCD treatment (n = 7). (C) Western blot and quantification of PPARα expression in the livers of male HFD- or MCD-fed control and Lyve1-Cre;Heg1^fl/fl mice (n = 6). (D) Representative H&E and Oil Red O histologic stains of livers sections from male control and Lyve1-Cre;Heg1^fl/fl mice treated with fenofibrate or vehicle. Scale bar, 100 μm. (E) mRNA expression of Ppara and PPARα target genes in the liver treated with fenofibrate (n = 3). Data are presented as mean ± SD using unpaired Student’s t test. ∗P < .05, ∗∗P < .01, ∗∗∗P < .001, ∗∗∗∗P < .0001. H&E, hematoxylin and eosin; SD, standard deviation.

To determine whether PPARα downregulation is functionally responsible for impaired FAO in Heg1^LECKO livers, we treated mice with a PPARα agonist, fenofibrate. Fenofibrate treatment effectively prevented lipid accumulation in the MCD-fed mice (Figure 5D). Moreover, it significantly upregulated the expression of key FAO genes. However, the induction of FAO genes such as Cpt1b, Cact, and Ech1 was substantially attenuated in Heg1^LECKO livers compared with controls after 2 weeks of MCD plus fenofibrate treatment (Figure 5E), indicating a blunted transcriptional response to PPARα activation.

Wnt Signaling Mediates Heg1-Depenent Regulation of Hepatic Lipid Metabolism

To identify potential liver EC–derived signals that regulate PPAR expression and subsequently FAO in hepatocytes, we adopted a 2-stage method,⁷ isolation of nonparenchymal cells with centrifugation separation followed by CD31 MicroBeads enrichment, to isolate liver ECs (Figure 6A). qPCR analysis was performed to examine the endothelial enrichment efficiency. The expression levels of EC marker genes (Pecam1, Lyve1, and Tie1) were approximately 100-fold higher in isolated CD31-positive ECs compared with starting dissociated liver cells (Figure 6B). We performed RNA-seq analysis of isolated liver ECs from neonatal (P3) and adult (P42) livers (Figure 6C and D). Gene expression profiles analysis identified a total of 37 genes encoding secreted proteins with differential expression at both developmental stages (Figure 6D). Among these genes, 28 were downregulated, and 9 were upregulated in Heg1-deficient ECs. Notably, Wnt ligands and enhancer (including Wnt2, Wnt9b, Rspo3), which were previously reported to be downregulated in Heg1-deficient liver ECs,¹² were among the top 10 differentially expressed secreted factors (Figure 6D). Consistently, qPCR analysis further validated the downregulated expression of Wnt2, Wnt9b, and Rspo3 in ECs isolated from Heg1^LECKO mice (Figure 6B).

Figure 6.

Deletion of Heg1 in liver ECs downregulates Wnt2/Wnt9b/Rspo3 expression. (A) Schematic of mouse liver EC isolation for transcriptomic analysis. (B) qPCR analysis verifying the mRNA expression levels of endothelial markers and in nonparenchymal cells and CD31-microbeads enriched liver ECs vs total cells (n = 3). (C) Volcano plot showing the top upregulated and downregulated genes in liver ECs of postnatal day 3 Heg1-deficient mice. Horizontal dashed line indicates −log1Pvalue = 1.3, and vertical dashed lines represent |log₂fold change| = .58. (D) Conjoint analysis of RNA-seq results of liver ECs isolated from P3 and P42 mice. The top 10 secretory protein coding genes were listed. Results shown are obtained from male mice. Data are presented as mean ± SD using unpaired Student’s t test. ∗P < .05, ∗∗P < .01. SD, standard deviation.

RSPO3 Knockin Reverses Liver Steatosis Phenotype in Heg1-Deficient Mice

Because we have found that loss of Heg1 in liver ECs downregulated the expression of secreted factors Wnt2, Wnt9b and Rspo3, recovering the expression levels of these factors in ECs would be necessary to confirm that these endothelial-derived Wnt ligands are the critical molecules transducing signal to regulating FAO and lipid accumulation in hepatocytes. Thus, we chose to generate conditional RSPO3 knockin mice by inserting coding sequence of human RSPO3 into the Rosa26 allele (RSPO3^KI) (Figure 7A) for the flexibility of inducing RSPO3 expression in a time- or cell type–specific manner. We crossed RSPO3^KI with Lyve1-Cre mice to generate mice with selective expression of RSPO3 in liver ECs. Lyve1-Cre;RSPO3^KI mice died shortly after birth due to the lung inflation defects. The patterning of liver sinusoid system appeared normal, as assessed by lymphatic vessel endothelial hyaluronan receptor 1 (Lyve1) immunostaining.

Figure 7.

Knockin expression of RSPO3 in ECs reverses hepatic steatosis and zonation. (A) Generation of the Rosa26-LSL-hRSPO3 mice with CAG-LSL-hRSPO3-IRES-tdTomato-WPRE-pA cassette knocked-in at Rosa26 gene locus via CRISPR/Cas9 technology. (B) qPCR analysis of hepatic Wnt pathway target gene expression after induced RSPO3 expression (n = 4). (C) The Rosa26-LSL-hRSPO3 Cas9 knockin mice were crossed with Cdh5CreERT2 or Cdh5CreERT2;Heg1^fl/fl mice to generate Cdh5CreERT2;RSPO3^KI/+ or Cdh5CreERT2;Heg1^fl/fl;RSPO3^KI/+ mice for tamoxifen induction and MCD treatment. (D) H&E and Oil Red O staining in liver sections of MCD-fed control, Cdh5CreERT2;Heg1^fl/fl and Cdh5CreERT2;Heg1^fl/fl;RSPO3^KI/+ mice, results shown are obtained from female mice. Scale bar, 100 μm. (E) Hepatic levels of TG and TC of MCD-fed mice (n = 5). (F) Western blots and quantification of PPARα expression in the livers of control, Cdh5CreERT2;Heg1^fl/fl, and Cdh5CreERT2;Heg1^fl/fl;RSPO3^KI/+ mice (n = 4). (G and H) Coimmunostaining and quantification of GS and ECAD, as well as immunostaining and quantification of PPARα, on liver sections of MCD-fed mice (n = 4). The dashed lines mark the edge of ECAD⁺ zones. Scale bar, 200 μm. (I and J) Coimmunostaining and quantification of GS and ECAD, as well as immunostaining and quantification of PPARα, on liver sections from control mice and Cdh5CreERT2;RSPO3^KI/+ mice following tamoxifen induction and MCD treatment (n = 3). The dashed lines mark the edge of ECAD⁺ zones. Scale bar, 200 μm. Data are presented as mean ± SD using unpaired Student’s t test or 1-way ANOVA. ∗P < .05, ∗∗P < .01, ∗∗∗P < .001, ∗∗∗∗P < .0001. H&E, hematoxylin and eosin; SD, standard deviation.

To bypass the neonatal lethality, we crossed RSPO3^KI mice with an inducible endothelial-specific Cre line (Cdh5-CreERT2) to generate Cdh5-CreERT2;RSPO3^KI mice. Following induction of RSPO3 expression with tamoxifen, the expression of Wnt targets genes were significantly upregulated in Cdh5-CreERT2;RSPO3^KI livers (Figure 7B).

To determine if the expression RSPO3 in ECs could reverse the phenotype caused by Heg1-dificiency, we crossed mice carrying RSPO3^KI allele with Cdh5-CreERT2;Heg1^fl/fl mice to generate Cdh5-CreERT2;Heg1^fl/fl;RSPO3^KI mice. In this model, we can induce Heg1 gene deletion and RSPO3 gene expression with tamoxifen treatment after mice reach maturity (Figure 7C). Under MCD diet conditions, induced RSPO3 expression in Heg1-deficiency background (Cdh5-CreERT2;Heg1^fl/fl;RSPO3^KI) reversed the lipid accumulation observed in Cdh5-CreERT2;Heg1^fl/fl (Heg1^iECKO) mice (Figure 7D and E). Western blot analysis further confirmed that PPARα protein levels, reduced in Heg1-deficient livers, were restored upon RSPO3 induction (Figure 7F). Notably, this rescue also extended to restoring metabolic zonation patterning and pericentral PPARα expression (Figure 7G and H). Induction of RSPO3 alone in the endothelium of Cdh5-CreERT2;RSPO3^KI mice, which were fed a MCD diet for 10 days, led to a decreased zone1 area and increased PPARα expression in zones 2/3 (Figure 7I and J), supporting roles for RSPO3 in modulating liver metabolic zonation and PPAR⍺ expression.

Loss of Heg1 Downregulates Bone Morphogenetic Protein Signaling in Endothelial Cells

To uncover the molecular mechanism by which Heg1 regulates Wnt2/Wnt9b/Rspo3 expression in liver ECs, we performed GO analysis of down-regulated genes from transcriptomic profiling of isolated liver ECs. The top enriched pathways include transforming growth factor β/BMP and extracellular regulated protein kinase (ERK)1/2 signaling (Figure 8A). The downregulated genes involved in BMP signaling include Bmp2, Ltbp1/4, Tgfbr2, and others (Figure 8B). The reduced expression of these genes in Heg1-deficient liver ECs can be confirmed with qPCR analysis (Figure 8C). In addition, the canonical BMP target genes Id1, Id2, and Id3 are all downregulated (Figure 8D), further supporting the conclusion that BMP signaling is impaired in Heg1-deficient liver ECs.

Figure 8.

Heg1 deficiency in liver ECs downregulates BMP and ERK signaling. (A) GO analysis of DEGs in liver endothelial cells of Lyve1-Cre;Heg1^fl/fl and littermate control mice at 3 days after birth. (B) Heatmap showed the downregulated genes in BMP/TGFβ receptor signaling pathway. (C) qPCR analysis confirms the downregulated genes in BMP/TGFβ receptor signaling pathway (n = 4). (D) mRNA expression of the inhibitor of DNA binding (ID)1-3, known BMP/TGFβ transcriptional targets (n = 4). (E) Western blots and quantification of ERK1/2, ERK5, and their phosphorylation expression levels in liver ECs of Lyve1-Cre;Heg1^fl/fl and control mice (n = 6). (F) qPCR analysis of mRNA levels of Mapk1/3 (genes encoding ERK1/2) (n = 5). (G) qPCR quantification of the expression of ERK1/2 target genes (n = 6). Results shown are obtained from male mice. Data are presented as mean ± SD using unpaired Student’s t test. ∗P < .05, ∗∗P < .01, ∗∗∗P < .001, ∗∗∗∗P < .0001. SD, standard deviation; TGFβ, transforming growth factor β.

For components of ERK1/2 signaling, total ERK1/2 protein levels were dramatically reduced in Heg1^LECKO liver ECs, whereas phosphorylated ERK1/2 (p-ERK1/2) levels showed a modest decrease (Figure 8E). In support of these findings, qPCR analysis verified that the mRNA expression of Mapk1/3 (encoding ERK1/2) was correspondingly downregulated in Heg1^LECKO liver ECs (Figure 8F). qPCR analysis also confirmed that several downstream targets of ERK1/2 signaling including Kit, Cav-1, Gas6, and Tiam1, were significantly downregulated in Heg1^LECKO liver ECs (Figure 8G).

Given that Heg1 is a type I transmembrane receptor, it may function in concert with other receptors in the cell membrane to modulate receptor-mediated signaling pathways, such as activin receptor-like kinase 1 (Alk1) mediated BMP signaling or Tie1/2 mediated angiopoietin signaling. It has been reported that Alk1 activation by BMP ligands promotes the expression of Wnt2, Wnt9b, and Rspo3 in liver ECs, whereas the deletion of Alk1 downregulates these genes and upregulates Prnd and Pgf expression.²¹ Consistently, small interfering RNA (siRNA)-mediated knockdown of HEG1 in cultured ECs (human umbilical vein endothelial cells [HUVECs]) decreased nuclear/cytoplasmic distribution ratio of SMAD family member (SMAD) 4, indicating reduced BMP signaling activation level (Figure 9A). HEG1 knockdown in HUVECs downregulates RSPO3 expression. The treatment of ECs with BMP9, but not BMP2, increased RSPO3 expression. The increase was attenuated in HEG1-knockdown ECs (Figure 9B and C). Treating HUVECs with SY-LB-35, a BMP signaling activator,²² induced nuclear levels of p-SMAD1/5/8 and SMAD4, and the effects were also attenuated with HEG1 knockdown (Figure 9D). Accordingly, HEG1 knockdown reduces the elevation of RSPO3 expression induced by SY-LB-35 treatment (Figure 9E). These data further support the role of HEG1 in facilitating BMP signaling.

Figure 9.

Heg1 deficiency in liver ECs downregulates Wnt ligand production through BMP signaling pathway. (A) Immunostaining and quantification of SMAD4 nuclear/cytoplasmic distribution in HUVEC following HEG1 gene knockdown with siRNA (n = 3). Scale bar, 10 μm. (B) HEG1 gene knockdown with siRNA attenuates the stimulated RSPO3 expression by BMP9 in HUVECs (n = 6). (C) BMP2 treatment does not affect RSPO3 expression in control or si-HEG1 knockdown HUVECs (n = 6). (D) Western blots and quantification of p-SMAD1/5/8, SMAD4, and Histone H3 in si-NC/si-HEG1-transfected HUVECs with or without SY-LB-35 treatment (n = 4). (E) qPCR analysis of HEG1 and RSPO3 mRNA levels in HUVECs treated with vehicle or SY-LB-35 (n = 5). (F) Western blots and quantification of p-Smad1/5/8 normalized to histone H3 in nuclear fractions of liver tissues from wild-type mice treated with vehicle or SY-LB-35 (n = 3). (G and H) Coimmunostaining and quantification of GS and E-Cadherin to indicate the zonal restoration in Cdh5-CreERT2;Heg1^fl/fl when treated with SY-LB-35 (n = 3). The dash lines mark the edge of E-Cad⁺ zones. Scale bar: 200 μm. Data are presented as mean ± SD using unpaired Student’s t test. ∗P < .05, ∗∗P < .01, ∗∗∗P < .001, ∗∗∗∗P < .0001. SD, standard deviation; si, small interfering; siRNA, small interfering RNA.

Consistently, in vivo administration of SY-LB-35 increased the level of p-Smad1/5/8 in nuclear fraction of liver cells (Figure 9F). In vivo administration of SY-LB-35 in Heg1^iECKO mice reversed the altered metabolic zonation phenotype, reducing the expanded zone 1 area (marked by E-cadherin [ECAD] staining) and restoring zone 2 area (area between ECAD and glutamine synthetase [GS] staining) in livers (Figure 9G and H). These findings suggest that Heg1 positively regulates BMP signaling in liver ECs, which, in turn, modulates Wnt ligand expression and hepatocyte metabolic zonation.

Liver Endothelial Deletion of Heg1 Exacerbates Fatty Liver in an Alcoholic Liver Disease Model

FAO is a central mechanism regulating hepatic metabolism, and its impairment is a key contributor to liver metabolic syndromes. Chronic alcohol consumption has been shown to suppress FAO and promote hepatic steatosis.^23,24 To determine if the role of Heg1-mediated signaling in FAO regulation and lipid accumulation extends beyond specific metabolic stresses, we subjected mice with the National Institute on Alcohol Abuse and Alcoholism model of alcoholic liver disease.²⁵ (Figure 10A). Similar to what we observed from the HFD and MCD models, Heg1^LECKO mice exhibited drastically increased lipid accumulation following ethanol exposure compared with littermate controls (Figure 10B–E). Importantly, this alcohol-induced increase of hepatic lipid accumulation in Heg1^iECKO mice could be reversed by increased expression of RSPO3 (Figure 10F and G).

Figure 10.

Deletion of Heg1 in liver ECs aggravates EtOH-induced lipid accumulation. (A) Chronic alcohol-fed alcoholic liver disease mouse model (NIAAA model). (B) PFA-fixed liver tissues of EtOH-fed mice in a 30% sucrose solution. (C) H&E and Oil Red O staining of liver sections from Lyve1Cre;Heg1^fl/fl and control mice. Scale bar, 100 μm. (D) Hepatic levels of TG and TC of EtOH-fed mice (n = 5). (E) Serum TG and TC level of EtOH-fed mice (n = 6). (F) Representative H&E and Oil Red O histologic stains of liver sections from control, Cdh5-CreERT2;Heg1^fl/fl, and Cdh5-CreERT2;Heg1^fl/fl;RSPO3^KI/+ mice fed with EtOH. Scale bar: 100 μm. (G) Hepatic levels of TG and TC of EtOH-fed mice (n = 4). Results shown are obtained from male mice. Data are presented as mean ± SD using unpaired Student’s t test or 1-way ANOVA. ∗P < .05, ∗∗P < .01, ∗∗∗P < .001, ∗∗∗∗P < .0001. EtOH, ethanol; H&E, hematoxylin and eosin; NIAAA, National Institute on Alcohol Abuse and Alcoholism.

Together, these findings demonstrate that Heg1-initiated signaling in liver ECs plays a critical role in maintaining lipid homeostasis in hepatocytes. Our study uncovered a previously unrecognized endothelial–hepatocyte communication axis essential for protecting the liver from lipid accumulation under metabolic stress.

Discussion

Our study investigated the cellular communication between ECs and hepatocytes in the development of fatty liver diseases (Figure 11). We found that the absence of endothelial Heg1 leads to increased diet-induced steatosis in hepatocytes, primarily due to reduced FAO. This effect is mediated by diminished secretion of Wnt ligands by ECs. Overexpression of RSPO3 in ECs or administration of a Wnt signaling activator in vivo significantly attenuated steatosis, revealing a novel intercellular communication pathway regulating lipid metabolism between ECs and hepatocytes.

Figure 11.

A schematic diagram depicts endothelial Heg1 modulating hepatocyte lipid metabolism via the TGFβ/BMP-WNT signaling axis. Endothelial Heg1 functions as a receptor that senses extracellular cues to potentiate TGFβ/BMP signaling, thereby sustaining the transcription of Wnt2, Wnt9b, and Rspo3. These endothelial-derived WNT ligands engage WNT receptors on hepatocytes to activate WNT signaling. In hepatocytes, WNT pathway activation upregulates PPARα, which in turn induces the expression of FAO enzymes. Deletion of Heg1 in liver ECs dampens TGFβ/BMP signaling and reduces transcription of Wnt2, Wnt9b, and Rspo3. The diminished secretion of these WNT ligands wakens hepatocyte WNT signaling, leading to reduced PPARα expression and suppression of downstream FAO gene programs. Consequently, impaired FAO capacity predisposes the liver to TG accumulation and lipid droplet (LD) formation, thereby promoting hepatic steatosis under metabolic stress. SD, standard deviation; TGFβ, transforming growth factor β.

Wnt2, Wnt9b, and Rspo3 are essential angiocrine or stellacrine factors for the establishment and maintenance of liver metabolic zonation and act as potent promoters of liver regeneration.^7,14,26,27 These Wnt ligands are predominantly expressed in central vein ECs.^{7,12,14,28,29} A recent report demonstrated RSPO3 from HSCs is also required for liver metabolic zonation.²⁶ Despite their importance in liver homeostasis and regeneration, the upstream signaling regulating their expression remain incompletely understood. Augustin et al have demonstrated that the activation state of Tie1 coordinates the balance of Stat3 and Foxo1 in liver ECs to regulate Wnt9b expression.⁷ Additionally, Alk1 mediates BMP signaling to promote Wnt ligands expression in ECs.²¹ We previously showed that Heg1 regulates Wnt expression in central vein ECs.¹² However, how Heg1, as a receptor, modulates Wnt transcription remains unclear. Here, we demonstrated that the loss of Heg1 can tune down BMP signaling in ECs, whereas treatment with BMP or BMP agonist increased Wnt/Rspo3 expression. These findings suggest that Heg1, a transmembrane protein, may regulate Wnt expression by modulating receptor-mediated BMP signaling. The precise interaction between Heg1 and BMP receptors requires further investigation.

In cultured HUVECs, BMP9, but not BMP2, induced RSPO3 expression, and this induction was attenuated by HEG1 knockdown, suggesting that HEG1 modulates BMP9-dependent endothelial transcriptional programs. In the liver, stellate cells are the major source of BMP9 and BMP10,³⁰ raising the possibility of paracrine crosstalk between stellate cells and liver ECs. In this context, stellate cell–derived BMP9/10 may sustain endothelial WNT/RSPO3 production, thereby contributing to the maintenance of metabolic zonation. Conversely, RSPO3 may function as a negative autocrine regulator of BMP signaling in liver ECs. Beyond its established role as a WNT pathway enhancer via formation of complexes with ZNRF3/RNF43 and LGR receptors, RSPO3 has been reported to act as a BMP antagonist by binding BMP receptors and ZNRF3, promoting BMP receptor turnover and attenuating downstream signaling.^31,32 These findings suggest a BMP–BMPR–RSPO3 feedback loop, in which BMP9 induces RSPO3 via BMP receptors and RSPO3, in turn, restrains BMP receptor availability, thereby fine-tuning signaling amplitude. HEG1 potentially functions as a key modulator of BMP responsiveness and signaling balance.

Although abundant evidence demonstrates that Wnt signaling is essential for liver metabolic zonation, its precise role in liver lipid metabolism remains unclear. β-Catenin knockout mice exhibit increased liver steatosis following ethanol exposure,³³ whereas transgenic expression of stabilized β-catenin in hepatocytes enhances steatosis under HFD conditions.³⁴ Here, we reported that under both HFD and alcohol consumption conditions, maintaining appropriate Wnt signaling levels is necessary to prevent excessive lipid accumulation during metabolic stress. In our Heg1 loss-of-function model, we found that endothelial-derived Wnt/Rspo were required to sustain PPARα expression and FAO under stress conditions. These findings underscore endothelial Wnt in supporting healthy hepatocyte lipid metabolism. Sugimoto et al reported that Rspo3 from HSCs is also important for liver zonation; however, deletion of Rspo3 in HSCs does not alter canonical target gene expression in hepatocytes adjacent to central veins, such as GS and ornithine aminotransferase.²⁶ This suggests distinct roles for EC- and HSC-derived Rspo3. It is possible that the combination of Wnt2, Wnt9b, and Rspo3, along with their interactions with specific hepatocyte receptors, are crucial. Further investigation is needed to define the receptors in hepatocyte that transduce Wnt signaling to regulate PPARα expression.

The metabolic processes are spatially organized along the portal–central axis, with hepatocytes arranged into distinct zones resemble specialized metabolic “assembly lines.”^29,35, 36, 37, 38 FAO predominantly occurred in zone 1 hepatocytes, whereas Wnt ligands are from central vein ECs surrounded by zone 3 hepatocytes. Notably, our Oil red O staining revealed that the increased lipid accumulation mainly occurred in zone 2 hepatocytes. Mouse liver lobules contain approximately ∼20 layers of hepatocytes, distributed as ∼10 layers in zone 1, ∼7 layers in zone 2, and ∼3 layers in zone 3. Understanding how the reduction of Wnt/Rspo expression in central vein ECs can be relayed across zone 3 to zone 2 and possibly zone 1 is an important and open question.

BMPs, members of the transforming growth factor β/BMP superfamily, are significant regulators of liver metabolism. For example, BMP8b promotes the progression of MASH by enhancing hepatic steatosis and inflammatory responses,³⁹ whereas BMP9 levels are decreased in patients with MASLD and model mice.⁴⁰ Bmp9-deficient mice exhibit reduced PPARα expression and impaired FAO.⁴⁰ In cultured hepatocytes, recombinant BMP9 activates Smad to enhance the PPARα promoter activity, thereby reducing TG accumulation. In vivo, recombinant BMP9 treatment alleviates hepatic steatosis and obesity in HFD-fed mice.⁴¹ BMP9 and BMP10 are primarily expressed in HSCs and Rspo3 expression in ECs via a paracrine mechanism.^30,42 This study demonstrates the role of Heg1 in regulating the Wnt/Rspo3 expression by modulating BMP9/10 signaling, most likely through Alk1 receptor on endothelial membranes. Detailed molecular and signaling interactions between Heg1 and BMP receptors could be important for developing targeted interventions to prevent hepatosteatosis. Furthermore, a BMP9–PPAR⍺–fibroblast growth factor 21 regulation network was proposed.^40,43 Given that fibroblast growth factor 21 is a key metabolic regulator, investigating its involvement in the Heg1–BMP9–PPARα axis, as well as its potential role in mediating crosstalk between Wnt/Rspo3 and BMP signaling pathways, could reveal novel insights into liver metabolic zonation and lipid homeostasis.

Here, we showed Heg1-deficiency in liver ECs impairs lipid homeostasis under stress conditions. A critical question remains regarding the ligand(s) of Heg1, or the extracellular signals it senses. Heg1 is a transmembrane receptor characterized by a relatively large, highly glycosylated extracellular domain and a short intracellular tail that interacts with Ccm1 and Rasip1.^44,45 The roles of these interacting proteins in liver lipid metabolism remain unreported. Recent studies on the cardiovascular system demonstrated that Heg1 functions as a mechanosensor for shear stress, regulating the expression of flow-inducible transcriptional factors KLF2/4.⁴⁶ In the liver, deletion of endothelial Piezo1 leads to the increased expression of Cyp7a1, a key enzyme for bile acid synthesis, resulting in elevated bile production and decreased hepatic TC and TG levels. This finding suggests that mechanic sensing by liver ECs influences liver lipid metabolism.¹⁸ Similarly, it is plausible that the flow conditions sensed by central vein ECs are transduced to hepatocytes, significantly contributing to lipid homeostasis. An important question is whether flow-induced signal detected by central vein ECs and transduced via Heg1-mediated mechanosensing synergize with metabolic pathways like Wnt/Rspo3 to maintain hepatic lipid homeostasis. For instance, whether shear stress–induced KLF2/4 regulates endothelial Wnt/Rspo expression or directly impacts key lipid metabolism regulators, such as hepatocyte PPARα, warrants further investigation. Future studies should focus on identifying Heg1’s endogenous ligands and exploring the crosstalk between mechanosensory and metabolic pathways.

In the cardiovascular system, Heg1 plays critical roles in endothelium by regulating cardiac development and vessel lumen formation.^44,47, 48, 49 The intracellular CCM complex and Rasip1 recruited by Heg1 mediate its function in cardiac development, whereas MEKK3/KLF2 were shown to mediate its role in sensing flow forces.^46,50 However, whether Heg1 interacts with other membrane proteins to regulate these essential endothelial functions remains unexplored. Our findings showed that Heg1 modulated BMP9-initiated signaling in liver ECs. The BMP9–Endogolin–Alk1 pathway has been implicated in hereditary hemorrhagic telangiectasia, which manifests as arteriovenous malformation.⁵¹ It will be intriguing to investigate whether Heg1 contributes to arteriovenous malformation pathogenesis by modulating BMP9/Alk1 signaling.

In summary, this study identified Heg1 as an essential endothelial membrane receptor that transduces signals to hepatocytes via Wnt ligands, which in turn sustain PPARα and FAO enzyme expression in hepatocytes, thereby controlling hepatic lipid homeostasis, particularly under metabolic stress conditions. The conclusions of this study were primarily based on in vivo results from 129/C57/Bl6 mice. However, mouse strain can profoundly influence MASLD development, with C57/Bl6 and FVB strains being particularly susceptible.⁵² Whether the signaling mechanisms identified in this study underlie fatty liver disease in humans requires further investigation using human subjects or human-derived organoids.

Materials and Methods

Mice

Lyve1-Cre mice were purchased from Jax Laboratories. Cdh5-CreERT2 and Heg1^fl/fl were described previously.^12,44,53 RSPO3^KI/+ mice were generated from Shanghai Model Organisms of China by inserting the human RSPO3 coding sequence into the Rosa26 allele; the designed schematic is shown in Figure 7A. All mice were maintained in C57Bl6/129 mixed strain background. Both female and male mice were used. Sex-matched littermates were used as controls for all experiments. To induce gene deletion or expression, Cdh5-CreERT2;Heg1^fl/fl, Cdh5-CreERT2;Heg1^fl/fl;RSPO3^KI/+, and Cdh5-CreERT2;RSPO3^KI/+ mice were given tamoxifen (Selleck) by intragastric gavage starting at 6 weeks of age, with a dose of 2 mg per mouse every other day, for a total of 5 doses. The mice were fed with a 60% fat diet (HFD; Research Diets) for 2 to 8 weeks to establish MASLD with simple steatosis. Mice were fed with an MCD diet (Research Diets) for 2 to 4 weeks to induce MASH with hepatic steatohepatitis and fibrosis. Mice were fed with liquid alcohol diet (Research Diets) to establish alcoholic fatty liver disease. Fenofibrate was dissolved in 0.5% carboxymethylcellulose and administered orally at a dose of 100 mg/kg/day. BMP agonist (SY-LB-35) was dissolved in DMSO and administered intraperitoneally at a dose of 1 mg/kg daily. All experiments were conducted under the guidelines/regulations of Tianjin Medical University and National Research Council of the National Academies.⁵⁴ The Institutional Animal Care and Use Committee of Tianjin Medical University approved all animal ethics and protocols.

In Vitro Cell Culture Studies

HUVECs were cultured in EC medium. Small interfering RNA targeting HEG1 (F-CGGCUCUUCAAGUAGAACATT; R-UGUUCUACUUGAAGAGCCGTT) was used to knock down HEG1 expression in HUVECs. After 24 hours, human BMP9 (Yeasen, 92059ES25, 100 ng/mL), human BMP2 (Yeasen, 92051ES10, 100 ng/mL), and BMP agonists (SY-LB-35, MCE, 5μM) were added and treated for another 24 hours. Total RNA was extracted from cells for RSPO3 gene expression analysis.

Biochemistry Analysis

The serum levels of ALT, AST, TC, and TG were determined using an automatic biochemical analyzer (AU5800, Beckman Colter Chemistry System analyzer). The liver TG and TC levels were measured using a commercially available assay kit (Applygen, E1013, E1015).

Glucose Tolerance Test

For the glucose tolerance test, mice were fasted for 12 hours (with free access to water) to ensure baseline fasting conditions. Prior to the test, mice were weighed, and baseline blood samples (from the tail vein) were collected to measure fasting blood glucose. A glucose solution (2 g/kg body weight, dissolved in sterile saline, prepared as a 20% solution) was then administered via intraperitoneal injection within 1 to 2 minutes. Blood samples were collected at 15, 30, 60, 90, and 120 minutes postinjection to measure glucose levels using a glucometer.

Histology

The livers and lungs were excised and immersed in an appropriate amount of 4% PFA. Tissues was then fixed overnight on a shaker at 4 °C. The adequately fixed liver and lung tissues were washed with phosphate-buffered saline (PBS) and subsequently dehydrated through a graded series of alcohol concentrations (30%, 50%, 70%, and 100%). Following dehydration, the tissue was embedded in paraffin, sectioned, and stained. The liver and lung tissues were cut into 6-μm sections and processed for hematoxylin and eosin staining to examine their morphology. Sirius Red staining was used to assess fibrosis. For Oil Red O staining, the liver tissues were fixed in formalin at least 24 hours, dehydrated by titrating in sucrose (30%), and frozen in plastic molds with optimal cutting temperature embedding media. Cryosections of the liver (10 μm) were stained with Oil Red O solution at 37 °C for 15 minutes to visualize lipid droplets.

Immunofluorescence and Immunohistochemistry

After dewaxing and hydration, the tissue sections were boiled in 10 mM citrate buffer solution (pH 6.0) for antigen retrieval, then blocked with antibody diluents at room temperature, and 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 substrate (Sigma), counterstained with hematoxylin. For immunofluorescence staining, sections were incubated with fluorescence-conjugated secondary antibodies for 2 hours at room temperature, then counterstained with 4,6-diamidino-2- phenylindole (DAPI; Solarbio) and images were acquired under a fluorescence microscope. The following primary antibodies were used: anti-CD45 (CST, 70,257, 1: 300 dilution), anti-PPARα (Proteintech, 15540-1-AP, 1: 300 dilution), anti-GS (Abcam, ab49873, 1: 5000 dilution), anti-ECAD (BD, BD610181, 1: 300 dilution), anti-Lyve1 (R&D, 216,021, 1: 300 dilution), anti-SMAD4 (Absin, abs156039, 1: 300 dilution), fluorescein phalloidin (Thermofisher, 2,406,491, 1: 500 dilution).

Electron Microscopy

Mice were anesthetized with avertin and perfused with cold 2.5% PFA solution. The livers were quickly cut into small pieces (<1 mm²) and fixed in 2.5% PFA. Then samples were postfixed with 1% (wt/vol) OsO4 and 1.5% (wt/vol) potassium ferricyanide aqueous solution at 4 °C for 2 hours, dehydrated through a graded ethanol series (30, 50, 70, 80, 90, and 100% × 2, 8 minutes) into pure acetone (2 × 8 minutes). Samples were infiltrated in graded mixture (3:1, 1:1, 1:3) of acetone and SPI-PON812 resin (21 mL SPI-PON812, 13 mL dodecenylsuccinic anhydride, and 11 mL nadic methyl anhydride), then changed to pure resin. Finally, samples were embedded in pure resin with 1.5% N,N-dimethylbenzylamine and polymerized for 12 hours at 45 °C and 48 hours at 60 °C. The ultrathin sections (70-nm thick) were sectioned with microtome (Leica EM UC6), double-stained by uranyl acetate and lead citrate, and examined by a transmission electron microscope (FEI Tecnai Spirit120kV).

Immunoblotting

Whole liver tissue homogenates were prepared in NP40 buffer (150 mM NaCl, 1% NP-40, 50 mM Tris [pH 8.0]) containing protease and phosphatase inhibitors. Protein was separated by electrophoresis using sodium dodecyl sulfate–polyacrylamide gels and transferred to polyvinylidene difluoride membranes. The membranes were blocked with 5% skim milk and probed with primary antibodies: anti-PPARα (Proteintech, 15540-1-AP, 1: 1000 dilution), anti-ERK1/2 (CST, 9102, 1: 1000 dilution), anti-P-ERK1/2 (CST, 9101, 1:1000 dilution), anti-ERK5 (CST, 3372, 1: 1000 dilution), anti-P-ERK5 (CST, 3371, 1: 1000 dilution), anti-P-SMAD1/5/8 (ABclonal, AP1518, 1:1000 dilution), anti-SMAD4 (Absin, abs156039, 1:1000 dilution), anti-Histone H3 (ABclonal, A17562, 1:1000 dilution), and anti-Actin (ABclonal, AC026, 1: 10,000 dilution) overnight at 4 °C and followed with the incubation of respective peroxidase-conjugated secondary antibodies for 2 hours at room temperature.

Tissue and Cell Nuclear Protein Extraction

For nuclear protein extraction, cultured cells in plates were placed on ice, washed with PBS, and digested with 0.25% trypsin for 5 minutes. Cells were then collected, washed again with PBS, and centrifuged at 800 g for 3 minutes at 4 °C to obtain cell pellets. The pellets were resuspended in hypotonic buffer A, vortexed vigorously for 5 seconds to ensure complete dispersion, and incubated on ice for 10 minutes. Nonidet P-40 (10%) was added to a final concentration of 0.5%, mixed by flicking, and incubated on ice for an additional 5 minutes, followed by centrifugation at 1000 g for 5 minutes at 4 °C. The pellet was washed with buffer A and centrifuged under the same conditions, with the supernatant carefully removed. The pellet was then resuspended in 3 to 5 volumes of hypertonic buffer C, vortexed, and incubated on ice for 30 minutes, with vigorous vortexing (maximum speed, 15–30 seconds) every 2 minutes. After centrifugation at 12,000 g for 15 minutes at 4 °C, the supernatant was collected as the nuclear extract, and a small aliquot was reserved for protein concentration determination. Loading buffer was added, and the extract was heated at 95 °C for 10 minutes to denature proteins. For tissue samples, 20 mg of tissue was thoroughly homogenized with 200 μL hypotonic buffer A on ice or at 4 °C, and subsequent steps were performed as described for cell samples. Hypotonic buffer A (pH 7.9) contained 10 mM HEPES, 10 mM KCl, 0.1 mM EDTA, 0.1 mM ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid (EGTA), 1 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride. Hypertonic buffer C (pH 7.9) consisted of 20 mM HEPES, 0.4 M NaCl, 1 mM EDTA, 1 mM EGTA, 25% glycerol, 1 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride.

Isolation of Liver Endothelial Cells

Mice were anesthetized with avertin and perfused with Krebs solution containing EGTA through the inferior vena cava, followed by perfusion with Krebs solution containing 0.2mg/mL of Collagenase, Type 1 (Worthington, LS004196). The liver was removed and placed in Krebs solution, and the cells were squeezed out. The cell suspensions were filtered through a 70-μm nylon mesh and centrifuged at 50 g for 5 minutes at 4 °C. The precipitations contained mostly hepatocytes. The supernatant was further centrifuged at 350 g for 5 minutes. Cell pellets were resuspended with PB buffer. Liver ECs were then isolated by magnetic-activated cell sorting using anti-CD31 MicroBeads (Miltenyi Biotec) according to the manufacturer’s instructions. The purity of liver ECs was confirmed by qPCR compared with whole liver, hepatocytes, supernatants, and flow-through using different cell markers.

RNA Extraction, Reverse Transcription, and Quantitative Polymerase Chain Reaction Analysis

Total RNA was extracted from tissue or cell homogenates using TRIzol reagent (Invitrogen), and complementary DNA (cDNA) was synthesized from 1 μg of total RNA using reverse transcription (Yeasen, China). qPCRs were performed using SYBR QPCR Master Mix (Vazyme), and the primer pairs used are listed in Tables 1 and 2. The level of target gene expression was normalized to glyceraldehyde-3-phosphate dehydrogenase expression in each sample.

Table 1.

List of Primer Sequences for Mouse Genes

Mouse genes	Forward sequence 5′–3′	Reverse sequence 5′–3′
Gapdh	GTCCCGTAGACAAAATGGTGA	TTTGATGTTAGTGGGGTCTCG
Ppara	AGACCTCAAATCTCTGGGCAAG	TGGGATAGCCTTGGCAAATTCT
Cpt1b	GCTGCTTGCACATTTGTGTT	TGAGTGACTGGTGGGAAGAA
Cact	GGTGGCTGTCCAGACAAACT	TCCGTTTAAGAACCTCCTGG
Hadh	GTTTGAGGACCTCGGTGTAAAGC	GAGAGCAGATGTGTTGCTGGCA
Ech1	CGTGACCTCATCAGCAAGTACC	GCAGTAGCGAATGTCACAGGCA
Acadvl	CCGGTTCTTTGAGGAAGTGAA	AGTGTCGTCCTCCACCTTCTC
Acadl	TTTCCGGGAGAGTGTAAGGA	ACTTCTCCAGCTTTCTCCCA
Acadm	GATCGCAATGGGTGCTTTTGATAGAA	AGCTGATTGGCAATGTCTCCAGCAAA
Ehhadh	CAGCACTGGATGTGGATGAC	CATGACTGTGGCGATGGTAG
Id1	GAACCGCAAAGTGAGCAAGG	GGAACACATGCCGCCTCA
Id2	CCTGCATCACCAGAGACCTG	GGGAATTCAGATGCCTGCAA
Id3	ACATGAACCACTGCTACTCGC	GTGAGCTCAGCTGTCTGGATCG
Ltbp1	CTGGTCGCATCAAGGTGGT	TGGGCATACTGGCACCAAG
Tgfb1i1	GCAGTCTGGACACCATGCT	TAGCACTCGGGGCAAAAGG
Tgfbr2	CCGCTGCATATCGTCCTGTG	AGTGGATGGATGGTCCTATTACA
Hipk2	ATGTGCAAGTTTTCTCCCCTC	CTCGTAAGGTAGGCTTGGGTT
Crebbp	GGCTTCTCCGCGAATGACAA	GTTTGGACGCAGCATCTGGA
Zmiz1	CCCCGCCAACTTCCACAAT	AGCCAAGAGTCTGTAGCCCA
Bmp2	GGGACCCGCTGTCTTCTAGT	TCAACTCAAATTCGCTGAGGAC
Cav2	TAGGCTTCGAGGATCTGATTGC	TCACAAAAGGCATCAGGATCCA
Arid4b	AACCTCCTTATTTGACAGTGGGC	ATAGCTCCTACCTTCAGTGGG
Kit	GAACAGGACCTCGGCTAACAA	CCTTTGCTCTGCTCCTGTACA
Cav1	GACGAGGTGACTGAGAAGCAA	TTGGGATGCCGAAGATCGTAG
Gas6	AGAGGAGGCCAGAGAGGTG	TGCAGCCCCCATTCTTCTG
Tiam1	CCCATCCAGAGGGTCCTCA	AAGAGCAGGTCACCCATGC
Ramp3	CTGCTTTGTGGTGAGTGTGC	CCACCGTGCAGTTGGAAAAG
Ackr3	AGGAAGCCCTGAGGTCACTT	CAATGCAGTCGCTGCTGTTAC
Cd36	ATCCTGTGACTGGGCAAGC	TCTACGTGGCCCGGTTCTA
Fatp1	CATCCGTCTGGTCAAGGTCAA	TTGAAGGTGCCTGTGGTATCC
Fatp4	GACTTCTCCAGCCGTTTCCA	TCTGTGCAAAGCTCTCCAGG
Chrebp	GAAGCCACCCTATAGCTCCC	CTGGGGACCTAAACAGGAGC
Srebp1c	TAACGTGGGCCTAGTCCGA	CAGGAGAGTTGGCACCTGG
Elovl5	CATTTCGATGCGTCACTCAGTAC	CACACCTGTCACCAACTCATAGA
Dgat2	TGTCACCTGGCTCAACAGATC	ATCTCCTGCCACCTTTCTTGG
Apob	TCGATTCAAGCACCTCCGAAA	GCTTGAGTTCGTACCTGGACA
Apoc3	ACCCTGAGGACCAACCAACT	AGGCACATCTGCAACACAGA
Ldlr	CGGAAAATGCATCGCTAGCAA	CTCGGTCTCCATCACACACAA
Wnt2	CCTGATGAACCTTCACAACAAC	TCTTGTTTCAAGAAGCGCTTTAC
Wnt9b	CTGGTGCTCACCTGAAGCAG	CCGTCTCCTTAAAGCCTCTCTG
Rspo3	CTTACACCTTGGAAAGTGCCTTG	ATACAAGTTCTTGTCTCGCTGGT
Pecam1	GAATGACACCCAAGCGTTTT	GGCTTCCACACTAGGCTCAG
Lyve1	CAGCACACTAGCCTGGTGTTA	TCCGAGGGATGACAGAGAACT
Tie1	CCAGTCAGGATCGGGTGAAG	ATCATGGCCCGGATCACTTG

Table 2.

List of Primer Sequences for Human Genes

Human genes	Forward sequence 5′–3′	Reverse sequence 5′–3′
RSPO3	GCACGCCTATCGGATGTGA	GATGCATTCTTCGCTGGCG
HEG1	AGCAGCTTCCTCTCCTCTCTTA	AGGATGAAGTCGTGGCAATTCT
GAPDH	GGATTTGGTCGTATTGGG	GGAAGATGGTGATGGGATT

RNA Sequencing

Total RNA was extracted from the tissue or isolated cells using TRIzol Reagent according to the manufacturer’s instructions. RNA purification, reverse transcription, library construction, and sequencing were performed at Shanghai Majorbio Bio-Pharm Biotechnology according to the manufacturer’s instructions. mRNA was isolated according to PolyA selection method by Oligo (dT) beads, and then fragmented by fragmentation buffer. Double-stranded cDNA was synthesized using a SuperScript double-stranded cDNA synthesis kit (Invitrogen) with random hexamer primers. Then the synthesized cDNA was subjected to end-repair, phosphorylation, and adapter addition according to library construction protocol. Libraries were size-selected for cDNA target fragments of 300 bp on 2% low range ultra agarose gel, followed by PCR amplified using Phusion DNA polymerase (NEB) for 15 cycles. After quantification using Qubit 4.0, the sequencing library was performed on NovaSeq X Plus platform (PE150) using NovaSeq Reagent Kit.

Raw paired end-reads were trimmed and quality controlled by fastp with default parameters. Then clean reads were separately aligned to the reference genome with orientation mode using HISAT2 software. The mapped reads of each sample were assembled by StringTie using a reference-based approach. To identify differentially expressed genes (DEGs) between 2 different conditions, the raw read count file generated by featureCounts was used as an input for DESeq2, followed by standard differential expression analysis. Genes with Q value ≤0.05 and |log2fold change| >1(DESeq2) were considered to be significantly DEGs. In addition, GO functional enrichment analysis was performed using the clusterProfiler to identify which DEGs were significantly enriched in GO terms and metabolic pathway enrichment with Bonferroni-corrected P value ≤ .05 compared with all expressed genes as background.

RNA samples from livers of 3 HFD-fed Heg1^LECKO mice and 2 littermate controls, isolated liver endothelial cells of 4 Heg1^LECKO mice and 3 littermate controls at P3, and 3 Heg1^LECKO mice and 3 littermate controls at P42 were used for RNA-seq. RNA-seq datasets were deposited in Gene Expression Omnibus (accession # GSE315829,GSE315356). The dataset comprising proteins predicted to be secreted (SPOCTOPUS class) was retrieved from The Human Protein Atlas, available at: https://www.proteinatlas.org/search/protein_class%3ASPOCTOPUS+predicted+secreted+proteins.

Acyl-Carnitine Analysis by Liquid Chromatography–Mass Spectrometry

The acylcarnitine content of the samples was analyzed using liquid chromatography-mass spectrometry. First, metabolites were extracted from samples and analyzed by liquid chromatography-mass spectrometry. TargetLynx quantitative software was employed to calculate the peak area of the targeted data, and identification results were obtained using the single-point internal standard method. A precise amount of tissue (approximately 50 mg) was accurately weighed and mixed with 100 μL of 75% methanol aqueous solution, followed by homogenization for 2 minutes. Subsequently, 900 μL of 75% methanol aqueous solution was added, and the mixture was oscillated and extracted at 4 °C for 1 hour. After centrifugation at 12, 000 rpm for 10 minutes, 95 μL of the supernatant was transferred to a 1.5-mL centrifuge tube, and 5 μL of internal standard (containing 5 internal standards: C0-D3 5000 ng/mL, C2-D3 5000 ng/mL, C4-D3 500 ng/mL, C8-D3 500 ng/mL, C16-D3 500 ng/mL) was added, followed by vortex oscillation for 1 minute. An aliquot of 80 μL was transferred to a 200-μL insert tube, and chromatographic separation was performed using Waters ACQUITY UPLC I-CLASS ultra-high-performance liquid chromatography. Mass spectrometry analysis was conducted using a Waters XEVO TQ-S tandem quadrupole mass spectrometry system. The positive ion source voltage was set to 3.0 kV, the cone voltage was 20 V, the ion source temperature was 150 °C, the desolvation temperature was 550 °C, the desolvation gas flow rate was 1000 L/h, and the cone gas flow rate was 10 L/h. TargetLynx quantitative software was utilized to calculate the peak area, allowing a retention time error of 15 seconds. Quantitative results were obtained using the single-point isotope internal standard method.

Statistical Analysis

All data are presented as mean ± standard deviation and analyzed with GraphPad Prism 9.0. Statistical comparisons between 2 groups were performed using 2-tailed Student’s t test, whereas 1-way analysis of variance was used for multigroup comparisons. Statistical significance was defined as ∗P < .05, ∗∗P < .01, ∗∗∗P < .001, and ∗∗∗∗P < .0001.