New aspects of hepatic endothelial cells in physiology and nonalcoholic fatty liver disease

Abstract

The liver is the central metabolic hub for carbohydrate, lipid, and protein metabolism. It is composed of four major types of cells, including hepatocytes, endothelial cells (ECs), Kupffer cells, and stellate cells. Hepatic ECs are highly heterogeneous in both mice and humans, representing the second largest population of cells in liver. The majority of them line hepatic sinusoids known as liver sinusoidal ECs (LSECs). The structure and biology of LSECs and their roles in physiology and liver disease were reviewed recently. Here, we do not give a comprehensive review of LSEC structure, function, or pathophysiology. Instead, we focus on the recent progress in LSEC research and other hepatic ECs in physiology and nonalcoholic fatty liver disease and other hepatic fibrosis-related conditions. We discuss several current areas of interest, including capillarization, scavenger function, autophagy, cellular senescence, paracrine effects, and mechanotransduction. In addition, we summarize the strengths and weaknesses of evidence for the potential role of endothelial-to-mesenchymal transition in liver fibrosis.

INTRODUCTION

The liver is the central metabolic hub for carbohydrate, lipid, and protein metabolism (160). It can uptake, store, synthesize, metabolize, and export glucose for distal tissues. It can absorb dietary lipids, convert carbohydrates into lipids, synthesize and secrete triglycerides, absorb and synthesize cholesterol, and produce ketone bodies from fatty acids. In addition, it can also produce glucose and ketone bodies using the carbon skeleton of specific amino acids while removing the excess nitrogen through the urea cycle. These metabolic processes are tightly controlled by hormones such as insulin and glucagon. Impaired liver function is involved in the pathogenesis of different diseases, including type 2 diabetes, cardiovascular disease, and cancer (16, 142).

The liver is composed of four major types of cells, including hepatocytes, endothelial cells (ECs), Kupffer cells (KCs), and stellate cells (87). Immunocytochemical staining and quantification revealed that hepatocytes and ECs constitute ∼52% and 22% of all labeled cells, respectively (6). In healthy liver, these cells cooperate to maintain normal liver homeostasis and function. Under pathological conditions, they undergo phenotypic changes at molecular and cellular levels. ECs represent a major subpopulation of nonparenchymal cells, and a majority of them are defined as noncontinuous ECs that line the hepatic sinusoids and are known as liver sinusoidal ECs (LSECs). Liver sinusoidal endothelium is structurally and functionally unique in multiple ways (34, 88, 130, 152): 1) discontinuous and the presence of open pores (see more details in the section of fenestrae and capillarization); 2) the lack of a basement membrane and tight junctions; 3) very high endocytotic capacity; 4) frequently exposed to the high levels of macromolecules, toxins, and waste products; and 5) potent immunological functions. The phenotypic changes of LSECs in nonalcoholic fatty liver disease (NAFLD) and their consequences on NAFLD progression and complications were elegantly reviewed recently (56). In addition to LSECs, there are other types of hepatic ECs such as periportal and pericentral ECs in mice (170) or hepatic artery, central vein, and lymphatic ECs in humans (136) clustered from single-cell RNA sequencing data. In this review, different aspects of hepatic ECs are discussed mainly in NAFLD as well as in other fibrotic liver-related conditions where it is appropriate (Table 1).

Table 1.

The key points of phenotypes and functions of hepatic ECs in pathophysiology

Phenotype and Function	Key Points	Reference No.
Fenestrae and capillarization	Fenestrae are very dynamic with a mean life span of about 18 minutes. The diameter and number of fenestrae vary from species to species and in response to different stimuli. Aging, a major risk factor for NAFLD, is associated with the loss of fenestrae in mice and humans. Fenestrae porosity did not differ in isolated LSECs between lean and obese mice.	(13, 75, 94, 107, 110, 176)
Scavenger function	Clearance of endogenous and exogenous material that contributes to inflammation is just beginning to be understood in a physiological context.	(58, 80, 109, 131, 151, 152)
Autophagy	Autophagy is impaired in NASH patients. The deficiency in endothelial autophagy induces liver injury and inflammation and promotes liver fibrosis in mouse NAFLD.	(55, 59, 141)
EndMT	A small percentage of hepatic ECs undergo mesenchymal transition in mice and humans under disease states. More studies are needed to reveal the functional significance of EndMT in liver fibrosis.	(44, 86, 138)
Cellular senescence	The knowledge about cellular senescence of hepatic endothelium is very limited in liver diseases; cGAS-STING signaling likely plays a key role; hepatic cellular senescence promotes aging or NAFLD-related liver steatosis.	(105, 119, 126, 175)
Mechanotransduction	Mechanotransduction in hepatic endothelium induces hepatocyte proliferation and liver growth in mice during development. Pathological mechanical stretching of hepatic endothelium induces the recruitment of neutrophils, which in turn induces microvascular thrombosis and fibrosis.	(69, 104)

cGAS-STING, cyclic GMP-AMP (cGAMP) synthase-stimulator of interferon genes; ECs, endothelial cells; EndMT, endothelial-to-mesenchymal transition; LSECs, liver sinusoidal endothelial cells; NAFLD, nonalcoholic fatty liver disease; NASH, nonalcoholic steatohepatitis.

NAFLD

NAFLD is a spectrum of liver diseases in which excess fat builds up in livers without secondary causes of hepatic fat accumulation, such as significant alcohol consumption (18). Scientists have proposed that the name of NAFLD should be changed to obesity-associated fatty liver disease to emphasize the major driver of NAFLD, obesity (150). The early stage of NAFLD is known as steatosis (simple fatty liver) which has been chemically defined as excessive intrahepatic triglyceride content more than 5% of liver volume or liver weight (46). At this stage of NAFLD, there is no evidence of hepatocellular injury in the form of hepatocytes ballooning and no evidence of fibrosis (18). The progressive subtype of NAFLD is nonalcoholic steatohepatitis (NASH). It is characterized by steatosis (≥5%), hepatocyte injury (e.g., ballooning), and lobular inflammation with or without any fibrosis (18). The most advanced stage of NAFLD is NASH cirrhosis, in which liver is scarred and permanently damaged. It is defined as the presence of cirrhosis with current or previous histological evidence of steatosis or steatohepatitis (18). NASH can progress to cirrhosis in about 22% NASH patients with advanced fibrosis (143). While lifestyle modifications can control risk factors, there are no FDA-approved pharmacological therapies available to treat NAFLD (18).

FENESTRAE AND CAPILLARIZATION

Capillarization, also known as defenestration, refers to the loss of fenestrae. Fenestrae are unique morphological features of LSECs in that they are defined as holes that go through the cytoplasm (Fig. 1) and are organized in clusters termed sieve plates (167). They are transmembrane pores with a diameter of 50–150 nm in mouse LSECs. The lifespan, formation, and disappearance of LSEC fenestrae were recently monitored in mouse LSEC culture using atomic force microscopy (176). The major findings are that 1) 75% of the newly formed fenestrae remain open for ≤20 min, whereas the mean fenestrae lifespan is ∼18 min, 2) fenestrae are not static but instead continue to change their positions (≤5 µm) and diameter (≤200%) during their lifespan, 3) some of the fenestrae can close for a while and open again, 4) three distinct ways (the process of formation of new fenestrae becomes slower, the coalescence of neighboring fenestrae lead to the formation of gaps, and fenestrae number may drop because of the formation of defenestration centers) can result in the defenestration of LSECs, and 5) cytochalasin B, which promotes actin depolymerization (106), can induce fenestrae in initially defenestrated cells (176). In addition to atomic force microscopy, fenestrae can be visualized by several other microscopy techniques (113, 114). Mönkemöller et al. (114) resolved structural details of fenestrae in rat LSECs using direct stochastic optical reconstruction microscopy. They demonstrated that this technique can identify individual fenestrae with a diameter of <50 nm. In a more recent study, Mönkemöller et al. (113) combined two superresolution microscopy techniques, direct stochastic optical reconstruction microscopy and three-dimensional structured illumination microscopy, to reveal the new insights of fenestrae. The combination enables the authors to observe the subdiffraction-sized LSEC fenestrae with great detail down to 20 nm. They also found that microtubules appear to surround and separate sieve plates, whereas actin filaments surround each fenestrae within a sieve plate (113). With advances in microscopy techniques, new physiological facets of fenestrae will be revealed in the future.

Fig. 1.

Scanning electron microscope image shows mouse liver sinusoids. Liver sinusoidal endothelial cells (LSEC) line hepatic sinusoids. The circle indicates a group of fenestrae that form a sieve plate in the LSEC. Two arrows indicate the space of Disse that is the space between sinusoidal endothelium and microvilli of hepatocytes. This image was obtained under protocol #607 and approved by the Institutional Animal Care and Use Committee on Nov. 23, 2010.

The diameter and number of fenestrae vary from species to species and in response to different stimuli (13). Although aging is not a disease, it is a major risk factor for liver disease (74, 85). Aging is associated with markedly diminished fenestrae porosity, increased hepatic vascular resistance, and elevated portal pressure (107). Isolated LSECs from old mice and rats have reduced fenestrae porosity (75, 96). Aging in humans is also associated with the loss of fenestrae in the sinusoidal endothelium (110). Although aging correlates with the loss of fenestrae, age-matched animals fed a high-fat diet (HFD) do not appear to affect the numbers of fenestrae. In a mouse model of diet-induced obesity, LSEC fenestrae are functionally preserved in either the early (2–8 wk) or late (15–20 wk) stages of NAFLD progression, despite LSECs displaying a proinflammatory phenotype (94). Overall, LSEC fenestrae porosity did not differ between lean and obese mice. However, the mean diameter of fenestrae was larger in the obese mice fed a HFD (60 kcal% fat) to induce NAFLD. LSEC morphology was visualized using scanning electron microscopy and atomic force microscopy in primary cells within 24 h from the isolation. It remains unknown whether fenestrae would change during cell isolation that may affect the overall results.

There are several signaling pathways involved in capillarization. Delta-like 4, a ligand of the Notch signaling pathway, was upregulated in the LSECs of human and carbon tetrachloride (CCl₄)-induced murine fibrotic liver (21). It mediates LSEC capillarization and the vicious cycle between fibrosis and pathological sinusoidal remodeling (21). Bone morphogenetic protein 9 (BMP9) is a circulating factor produced by hepatic stellate cells. Loss of BMP9 led to the expression of several terminal differentiation markers such as LYVE1 and Stab2 and triggered hepatic perisinusoidal fibrosis (38). In addition, LSECs from BMP9 knockout mice had a reduced number of fenestrae, indicating that they are capillarized (38). These data demonstrated that BMP9 protects against hepatic fibrosis by influencing LSEC fenestration.

The strategies to induce the reformation of fenestrae have been explored by several groups. Metformin treatment caused an increase in fenestrae porosity in LSECs isolated from both young and old mice without altering fenestrae diameter (75). The use of 0.1% metformin diet for 12 mo reduced the age-related loss of fenestrae porosity and frequency by 50% in mice (75). Agents that target nitric oxide, actin, or lipid rafts promote changes in fenestrae in murine LSECs (76). Interestingly, LSECs from young and old mice respond differently to these agents. Bosentan, an endothelin receptor agonist used in the treatment for pulmonary hypertension, and 2,5-dimethoxy-4-iodoamphetamine, a hallucinogen that activates neurotrophin receptors, can increase fenestration porosity and/or frequency in old mice only (76). LSECs lose their fenestrae in cell culture in a time-dependent manner once they are isolated from rodents (12). Interestingly, actin depolymerization by cytochalasin D can induce fenestrae reformation in dedifferentiated LSECs devoid of fenestrate (39). These studies demonstrated that age-related defenestration can be reversed pharmacologically.

LSECs SCAVENGER FUNCTION AND THE ROLE OF LSECs IN INFLAMMATORY RESPONSE

Fenestrated LSECs allow large macromolecules into the Space of Disse; however, LSECs also have a high surface area and are covered with various scavenger receptors to recognize danger-associated molecular patterns along with a host of indigenous and foreign molecules. LSECs are among of the most endocytic cells of the mammalian system (146) and are unique in that their endocytic vesicles retain their clathrin coat longer than in any other cells. Endocytic vesicles are conveyed along a network of microtubules in which cargo is delivered to a robust lysosomal network to degrade cargo (47). The scavenger receptors at the cellular surface include SR-A (MSR1, SR-A1), SR-B (SR-B1, SCARB1), SR-E1 (LOX1), SR-H (stabilin-1, stabilin-2/HARE), FcγRII/CD32b, CD206/mannose receptor, and C-type lectin receptors (L-SIGN, LSECTin). Some of these receptors, such as SR-H, are constitutive and do not change significantly in their expression, even in chronic diseased states (8). However, some diseased states induce higher expression of these receptor and some of their splice variants. These would include SR-B, which increases in expression during steatohepatitis (134, 137), and SR-E under similar conditions in which there is hypercholesterolemia and hyperlipidemia (19, 22). It should be noted that LSECs take up molecules and small particles amenable for pinocytosis (<100 nm), and larger particles are primarily taken up by KCs (149, 151) in a coordinated fashion. Ligands for the known scavenger receptors expressed by LSECs are extensive and beyond the scope of this review. For suitable references on these ligands, the reader is invited to obtain this information from Refs. 34, 58, 80, 109, 131, and 152.

During NAFLD or lipotoxic conditions, many cells of the liver increase their expression of tumor necrosis factor receptors (31, 171); however, a coordinated effort between the nonparenchymal cells of the liver and immune cells promotes tolerance for foreign antigens (158). Fas receptor, which is normally absent on LSECs, is expressed during lipotoxic conditions to bind with activated immune cells expressing Fas ligand (37, 168). LSECs have a higher tolerance than hepatocytes and stellate cells to resist undergoing apoptosis under these conditions and instead start expressing ICAM-1 and VCAM-1 (7, 17) in addition to the constitutive, vascular adhesion protein-1 (VAP-1) (95). These adhesion molecules are thought to recruit KCs to clean up cellular debris from apoptotic hepatocytes and other cells. Without this coordinated effort between KCs and LSECs, the viability of LSECs decreases (70, 77). Overall, LSECs promote tolerance and immune dampening through their expression of programmed death ligand-1 (PD-1L). The interaction of programmed cell death protein 1 (PD-1) expression on KCs and CD8+ cells promotes immune tolerance, as the blood supply for the liver is coming from the gut circulation via the portal vein and contains a host of immune stimulatory molecules such as lipopolysaccharides and food antigens (82). PD-1L/PD-1 interactions between LSECs and T_H1/T_H17 cells, respectively, promote suppression of cytokine release from these cells. However, immunotolerance does have its limits, and a breach beyond this undefined threshold stimulates activation of the SHP-1 and SHP-2 phosphatases through the FAS and PD-1 pathways to promote activation of immune cells and cytokine release (120, 148). More detailed and in-depth information on LSEC activation in NAFLD was reviewed by Hammoutene and Rautou (56) and neutrophil recruitment in the liver by Liew and Kubes (102) and Zindel and Kubes (181).

Quiescence in the liver is promoted by the production of nitric oxide and hepatocyte growth factor (HGF) secreted by LSECs and VEGF secreted by hepatocytes (33, 93, 159). Cross-talk between liver cell types typically dampens inflammatory responses, but it may also stimulate repair and cleanup operations by immune cells. For example, lipopolysaccharide-stimulated KCs secrete prostaglandin E2, which is a suppressor of T cell receptor signaling. In this setting, it helps protect and increase the survival of LSECs against the harmful effects of ischemia-reperfusion (I/R) (2, 172). Naturally, during an acute injury such as I/R, which leads to increased reactive oxygen species and inflammation (79) and reduced nitric oxide (122), the endocannabinoid system is involved and upregulated in both parenchymal and nonparenchymal cells. The cannabinoid CB₂ receptor acts to dampen the damage of the inflammatory response. Activation of CB₂ through a synthetic agonist, JWH133, decreased levels of alanine aminotransferase and aspartate aminotransferase, hematopoietic markers of liver damage. Similarly, activation of CB₂ counteracted TNFα and lowered ICAM-1 and VCAM-1 expression in LSECs, which lowered the levels of neutrophil adhesion and possibly invasion within the parenchyma (7). On the other end, an antagonist of CB₁ with rimonabant (SR141716) was hepatoprotective by decreasing inflammation through TNFα in a hepatic steatosis rat model (51). Although CB₁ is found predominantly in the brain, it still has effects in liver that may be induced during specific injuries. For example, CB₁ is involved with liver regeneration and may cause scarring or fibrosis if inflammation is not controlled (115, 132). Generally, agonists of CB₂ and antagonists of CB₁ are beneficial for liver health. Effective therapeutic treatment of chronic (steatosis) or acute (I/R) liver disease may be managed by stimulating the endocannabinoid system, although such efforts are still controversial due to side effects in other organ systems and unintended behavioral consequences (5, 68, 108, 157).

AUTOPHAGY

Autophagy (the elimination/recycling of organelles) is critical for the proper functionality of the liver in physiological conditions, and defects in autophagy cause liver disease (59). The important role of autophagy in hepatic adaptation to stress was reviewed recently (59). In this section, we will focus on autophagy in hepatic ECs (Fig. 2A).

Fig. 2.

Hepatic endothelial cells (ECs) undergo phenotypic changes under disease condition. A: liver sinusoidal endothelial cells (LSEC) have impaired autophagy in nonalcoholic fatty liver disease (NAFLD). The loss of fenestrae is also known as capillarization or defenestration. B: hepatic ECs undergo cellular senescence in different liver diseases. The role of the cCAG-STING axis in the cellular senescence of hepatic ECs has not been examined, while it plays an important role in the cellular senescence of other cell types. It is very likely that cyclic GMP-AMP (cGAMP) synthase (cGAS)-stimulator of interferon genes (STING) axis is activated in hepatic ECs upon DNA damage or mitochondrial stress in different liver diseases. C: hepatic ECs undergo endothelial to mesenchymal transition (EndMT) under disease conditions. It is interesting to compare mitochondrial respiration and substrate preference for ATP production before and after EndMT. A–C: hepatic ECs can affect the function of hepatocytes and hepatic stellate cells through paracrine effects.

Endothelial autophagy plays a protective role in both acute and chronic liver injury. In patients with nonalcoholic steatohepatitis (a severe form of NAFLD), LSECs contain fewer and smaller autophagic vacuoles than in patients with no or mild histological abnormalities or with simple steatosis (55). In vitro mechanistic studies revealed that the deficiency in autophagy in hepatic endothelium induces endothelial inflammation, endothelial-to-mesenchymal transition (EndMT), and EC apoptosis. These in vitro experiments were performed in the immortalized murine liver EC line (transformed LSECs). Evidence from primary LSECs will strengthen the results (55). In a mouse model of NAFLD, the deficiency in endothelial autophagy induces liver injury and inflammation and promotes liver fibrosis (55). In rats with mild acute liver injury induced by CCl₄, endothelial autophagy was upregulated (141). To examine the function of endothelial autophagy during acute liver injury, EC-specific Atg7 knockout mice were treated with CCl₄. Their LSECs exhibited a reduction in the number of fenestrae, increased oxidative stress, decreased nitric oxide bioavailability, and aggravated liver fibrosis. In an acute liver injury induced by cold ischemia-reperfusion, endothelial autophagy improved cell viability and ameliorated hepatic damage and microvascular function (54). These data demonstrate that endothelial autophagy protects liver from injury during the early stages of liver diseases.

The autophagy of mitochondria, also known as mitophagy, has not been examined in hepatic endothelium. However, mitochondrial energetics was recently examined in LSECs isolated from lean and obese mice (94). Basal glycolysis and maximal glycolysis capacity were increased in LSECs from obese mice after 2 wk of HFD, while mitochondrial respiration was impaired in LSECs from mice after 4 wk of HFD. Mitochondrial oxidative phosphorylation was impaired, as evidenced by decreased ATP production, increased maximal mitochondrial respiration, and increased proton leak. Interestingly, mitochondrial bioenergetics were not changed at any other examined time points, including 8, 15, and 20 wk of HFD. The relative contribution of mitochondrial respiration and glycolysis to ATP production was not compared in this study, although it is known that LSECs contain few mitochondria (11) and are not highly glycolytic like other ECs (30, 140, 154). The experiments were conducted in LSECs within 24 h from the isolation using CD146 immunomagnetic sorting. However, the in vitro culture condition may not reflect in vivo status. Improving our understanding of mitophagy in hepatic ECs may provide new insights for targets against liver disease.

ENDOTHELIAL TO MESENCHYMAL TRANSITION

Liver fibrosis is a pathology caused by chronic infection or metabolic stress that is observed in different liver diseases such as alcoholic liver disease, NAFLD, and viral hepatitis. It may lead to liver cirrhosis or hepatocellular carcinoma that eventually causes the death of patients. Liver stellate cells are traditionally considered as the major cellular source of myofibroblasts that drive liver fibrosis during liver injury (67, 133). In the CCl₄-induced liver injury model, more than 87% of myofibroblasts originate from hepatic stellate cells (78). In a separate study, it was demonstrated that hepatic stellate cells give rise to 82–96% of myofibroblasts in bile duct ligation-induced, CCl₄-induced, or methionine- and choline-deficient diet-induced liver fibrosis (111). Recently, it has become appreciated that ECs also contribute to liver fibrosis through a process known as EndMT (Fig. 2C). It describes a process in which an EC undergoes transcriptomic reprograming that leads to phenotypic changes toward a mesenchymal cell (90, 129). It is a very complex biological process in which ECs lose EC phenotype and functions and acquire mesenchymal cell morphology and functions. At molecular levels, ECs lose the expression of EC markers such as platelet-EC adhesion molecule, von Willebrand factor, and vascular-endothelial cadherin, while they gain the ability to produce mesenchymal proteins, including α-smooth muscle actin (α-SMA), N-cadherin, vimentin, fibroblast-specific protein-1 (also known as S100A4 protein), fibroblast activating protein, and fibrillar collagens type I and type III. Many transcription factors, noncoding RNAs, and different signaling pathways participate in EndMT. It plays an important role in organ development and remodeling as well as the pathogenesis of human diseases involving fibrosis.

There are different methods that can be used to detect EndMT in vivo, such as simple immunostaining and genetic lineage tracing with inherent limitations (90). Immunostaining cannot identify cells that have substantially reduced or lost EC marker expression, while genetic lineage tracing such as Tie2-lineage cells can label a very small population of nonendothelial hematopoietic cells. These methods revealed that EndMT is present in mice and humans during liver injury. A very small subpopulation of liver ECs from cirrhotic patients and mice with liver fibrosis undergo EndMT (86, 138). Liver ECs from cirrhotic patients with alcohol-induced liver disease and hepatitis C virus are α-SMA positive, indicating an EndMT phenotype. In patients with idiopathic portal hypertension, double immunofluorescence staining revealed the co-localization of CD34 and S100A4 or CD34 and collagen (COL1A1) in the portal venous endothelium, indicating the presence of EndMT (86). In a mouse model of liver fibrosis induced by intraperitoneal injection of CCl₄, α-SMA expression was detected in CD31+ cells in parenchymatous areas (138). In the same study, genetic lineage-tracing revealed that about 3.8% of the tdTomato positive cells express mesenchymal markers in the livers of Tie2-tdTomato mice (138). This demonstrated the direct evidence of in vivo EndMT in mouse fibrotic livers. However, the percentage of myofibroblasts originating from hepatic ECs remains unknown, which very likely is not high because hepatic stellate cells account for 87% or 82–96% of myofibroblasts in different mouse models of fibrosis (78, 111). In other mouse models of fibrotic diseases, where EndMT is a key contributor to fibrosis, the percentage of EndMT-derived cells varies among the fibroblast population (90). In advanced atherosclerosis, EndMT-derived cells comprise almost half of the fibroblast population (45). In cardiac overload and fibrosis, EndMT-derived cells (Tie1 was used as endothelial marker) contribute to about 27–33% of cardiac fibroblasts (177).

Despite the small percentage of hepatic ECs that undergo mesenchymal transition, hepatic ECs could play important roles in liver fibrosis. For example, the endothelial transcription factor ETS-related gene (ERG) promotes the SMAD1 pathway, while it represses SMAD2/3 activity in vitro and in vivo (44), loss of endothelial ERG results in EndMT and spontaneous liver fibrosis in mouse models, ERG expression was reduced in CD31-positive cells from human fibrotic liver specimens, and importantly, loss of endothelial ERG correlates with EndMT and liver fibrosis in patients (44), indicating that ERG likely inhibits EndMT and liver fibrosis in end-stage human liver disease. Second, chromatin remodeler controls EndMT in liver fibrosis, brahma related gene 1 (Brg1) is a chromatin remodeling protein, and endothelial Brg1 deficiency attenuates bile duct ligation-induced EndMT and liver fibrosis in mice by reducing NADPH oxidase 4 expression and reactive oxygen species production (101). Third, the transcriptional modulator megakaryocytic leukemia 1 (MKL1) activates EndMT by activating TWIST1 expression in ECs, a master regulator of EndMT (100); EC-specific deletion of MKL1 suppresses EndMT and attenuates liver fibrosis in mice (100). In the last two studies (100, 101), primary LSECs isolated from Brg1 or MKL1 EC knockout mice display higher expression of EC markers and lower expression of mesenchymal markers at mRNA levels in liver fibrosis induced by bile duct ligation, and there was no other in vivo evidence examined for the role of Brg1 or MKL1 in EndMT. In all three studies (44, 100, 101), the potent effects of different genes on liver fibrosis could be attributed to EndMT-independent mechanisms or the paracrine effects of these EndMT-derived cells. Further studies are required to tease it out. In summary, there is evidence supporting that hepatic ECs undergo EndMT during fibrosis in human liver disease and animal studies. However, more studies are needed to demonstrate the significant functional effects of EndMT on liver fibrosis development. Improving our understanding of EndMT in liver fibrosis could provide new therapeutic targets for liver disease.

CELLULAR SENESCENCE

Cellular senescence is a stable form of cell cycle arrest characterized by different features such as DNA damage response, mitochondrial dysfunction, an increase in the expression of cyclin-dependent kinase inhibitors like p16 and p21, an increase in senescence-associated secretory phenotype (SASP), and an increase in senescence-associated β-galactosidase activity (SA-β-Gal) (4, 53, 63, 65). Cells become senescent in response to different stresses such as persistent DNA damage and oxidative stress (50, 63, 139). This is known as premature senescence, which is different from replicative senescence caused by telomere shortening (92). Although cell senescence can exert beneficial effects in human health and disease (36, 48, 116, 144, 155), it plays a causative role in the pathogenesis of obesity and associated vascular complications (14, 23, 84, 112, 117, 119, 123, 124, 164). Senescent cells are found in obesity (14, 112, 119, 164), diabetes (14, 123), and atherosclerosis (23). How to safely counter the detrimental effects of cellular senescence is an emerging research area of great interest (15, 24, 81, 121, 124, 162). For example, elimination of senescent cells improved glucose tolerance, enhanced insulin sensitivity, reduced macrophage accumulation in adipose tissue, lowered circulating inflammatory mediators, and promoted adipogenesis in obese mice (124). In the first clinical trial of senolytics, the combination of Dasatinib and Quercetin improved physical function in patients with idiopathic pulmonary fibrosis (81) and reduced adipose tissue senescent cell burden within 11 days in patients with diabetic kidney disease (66).

The role of senescent cells in liver diseases was recently summarized in two review papers with minimal information about the cellular senescence of hepatic endothelium (73, 126). Hepatocyte senescence was observed in acute live injury induced by acetaminophen or CCl₄ (10) and in nonalcoholic fatty liver disease (119) in mice. The accumulation of senescent hepatocytes promotes hepatic fat accumulation and steatosis, because mitochondria lose the ability to metabolize fatty acids efficiently (119). Hepatocyte senescence was also observed in patients with NAFLD (3). Importantly, hepatocyte senescence correlated closely with fibrosis stage, diabetes mellitus, and clinical outcome; an adverse liver-related outcome was strongly associated with higher hepatocyte p21 expression (3). Whereas the cellular senescence of hepatocytes aggravates fatty liver, senescent hepatic stellate cells appear to limit liver fibrosis. Krizhanovsky et al. (91) reported that in response to liver injury, senescent hepatic stellate cells accumulate in fibrotic livers to limit the extent of fibrosis. They found that if hepatic stellate cells were deficient in key senescent genes in the p53 or retinoblastoma protein pathways, these cells continued to proliferate and secrete excessive extracellular matrix components. The senescence of activated hepatic stellate cells also facilitated the resolution of fibrosis upon withdrawal of CCl₄. In this study, the authors also demonstrated that these senescent hepatic stellate cells stimulate immune surveillance, leading to the clearance of these senescent cells by natural killer cells. These findings suggest that senescent hepatic stellate cells are beneficial for the injured liver when the production of senescent cells does not outpace their clearance.

As discussed above and reviewed by others, the important roles of cellular senescence of hepatocytes and hepatic stellate cells have been revealed by many studies in liver diseases. In contrast, we barely know anything about the role of cellular senescence of hepatic endothelium or LSECs, although cellular senescence has been broadly studied in ECs. Aging-related senescent cells are observed in many organs, including liver (163). γ‐H2AX (phosphorylated H2AX at Ser¹³⁹) is a molecular marker of DNA damage and used to identify senescent cells. In liver, the frequency of γ‐H2AX foci‐positive cells increases significantly with age, and the vast majority of foci‐positive cells in liver are hepatocytes, as judged by morphological criteria (163). Senescent cells induce senescence in bystander cells in vitro (118) and in vivo (27). We speculate that a subpopulation of hepatic ECs or LSECs is senescent with aging. The LSECs-related phenotypic changes during aging were reviewed previously (74). These changes are reduction in the number and size of fenestrations, thickening of the endothelium, deposition of basal lamina and collagen, altered expression of antigens such as von Willebrand factor, CD31 and collagen, and increased perisinusoidal staining with Masson’s Trichrome and Sirius Red. These changes are observed in rodents, nonhuman primates, and humans (25, 26, 74, 107, 110). In addition, the endocytotic activity is impaired with aging. In vitro culture of rodent LSECs causes a rapid loss of the EC phenotype and an onset of spontaneous cellular senescence several days after isolation, which is blocked by the deletion of p19^ARF (89). Whether p19^ARF is involved in the cellular senescence of hepatic endothelium remains unknown in liver diseases.

What signaling pathways promote senescence in hepatic ECs? The role of p53‐p21 and p16‐retinoblastoma protein pathways in cellular senescence has been well established (20, 126). Below, we briefly review the cytoplasmic DNA-sensing pathway in cellular senescence (Fig. 2B). Cyclic GMP-AMP (cGAMP) synthase (cGAS) is a cytosolic DNA sensor that is activated by binding to double-stranded DNA (156). Activated cGAS catalyzes the synthesis of cGAMP by using ATP and GTP as substrates. cGAMP acts as a second messenger to bind and activate the adaptor protein stimulator of interferon genes (STING). This triggers a signaling cascade that leads to the production of type I interferons. Both DNA damage and mitochondrial damage can activate cGAS-STING signaling (99). Nuclear DNA damage causes the accumulation of cytoplasmic DNA, notably in the form of micronuclei. The rupture of micronuclei exposes genomic DNA to cytosolic cGAS. This activates cGAS-STING signaling and type I interferon response. Indeed, DNA damage activates cell-autonomous induction of IFN-β (174). DNA damage-induced IFN-β promotes cell growth arrest and senescence, which is attenuated by anti-IFN-β antibody or knockdown of IFN-β receptor. Mitochondrial stress leads to the release of mitochondrial DNA into the cytosol (166) and activates DNA sensor cGAS that potentiates type I interferon responses and elevates interferon-stimulated gene expression (166). This gene expression profile confers broad viral resistance, indicating that mitochondria play important roles in innate immunity.

The role of cGAS in mediating cellular senescence has been demonstrated in vitro and in vivo (43, 52, 173). In human and mice fibroblasts, cGAS deletion abrogates cell senescence phenotypes induced by replication, DNA damage, or ionizing radiation (173). Increased SA-β-Gal activity and SASP are two important features of senescence, among others, and both are reduced in cGAS-deficient cells (173). Studies from two more groups also demonstrate that different stressors such as oxidative stress and activation of oncogenes can trigger cellular senescence, depending upon cGAS-STING signaling to drive the production of SASP components not only in cells but also in mice (43, 52). Interestingly, cGAS can be recruited to chromosomes during mitosis (57, 173). The functional consequences of exposing cGAS to chromosomes upon mitotic nuclear envelope breakdown were examined recently (180). In normal mitosis, nucleosomes suppress cGAS, so cells divide without cGAS activation. When cells are arrested in mitosis, cGAS-dependent IRF3 activation triggers intrinsic apoptosis by inducing mitochondrial outer membrane permeabilization.

STING-deficient mice carry a missense mutation (I199N) in the STING protein rendering the protein inactive. These HFD-fed STING-deficient mice display higher insulin sensitivity, less hepatic steatosis, and fibrosis (175). A similar study showed that STING-deficient mice developed less severe hepatic steatosis, inflammation, and/or fibrosis in mice with NAFLD or NASH than control mice (105). Bone marrow transplantation revealed that myeloid cell-specific STING disruption leads to similar phenotypes, while myeloid reconstitution of STING has an opposite effect. Mechanistically, STING disruption in macrophages reduces macrophage pro-inflammatory activation, which exerts paracrine effects on hepatocytes by decreasing hepatocyte fat accumulation, pro-inflammatory responses, and the expression of genes involved in fatty acid synthesis. Moreover, cell-autonomous and cell-nonautonomous STING signaling promote the activation of hepatic stellate cells. These data demonstrate that STING plays an important role in regulating diet-induced hepatic steatosis, inflammation, and liver fibrosis. STING expression is increased in livers of human patients with NAFLD and in livers of mice with HFD-induced steatosis (105). Interestingly, it was found that hepatocytes in livers of adult mice did not express STING protein (175). In contrast, STING was expressed in KCs and hepatic ECs (175). Thus, the role of endothelial cGAS-STING signaling axis in hepatic senescence, steatosis, and fibrosis warrants further investigation (Fig. 2B).

PARACRINE EFFECTS OF LSEC

Perturbed EC communication with other cell types plays a key role in the development of diseases such as obesity (98, 127, 128). In liver, ECs can impair insulin action in hepatocytes via tyrosine nitration of insulin receptors, leading to impaired ability of insulin to suppress glucose production (161). Here, we will briefly discuss how LSECs exert paracrine effects on the function of hepatocytes and hepatic stellate cells.

LSECs protect hepatocytes and promote hepatocyte proliferation in partially hepatectomized mice through the release of angiocrine factors Wnt2 and hepatocyte growth factor (42). In a separate study, LSEC-derived angiopoietin-2 displays a dynamic expression pattern after partial hepatectomy (72). In the early inductive phase of liver regeneration, the reduction of angiopoietin-2 enables hepatocyte proliferation through the release of an angiocrine proliferative brake. In the later angiogenic phase of liver regeneration, the expression of angiopoietin-2 was restored in LSEC, which enables regenerative angiogenesis by controlling LSEC vascular endothelial growth factor receptor 2 expression and signaling (72).

Healthy LSECs prevent hepatic stellate cells activation through nitric oxide (35). Moreover, healthy LSECs promote the reversion of activated hepatic stellate cells to quiescence (169), although the paracrine mediator has not been identified. Divergent angiocrine signals from LSECs stimulate regeneration after immediate injury and provoke fibrosis after chronic insult (41). After chronic injury, LSECs promote liver fibrosis by activating hepatic stellate cells. This profibrotic pathway is driven by fibroblast growth factor receptor 1 and C-X-C motif chemokine receptor 4 angiocrine pathway that counteracts the pro-regenerative response (41). EC-derived exosomes regulate hepatic stellate cell function in a paracrine manner (165). Exosomes are cell-derived extracellular vesicles that, in some cases, promote intercellular or interorgan communication. EC-derived exosomes contain the sphingosine kinase 1 protein (165). After being delivered to hepatic stellate cells, the sphingosine kinase 1 protein increased the levels of sphingosine-1-phosphate and Akt phosphorylation, resulting in HSC activation and migration in vitro. In a murine model of liver fibrosis, sphingosine kinase 1 was upregulated in exosomes derived from CCl₄-treated mice compared with the control group, which may activate HSC and contribute to liver fibrosis in vivo (165).

MECHANOTRANSDUCTION

The effects of mechanical force on the functional properties of extrahepatic vascular endothelium have been well demonstrated, as previously reviewed (29, 60, 179), while the role of hepatic endothelium in mechanical sensing is less well understood (83, 153). LSECs are subjected to two major forces (135): 1) cyclic stretch resulted from mechanical distortion and tension of the vessel wall and 2) fluid shear stress caused by blood flow. Studies have demonstrated LSECs as important sensors that link mechanical forces to LSEC function and liver homeostasis. Lorenz et al. (104) found that vascular perfusion induces the proliferation of hepatocytes and the growth of liver in mice during development (Fig. 3A). Mechanistically, vascular perfusion leads to the mechanotransduction in hepatic endothelium, which in turn induces production of HGF through the β1-integrin-VEGFR3 signaling axis. HGF is one of the key angiocrine signals required for liver growth and survival. Mechanical stretching also induces HGF expression by activating the β1-integrin-VEGFR3 signaling axis in hepatic ECs in vitro, which is necessary and sufficient to promote hepatocyte proliferation. The authors also found that mechanical forces in the blood vascular system are associated with liver size in metabolically healthy individuals. These data demonstrated that mechanical forces generated by vascular perfusion in liver vessels play key roles in promoting embryonic liver development (Fig. 3A). Whether it also contributes to regeneration in adult liver remains unexamined. Recently, the impact of pathologic levels of mechanical stretch on LSEC signaling was examined in portal hypertension (Fig. 3B) (69), a common sequelae of chronic liver disease. Mechanical stretch of LSECs induces Notch-dependent upregulation and secretion of the neutrophil chemotactic chemokine CXCL1, which in turn recruit neutrophils in vitro and in vivo. These neutrophils interact with platelets in liver sinusoids, leading to the formation of prothrombotic neutrophil extracellular traps that contribute to microvascular thrombosis, fibrosis, and portal hypertension in different mouse models of liver disease (Fig. 3B). These results revealed a novel pathway of endothelial mechanocrine signaling upon pathological mechanical stretching within hepatic sinusoids that participates in the liver pathophysiology.

Fig. 3.

Mechanotransduction in liver sinusoidal endothelial cells (LSECs). A: mechanotransduction in LSECs induces hepatocyte proliferation and liver growth in mice during development. B: pathological mechanical stretching of LSECs induces the recruitment of neutrophils, which in turn induces microvascular thrombosis and fibrosis and promotes portal hypertension. CXCL1, C-X-C motif chemokine ligand 1; HGF, hepatocyte growth factor; NETs, neutrophil extracellular traps; VEGFR3, vascular endothelial growth factor receptor 3.

CHALLENGES AND FUTURE DIRECTIONS

The progression of chronic liver disease involves a complex interplay among the hepatic ECs, KCs, stellate cells, and hepatocytes. In rodent models, hepatic ECs are known to regulate fibrogenesis and cirrhosis. In human livers, certain subpopulations of ECs express pro-fibrogenic genes and inhabit the fibrotic niche in cirrhotic livers (136). This indicates that hepatic ECs are important regulators of human liver fibrosis and cirrhosis. However, our understanding of hepatic ECs in liver disease is very limited. Below, we discuss some challenges and future directions about the heterogeneity of hepatic ECs, cellular senescence, mechanotransduction, and the role of noncoding RNAs.

Hepatic ECs are highly heterogeneous in both mice and humans, representing the second largest population of cells in liver revealed by single-cell RNA sequencing (136, 170). In mice, liver ECs can be divided into four sub-clusters representing periportal and pericentral ECs and two clusters of LSECs that line the surface of liver sinusoids (170). In humans, clustering of liver ECs identified seven subpopulations, including LSEC, hepatic artery, central vein, lymphatic, two clusters of disease-specific ECs, and one unannotated cluster (136). The heterogeneity of liver ECs revealed by the studies raise two challenges for future studies. First, identification of common molecular signatures that can be used to cluster mouse and human liver ECs is important for improving our understanding about liver EC function in chronic liver disease. Second, generation of LSEC-specific Cre mouse line is important for loss-of-function studies in mice. The common approach to knock out a floxed gene of interest in hepatic endothelium is to breed floxed mice with VE-cadherin promoter-driven Cre mouse line (55, 141). However, this strategy cannot differentiate the heterogeneous population of hepatic ECs.

It is increasingly recognized that cell senescence is heterogeneous and cell specific with pleiotropic function (61–64, 97). There are no specific markers of cellular senescence, and a combination of several markers are often required to identify senescent cells (53). SA-β-Gal activity and expression of p16 are the most commonly used senescence markers (32, 40, 103, 147). However, p16 is not expressed by all senescent cells, while it is also expressed in nonsenescent cells and SA-β-Gal is neither required nor a determinant of the senescent phenotype (53). It is particularly challenging to identify, quantify, and characterize senescent cells in human tissues by SA-β-Gal staining because it is not suitable for fixed tissues or by p16 staining because of the lack of reliable antibodies. Recently, a quantitative approach has been established to identify and characterize senescent cells in aging and disease using ImageStreamX (9, 49). This new approach combines the quantitative power of flow cytometry with high content image analysis. It detects SA-β-gal activities using the bright-field channel and additional senescence-related markers such as p16 and Ki67 (cell proliferation markers) via multiple fluorescence channels (49). It overcomes an important limitation of standard methods by detecting several senescence markers at a time. This also helps identify the specific cell origin of senescent cells (49). However, it requires the specific instrument that is the ImageStreamX imaging flow cytometer.

The mechanobiology in LSECs is an exciting research topic that can potentially provide novel therapeutic targets for chronic liver disease (83, 153). There are several knowledge gaps in this research field. First, it is important to improve our understanding about how to specifically target mechano-sensing pathways in LSECs. The reasons are: 1) several mechano-signaling pathways, including Notch signaling, PIEZO channels, and YAP1 all mediate signaling transduction upon mechanical force in not only LSECs but also in ECs of other vascular beds (83, 153) and 2) although mechano-sensing pathways are involved in the pathogenesis of chronic liver disease, they are also important in maintaining tissue homeostasis and organ regeneration. Second, we need to study the synergistic effects of mechanotransduction and other biochemical signaling on LSEC function, which could lead to the development of combined strategies and agents that can improve the clinical outcomes of chronic liver disease (83). Third, the knowledge about mechanical force in mitochondrial function is limited in hepatic ECs. It has been proposed that mechanical forces regulate mitochondrial calcium signaling in the endothelium (1). Further studies are required to examine the role of mechanical forces in mitochondrial structure and function in hepatic ECs as well as their role in cell metabolism.

Not only protein-coding genes but also noncoding genes control many aspects of physiology and disease. Long noncoding RNAs (lncRNAs) are an important class of RNAs lacking protein coding capacity, and they are typically longer than 200 nucleotides in length. There are 19,175 potentially functional lncRNAs in the human genome, and the vast majority of them are functionally uncharacterized (71). Recently, several lncRNAs have emerged as crucial regulators of obesity and obesity-related pathologies (28, 145, 178). However, the role of lncRNAs in hepatic ECs is completely unknown. By integrating the lncRNA research into the existing body of biological knowledge, we will improve our understanding about how hepatic ECs respond to stress by changing gene expression and signal transduction in liver diseases. This may lead to the development of more effective therapies for liver disease and other chronic metabolic diseases such as obesity, diabetes, and cardiovascular disease.

GRANTS

This work was supported by the National Institutes of Health Funded COBRE Grant 1P20GM104320 (X.S. through the Nebraska Center for the Prevention of Obesity-related Diseases through Dietary Molecules), the American Heart Association SDG#15SDG25400012 (to X.S.), and the National Institutes of Health National Heart, Lung, and Blood Institute Grant R01 HL130864 to E.N.H.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

X.S. and E.N.H. prepared figures; drafted manuscript; edited and revised manuscript; and approved final version of manuscript.