Anandamide induces necrosis in primary hepatic stellate cells

Authors


  • See Editorial on Page 983

  • Potential conflict of interest: Nothing to report.

Abstract

The endogenous cannabinoid anandamide (AEA) is a lipid mediator that blocks proliferation and induces apoptosis in many cell types. Although AEA levels are elevated in liver fibrosis, its role in fibrogenesis remains unclear. This study investigated effects of AEA in primary hepatic stellate cells (HSCs). Anandamide blocked HSC proliferation at concentrations of 1 to 10 μmol/L but did not affect HSC proliferation or activation at nanomolar concentrations. At higher concentrations (25–100 μmol/L), AEA rapidly and dose-dependently induced cell death in primary culture-activated and in vivo-activated HSCs, with over 70% cell death after 4 hours at 25 μmol/L. In contrast to treatment with Fas ligand or gliotoxin, AEA-mediated death was caspase independent and showed typical features of necrosis such as rapid adenosine triphosphate depletion and propidium iodide uptake. Anandamide-induced reactive oxygen species (ROS) formation, and an increase in intracellular Ca2+. Pretreatment with the antioxidant glutathione or Ca2+-chelation attenuated AEA-induced cell death. Although the putative endocannabinoid receptors CB1, CB2, and VR1 were expressed in HSCs, specific receptor blockade failed to block cell death. Depletion of membrane cholesterol by methyl-β-cyclodextrin inhibited AEA binding, blocked ROS formation and intracellular Ca2+-increase, and prevented cell death. In primary hepatocytes, AEA showed significantly lower binding and failed to induce cell death even after prolonged treatment. In conclusion, AEA efficiently induces necrosis in activated HSCs, an effect that depends on membrane cholesterol and a subsequent increase in intracellular Ca2+ and ROS. The anti-proliferative effects and the selective killing of HSCs, but not hepatocytes, indicate that AEA may be used as a potential anti-fibrogenic tool. Supplementary material for this article can be found on the HEPATOLOGYwebsite (http://www.interscience.wiley.com/jpages/0270-9139/suppmat/index.html). (HEPATOLOGY 2005;41:1085–1095.)

Liver fibrosis is a common response to chronic hepatic damage induced by a variety of insults, including viral and parasitic infection, drugs, alcohol abuse, iron and copper overload, nonalcoholic steatohepatitis, and autoimmune diseases. Hepatic stellate cells (HSCs) are believed to play a key role in the development and resolution of liver fibrosis. In the injured liver, HSCs undergo an activation process that results in a phenotypic change from retinoid-storing quiescent cells to activated, collagen-producing HSCs with a myofibroblast phenotype.1 The resolution of liver fibrosis correlates with an increase in HSC apoptosis,2, 3 and current treatment strategies for liver fibrosis include the induction of cell death in HSCs.4 In particular, members of the tumor necrosis factor (TNF) receptor-family including TNFα, TNF-related apoptosis-inducing ligand, nerve growth factor, and Fas ligand (FasL), have been shown to induce apoptosis in HSCs.5–8 Moreover, gliotoxin, a toxin produced by Aspergillus fumigatus, causes apoptosis in HSCs.9

Anandamide (N-arachidonoylethanolamine; AEA) is the main endogenous agonist among a recently discovered class of lipid mediators termed endocannabinoids acting on the cannabinoid receptors CB1, CB2, and vanilloid receptor (VR) 1.10–12 AEA evokes a wide spectrum of physiological actions, such as analgesia, vasodilation, cell proliferation, growth arrest, and cell death, through these receptors.13–19 Anandamide also has been shown to induce cell death independently of these receptors.20, 21 Anandamide is produced in neurons, endothelial cells, platelets, or macrophages11, 19, 22–24 and is believed to reach local concentrations up to 50 μmol/L.20 In hepatic diseases such as acute hepatitis and compensated liver cirrhosis, serum levels of AEA are elevated,20, 25 but the role of AEA in fibrogenesis remains unknown.

In this study, we investigate potential effects of AEA in primary HSCs and hepatocytes. Our study demonstrates that AEA induces necrotic cell death in culture- and in vivo-activated HSCs, but not in hepatocytes. Anandamide-mediated death in HSCs occurs independently of cannabinoid receptors CB1, CB2, and VR1, but requires membrane cholesterol and is mediated by intracellular reactive oxygen species (ROS) and Ca2+. Anandamide may exert antifibrogenic effects in liver fibrosis.

Abbreviations

HSCs, hepatic stellate cells; TNF, tumor necrosis factor; ; FasL, Fas ligand; AEA, N-arachidonoylethanolamide; VR, vanilloid receptor; ROS, reactive oxygen species; GFP, green fluorescent protein; BDL, bile duct ligation; GSH, glutathione; CM-H2DCFDA, 5-(and-6)-chloromethyl-2′,7′-dichlorodihydrofluorescein diacetate; rhTNFα, recombinant human tumor necrosis factor alpha; rmTNFα, recombinant murine tumor necrosis factor alpha; BAPTA-AM, 1,2-bis(2-aminophenoxy)ethane-N, N, N′, N′-tetraacetic acid acetoxymethyl ester; CB, cannabinoid receptor; MCD, methyl-β-cyclodextrin; NFκB, nuclear factor κB; JNK, c-Jun N-terminal kinase; αSMA, alpha smooth muscle actin; ATP, adenosine triphosphate; Z-VAD-FMK, Val-Ala-Asp-fluoromethylketone.

Material and Methods

Hepatic Stellate Cell and Hepatocyte Isolation and Culture.

Primary HSCs were isolated by a 2-step collagenase perfusion from surgical specimens of healthy human livers (n = 3) or from livers of male Sprague-Dawley rats (300–450 g, n = 20) followed by Nycodenz (Axis-Shield, Oslo, Norway) two-layer discontinuous density gradient centrifugation as previously described.26–28 All tissues were obtained by qualified medical staff, with donor consent and the approval of the Ethics Committee of Columbia University. Quiescent and in vivo-activated HSCs were isolated from transgenic mice (n = 12) expressing green fluorescent protein (GFP) under the control of the collagen α1(I) promoter 3 weeks after undergoing bile duct ligation (BDL).29, 30 Purity of human, rat, and mouse HSC preparations was 90%, 96%, and 94%, respectively, as assessed by autofluorescence at day 2 after isolation. Hepatic stellate cells were cultured on uncoated plastic tissue culture dishes as described.26–28 Culture-activated human HSCs were used between passages 2 and 7. Rat and mouse HSCs were not passaged and considered culture-activated between day 7 and 14 after isolation. Rat hepatocytes were isolated from male Sprague-Dawley rats (225–250 g, n = 15) and cultured as previously described.31, 32

Treatment of Cells and Detection of Cell Death.

Cells were serum-starved with Dulbecco's modified eagle medium containing 0.5% fetal calf serum for 12 to 24 hours and treated with either AEA or vehicle (ethanol; 0.1% final concentration), arachidonic acid, ethanolamine, gliotoxin (all from Sigma, St. Louis, MO), actinomycin D (Sigma) plus rhTNFα or rmTNFα (both from R&D Systems, Minneapolis, MN). For some experiments, HSCs were co-cultured with FasL-expressing 3T3 fibroblasts or control 3T3 fibroblasts33 at a ratio of 1:5 (HSC : 3T3). Where indicated, cells were pretreated with the pan-caspase inhibitor Z-VAD-FMK (R&D), the JNK inhibitor SP600125 (Celgene, San Diego, CA), CB1 antagonist SR141716, CB2 antagonist SR144528 (Sanofi-Synthélabo, Montpellier, France), VR1 antagonist capsazepine (Sigma), AEA transport inhibitor AM404 (Cayman Chemicals, Ann Arbor, MI), methyl-β-cyclodextrin (MCD; Sigma), glutathione (GSH; Sigma), BAPTA-AM (1,2-bis(2-aminophenoxy)ethane-N, N, N′, N′-tetraacetic acid acetoxymethyl ester; Molecular Probes, Eugene, OR), BAPTA-tetrapotassium salt (Molecular Probes), or EDTA (Sigma). Cell death in HSCs was measured by lactate dehydrogenase (LDH) release into the culture medium according to the manufacturer's instructions (Roche, Indianapolis, IN). Hepatocyte cell death was measured by propidium iodide (PI; Sigma) fluorescence.34 Caspase 3–like activity was detected by the amino-4-trifluoromethyl coumarin (AFC) release method using DEVD-AFC (MP Biomedicals, Irvine, CA) as substrate.32 Apoptotis and necrosis were visualized by fluorescent microscopy of PI- and Annexin V-staining (Roche) or Hoechst 33258 (Molecular Probes) according to the manufacturer's instructions. DNA laddering was performed as previously described.32

Adenoviral Infection.

Human HSCs were infected with adenoviruses expressing nuclear factor-kappaB (NFκB)-driven luciferase (Ad5NFκBLuc),35 bacterial β-galactosidase (Ad5LacZ),28 NFκB-inducing kinase (AdNIK),36 or green fluorescent protein (Ad5GFP)36 at a multiplicity of infection of 250 (Ad5LacZ) or 500 (all others), achieving transduction rates of greater than 80%. Rat hepatocytes were infected with an adenovirus encoding IκB superrepressor (Ad5IκBsr) at an multiplicity of infection of 50.28

NFκB Responsive Luciferase Assay.

HSCs were coinfected with AdNFκBLuc and the cytomegalovirus-promoter driven Ad5LacZ, treated with agonists for 6 hours, and lysed. Luciferase activity was measured in a multiwell platereader (Fluostar Optima, BMG, Offenburg, Germany) and normalized to β-galactosidase activity as determined by the chlorophenol red-β-D-galactopyranoside-method.

Detection of Reactive Oxygen Species and Intracellular Ca2+.

Serum-starved HSCs (2 × 104 cells/well) were loaded with either the redox-sensitive dye 5-(and-6)-chloromethyl-2′,7′-dichlorodihydrofluorescein diacetate (CM-H2DCFDA; Molecular Probes) or the Ca2+-sensitive dye fluo-4 (Molecular Probes) for 30 minutes at 37°C, washed, and stimulated with AEA. Reactive oxygen species formation or changes in intracellular Ca2+ were measured for the indicated time in a multiwell fluorescence plate reader by using excitation and emission filters of 485 nm and 535 nm, respectively.

[3H]-Anandamide Binding Assay.

Activated HSCs (0.4 × 105/well) or hepatocytes (2 × 105/well) were plated in 12-well dishes and serum-starved for 12 hours. The cells were incubated with [3H]-AEA (205 Ci/mmol; Perkin-Elmer, Boston, MA) in triplicates for 10 minutes and washed extensively with phosphate-buffered saline at 4°C. Extracts were prepared with 0.5 N NaOH containing 0.1% sodium dodecyl sulfate, and measured in a scintillation counter (Perkin-Elmer).

[3H]-Thymidine Incorporation Assay.

Serum-starved HSCs (0.4 × 105/well) were incubated with AEA in the presence or absence of platelet-derived growth factor BB (Sigma). Sixteen hours later, the cells were pulsed with 1 μCi/mL [3H]-thymidine (Amersham Biosciences, Piscataway, NJ) for 8 hours, followed by TCA-precipitation, lysis, and measurement in a scintillation counter.

Reverse Transcription Polymerase Chain Reaction Analysis.

For detection of VR1 mRNA expression, total RNA was isolated from activated human HSCs and Hela cells with TRIzol (Invitrogen, Carlsbad, CA), according to the manufacturer's instructions, followed by reverse transcription,27 and polymerase chain reaction (PCR), using previously described primers.37

Western Blot Analysis.

Electrophoresis of protein extracts and subsequent blotting were performed as described.27 Blots were incubated with anti-caspase-3, anti- poly(ADP-ribose) polymerase (PARP), anti-phospho-Erk (all Cell Signaling Technology, Beverly, MA), anti-IκBα, anti-phospho-c-Jun (both Santa Cruz Biotechnology, Santa Cruz, CA), anti-α-smooth muscle actin (αSMA, Sigma), anti-CB1, or anti-CB2 (both Cayman Chemicals) overnight at 4°C. After incubation with secondary horseradish-peroxidase–conjugated antibodies (Santa Cruz Biotechnology), the bands were visualized by the enhanced chemiluminescence light method (Amersham Biosciences) and exposed to X-omat film (Eastman Kodak Co., New Haven, CT) or a chemiluminescence imager (Image Station 2000R, Eastman Kodak Co.). Blots were reprobed with monoclonal anti-actin (MP Biomedicals) to demonstrate equal loading.

Adenosine Triphosphate Measurement.

Human HSCs (4 × 104/well) were treated with AEA or gliotoxin for 20 minutes. Cellular adenosine triphosphate (ATP) was measured by the luciferin method according to the manufacturer's instructions (Sigma) and normalized to protein content.

Cholesterol Assay.

Serum-starved HSCs were treated with 1 mmol/L MCD or vehicle for 1 hour and lysed as described20 to yield cell membrane–enriched lysates. Unesterified membrane cholesterol was measured in the supernatant applying the fluorometric cholesterol oxidase-resorufin method (Molecular Probes) in a multiwell plate reader at excitation of 544 nm and emission of 590 nm. For cholesterol reloading, HSCs were incubated with a mixture of cholesterol and MCD at a molar ratio of 1:5.

Statistical Analysis.

All data represent the mean of 3 independent experiments ± SEM, if not otherwise stated. For the determination of statistical significance, unpaired Student t tests were performed using SigmaStat (SPSS, Chicago, IL). The correlation coefficient was determined by Pearson correlation using SigmaStat. P values of less than .05 were considered statistically significant.

Results

Anandamide Mediates Cell Death in Activated Human and Rat HSCs.

After addition to the culture media, AEA rapidly and dose-dependently induced cell death in human HSCs (Fig. 1A). After 2 hours of AEA treatment, 24%, 39%, and 58% of human HSCs underwent cell death at concentrations of 25, 50, and 100 μmol/L, respectively. Cell death reached plateau levels after 4 hours at AEA concentrations of 25 μmol/L and higher, achieving a maximum of 83%. Concentrations of 10 μmol/L and lower did not induce significant death in HSCs. Comparable results were obtained in rat HSCs (Fig. 1B). To investigate the possibility that this effect of AEA was mediated by its metabolites ethanolamine and arachidonic acid, we incubated human HSCs with 25 μmol/L of each substance for 4 hours (Fig. 1C). Ethanolamine did not cause any cell death, whereas arachidonic acid induced 19% cell death compared with 78% induced by AEA.

Figure 1.

Anandamide induces cell death in activated human and rat hepatic stellate cells. (A, B) Human (A) and rat (B) HSCs were serum-starved for 12 hours and treated with the indicated concentrations of AEA or vehicle up to 4 hours. Gliotoxin (Glio) was used as control to induce early apoptotic cell death in HSCs. Media was collected after 2 or 4 hours of treatment, and cell death was determined by measuring the release of LDH into the media. Cell death is displayed as a percentage of the maximum LDH release achieved by complete cell lysis with Triton-X 100. (C) Cells were treated with 25 μmol/L arachidonic acid (AA), ethanolamine (EtAm), or AEA for 4 hours. Cell death was assessed by LDH-assay. *P < .05, **P < .001 vs. vehicle treatment. All figures are representative of at least 3 independent experiments performed in duplicates. HSC, hepatic stellate cells; AEA, N-arachidonoylethanolamide; LDH, lactate dehydrogenase.

Anandamide Induces Necrosis, But Not Apoptosis, in Activated Hepatic Stellate Cells.

Anandamide treatment induced a considerably higher degree of LDH release in HSCs than treatment with FasL-expressing 3T3 cells or gliotoxin (Figs. 1A–B, 2D), suggesting the occurrence of necrosis. To determine whether cell death induced by AEA was indeed necrotic, we checked for cleavage of caspase 3 and the 89-kd cleavage product of PARP, which are considered hallmarks of apoptotic cell death. In contrast to treatment with FasL-expressing 3T3 cells or gliotoxin, AEA did not induce caspase 3- and PARP-cleavage, nor caspase 3 activity (Fig. 2A–C). Furthermore, AEA-induced cell death was not blocked by the pan-caspase inhibitor Z-VAD-FMK, which efficiently blocked cell death induced by FasL-expressing 3T3 cells (Fig. 2D). In contrast to HSCs treated with actinomycin D and TNFα, AEA-treated HSCs displayed no Annexin V staining, but showed strong nuclear PI staining, confirming that AEA-induced death is indeed necrotic (Fig. 2E). Furthermore, early depletion of ATP in AEA-treated cells in comparison to the apoptosis-inducing substance gliotoxin underlined the necrotic nature of AEA-caused cell death in HSCs (Fig. 2F).

Figure 2.

Anandamide-induced cell death in activated HSCs is purely necrotic. (A, B) Human HSCs were serum-starved for 12 hours and treated with AEA (25 μmol/L), vehicle, gliotoxin (Glio; 0.75 μmol/L), or FasL-expressing 3T3 fibroblasts for the indicated time. Western blotting was performed with antibodies directed against cleaved caspase 3 (A), cleaved PARP (B), and anti-actin. Blots are representative of 3 independent experiments. (C) Cells were incubated with FasL-expressing 3T3 fibroblasts or treated with AEA (25 μmol/L) for 4 hours. The enzymatic caspase 3 activity assay was performed by fluorometric measurement of AFC release. **P < .001 vs. vehicle treatment. (D) Human HSCs were pretreated with the pan-caspase inhibitor Z-VAD-FMK (20 μmol/L) for 30 minutes followed either by incubation with FasL-expressing 3T3 fibroblasts or treatment with AEA (25 μmol/L) for 6 hours. Cell death was measured by LDH assay. *P < .05 vs. FasL alone. (E) Human HSCs were treated with AEA (25 μmol/L) or vehicle for 4 hours, or with actinomycin D (ActD; 0.2 μg/mL) plus rhTNFα (10 ng/mL) for 18 hours. Apoptotic cell death is indicated by green fluorescence of annexin V, necrotic cell death is shown by red staining of the nuclei by propidium iodide (PI). (F) ATP content of human HSCs was analyzed by the luciferin method in triplicates as described. *P < .05 vs. vehicle treatment. All figures are representative of 3 independent experiments. HSC, hepatic stellate cells; AEA, N-arachidonoylethanolamide; FasL, Fas ligand; PARP, anti-poly(ADP-ribose) polymerase; LDH, lactate dehydrogenase; rhTNFα, recombinant human tumor necrosis factor; ATP, adenosine triphosphate.

Anandamide Inhibits TNFα-Induced NFκB Activation and Stimulates the JNK Pathway in Human HSCs.

To assess whether the NFκB pathway was involved in AEA-induced signaling as suggested in previous studies,38 we measured NFκB-dependent transcription in HSCs in a luciferase reporter assay and early steps of NFκB activation by western blotting for IκBα. Sublethal concentrations of AEA (1–10 μmol/L) did not significantly alter NFκB-dependent transcription or IκBα degradation (Supplementary Fig. 1A–B available on the HEPATOLOGY website: http://interscience.wiley.com/jpages/0270-9139/suppmat/index.html). However, when HSCs were treated with rhTNFα, a strong inducer of NFκB, AEA pretreatment dose-dependently inhibited NFκB activation and IκBα degradation (Supplementary Fig. 1A–B). Anandamide dose-dependently led to a sustained activation of the JNK pathway, as shown by western blotting for phospho-c-Jun (Supplementary Fig. 1C). Because a sustained activation of the JNK pathway in combination with inhibition of the NFκB pathway may contribute to cell death,39 we investigated whether the NFκB and JNK pathways were involved in AEA-mediated death. Interestingly, neither constitutive activation of the NFκB-pathway by adenoviral overexpression of NFκB-inducing kinase, nor inhibition of the JNK pathway by SP600125 were able to prevent the AEA-induced killing of HSCs (Supplementary Fig. 1D–E) suggesting that these pathways are not causatively linked to AEA-induced cell death.

Anandamide-Induced Necrosis Depends on Increases in Intracellular Calcium and ROS.

Anandamide treatment caused a dose-dependent increase in intracellular Ca2+, which was detectable as early as 3 minutes after addition of AEA (Fig. 3A). Pretreatment with either the extracellular Ca2+-chelators EDTA and BAPTA-tetrapotassium salt or the intracellular chelator BAPTA-AM diminished AEA-induced cell death by 70%, 61%, and 62%, respectively (Fig. 3B, P < .001). AEA also caused a marked, dose-dependent, and extremely rapid increase in ROS formation (Fig. 3C). Pretreatment with the antioxidant GSH completely abrogated the generation of ROS by AEA (Fig. 3C) and reduced AEA-mediated cell death by 40% (Fig. 3D; P < .05), indicating that ROS also contribute to AEA-induced cell death.

Figure 3.

Intracellular calcium and ROS contribute to AEA-induced cell death. (A) HSCs were incubated with the Ca2+-sensitive dye Fluo-4 (2 μmol/L) for 30 minutes followed by AEA treatment (25–100 μmol/L). Increase of Ca2+ was analyzed in a multiwell plate reader in quadruplicates as described. (B) Cells were incubated for 30 minutes with EDTA (5 mmol/L), BAPTA-tetrapotassium salt (1 mmol/L) or BAPTA-AM (10 μmol/L) followed by addition of AEA (25 μmol/L) and cell death measurement by LDH assay. (C) Activated human HSCs were loaded with CM-H2DCFDA (5 μmol/L) for 30 minutes and treated with AEA (25–100 μmol/L) alone or pretreated with GSH (4 mmol/L) 1 hour before addition of 25 μmol/L AEA. Reactive oxygen species formation was measured in a multiwell platereader in quadruplicates. (D) After 1 hour of pretreatment with GSH, cells were exposed to AEA (25 μmol/L) for 4 hours. Cell death was analyzed by LDH assay. *P < .05, **P < .001 vs. AEA alone. All figures are representative of at least 3 independent experiments. ROS, reactive oxygen species; AEA, N-arachidonoylethanolamide; HSC, hepatic stellate cells; GSH, glutathione.

Anandamide-Induced Death in Human HSCs Occurs Independently of the Endocannabinoid Receptors CB1, CB2, and VR1.

Because many biological effects of AEA are mediated by its receptors CB1, CB2, or VR1, we checked for the expression of these receptors in human HSCs. We found that all receptors were expressed in HSCs (Fig. 4A). However, blockade of CB1, CB2, and VR1 by the antagonists SR141716, SR144528, and capsazepine, respectively, did not inhibit AEA-induced cell death (Fig. 4B), indicating that AEA induces cell death independently of these receptors. To exclude a facilitated uptake process by the putative AEA membrane transporter (AMT),40 HSCs were treated with the AMT inhibitor AM404. AM404 did not reduce AEA-induced cell death (data not shown).

Figure 4.

AEA-induced killing of HSCs is not mediated by CB1, CB2, and VR1 receptors. (A) The expression of CB1 and CB2 receptors in human HSCs was determined by western blotting, expression of VR1 was assessed by RT-PCR. Human U373 glioma cells (for CB1), Jurkat T-cells (for CB2) and Hela cells (for VR1) were used as positive controls. (B) HSCs were incubated with the specific receptor antagonists SR141716 (SR1; 1 μmol/L), SR144528 (SR2; 1 μmol/L) or capsazepine (Cpz; 25 μmol/L) for CB1, CB2, and VR1 blockade, respectively, 30 minutes before addition of AEA (25 μmol/L for 4 hours). Cell death was analyzed by LDH assay. All figures are representative of 3 independent experiments. AEA, N-arachidonoylethanolamide; HSC, hepatic stellate cells; RT-PCR, reverse transcription polymerase chain reaction; LDH, lactate dehydrogenase.

Anandamide Binding and Anandamide-Induced Cell Death Depend on Membrane Cholesterol.

Previous studies in other cell types have shown that AEA may induce cell death independently of CB1 and CB2 through interaction with membrane cholesterol.20, 21 Pretreatment of HSCs with MCD (1 mmol/L) for 1 hour decreased the membrane cholesterol content by 56% (Fig. 5A; P < .001). MCD preincubation completely inhibited necrosis induced by 25 μmol/L AEA and significantly (P < .001) decreased cell death induced by 50 μmol/L AEA by 75% (Fig. 5B). However, at 100 μmol/L AEA, MCD reduced cell death only by 10%. Similar results were obtained in rat HSCs (data not shown). In contrast to AEA-induced necrosis, apoptosis caused by either gliotoxin (Fig. 5B), TNFα, or FasL (data not shown) was not inhibited by MCD preincubation. To further demonstrate the causative role of cholesterol in AEA-mediated necrosis, HSCs were reloaded with cholesterol (Fig. 5C, left panel), which completely restored HSC sensitivity to AEA-mediated cell death (Fig. 5C, right panel). MCD pretreatment also inhibited downstream events of AEA signaling, including ROS formation (Fig. 5D, left panel) and intracellular Ca2+ increase (Fig. 5D, right panel). [3H]-AEA binding assays showed that AEA was taken up by the cells in a dose-dependent manner (Fig. 5E, left panel) and AEA binding significantly correlated with its cytotoxic effects (R = 0.970, P < .01). Interestingly, MCD decreased [3H]-AEA binding to HSCs and its subsequent uptake by 71% (Fig. 5E, right panel; P < .05), suggesting either that membrane cholesterol directly interacts with AEA or that its presence is required to allow AEA to bind to the cell membrane.

Figure 5.

Membrane cholesterol depletion by methyl-β-cyclodextrin (MCD) prevents binding of AEA to activated HSCs and inhibits AEA-induced necrosis. (A) Unesterified membrane cholesterol was measured in membrane-rich cell lysates of human HSCs in the absence or presence of MCD (1 mmol/L). **P < .001 vs. untreated. (B) Human HSCs were pretreated with MCD (1 mmol/L) for 1 hour followed by treatment with the indicated concentrations of AEA for 4 hours. Cell death was determined by LDH assay (**P < .001 vs. AEA alone). (C) HSCs were either left untreated, incubated with MCD, or incubated with MCD followed by cholesterol reloading. Unesterified membrane cholesterol was analyzed as described (left panel; *P < .05 vs. vehicle treatment; # P < .05 vs. cholesterol reloaded). Cell death was determined by LDH assay in cells after 4 hours of 25 μmol/L AEA (right panel). **P < .001 vs. vehicle treatment. (D) Human HSCs were preincubated with MCD for 1 hour followed by loading with CM-H2DCFDA or fluo-4. Immediately after AEA addition, ROS formation (left panel) and intracellular Ca2+-increase (right panel) were determined in quadruplicates in a multiwell platereader. (E) Human HSCs were treated with the indicated concentrations of [3H]-AEA for 10 minutes, washed, and lysed as described. [3H]-binding was measured in a scintillation counter (left panel) and normalized to the number of cells per dish (expressed as cpm/cell). In the right panel, HSCs were preincubated with MCD 1 hour before addition of [3H]-AEA (25 μmol/L) and the binding assay was performed as described (*P < .05 vs. [3H]-AEA alone). All figures are representative of 3 independent experiments. AEA, N-arachidonoylethanolamide; HSC, hepatic stellate cells; LDH, lactate dehydrogenase; ROS, reactive oxygen species.

Anandamide Does Not Induce Cell Death in Primary Rat Hepatocytes.

Anandamide has been shown to induce apoptosis in HepG2 cells and primary rat hepatocytes at concentrations from 0.5 to 10 μmol/L.20 However, even at concentrations of 100 μmol/L, we did not observe cell death in primary rat hepatocytes (Fig. 6A). Additionally, AEA did not cause caspase 3 or PARP cleavage in primary rat hepatocytes (Fig. 6B). AEA-treated hepatocytes did not display nuclear condensation/fragmentation (Fig. 6C, left panel), Annexin V/PI staining (Fig. 6C, right panel) or DNA-laddering (Fig. 6D), whereas hepatocytes treated with recombinant murine TNFα (rmTNFα) plus actinomycin D showed all these hallmarks of apoptotic cell death. Because cell death in hepatocytes may require inhibition of NFκB,41 we also infected hepatocytes with Ad5IκBsr followed by AEA or rmTNFα treatment. Even in Ad5IκBsr-infected hepatocytes, AEA did not cause cell death (data not shown). To further analyze this remarkable difference in susceptibility to AEA-induced cell death, we performed [3H]-AEA binding experiments in rat HSCs and hepatocytes. Although [3H]-AEA dose-dependently bound to both HSCs and hepatocytes, [3H]-AEA binding to hepatocytes was significantly lower and even at 100 μmol/L never reached the binding levels that were associated with cell death in HSCs (Fig. 6E). Interestingly, membrane cholesterol was significantly lower in rat hepatocytes (Fig. 6F), suggesting that this may be a factor responsible for lower AEA binding and cell death in hepatocytes.

Figure 6.

Anandamide does not induce cell death in primary rat hepatocytes. (A) Hepatocytes were loaded with PI and treated with AEA (5–100 μmol/L) or actinomycin D plus rmTNFα for 22 hours. A fluorometric PI-assay indicating necrotic cell death was performed as described in Material and Methods (**P < .001 vs. vehicle treatment). (B) Primary rat hepatocytes were treated with AEA (25 μmol/L), actinomycin D plus rmTNFα, or vehicle only. Western blotting was performed with antibodies directed against cleaved caspase 3, PARP, or actin. (C) Primary rat hepatocytes were treated with vehicle or AEA (100 μmol/L) for 24 hours or with actinomycin D (ActD; 0.2 μg/mL) plus rmTNFα (10 ng/mL) for 10 hours. The left panel (×80 magnification) shows nuclei stained with Hoechst 33258 (10 μg/mL) displaying nuclear condensation and fragmentation where indicated (white arrows). In the right panel (×40 magnification), apoptotic cell death is indicated by green fluorescence of annexin V (Ann V) and necrotic cell death is shown by red fluorescence of nuclei stained by PI. (D) Primary rat hepatocytes were incubated with vehicle or AEA (10–100 μmol/L) for 24 hours or with actinomycin D plus rmTNFα for 10 hours. DNA was extracted and analyzed on a 1.8% agarose gel. (E) Primary rat HSCs and primary rat hepatocytes were treated with the indicated concentrations of [3H]-AEA for 10 minutes. [3H]-AEA binding was performed as described. (F) Unesterified membrane cholesterol was measured in primary rat hepatocytes and HSCs as described. *P < .05 vs. rat hepatocytes. All figures are representative of at least 3 independent experiments. PI, propidium iodide; AEA, N-arachidonoylethanolamide; rmTNFα, recombinant murine tumor necrosis factor alpha; PARP, anti-poly(ADP-ribose) polymerase.

Sublethal Doses of AEA Reduce DNA-Synthesis in Activated Human HSCs.

Previous studies have shown that AEA may either induce or inhibit proliferation.15–17 In human HSCs, sublethal doses of AEA (1, 5, and 10 μmol/L) reduced the amount of incorporated [3H]-thymidine by 38%, 35%, and 46%, respectively (Supplementary Fig. 2A, left panel; P < .05). When we stimulated human HSCs with the complete mitogen platelet-derived growth factor BB, the inhibitory effect of AEA on DNA synthesis became even more apparent: 1, 5, and 10 μmol/L AEA inhibited DNA synthesis by 54%, 35%, and 57%, respectively (Supplementary Fig. 2A, right panel; P < .05). Because nanomolar doses of AEA may increase cell proliferation, we tested the effects of AEA (1–1,000 nmol/L) on HSC proliferation and activation. Nanomolar concentrations of AEA did not affect [3H]-thymidine uptake (Supplementary Fig. 2A) or Erk activation (Supplementary Fig. 2B). Furthermore, nanomolar concentrations of AEA did not affect HSC activation as judged by the expression of collagen α1(I)-driven GFP and western blotting for αSMA (Supplementary Fig. 2C–D).

Anandamide Induces Necrosis in In Vivo-Activated HSCs.

To exclude the possibility that culture-activation of HSCs artificially modulated their sensitivity to AEA-induced death, we performed BDL in transgenic mice expressing the collagen α1(I)-driven GFP reporter gene. Three weeks after BDL, HSCs were isolated and showed an activated phenotype including the expression of the collagen-driven GFP reporter (Fig. 7A). In vivo-activated HSCs were treated with AEA at the day of isolation and were extremely sensitive to AEA-mediated cell death: More than 60% of in vivo-actived HSC underwent cell death after 2 hours of 25 μmol/L or 50 μmol/L AEA (Fig. 7B). AEA treatment (25 μmol/L) of in vivo-activated HSCs induced the uptake of PI within minutes, demonstrating the rapid onset of necrosis, whereas vehicle-treated in vivo-activated HSCs displayed no PI uptake (Fig. 7C). After 2 hours, almost all in vivo-activated HSCs were PI-positive.

Figure 7.

AEA induces cell death in in vivo-activated mouse HSCs. (A) Hepatic stellate cells were isolated from transgenic mice expressing GFP under the control of the collagen α1(I) promoter 3 weeks after sham operation (left panel, day 0) or bile duct ligation (right panel, day 0). (B) On the day of isolation (day 0), in vivo-activated mouse HSCs were treated with vehicle, 25 μmol/L or 50 μmol/L AEA for 2 or 4 hours. Cell death was analyzed by LDH assay (**P < .001 vs. vehicle treatment). (C) In vivo-activated HSCs (day 0) were incubated with PI followed by treatment with 25 μmol/L AEA or vehicle for the indicated times. The pictures show overlays of GFP indicating collagen α1(I) expression and red fluorescence of nuclei stained by PI demonstrating necrotic cell death. All figures are representative of 3 independent experiments. AEA, N-arachidonoylethanolamide; HSC, hepatic stellate cells; GFP, green fluorescent protein; LDH, lactate dehydrogenase; PI, propidium iodide.

Discussion

Although liver fibrosis has been regarded as irreversible, recent studies in animal models and patients have suggested that hepatic fibrosis is, at least to some degree, a reversible process.2, 3 Elimination of activated HSCs has been linked to the reversal of liver fibrosis, and treatments that induce HSC apoptosis, such as gliotoxin, are currently under investigation as potential treatments for liver fibrosis.9, 42, 43

Our study investigates the effects of the endocannabinoid AEA on cell death, activation, and proliferation of HSCs. We demonstrate that AEA blocks HSC proliferation at concentrations from 1 to 10 μmol/L and that AEA induces cell death at concentrations of 25 to 100 μmol/L in a dose-dependent manner. Nanomolar concentrations of AEA do not affect HSC proliferation or activation. Anandamide-mediated cell death is highly efficient, with over 80% cell death in culture-activated HSCs after 4 hours at 25 μmol/L. Moreover, AEA also induces cell death in in vivo-activated HSCs with over 60% cell death after 2 hours at 25 μmol/L. Although this rapid onset of cell death is similar to the effect of gliotoxin on HSCs, our study shows substantial differences between gliotoxin- and AEA-mediated cell death. Whereas gliotoxin induces apoptotic cell death unless used in very high doses,42 AEA induces a purely necrotic form of cell death. Anandamide-treated HSCs show no signs of apoptosis such as cellular shrinkage, caspase activation/PARP cleavage, and positive Annexin V staining. Moreover, the caspase inhibitor Z-VAD-FMK does not rescue HSCs from AEA-induced death. Because endocannabinoid-induced cell death in neurons and other cell types is apoptotic and occurs much later,16–18, 20, 21, 37 we attempted to further characterize the signaling events that lead to AEA-induced necrosis in HSCs.

Endocannabinoids mediate many of their effects, including cell death, through CB1, CB2, and VR1 receptors.13–19, 22–24 Although HSCs express CB1, CB2, and VR1, AEA-induced cell death still occurs in the presence of antagonists of these receptors. Previous studies have demonstrated that membrane cholesterol may mediate AEA-induced cell death.20, 21 In our study, cell membrane cholesterol depletion not only prevents binding of AEA to HSCs, but also effectively inhibits ROS formation, intracellular Ca2+-release, and cell death. Moreover, reloading of HSCs with cholesterol restores the sensitivity toward AEA-mediated death. AEA-induced cell death may be mediated by its polyunsaturated acyl chain, which it shares with its precursor and metabolite arachidonic acid. Arachidonic acid is known to induce necrotic cell death because of detergent-like perturbation of the cell membrane and increased ROS formation followed by lipid peroxidation at different threshold concentrations, depending on the examined cell type.44 In our study, arachidonic acid also induces ROS formation (data not shown) and cell death in HSCs, but to a lesser extent than AEA. AEA and arachidonic acid likely share a common mechanism to induce cell death in HSCs, but structural differences between these substances may be responsible for lower binding of arachidonic acid to HSCs or lower toxicity.

The analysis of downstream events shows that AEA induces an immediate burst of ROS and an early increase in intracellular Ca2+ in HSCs and that both these events contribute to AEA-induced cell death. High amounts of suddenly generated ROS are known to decrease the cellular oxidative defense, leading to lipid peroxidation and causing rapid damage to vitally important intracellular macromolecules, thus initiating necrosis.45 Additionally, ROS trigger the depletion of ATP by directly damaging mitochondrial DNA, enzymes such as ATP synthase, and cell organelle membranes.45, 46 We observe an early depletion of ATP in AEA-treated HSCs, one of the key features of necrotic cell death.47, 48 Loss of ATP leads to inactivation of Na+/K+-ATPase with subsequent increase of intracellular Na+ and loss of K+, accompanied by Cl increase and water influx entailing cell swelling and membrane blebbing,45, 49 features that occur in AEA-treated HSCs in our study. Increased intracellular Ca2+ is another pivotal feature of necrosis that we observed in HSC. Elevation of intracellular Ca2+ is believed to trigger the activation of proteases, endonucleases, and phospholipases, all of which contribute to the cellular changes during the prelethal phase.45, 49 Although our study demonstrates that AEA inhibits TNFα-mediated NFκB activity and concomitantly activates the JNK pathway, a constellation that promotes cell death in many cell types,39 we do not find evidence that these pathways contribute to AEA-mediated death.

Our study does not provide data on the effects of endocannabinoids in animal models of hepatic fibrosis. Because AEA is elevated in patients with liver disease and liver fibrosis and is believed to reach local levels of up to 50 μmol/L,20, 25 it is possible that AEA functions as an endogenous anti-fibrogenic mediator. Conversely, we observed that AEA was only moderately elevated in mice 3 weeks after BDL, whereas the endocannabionids 1-arachidonoyl glycerol and 2-arachidonoyl glycerol were elevated 2.5-fold (data not shown). Although AEA is possibly more substantially elevated in earlier stages of fibrosis or in more inflammatory models of fibrosis, it seems unlikely that AEA concentrations constantly reach levels that are sufficient to induce HSC death and prevent fibrosis. Therefore, AEA should be viewed rather as a target for anti-fibrotic therapy than as an endogenous anti-fibrogenic mediator. The potential use of endocannabinoids in hepatic fibrosis is supported by our findings that (i) in vivo-activated HSCs are extremely sensitive to AEA-mediated necrosis and that (ii) AEA does not induce apoptosis or necrosis in primary hepatocytes as demonstrated by a multitude of assays. This striking difference in the response to AEA is, at least in part, explained by the significantly lower membrane cholesterol content and the lower AEA binding in hepatocytes, but intracellular factors such as the AEA-degrading enzyme fatty acid amide hydrolase are also likely to play a role. Our results are in contrast with those of a previous study showing that AEA mediates cell death in hepatocytes and HepG2.20 We have confirmed that hepatoma cell lines including HepG2 are susceptible to AEA-mediated death (data not shown), and this finding may be of interest in view of the increased rate of premalignant and malignant lesions in the fibrotic liver.

In conclusion, AEA may exert anti-fibrogenic effects by inhibiting HSC proliferation at low concentrations and by inducing HSC death at concentrations above 10 μmol/L. Further studies are needed to investigate in detail (i) whether endocannabinoids contribute to HSC death during hepatic fibrogenesis and its resolution in vivo and (ii) whether AEA may be useful for the treatment of hepatic fibrosis. Possibly AEA derivatives with a longer intracellular half-life such as methanandamide may be more useful for in vivo use.50 Because AEA-induced cell death is necrotic and may therefore increase inflammation and tissue damage, future experiments need to consider potential proinflammatory effects. Moreover, the vasodilative properties of endocannabinoids have to be taken into account in advanced liver cirrhosis.14, 19

Acknowledgements

The authors thank Sanofi-Synthélabo (Montpellier, France) for SR141716 and SR144528. We are grateful to George Kunos (NIH-NIAAA, Bethesda, MD) for helpful discussion. We thank Judith Harvey-White for technical support (NIH-NIAAA).

Ancillary