Endocannabinoid System Contributes to Liver Injury and Inflammation by Activation of Bone Marrow–Derived Monocytes/Macrophages in a CB1-Dependent Manner

The endocannabinoid system (ECS) is an extremely complex system. It is comprised of endogenous ligands N-arachidonoylethanolamine (anandamide, AEA) and 2-arachidonoylglycerol (2-AG), two types of G protein–coupled cannabinoid receptors (CBs; CB1 and CB2), and the corresponding biosynthetic enzymes (N-acylphosphatidyl-ethanolamine selective phospholipase D [NAPE-PLD] for AEA, diacylglycerol lipase [DAGL] for 2-AG) and degradative enzymes (fatty acid amidohydrolase [FAAH] for AEA, monoacylglycerol lipase [MAGL] for 2-AG) (1). CB1 is highly expressed in the CNS, and lower yet functionally relevant in the periphery under physiological conditions (1), whereas CB2 is predominantly expressed in immune cells and participates in modulating the immune system (2).

CBs mediate various physiological and pathological processes ranging from memory and cognition, lipid metabolism, and inflammatory regulation to gastrointestinal diseases, obesity and metabolic disorders, cardiovascular diseases, and almost all liver diseases (3, 4). The CB1 signal was involved in maintaining food intake and promoting lipogenesis in adipocytes (5, 6). Activation of CB1 increased the hepatic gene expression of the lipogenic transcription factor SREBP-1c and its target acetyl-CoA carboxylase-1 in mice, and it significantly increased de novo fatty acid synthesis in the liver or in hepatocytes simultaneously (7). Blockade of CB1 reduced the basal rates of fatty acid synthesis. In alcohol-induced fatty liver, activation of hepatic CB1 via 2-AG, which was secreted by stellate cells, mediated ethanol-induced steatosis through increasing lipogenesis and decreasing fatty acid oxidation (8). Rimonabant (CB1 antagonist) reduced the high levels of enzyme markers of hepatic damage in plasma and the level of proinflammatory cytokine TNF-α (9). Moreover, activation of CB1 contributed to the fibrogenic process in liver injury by affecting growth and apoptosis toward hepatic myofibroblasts (10) and aggravated the hemodynamic dysfunction and portal hypertension by mediating vasodilation in advanced cirrhosis (11). Alternatively, CB2 displayed anti-inflammatory and anti-fibrogenic properties by affecting inflammatory and fibrogenic cells in liver injury (12). In the cirrhotic model, activation of CB2 decreased inflammatory cell infiltration and the number of activated stellate cells (13). Additionally, CB2 protected against hepatic ischemia/reperfusion (I/R) injury by reducing inflammatory response and oxidative stress (14). In alcohol-induced liver injury, activation of CB2 reduced hepatic steatosis and alcohol-induced inflammation by regulating M1/M2 balance in hepatic macrophages (15).

Macrophages are found in various tissues and play significant roles in development, homeostasis, and diseases (16). Macrophages are not only significant members of the immune system, but they also contribute to fibrogenic disorders such as liver fibrosis through driving myofibroblast activation via profibrogenic cytokines such as TGF-β1 (17, 18). When tissues undergo serious injury, there are not enough indigenous macrophages (Kupffer cells in the liver) to cope with the violent destructive agents. Therefore, copious amount of macrophages would be recruited from bone marrow (BM), aggregate in inflammatory sites, and play a crucial role in inflammation and fibrosis (16). Indeed, it has been reported that the inflammation and fibrosis of injured liver were ameliorated after the depletion of macrophages (19). We have also found that reducing the recruitment of BM-derived monocytes/macrophages (BMM) can attenuate hepatic inflammation and fibrosis (20). Furthermore, in alcoholic liver disease, CB2 played a protective role by regulating macrophage polarization (15). Interestingly, recent studies found that CB1 exhibited proinflammatory effects on macrophages in type 2 diabetes (21) and atherogenesis of coronary artery disease (22). However, few studies have focused on the relationship between CB1 and macrophages in liver diseases.

In this study, we investigated the effects of ECS on BMM in liver injury. We found that the biosynthetic enzymes of endocannabinoids were significantly elevated in human fibrotic livers. Meanwhile, we detected changes in the level of the synthetase and catabolic enzymes of endocannabinoids in different mouse tissues after carbon tetrachloride (CCl₄) administration and found that synthetases (NAPE-PLD and DAGL-α) showed positive correlations with fibrosis parameters (procollagen α1(III) [Col α1(III)]) in injured liver. Focusing on BMM, we investigated the expressions of CB1 and CB2, as well as the roles of CB1 and CB2, in BMM migration and secretion of inflammatory cytokines by BMM. We discovered that CB1 but not CB2 was involved in the migration and activation of BMM. Furthermore, the blockade of CB1 suppressed BMM infiltration and ameliorated the inflammation and fibrosis in CCl₄-induced liver injury.

RPMI 1640 was from Life Technologies/Invitrogen (Grand Island, NY). FBS was from HyClone/Thermo Scientific (Scoresby, VIC, Australia). PCR reagents were from Applied Biosystems (Foster City, CA). Methanandamide (mAEA) and ACEA (both special CB1 agonists), JWH133 (special CB2 agonist), AM281 (CB1 antagonist), and AM630 (CB2 antangonist) were from Tocris Bioscience/R&D Systems (Minneapolis, MN). The pHrodo Zymosan A BioParticles (P35364 and P35365) were from Life Technologies (Carlsbad, CA). Microparticles based on polystyrene (dark red, 3 μm), Y27632 (the inhibitor of Rho-associated protein kinase ROCK), pertussis toxin (PTX), and other common reagents were from Sigma-Aldrich (St. Louis, MO).

We retrospectively studied snap-frozen surgical liver resections from 21 patients (13 men, 8 women; mean age, 56 y; range, 42–69 y). Normal liver samples were collected from five patients undergoing hepatic resection for colorectal metastasis (n = 5). Fibrotic samples (fibrosis stage F2–4) were obtained from 16 livers of patients undergoing liver transplantation. Fibrosis was consecutive to chronic hepatitis C virus (n = 4) or hepatitis B virus (n = 10) infections, as well as alcohol-induced liver disease (n = 2). All tissues were obtained with donor consent and the approval of the Capital Medical University Ethics Committee (2011SY08).

ICR mice aged 3 wk were sacrificed by cervical dislocation at the time of BM harvest. BM cells were extracted from the tibias and femurs by flushing with culture medium using a 25-gauge needle. The cells were then passed through a 70-mm nylon mesh (BD Biosciences, Franklin Lakes, NJ) and were washed three times with PBS containing 2% FBS. Extracted BM cells were implanted with 2.4 × 10⁷/100 mm culture dish (BD Falcon) and cultured for 7 d in the presence of L929-conditioned medium (replacing culture medium at days 3 and 5). All animal work was performed under the ethical guidelines of the Ethics Committee of Capital Medical University.

ICR male mice aged 6 wk received lethal irradiation (8 Gy) and immediately received transplantation by a tail vein injection of 1.5 × 10⁷ whole BM cells obtained from 3-wk-old enhanced GFP (EGFP) transgenic mice. Four weeks later, mice of BM rebuild were subjected to CCl₄-induced liver injury. After another 4 wk, mice were sacrificed and liver tissues were harvested.

A CCl₄ (1 μl/g body weight)/olive oil (OO) mixture (1:9 [v/v]) was injected into abdominal cavity of mice twice per week. Mice were sacrificed at 1 d, 3 d, and 1, 2, 4, and 8 wk. The liver tissues were harvested, as were brains, hearts, BM, spleens, and kidneys, at 4 wk. The i.p. injection of AM281 (2.5 mg/kg body weight) was performed 4 h before CCl₄ administration (n = 6/group).

BMM that had been cultivated for 7 d were fixed by 4% paraformaldehyde for 30 min and penetrated by 0.5% Triton X-100 (Amresco, Solon, OH) for 15 min. After blocking with 3% BSA (Roche, Basel, Switzerland), they were incubated with anti-F4/80 rat mAb (1:100, Santa Cruz Biotechnology, Santa Cruz, CA), anti-CB1, or anti-CB2 rabbit polyclonal Abs (1:50, Cayman Chemical, Ann Arbor, MI). FITC-conjugated AffiniPure goat anti-rat IgG or Cy3-conjugated AffiniPure goat anti-rabbit IgG (both 1:100, Jackson ImmunoResearch Laboratories, West Grove, PA) was as a secondary Ab. Finally, nuclei were stained with DAPI.

BM samples were extracted from the tibias and femurs of mice. The cell suspension was subjected to the lysis buffer of RBCs for 15 min and then centrifuged at 1200 rpm for 10 min. After being washed by PBS, the BM cells were fixed by 4% paraformaldehyde for 30 min and made into BM smear. The immunofluorescence staining of CB1 (1:50, Cayman Chemical), F4/80 (1:100, Santa Cruz Biotechnology), CD3e (1:100, BD Biosciences), and CD11c (1:100, BD Biosciences) on BM smear was the same as that in BMM.

The liver specimen was fixed in 4% paraformaldehyde and frozen sections of 6 μm were used for immunofluorescence and paraffin sections for H&E and sirius red stain. Frozen sections were incubated with F4/80 Ab as the first Ab and Cy3-conjugated goat anti-rat IgG (1:100, Jackson ImmunoResearch Laboratories) as the secondary Ab. Finally, the sections were stained with DAPI and observed under a confocal microscope (LSM510, Carl Zeiss MicroImaging, Jena, Germany). Image-Pro Plus was used for the quantitation. For costaining of F4/80 and sirius red, the same liver sections were stained with sirius red for 20 s. Sirius red staining was performed to detect collagen deposition and observed under a light microscope.

BMM cultivating for 7 d were starved for 24 h. Then, 4 × 10⁴ BMM with or without pretreatment of AM281 for 1 h were seeded to the upper chamber of a Boyden chamber system and mAEA or JWH133 of different concentration was placed in the lower wells. The migration system was put in the incubator of 37°C and 5% CO₂ for 4 h. Next, BMM that migrated to the lower surface of the filter were fixed with ice-cold methanol for 20 min and were stained for 30 min with hematoxylin. At last, the cells that did not migrate to the lower surface were erasured with cotton swabs and the cells that did migrate were imaged with a microscope and quantified by cell counting.

Total RNA was extracted from liver frozen specimens or cultured BMM with or without treatments using an RNeasy kit (Qiagen, Hilden, Germany). Real-time RT-PCR was performed in an ABI Prism 7300 sequence detecting system (Applied Biosystems, Foster City, CA). Primers (MWG Biotech, Ebersberg, Germany) used for real-time RT-PCR were as follows: 18S rRNA, sense, 5′-GTA ACC CGT TGA ACC CCA TT-3′, antisense, 5′-CCA TCC AAT CGG TAG TAG CG-3′; mouse α-SMA, sense, 5′-ATG CTC CCA GGG CTG TTT T-3′, antisense, 5′-TTC CAA CCA TTA CTC CCT GAT GT-3′; mouse Col α1(I), sense, 5′-AGG GCG AGT GCT GTG CTT T-3′, antisense, 5′-CCC TCG ACT CCT ACA TCT TCT GA-3′; mouse Col α1(III), sense, 5′-TGA AAC CCC AGC AAA ACA AAA-3′, antisense, 5′-TCA CTT GCA CTG GTT GAT AAG ATT AA-3′; mouse NAPE-PLD, sense, 5′-CGA CGC TGA TGG TGG AAA T-3′, antisense, 5′-AGA GGC ACG GGA GCT GAA-3′; mouse FAAH, sense, 5′-TGG GAA CTG CAG CAT GAG ATT-3′, antisense, 5′-GGG TTA GCA CCA CGT CCA A-3′; mouse DAGL-α, sense, 5′-TCC CGC CGT CTA AAA GTG TT-3′, antisense, 5′-GCG ATT TCT GAG TAG GCA TCT GA; mouse DAGL-β, sense, 5′-TCT CCA CCA GCA ACA AGA CAA TA-3′, antisense, 5′-CAC TTC TGC ATC TAA CTC AGT TTC CT-3′; mouse MAGL, sense, 5′-CAA GAG TGG AGC GAG CAA TG-3′, antisense, 5′-TGA CTC CGG GAT GAT TCC AT-3′; mouse IL-6, sense, 5′-CTC TGG GAA ATC GTG GAA ATG-3′, antisense, 5′-AAG TGC ATC ATC GTT GTT CAT ACA-3′; mouse IL-10, sense, 5′-CAG TAC AGC CGG GAA GAC AAT AA-3′, antisense, 5′-CCG CAG CTC TAG GAG CAT GT-3′; mouse IFN-γ, sense, 5′-TCT GAG ACA ATG AAC GCT ACA CAC T-3′, antisense, 5′-TGG CAG TAA CAG CCA GAA ACA-3′; mouse IL-12, sense, 5′-CCT GAA GTG TGA AGC ACC AAA TT-3′, antisense, 5′-CTT CAA GTC CAT GTT TCT TTG CA-3′; mouse F4/80, sense, 5′-AGC ACA TCC AGC CAA AGC A-3′, antisense, 5′-CCA TCT CCC ATC CTC CAC AT-3′; mouse TNF-α, sense, 5′-GGC AGG TTC TGT CCC TTT CA-3′, antisense, 5′-CTGTGCTCATGGTGTCTTTTCTG-3′; mouse MCP-1, sense, 5′-TCT GGG CCT GCT GTT CAC A-3′, antisense, 5′-GGA TCA TCT TGC TGG TGA ATG A-3′; human NAPE-PLD, sense, 5′-CTG GGA CAT GCC ACG GTA AT-3, antisense, 5′-TCG CTT TGG ACC CAT GTA CTG-3′; human FAAH, sense, 5′-GCA CAC GCT GGT TCC CTT CTT G-3′, antisense, 5′-GTT TTC CAG CCG AAC GAG ACT TCA TGT TG-3; human DAGL-α, sense, 5′-TGG GTT TGG GAA GCA GTT AGG-3′, antisense, 5′-TGC ATA CGC GCC AAC ACT A-3′; human DAGL-β, sense, 5′-GCG GAA TTC AGC AAA ATA CTC A-3′, antisense, 5′-TCC AAG GCC CGC ATC A-3′; human MAGL, sense, 5′-AGT CAC CAA CTC CGT CTT CCA T-3′, antisense, 5′-ACC GGC CAA TGC ATT CAG-3′; human Col α1(III), sense, 5′-AGC TGG AAA GAG TGG TGA CAG-3′, antisense, 5′-CCT TGA GGA CCA GGA GCAC-3′; and mouse CB1 (Mm00432621_s1) and mouse CB2 (Mm00438286_m1) probes were from Applied Biosystems.

Liver tissues were fixed in PBS containing 4% paraformaldehyde for 24 h and embedded in paraffin. Sections (5 μm) were stained with H&E for analysis of necrotic area and sirius red for collagen visualization. The necrotic area and fibrotic area were assessed by computer-assisted image analysis with MetaMorph software (Universal Imaging, Downingtown, PA). The mean value of 15 randomly selected areas per sample was used as the expressed percentage of fibrosis or necrosis area.

Western blot analysis was performed with 50 μg protein extract. Abs were as follows: rabbit anti–NAPE-PLD polyclonal Ab (1:1000, Novus Biologicals, Littleton, CO); rabbit anti-FAAH affinity-purified IgG (1:1000), anti-CB1 rabbit polyclonal Ab (1:100), and anti-CB2 rabbit polyclonal Ab (1:100) (Cayman Chemical); and rabbit anti–β-actin and anti-GAPDH mAbs (1:1000, Abcam, Cambridge, U.K.). Odyssey goat anti-rabbit IRDye 800 CW Ab (1:10000, LI-COR Biosciences, Lincoln, NE) was used as secondary Ab. The bands were displayed using Odyssey and quantified by Odyssey v3.0 software. β-actin or GAPDH was used as reference.

The ON-TARGETplus mouse CB1 small interfering RNA (siRNA) SMARTpool (L-042461-00), CB2 siRNA SMARTpool (L-062503-00), and nontargeting control pool (D-001810-10-05) were from Dharmacon (Thermo Scientific, Pittsburgh, PA). Forty to 50% confluent BMMs were prepared in 60-mm dishes. Transient transfection of siRNA (40 nmol/l) was performed by using Lipofectamine RNAiMAX (Invitrogen, Carlsbad, CA) as recommended by the manufacturer. Control cells were treated with 40 nmol/l RNA interference negative control duplexes (scramble siRNA). After 48 h, cells were used to perform the following assay.

Livers were perfused with 20 ml PBS, minced with scissors, and digested for 24 min with collagenase type IV at 37°C. Digested extracts were filtrated through 70-mm cell strainers to achieve single-cell suspensions. The cell suspension was subjected to density gradient (Histopaque-1077) centrifugation at 2000 rpm for 20 min. The cells were collected from the interface after centrifugation, washed twice with PBS, and resuspended in PBS at 1.5 × 10⁶ cells/100 μl. Subsequently, Abs, including allophycocyanin-F4/80 (eBioscience, San Diego, CA), allophycocyanin-CD11c, allophycocyanin-CD3e (BD Biosciences), and their isotype-matched negative control Abs, were added to the cell suspension. After 15 min of incubation in the dark, the cells were washed with PBS and subjected to FACS. FACS was performed on a FACSAria and analyzed with FACSDiva 4.1 (BD Biosciences).

To make gating strategies for analysis of BMM and Kupffer cells, liver nonparenchymal cells were stained with allophycocyanin or PE-F4/80. Cells with EGFP and F4/80 coexpression were BMM and cells with EGFP⁻ and F4/80⁺ were Kupffer cells. For IL-6, MCP-1, TNF, and CB1 staining, cells were stained in 50 μl staining buffer (BD Biosciences) with allophycocyanin or PE-F4/80. After 15 min of incubation in the dark, the cells were washed twice with staining buffer. Then, cells were suspended with fixation/permeabilization solution (BD Biosciences) for 30 min. Cells were washed twice and resuspended in PBS. Subsequently, Abs, including PE–IL-6, PE–MCP-1, allophycocyanin-TNF, anti-CB1 rabbit polyclonal Ab, and their isotype-matched negative control Abs, were added to the cell suspension for 20 min. For the cytokines, the cells were washed with PBS and subjected to FACS. FACS was performed on a FACSAria and analyzed with FACSDiva 4.1 (BD Biosciences). For CB1 staining, after 20 min of primary Ab incubation in the dark, cells were then washed twice and stained with the appropriate PE-conjugated secondary Abs, that is, PE-F(ab′)₂ donkey anti-rabbit IgG (BD Biosciences). After 30 min of incubation in the dark, the cells were washed with PBS and subjected to FACS. FACS was performed on a FACSAria and analyzed with FACSDiva 4.1 (BD Biosciences).

BMM with or without AM281 or PTX pretreatment were exposed to mAEA for different times, and then BMM were lysed in lysis buffer and active Rho and Rac were extracted using a pull-down and detection kit, Thermo Scientific/Pierce Biotechnology, Rockford, IL, catalog nos. 16116 and 16118 ), and finally active and total Rho and Rac were shown by Western blot.

Liver tissues (40 mg) were homogenized and lysed in 30 μl lysis buffer, and IL-6, IL-10, MCP-1, IFN-γ, TNF, and IL-12p70 in liver homogenates were measured using a cytometric bead array (CBA) mouse inflammation kit (BD Biosciences, catalog no. 552364) and FACS.

Results are expressed as means ± SEM. Statistical significance was determined by Student t test or ANOVA. Correlation coefficients were calculated by a Pearson test. A p value <0.05 was considered significant.

Human fibrotic samples were obtained from livers of patients undergoing liver transplantation. The mRNA levels of biosynthetic enzymes for endocannabinoids (NAPE-PLD for AEA, DAGL-α and DAGL-β for 2-AG) were elevated in human fibrotic livers. However, the degradative enzymes (FAAH for AEA, MAGL for 2-AG) had few changes (Table I). These strongly prompted an increase of endocannabinoids in human fibrotic livers. To further reveal the global changes of ECS in mice after CCl₄ i.p. injection, we measured the mRNA levels of enzymes in different organs, including the celiac liver, the kidney, and the spleen, as well as in the heart, the BM, and the brain, which have close associations with ECS (3). We found that ECS displayed predominant changes in liver. The mRNA levels of NAPE-PLD and DAGL-α increased >2-fold, whereas FAAH and MAGL dropped to around half of their normal levels, suggesting that there was a net increase of endocannabinoids, AEA, and 2-AG. The elevated levels of endocannabinoids have already been reported during liver regeneration and injury (23, 24). In liver, DAGL-β has been previously shown to be much more important on 2-AG synthesis, but our study found that DAGL-β was hardly upregulated. Therefore, during mouse CCl₄-induced liver injury, the upregulation of 2-AG was probably due to the decrease of degradative enzyme MAGL. In contrast, few changes of ECS were found in other organs, indicating that ECS was mainly involved in liver injury induced by CCl₄ treatment in mice (Table I).

We further detected the dynamic changes of ECS in mouse liver at different times after CCl₄ treatment. The mRNA level of NAPE-PLD rose rapidly at the early injured period and reached the peak at 1 wk. After that, NAPE-PLD mRNA sustained the level >2-fold. DAGL-α mRNA also increased and got to the topmost level after 4 wk, whereas DAGL-β mRNA had no changes. In contrast, FAAH and MAGL mRNA had both generated a violent decline at the first day of injury and kept at low levels during chronic stage (Fig. 1A, 1B). Especially, CB1 mRNA was significantly upregulated from 1 d of CCl₄ administration with a maximal increase after 4 wk. Similar results were obtained with CB2 (Fig. 1C). The protein changes of NAPE-PLD, FAAH, CB1, and CB2 basically corresponded to the changes of their mRNA levels (Fig. 1D).

Additionally, we undertook correlation analyses between ECS and the fibrosis parameter Col α1(III) in liver. In human fibrotic liver, the mRNA levels of biosynthetic enzymes for endocannabinoids (NAPE-PLD, DAGL-α, and DAGL-β) displayed positive correlations with Col α1(III), whereas the degradative enzymes showed no correlation with it (Table II). In mouse CCl₄-injured liver, the mRNA levels of NAPE-PLD and DAGL-α had positive correlations with Col α1(III), whereas the mRNA levels of FAAH and MAGL showed negative correlations with it (Table II). Collectively, these results suggested that ECS, which changed significantly in fibrotic liver, might play an important role in liver injury.

Our preceding study confirmed that various liver injuries were accompanied by the infiltration of BMM (20). Furthermore, BMM infiltration was generated at the early stage of liver injury and persisted through the whole insulted process (20). Thus, we wonder whether ECS is involved in the infiltration of BMM. We undertook immunofluorescence (Fig. 2A), DNA electrophoresis (RT-PCR production) (Fig. 2B), and Western blot (Fig. 2F, insert), proving that BMM express CB1 and CB2. The results of a Transwell migration assay revealed that JWH133 (CB2 agonist) had no influence on BMM migration at common doses (0.1, 0.5, and 1 μmol/l) and a flushing dose (10 μmol/l), whereas mAEA (CB1 agonist) caused a concentration-dependent increase in BMM migration and had the strongest promigratory activity at 1 μmol/l concentration (Fig. 2C).

Furthermore, we applied pharmacological inhibition and genetic knockdown of CB1 or CB2 in BMM. The silencing effects of CB1 and CB2 siRNA were showed in Western blot (Fig. 2F, insert). AM281 (CB1 antagonist, 1 and 10 μmol/l) markedly attenuated mAEA-mediated migration with a complete inhibition at 10 μmol/l concentration (Fig. 2E). As expected, CB1 knockdown restrained mAEA-induced migration but had no effects on spontaneous migration of BMM (Fig. 2F). Actually, mAEA also showed low CB2 affinity, but AM630 (10 μmol/l) and CB2 knockdown hardly inhibited mAEA-mediated migration (Fig. 2F). These results suggested that CB1 but not CB2 mediated BMM migration.

The small GTPase family, primarily including Rho, Rac, and Cdc42, is considered closely relevant to cell migration for its molecular switch role (GTP bound for active) to reconstruct the cytoskeleton (25). Pull-down analysis showed that mAEA promoted active GTP-bound Rho protein levels of BMM in 1, 2, and 3 h but that Rac-GTP changed slightly (Fig. 2D), and this added GTP-bound Rho conformation by mAEA was inhibited by AM281 (Fig. 2D). Additionally, CB1 is a G_(α)i/o protein–coupled receptor (1). We noted that PTX (G_(α)i/o protein inhibitor) weakened GTP-bound Rho protein with mAEA (Fig. 2F). These findings imply that mAEA activated G(α)_i/o-coupled CB1, and then enlarged GTP-bound Rho signal, finally promoting cell migration. Moreover, mAEA-mediated migration was impaired by PTX and Y27632 (ROCK inhibitor) pretreatment (Fig. 2D).

To investigate the role of CB1 on BMM infiltration in insulted liver, we performed an EGFP⁺ BM cell transplantation to the mice whose BM had been entirely destroyed. Then, the BM rebuilding mice had liver injury induced through CCl₄ administration. Cells that originated from BM (EGFP⁺) emerged in liver sections of OO plus vehicle and OO plus AM281 groups (Fig. 3A, 3B). However, after CCl₄ administration for 4 wk, a mass of cells with EGFP and F4/80 coexpression (indicating they were BMM) aggregated in liver-injured sites. AM281 pretreatment decreased markedly BMM numbers (Fig. 3C, 3D). BMM accounted for >70% of total macrophages in injured liver with CCl₄ administration. However, the proportion of BMM was significantly decreased by AM281 pretreatment (Fig. 3E).

To further confirm the pivotal role of CB1 in BMM infiltration in liver injury, we performed FACS on liver nonparenchymal cells. The proportion of total macrophages in nonparenchymal cells in CCl₄-induced damaged liver increased markedly compared with that in OO-treated liver. AM281 pretreatment decreased the total proportion of macrophages (Fig. 4A, 4B). Additionally, BMM proportion in nonparenchymal cells of damaged liver with AM281 pretreatment was also significantly downregulated (Fig. 4C, 4D). These results indicate that the decrease of total macrophage number was due to the downregulation of BMM by AM281.

T cells and dendritic cells are also important inflammatory cells that participate in hepatic injury (26, 27). In injured liver by CCl₄, BM-originated T cells (CD3e⁺) and dendritic cells (CD11c⁺) also increased. Unlike BMM, their quantities did not change by AM281 pretreatment (Fig. 4C, 4D). Furthermore, we performed a double immunofluorescence staining (F4/80-CB1, CD3e-CB1, or CD11c-CB1) on BM smears, and we found F4/80⁺ BMM highly expressed CB1, but CD3e⁺ and CD11c⁺ cells hardly expressed any CB1 (Fig. 4E). These results might explain the reason why AM281 had no influence on the infiltration of BM-originated T cells and dendritic cells.

Inflammatory cytokines and chemokines are crucial in immune responses. We investigated the effects of CB1 on cytokines and chemokines, including TNF-α, INF-γ, IL-12, IL-6, IL-10, and MCP-1. mAEA (1 μmol/l) promoted the mRNA levels of TNFα at 5, 12, and 24 h and IL-6 and MCP-1 at 2, 5, and 12 h (Fig. 5A) in BMM, implying that the activation of CB1 had a direct effect on TNFα, IL-6, and MCP-1 mRNA generation, and kept longer excitation to TNFα mRNA relatively. However, mAEA had no effects on IFN-γ, IL-12, and IL-10 mRNA (Fig. 5A). When BMM were pretreated with AM281 or CB1 siRNA, mAEA-mediated elevations of TNFα, IL-6, and MCP-1 mRNA levels were inhibited (Fig. 5B, 5C).

CBA was used to measure the protein levels of inflammatory cytokines in liver. AM281 administration significantly decreased both the mRNA and protein levels of inflammatory cytokines in CCl₄-damaged liver (Fig. 5D–F). The reductions of IFN-γ, IL-12, and IL-10, which were hardly influenced by the activation of BMM-expressed CB1 in vitro, might be due to the quantitative decrease of macrophages (Fig. 5D, F4/80) and/or AM281 producing some effect on other immune cells with CB1 expression.

Furthermore, we detected the effects of CB1 on the expression changes of cytokines and CB1 of liver macrophages (including BMM and Kupffer cells) by FACS in vivo. At first, we performed an EGFP⁺ BM cell transplantation experiment followed by CCl₄-induced mouse liver injury with or without AM281 administration. As expected, over 4 wk of AM281 pretreatment, the proportion of cells with EGFP and F4/80 coexpression (indicating they were BMM) in the injured liver decreased markedly compared with that in the liver without AM281 pretreatment, whereas the number of cells that were EGFP⁻ and F4/80⁺ (indicating they were Kupffer cells) seemed unchanged (Fig. 6A). These data demonstrated that systemic treatment of CB1 antagonist (AM281) could decrease BMM infiltration.

Furthermore, to investigate the role of CB1 on the expression changes of cytokines in liver macrophages, including BMM and Kupffer cells, we further performed FACS on liver nonparenchymal cells. We measured the expression changes of IL-6, MCP-1, and TNF in BMM and Kupffer cells. We observed that the proportion of Kupffer cells (EGFP⁻ and F4/80⁺) that expressed IL-6 and MCP-1 in the damaged liver were decreased compared with that in the liver without AM281 pretreatment. However, AM281 administration had no change on the number of the Kupffer cells (EGFP⁻ and F4/80⁺) that expressed TNF (Fig. 6B, 6C). However, for BMM (EGFP⁺ and F4/80⁺), AM281 pretreatment significantly decreased the number of BMM (EGFP⁺ and F4/80⁺) that expressed IL-6, MCP-1, and TNF in injured liver (Fig. 6D, 6E).

Moreover, we measured the CB1 expression in liver nonparenchymal cells by FACS. Interestingly, the proportion of both CB1⁺ BMM and CB1⁺ Kupffer cells were downregulated by AM281 pretreatment (Fig. 6B–E). These results showed that blocking CB1 could downregulate the expression of CB1 in BMM and Kupffer cells in liver injury induced by CCl₄.

Studies have demonstrated that reducing macrophages would improve hepatic fibrosis (19). To confirm this, we stained macrophages with F4/80 Ab by immunofluorescence assay and then performed sirius red staining for collagen in the same liver sample. As expected, immunoreactivity for F4/80 was predominantly colocated in the fibrotic areas (positive of sirius red staining) of the liver in CCl₄-treated mice for 2 wk (Supplemental Fig. 1). In representative H&E-stained images, large inflammatory area emerged in CCl₄-injured liver. The inflammatory area was significantly reduced by AM281 pretreatment (Fig. 7A, 7B). AM281 administration also decreased the levels of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) in serum of mice with liver injury (Fig. 7C, 7D). Sirius red can stain collagen into scarlet. Our data showed that collagen deposition was significantly attenuated by AM281 treatment (Fig. 7E, 7F). Meanwhile, mRNA levels of procollagen α1(I) (Col α1(I)), Col α1(III), and α-smooth muscle actin (α-SMA), which mainly reflect fibrotic information, were downregulated by AM281 (Fig. 7G). Hydroxyproline content in the liver also showed a similar result (Fig. 7H). All these results implied that inhibition of CB1 exerted anti-inflammatory and antifibrotic effects and protected hepatocytes from liver injury.

In this study, we investigated the effects of CB1 on BMM in liver injury. We found that biosynthetic enzymes of endocannabinoids were significantly elevated in human fibrotic livers. Meanwhile, we detected the changes in the level of biosynthetic and degradative enzymes of endocannabinoids, CB1 and CB2 in different tissues after CCl₄ administration and found that synthetases (NAPE-PLD and DAGL-α) and CBs showed positive correlation with inflammation/fibrosis parameters [TNF-α and Col α1(III)] in injured liver. Focusing on BMM, we investigated the expressions of CB1 and CB2, as well as the roles of CB1 and CB2 on migration and production of inflammatory cytokines by BMM. We discovered that CB1 but not CB2 was involved in the migration and activation of BMM. Furthermore, the blockade of CB1 suppressed BMM infiltration and ameliorated the inflammation and fibrosis in CCl₄-induced liver injury.

Studies about ECS contributions to inflammation focus on CB2 rather than CB1 because CB2 is predominantly expressed in the immune cells (1), whereas the relationship of CB1 and inflammatory cells has been poorly studied. So far, the research on CB1 in liver disorders mainly focused on hepatocytes, fibroblasts, myofibroblasts, and hepatic stellate cells (28). Animal studies showed that AEA and CB1 were upregulated in the condition of high-fat diet, and basal rates of fatty acid synthesis were reduced by CB1 blockade (7). CB1-deficient mice (CB1^−/−) and liver-specific CB1 knockout (LCB1^−/−) mice with a high-fat diet both had less, hyperglycemia, dyslipidemia, and insulin and leptin resistance compared with wild-type mice. An increase in de novo hepatic lipogenesis and a decrease of total energy expenditure by CB1 agonists were absent in both CB1^−/− and LCB1^−/− mice (29). Additionally, CB1^−/− and LCB1^−/− mice posed resistance to ethanol-induced steatosis (8). CB1 expression also increased in stellate cells and in hepatic myofibroblasts in patient cirrhotic livers and mice livers of fibrotic models. Pharmacological or genetic ablation of CB1 reduced fibrogenesis of injured mice livers by decreasing hepatic TGF-β levels and reducing the accumulation of fibrogenic cells (10). Cirrhosis was accompanied by elevated CB1 in endothelial cells and mesenteric arteries. Blockade of CB1 improved hemodynamic abnormalities and prevented progression of ascites (30). Liver regeneration and I/R injury were associated with a marked increase of CB1 expression in hepatocytes (23, 31). Blockade or ablation of CB1 showed a protective role in I/R injury and delayed liver regeneration. These studies showed that CB1 contributed to a variety of hepatic pathogenesis. In the present study, we showed the changes of endocannabinoid-associated enzymes and receptors, which correlated positively or negatively with fibrosis parameters. Furthermore, CB1 but not CB2 mediated the recruitment of BMM through G(α)_i/o and Rho/ROCK signaling pathways. Additionally, CB1 intensified secretion of some inflammatory cytokines of BMM. Hepatic inflammation and fibrosis were remarkably ameliorated by blocking CB1.

An increasing body of evidence has shown the changes of ECS in various liver injuries (32). For example, Mallat and Lotersztajn (33) reported the rising levels of endocannabinoids during liver injury and fibrogenesis. However, our data provide the clear and dynamic analysis for changes of ECS in CCl₄-induced liver injury. The comprehensive changes of four primary enzymes and two receptors indicate the importance of ECS. These elevated endocannabinoids and their receptors play different roles in liver diseases through connecting with multiple cells, including macrophages (32).

The role of CB2 in regulating immune function, including cell migration, has been widely investigated (34, 35). Recent studies have found that CB1 blockade delayed the progression of type 2 diabetes by reducing macrophage infiltration and causing an M1–M2 shift in Zucker diabetic fatty rat islets (21). Lesional macrophages of human coronary atheroma expressed CB1, and CB1 blockade reduced the LPS-induced production of inflammatory cytokines (22). Nevertheless, the relationship of CB1 to macrophages is rarely mentioned in hepatic pathology. Employing a Boyden chamber assay, we found that mAEA promoted BMM migration, whereas JWH133 had no such effects. By giving AM281 or silencing CB1 using siRNA, mAEA-mediated migration was reduced. These findings indicated that the elevated endocannabinoids in insulted liver induced BMM infiltration by CB1 but not by CB2. To verify the contribution of CB1 to BMM migration in vivo, we performed EGFP⁺ BM cell transplantation to the BM-destroyed mice. After BM rebuilding, these mice received liver injury by CCl₄ injection. In control liver sections, we found that there were scattered macrophages, and a part of them came from BM (expressing EGFP); the others may be indigenous macrophages that originated from yolk sac and fetal liver progenitors (36). After CCl₄ treatment, profuse macrophages aggregated to the liver inflammatory region and the number of macrophages in liver increased significantly. In this study, we proved that the increased macrophages were mainly recruited from BM.

The macrophages recruited from BM can produce excessive inflammatory cytokines and chemokines and induce damaged effects to remnant normal liver tissue and exacerbate tissue injury, which ultimately lead to liver fibrosis or sclerosis (37). Thus, if the excessive macrophages could be reduced in time, inflammation and fibrosis could be ameliorated. In fact, the relationship between CB1 and BMM is still undefined. In this study, we demonstrated that the activation of CB1 promoted BMM migration in vitro; the blockade of CB1 inhibited BMM infiltration in vivo. Furthermore, blockage of CB1 decreased the proportion of BMM and Kupffer cells that expressed IL-6 and MCP-1. CBA analysis results showed that AM281 administration significantly decreased the mRNA and protein levels of inflammatory cytokines, including IL-6, IL-10, MCP-1, IFN-γ, TNF, and IL-12. Because T cells (Th1 cells) and NK cells are also the major source of IFN-γ, further studies will be needed to explore the effect of CB1 on T cells or NK cells in vitro and in vivo.

It has been reported that CB1 was weakly expressed in T lymphocytes and dendritic cells in vitro (38, 39). However, little is known about CB1 expression of BM-derived T cells and dendritic cells in CCl₄-induced murine liver injury. We found that expressions of CB1 in these cells were at low levels in damaged liver. Furthermore, blocking CB1 by AM281 has no effect on the migration of BM-derived T cells and dendritic cells in injured liver. It may be the reason why blockage of CB1 cannot completely rescue liver fibrosis and inflammation. Further studies are necessary to elucidate the effects of CB1 on BM-derived T cells and dendritic cells in vivo and in vitro.

Although CB1 antagonist rimonabant for treatment of obesity and related comorbidities has been withdrawn from clinical application due to its severe neuropsychiatric risks, it is important to continue the CB1-associated studies and further explore peripherally restricted CB1 antagonists for their crucial roles in diverse diseases (40). Evidence has shown that inactivation of CB1 ameliorated hepatic fibrogenesis (10). In the present study, we have also observed that the liver inflammatory area was markedly reduced by AM281 pretreatment. The levels of ALT and AST were downregulated, indicating the blockade of CB1-protected hepatocytes from damage. Additionally, fibrosis was markedly attenuated, possibly due to the fact that blockade of CB1 reduced the levels of profibrogenic cytokines and growth factors such as TNF-α, IL-6, and TGF-β1, which can drive hepatic stellate cell activation or promote the survival of activated hepatic stellate cells (19, 41). However, other studies have also shown that macrophages revealed distinct roles during liver injury (42) and their depletion exacerbated liver injury (43). The interference of macrophages in different stage of liver inflammation and fibrosis may cause the opposite outcome. The remaining partial or specified phenotype macrophages will benefit fibrosis resolution. More research about BMM needs to be performed.

In summary, our data show that the ECS contributed to liver fibrosis by activation of BMM in a CB1-dependent manner. The blockade of CB1 can reduce the quantity of BMM infiltrating the damaged liver, silence overactive BMM, improve liver function, and finally attenuate the extent of inflammation/fibrosis. This study demonstrates a novel biological mechanism of treating liver fibrosis by aiming at CB1 of BMM.

The authors have no financial conflicts of interest.