Potential conflict of interest: Nothing to report.
Studies of the molecular and cellular mechanisms of concanavalin A (ConA)-induced liver injury have provided important knowledge on the pathogenesis of many liver diseases involving hepatic inflammation. However, studies identifying hepato-protective factors based on the mechanistic understanding of this model are lacking. Evidence suggests that certain prostaglandin (PG) products of cyclooxygenase (COX)-1 and COX-2 provide important anti-inflammatory and cytoprotective functions in some pathophysiological states. In the present study, we demonstrate a protective role of COX-2 derived PGs in ConA-induced liver injury. COX-2−/− mice developed much more severe liver damage upon ConA treatment compared with wild-type and COX-1−/− mice. Treatment of COX-2−/− mice with misoprostol (a PGE1/2 analog) or beraprost (a PGI2 analog) significantly decreased ConA-induced liver injury. Data from both in vivo and in vitro experiments demonstrated that misoprostol and beraprost acted directly on hepatic leukocytes, including natural killer (NK)T and T cells, and down-regulated their production of interferon (IFN)-γ, which are critical in mediating ConA-induced tissue damage. Collectively, the results provide strong evidence that the protective effects of COX-2 within the liver are mediated through the production of PGE2 and PGI2, which exert anti-inflammatory functions. These findings suggest that COX-2-derived PGs may have great therapeutic potentials in treating patients with inflammatory liver diseases. (HEPATOLOGY 2007;45:159–169.)
Liver diseases, caused by hepatitis A, B and C viral infection, autoimmune conditions, alcohol ingestion, or the use of certain drugs, are a significant health issue, as many can develop into liver cirrhosis and cancer. Understanding the complex pathogenesis of liver diseases is important for designing better therapies. The mouse model of concanavalin A (ConA)-induced liver injury closely resembles the pathophysiology of many liver diseases and is used widely as an experimental tool.1 Intravenous injection of ConA induces hepatitis, which is manifested by the elevation of serum alanine aminotransferase (ALT) activities and the histological exhibition of T cell infiltration, massive granulocyte accumulation, and hepatocyte apoptosis and necrosis. Evidence suggests that activation of the innate and adaptive immune cells, including natural killer T (NKT) cells,2 Kupffer cells,3 and CD4+ T cells,1 and their production of inflammatory cytokines, such as interferon (IFN)-γ4, 5 and tumor necrosis factor (TNF)-α,6, 7 play essential roles in the pathogenesis of liver injury in this model. Despite the extensive investigation of the molecular and cellular mechanisms of ConA-induced hepatitis, studies identifying hepato-protective factors based on the mechanistic understanding of this model are lacking.
Cyclooxygenase (COX)-1 and 2 catalyze the formation of prostaglandin H2 (PGH2), which is further metabolized to form prostacyclin and various PGs through downstream synthetases. COX-1 is constitutively expressed in most cells, whereas COX-2 expression is low or undetectable, but increases drastically upon stimulation. Both COX-1 and COX-2 have been implicated to play important roles in inflammation, and the differential contributions by each COX isoform appear to depend on the inflammatory stimulus, the time point examined, and the tissue or organ in which the insult occurs.8–10 Although it is widely accepted that PGs contribute to the pathology of various inflammatory diseases, which provided the rational for the development of COX inhibitors and COX-2 selective drugs for the treatment of chronic inflammatory diseases,11, 12 the beneficial functions, perhaps tissue-specific, of PG products from either COX-1 or COX-2 cannot be overlooked. Many case reports revealed the association of COX-2-specific inhibitors with liver injury,13–17 suggesting that suppression of COX-2 may directly or indirectly result in liver pathology. It has also been demonstrated that exogenous PGs are protective against lipopolysaccharide (LPS)-, carbon tetrachloride-, virus- and ischemia-reperfusion induced liver injury.18 Although ConA-induced liver injury is a model of immune-mediated inflammatory disease of the liver, the roles of the COX-1, COX-2, and their PG products in the mechanism of the tissue damage or protection have not been investigated.
The functions of specific PGs have been investigated in many studies. It is known that PGI2 (prostacyclin) has profound anti-platelet aggregation, vasodilatory, and myocardial protective effects.19, 20 A stable PGI2 analog, beraprost, has been used in clinical treatments of peripheral arterial occlusive disease, primary pulmonary hypertension, as well as vascular complications in diabetes mellitus.21–23 Moreover, in animal models of liver injury caused by ischemia/reperfusion and ConA treatment, beraprost was shown to play a protective role, although the mechanism was not clearly understood.24, 25 PGE2, another major product of COX, is known to play an important role in acute inflammation, and high levels of PGE2 are found in numerous disease states.26 However, studies have also demonstrated that PGE2 plays a protective role in various animal models of liver injury.18, 27 Further, clinical trials for the hepato-protective effect of a stable analog of PGE1/2, misoprostol, have been conducted in hepatitis B patients.28
Based on these data, the present studies aim to investigate (1) the role of COX in ConA-induced liver injury, (2) the specific PGs that may mediate the effects of COX in this model, and (3) the mechanisms of PG-mediated effects in ConA-induced hepatitis. Using transgenic knock-out mice, the current study is the first to demonstrate a protective role of COX-2 in ConA-induced liver injury. Our data provide strong evidence that the increased susceptibility of COX-2−/− mice to ConA-induced hepatitis is due to the lack of production of PGE2 and PGI2, which play an anti-inflammatory role in this model.
Animal treatment and Assessment of Hepatotoxicity.
Male COX-1−/− and COX-2−/− mice [129/Ola X C57BL/6 (129Ola/C57) background, 13-15 weeks of age] were provided by Dr. Robert Langenbach (NIEHS) through Taconic Farms (Germantown, NY). These animals were kept in the Center for Laboratory Animal Care at the University of Colorado Health Sciences Center (UCHSC) for one week before treatment. All following described animal experiments were performed in accordance with guidelines from the UCHSC Institutional Animal Care and Use Committee. Genotypes were determined using PCR analysis of tail DNA.29 To induce hepatitis, mice were intravenously (i.v.) injected with 10 or 20 mg/kg of ConA (cell culture tested, Sigma, St. Louis, MO). After 8 and 20 hours, blood was collected by retro-orbital puncture after the mice were anesthetized. Blood samples were allowed to clot at 4°C overnight before sera were prepared by centrifugation at 10,000×g for 20 minutes. ALT levels were measured using a diagnostic assay kit (Teco Dignostics, Anaheim CA) according to the manufacturer's instruction. Twenty hours after ConA treatment, the animals were sacrificed and the livers removed. A portion of each excised liver was fixed in 10% formaldehyde overnight before being transferred into 70% ethanol solution and the reminder was snap-frozen and stored at −80°C for subsequent RNA isolation. Tissue sections were embedded in paraffin and stained with hematoxylin and eosin (H&E, Department of Pathology, UCHSC).
For misoprostol pretreatment, male COX-2−/− mice were subcutaneously (s.c.) injected with misoprostol (Sigma, 0.125 mg/kg, dissolve in 100 μl olive oil) at 2 and 24 hours prior to ConA administration. Beraprost (Cayman Chemical, Ann Arbor MI) working solution was prepared from an ethanol-based stock solution by evaporating ethanol under a gentle stream of nitrogen followed by dissolving beraprost in PBS (pH = 7.2). For beraprost treatment, male COX-2−/−mice were injected intraperitoneally (i.p.) with 0.125 mg/kg beraprost 1 hour prior to and 3 hours after ConA treatment.
Wild-type and COX-2−/− mice were treated with ConA (10 mg/kg), and 1, 3, and 5 hours later, blood was collected by retro-orbital puncture after the mice were anesthetized. Sera were prepared immediately by centrifugation at 10,000×g for 20 minutes. Serum PGI2 levels were determined using an enzyme immunoassay kit (Assay Designs, Ann Arbor, MI) according to the manufacturer's instruction. The lower limit of detection for PGI2 was 3.2 pg/ml.
Total RNA was isolated from 15 mg of frozen liver tissue or 5 × 105 cells using RNeasy Mini Kit or RNeasy Micro Kit (Qiagen, Valencia, CA), respectively, as described by the manufacturer. One or 0.5 μg of RNA were reverse transcribed to cDNA at 42°C for 60 minutes using Superscript II Reverse Transcriptase (Invitrogen, Carlsbad, CA) with oligo (dT) Primers (Invitrogen). The resultant cDNA was amplified using Jump-Start Taq polymerase (Sigma) and gene specific primers for TNF-α (sense 5′-TTCTGTCTACTGAACTTCGGGGTGATCGGTCC-3′, anti-sense 5′-GTATGAGATAGCAAATCGGCTGACGGTGTGGG-3′), IFN-γ (sense 5′-TG CATCTTGGCTTTGCAGCTCTTCCTCATGGC-3′, anti-sense 5′-TGGACCTGTGGGTTG ACCTCAAACTTGGC-3′), Bcl-xL (sense 5′-AGAGAGGCAGGCGATGAGTTTGAA-3′, anti-sense 5′-TCCAACTTGCAATCCGACTCACCA-3′) and β-actin (a housekeeping gene for control purposes, sense 5′-TCTTGGGTATGGAATCCTGTGGCA-3′, anti-sense 5′-ACTCCTGCTTGCTGATCCACATCT-3′). All PCR products were resolved on 1.2% agarose gels and visualized using ethidium bromide staining. RNA expression levels were determined by normalizing band intensities relative to the levels of β-actin expression using Adobe Photoshop 6.0.
Preparation of Hepatic Leukocytes and In Vitro Treatment with ConA, Misoprostol and Beraprost.
Hepatic leukocytes were isolated following a previously described method with slight modification.30 In brief, the animals were anesthetized, and liver tissues perfused by insertion of a 20 G catheter into the superior vena cava. The liver was perfused in situ at 37°C with Hank's balanced salt solution (HBSS) for 5 minutes. Single cell suspensions were filtered through a 100 μm cell strainer (BD Falcon, Bedford, MA), and centrifuged at 300×g for 5 minutes. The pellet was resuspended in 15 ml of 34% Percoll (Sigma) containing 100 U/ml of heparin (Baxter Healthcare Corporation, Deerfield, IL) and centrifuged at 500×g for 15 minutes. The resulting pellet was collected and resuspended in 2 ml of red blood cell lysing buffer (Sigma). After 5 minutes, the cells were washed in 0.6% acid citrate-dextrose (ACD-A, Sigma) solution containing 0.5% BSA. Total hepatic leukocytes were counted by trypan blue exclusion. To obtain purified NKT and T cells for some experiments, total hepatic leokocytes were stained with allophycocyanin (APC)-conjugated anti-CD3 and phycoerythrin (PE)-conjugated anti-NK1.1 antibodies (eBioscience). Subsequently, NKT (CD3+NK1.1+) and T (CD3+NK1.1−) cells were sorted with a Moflo cell sorter (Cytomation, Fort Collins, CO), yielding a purity of each subsets of >90%.
Hepatic leukocytes (0.5×106/well), NKT (0.57×106/well), or T cells (0.5 × 106/well) were treated in vitro with 4 μg/ml ConA for 20 hours in the presence or absence of various concentrations of misoprostol or beraprost. In the NKT and T cell cultures, purified naïve peritoneal macrophages (0.15×106/well) were added to increase the levels of IFN-γ production. Misoprostol solutions were prepared from a DMSO-based stock solution by evaporating DMSO under a gentle stream of nitrogen followed by dissolving misoprostol in PBS (pH = 7.2). The preparation of beraprost solutions is the same as previously described. At the end of the treatment, the cells were collected and RNA extracted for RT-PCR analysis of IFN-γ expression. IFN-γ release into the culture supernatant was determined by sandwich ELISA using capture and detection antibody pairs according to the manufacturer's instructions (R&D Systems). Protein standard curves were made using recombinant mouse IFN-γ (R&D Systems). The lower limit of detection was 7 pg/ml.
Flow Cytometric Analysis.
For IFN-γ intracellular staining, freshly isolated hepatic leukocytes were incubated at 37°C with 1 μl/ml of GolgiPlug, which contains Brefeldin A, an inhibitor of cytokine secretion (BD PharMingen, San Diego, CA). Three hours later, the cells were harvested and incubated with anti-FcγR II/III antibody (10 μg/ml, eBioscience, San Diego, CA) plus normal rat serum (1:10, Sigma) for 10 minutes on ice to prevent nonspecific binding. NK and T cell populations in freshly isolated hepatic leukocytes were characterized by staining the cells with fluoresceinisothiocyanate (FITC)-conjugated anti-CD3 and PE-conjugated anti-NK1.1 antibodies (eBioscience) for 30 minutes on ice. The cells were then fixed and permeablized with Cytofix/Cytoperm buffer (BD PharMingen) for 20 minutes at 4°C. Intracellular IFN-γ was detected by staining the cells for 30 minutes on ice with APC-conjugated anti-IFN-γ monoclonal antibody (eBioscience) in Perm/Wash buffer (BD PharMingen). The cells were analyzed on a FACS Calibur using CellQuest software (BD Biosciences, San Jose, CA). The data were further analyzed using FlowJo software (Tree Star Inc., Ashland, OR).
Caspase-3 Activity Assay.
The liver tissue samples were homogenized in ice-cold Tris buffer (100 mM, pH 7.5), containing 250 mM sucrose, 2 mM EDTA, and a cocktail of protease inhibitors (1:100, Sigma). Caspase-3 activities were measured by using a fluorogenic substrate Ac-DEVD-AMC (Biomol; Plymouth Meeting, PA) as previous described.31 Reactions were performed at 37°C, and fluorescence intensity monitored using a Packard FluoroCount plate reader at 30 min intervals (Packard Instrument, CT). Substrate autofluorescence was subtracted from each value and specific activities were calculated based on a standard curve of aminomethyl coumarin (Sigma).
Data are presented as mean ± SEM. Comparisons between two groups were carried out using two-tailed Student t test, and between multiple groups using one-way analysis of variance (ANOVA) with a post hoc test of significance between individual groups. Differences were considered significant when P < 0.05.
COX-2−/− Mice Are More Susceptible to ConA-Induced Liver Injury.
Both pro-inflammatory and cytoprotective functions of COX-1 and COX-2 have been demonstrated in animal models of inflammation in various tissues, however, their roles in liver diseases remain unclear.8, 10, 29 We set out to investigate the effects of COX-1 and COX-2 on ConA-induced liver injury using COX-1−/− and COX-2−/− mice. Following i.v. injection of 10 mg/kg of ConA to male COX-1−/−, COX-2−/−, and wild-type mice, serum ALT activities revealed that COX-2−/− mice were much more susceptible to ConA-induced liver damage. The ALT levels in COX-2−/− mice reached as high as 1,300 and 1,600 IU/l at 8 and 20 hours, respectively, after ConA treatment; whereas the levels in COX-1−/− mice were less than 200 IU/l (Fig. 1A). No significant liver injury was observed in wild-type mice (Fig. 1 A). In line with the elevated ALT activities, histological evaluation of the liver sections obtained from COX-2−/− mice at 20 hours after ConA treatment revealed massive hepatocyte necrosis and significant leukocyte infiltration (Fig. 1B). However, no obvious tissue damage was observed in liver sections of wild-type and COX-1−/− mice (Fig. 1B).
Misoprostol Pretreatment Protects COX-2−/− Mice Against ConA-Induced Liver Injury.
It is known that, in inflammatory conditions, COX-2 is induced and predominantly responsible for the production of a variety of PGs, which may cause or protect against inflammation-induced pathology.12 One major PG produced by COX-2 during inflammation is PGE2. We investigated whether or not the lack of PGE2 production in COX-2−/− mice could explain why these mice were more susceptible to ConA-induced liver injury. A widely used stable analog, misoprostol, was administered to mice by subcutaneous injection (0.125 mg/kg dissolved in 100 μl of olive oil) at 2 and 24 hours prior to ConA treatment. The ALT levels decreased significantly in misoprostol plus ConA-treated mice compared with that in ConA (20 mg/kg) alone-treated mice (Fig. 2A). Similar protective effect of misoprostol was observed when both groups of mice were treated with 10 mg/kg of ConA (data not shown). Consistent with the result from biochemical analysis, histological evaluation demonstrated that ConA-induced massive hepatocyte necrosis in COX-2−/− mice was ameliorated when the animals were pretreated with misoprostol (Fig. 2B).
Furthermore, hepatic message levels of IFN-γ and TNF-α were measured in this study to determine whether they could be down-regulated by misoprostol treatment. The data revealed that the expression levels of both IFN-γ and TNF-α were significantly decreased in mice treated with misoprostol plus ConA compared with that in ConA alone-treated mice (Fig. 2C,D). It is possible that the inhibition of IFN-γ and TNF-α expression is merely a result of misoprostol-induced alleviation of tissue damage, rather than a direct effect of misoprostol on the cellular source of these cytokines. To address whether or not misoprostol has direct effects on hepatic leukocytes, these cells were isolated from naïve COX-2−/− mice and treated with ConA in vitro in the presence and absence of misoprostol. Flow cytometric analysis revealed that the major populations in the mixture of isolated hepatic leukocytes were NK (10%), NKT (28%) and T cells (25%, data not shown), which are known to produce IFN-γ, rather than TNF-α, upon activation by ConA. Our data demonstrated that both the mRNA and protein levels of IFN-γ produced by hepatic leukocytes in response to in vitro ConA stimulation were significantly down-regulated by misoprostol treatment (Fig. 3A -C). These results indicate that the hepato-protective effect of misoprostol is due to its direct anti-inflammatory action on hepatic leukocytes. Moreover, hepatic leukocytes were isolated from wild-type mice and treated with ConA in vitro in the presence or absence of misoprostol. Similarly to what was observed with COX-2−/− cells, IFN-γ production by wild-type cells was inhibited by misorpostol, suggesting that hepatic leukocytes are not the source, but rather a target of PGs.
The Hepato-Protective Effect of COX-2 Is Predominantly Mediated by PGI2.
Our data demonstrated that misoprostol significantly, but not completely, attenuated ConA-induced liver injury in COX-2−/− mice. This finding suggests that, aside from PGE2, other PG product(s) of COX-2 may play an important role in mediating its protective effect. We measured serum levels of PGI2 at 1, 3 and 5 hours after ConA challenge and compared the levels between wild-type and COX-2−/− mice. The data revealed that, at all 3 time points (data collected at 1 and 3 hours shown in Fig. 4A; data at 5 hours not shown), PGI2 levels were significantly higher in wild-type mice, compared with those in COX-2−/− mice, suggesting that PGI2, as another major COX-2-derived product may play a role in this model. Subsequently, we investigated whether or not PGI2 is important in mediating the hepato-protective role of COX-2. A stable analog of PGI2, beraprost, was administered to COX-2−/− mice at 1 hour prior to and 3 hour after ConA treatment. The ALT levels decreased drastically in beraprost plus ConA-treated mice compared with that of ConA (20 mg/kg) alone treated mice (Fig. 4B). The data demonstrate that beraprost, compared with misoprostol, is much more potent in inhibiting ConA hepatotoxicity. Similar degree of protection by beraprost was observed when both groups of mice were treated with 10 mg/kg of ConA (data not shown). Consistent with the inhibition of ALT activities, histological evaluation revealed that beraprost treatment caused a nearly complete protection from ConA-induced liver damage in COX-2−/− mice (Fig. 4C).
PGI2 is known to have profound vasodilatory effect, which is thought to be involved in the mechanism of exogenous PGI2-mediated hepato-protection observed in animal models of liver injury.24, 25 We hypothesized that, aside from the vasodilatory effect of beraprost, other functions of PGI2 may be important in its hepato-protective mechanism. The effect of beraprost treatment on IFN-γ and TNF-α expression within the liver was investigated to determine whether beraprost could counteract ConA-induced inflammation. Similar to misoprostol, beraprost treatment in vivo significantly decreased mRNA levels of IFN-γ and TNF-α in the livers of COX-2−/− mice compared with those of ConA alone-treated mice (Fig. 4D,E). Furthermore, the levels of IFN-γ protein expression by hepatic leukocytes were determined by flow cytometric analysis using intracellular staining method. The data demonstrated that IFN-γ expression by both NK and T cells was significantly down-regulated by beraprost treatment (Fig. 5). It has been demonstrated that NKT cells disappear from the liver as early as 2 hours after ConA treatment.2 This rapid elimination of NKT cells from the liver could explain why we did not detect this population of cells in our study.
To determine whether the anti-inflammatory effect of beraprost is mediated through its direct action on hepatic leukocytes, these cells were isolated from naïve COX-2−/− mice and treated with ConA in vitro in the presence and absence of beraprost. The data demonstrated that both mRNA and protein levels of IFN-γ produced by hepatic leukocytes in response to ConA stimulation were down-regulated significantly by beraprost treatment (Fig. 6A -C). Further, the effects of beraprost on NKT and/or T cells were investigated by purifying each subpopulations using FACSorting. We found that purified NKT or T cells alone did not produce significant amount of IFN-γ after ConA stimulation in vitro (60 pg/ml for NKT cells and 125 pg/ml for T cells). Previous studies32, 33 and our unpublished data also demonstrated that cell-cell contact interaction between macrophages and NKT or T cells was necessary for sufficient activation of NKT or T cells. Therefore, peritoneal macrophages were isolated and purified from COX-2−/− mice and added (0.15 × 106/well) to cultures of NKT and T cells to increase the levels of IFN-γ production by these cells. The results demonstrated that beraprost inhibited INF-γ production by both NKT and T cells, and that the suppressive effects on these two cell types are comparable (Fig. 6D). These results indicate that beraprost directly targets hepatic leukocytes, including NKT and T cells, and inhibits pro-inflammatory cytokine production. Moreover, hepatic leukocytes were isolated from wild-type mice and treated with ConA in vitro in the presence or absence of beraprost. Similarly to what was observed with COX-2−/− cells, IFN-γ production by wild-type cells was inhibited by beraprost (Fig. 6C), suggesting that hepatic leukocytes are not the source, but rather a target of PGs.
Although beraprost had similar anti-inflammatory effect as misoprostol, the in vivo hepato-protection mediated by beraprost was much stronger, suggesting the possible involvement of other mechanism(s). One possibility is that beraprost may prevent hepatocyte apoptosis induced by ConA treatment. It has been shown that caspase-3 is up-regulated in the mouse liver after ConA administration.34–36 Caspase-3 activities were measured in the liver homogenates prepared from beraprost plus ConA-treated mice and from ConA alone-treated mice, and the activities were significantly decreased in the former samples (Fig. 7A). Furthermore, hepatic message levels of BcL-xL, which is an important anti-apoptotic factor within the liver, were determined. Low levels of BcL-xL expression were detected in the liver of COX-2−/− mice treated with ConA alone, however, the expression levels were significantly increased in mice treated with beraprost plus ConA (Fig. 7B,C). These data suggest that beraprost may play a role in counteracting events leading to hepatocyte apoptosis.
Although studies using ConA-induced liver injury as a model for viral- and autoimmune-hepatitis have yielded important knowledge on the complex pathogenesis of liver diseases, investigations of potential regulatory and protective factors that counteract events leading to tissue damage have been lacking. The current study hypothesized that COX-2, through the production of certain PGs, represents a key player in the protective mechanisms against liver injury.
Our data demonstrated that COX-2−/− mice, compared with their wild-type counterparts and COX-1−/− mice, were much more susceptible to ConA-induced liver injury (Fig. 1). This observation of a protective role of COX-2 in ConA-induced liver injury is consistent with the result from a previous study, which demonstrated that COX-2−/− mice were more susceptible to acetaminophen (APAP)-induced hepatotoxicity.37 The data suggested that COX-2-derived PG products may induce a stress response that limits APAP-induced hepatocyte damage. However, specific PGs that may mediate such protective effect were not identified. In contrast to the protective effect of COX-2 demonstrated in our study and in APAP-induced liver injury, it has also been reported that LPS-induced liver injury was actually mitigated in COX-2−/− mice.38 Although the reasons for these discrepancies are not clear, they may be attributed to mechanistic differences in tissue damage caused by various hepatotoxins and perhaps, variations in the production and functions of specific PGs in various liver injury models.
The present study demonstrated that PGI2 and PGE2 play important roles in mediating the protective functions of COX-2. Administration of a PGI2 analog, beraprost, drastically inhibited ConA-induced liver injury in COX-2−/− mice, with a nearly 10-fold reduction in ALT levels, compared with those in ConA alone-treated mice (Fig. 4). Similar hepatoprotective effect was previously reported in wild-type C57BL/6J mice given beraprost plus ConA,25 although ALT levels were only decreased two-fold compared with those in control mice treated with ConA only. We measured serum levels of PGI2 in both wild-type and COX-2−/− mice after ConA challenge (Fig. 4A). The data demonstrated that PGI2 levels were more than 7-fold higher in wild-type mice compared with that in COX-2−/− mice, suggesting that PGI2 production was substantially compromised in COX-2-deficient mice. This result could also explain why more dramatic protection by exogenous beraprost treatment was observed in COX-2−/− mice in our study compared with that in wild-type C57Bl/6J mice reported previously.25
In the above-mentioned study, the mechanism of protection was thought to be mainly associated with beraprost-mediated maintenance of hepatic blood flow through its vasodilatory function.25 Numerous studies of the mechanisms of ConA-induced liver injury suggest that IFN-γ production by activated NKT and T cells in the liver plays an indispensable role in causing massive hepatocellular apoptosis and necrosis.4, 5 Whether or not TNF-α plays a critical role in ConA-induced liver injury remains a controversial issue,39 however, large amount of evidence suggests that TNF-α is important in the aggravation of hepatotoxicity initiated by IFN-γ and various immune cells of the liver.6, 7 In the above-mentioned study using wild-type C57BL/6J mice, decreases in serum levels of IFN-γ and TNF-α were observed after beraprost treatment; however, their expression by splenocytes was increased, and the hepatic expression levels of these cytokines were not determined.25 Hence, it was not clear whether the decreases in serum levels of TNF-α and IFN-γ were the result, or the actual cause, of attenuated liver toxicity after beraprost treatment.
Our results obtained from both in vivo and in vitro experiments suggest that beraprost-mediated inhibition of TNF-α and IFN-γ expression by hepatic leukocytes was not a result of attenuated liver toxicity, but rather a direct effect of beraprost on these cells. Our data obtained from in vivo experiments demonstrated a significant reduction in hepatic message levels of TNF-α and IFN-γ in COX-2−/− mice treated with beraprost plus ConA, compared with those in ConA alone-treated control mice (Fig. 4D,E). Moreover, intracellular staining experiments revealed that ConA-induced expression of IFN-γ by NK and T cells in the liver was inhibited by beraprost treatment (Fig. 5). Although ample evidence suggests that hepatic NKT cells are the initial targets of ConA toxicity and that these cells are critical in contributing to tissue damage, it has also been demonstrated that NKT cells are depleted rapidly from the liver, with greater than 70% decrease in number within 4 hours after ConA treatment.2 This explains why we did not detect a NKT cell population by flow cytometry (Fig. 5). The inhibitory effect of beraprost on hepatic leukocytes observed in vivo was corroborated by our in vitro experiments, in which naïve hepatic leukocytes, as well as NKT and T cells, were isolated from COX-2−/− mice and treated with ConA in the presence and absence of various concentrations of beraprost. The data demonstrated that beraprost could significantly reduce both message and protein levels of IFN-γ expression by hepatic leukocytes (Fig. 6A-C). We found that hepatic leukocytes from wild-type and COX-2−/− mice responded similarly to misoprost- or beraprost-mediated inhibition of IFN-γ production (Fig. 3C; Fig. 6C). This result suggests that hepatic leukocytes are a target of PGs produced by other cells within the liver, such as Kupffer cells and endothelial cells; however, the leukocytes themselves may not produce PGs to allow an autocrine effect, as it has been demonstrated that blood T cells do not synthesize PGs.40
Moreover, we found that beraprost inhibited INF-γ production by both NKT and T cells, and that the suppressive effects on both cell types are comparable (as shown in Fig. 6D). These data suggest that the in vitro effects of beraprost are predominantly targeted on NKT and T cells. It has been demonstrated that many cells of the immune system express multiple PG receptors that couple to a range of intracellular signaling pathways that mediate the effects PGs on leukocyte function.41 PGI2 receptor (IP) can couple to Gs and lead to an increase of intracellular cAMP, cAMP accumulation is generally associated with inhibition of effector cell functions. For example, recent studies demonstrated that PGI2, acting through IP receptor, suppresses T cell-mediated allergic inflammatory responses.42 The above-described evidence and our data suggest that beraprost, though binding to IP receptor, exerts an anti-inflammatory effect directly on certain inflammatory leukocytes.
It has been shown that hepatocyte apoptosis occurs in ConA-induced liver injury, and that caspase-3 activity increases in this experimental setting.34–36 Our data showed that, comparing with those of ConA alone-treated mice, COX-2−/− mice treated with beraprost plus ConA exhibited significant increases in hepatic message levels of Bcl-xL and marked decreases in caspase-3 activities in the liver (Fig. 7). Anti-apoptotic effect of PGI2 on hepatocytes has not been reported, however, several in vitro studies demonstrated that PGI2 caused significant attenuation of kidney cells through up-regulation of Bcl-2 expression.43–45 Although our data raised an interesting possibility that the protective effect of beraprost on ConA-induced hepatotoxicity may be attributable to an anti-apoptotic function, the mechanism by which beraprost exerts anti-apoptotic effects is unknown and the possible induction and activation of cAMP and NF-κB by beraprost warrant further investigation. Beraprost is currently used in clinics to treat vascular complications and pulmonary hypertension.21, 22 Interestingly, several clinical reports revealed significant success in liver transplantation procedures, in which beraprost was used to treat pulmonary hypertension.46, 47 These reports suggest a possible protective role of beraprost against liver injury in clinical settings. Although clinical benefits of the usage of PGI2 analogs in the amelioration of liver diseases have not been established, the data from our studies provide clear evidence that beraprost inhibits inflammatory cytokine production by hepatic leukocytes and counteracts events leading to hepatocyte apoptosis.
Another potential hepatoprotective prostaglandin is PGE2, which has been shown to inhibit tissue damage in various animal models of liver injury.18, 27 Our data demonstrated that misoprostol decreased ConA-induced liver injury in COX-2−/− mice, although its protective effect was not as profound as that of beraprost (Fig. 2). Similar to beraprost, both in vivo and in vitro experiments suggest that misoprostol acts directly on hepatic leukocytes and inhibits TNF-α and IFN-γ production by these cells (Fig. 2; Fig. 3). The anti-inflammatory function of PGE2 has been suggested by studies demonstrating its inhibitory effect on T cell proliferation and the production of interleukin-2 and IFN-γ by T cells.41 Moreover, it has been shown that misoprostol inhibits a variety of inflammatory processes and the release of tissue damaging cytokines involved in conditions as diverse as asthma and osteoarthritis.48 The data from these studies and ours support that the hepatoprotective role of misoprostol in ConA-induced liver injury is likely attributed to its anti-inflammatory function.
In summary, our results demonstrated a protective role for COX-2 in ConA-induced liver injury. This observation may suggest that patients with existing acute or chronic liver diseases are prone to developing aggravated tissue damage when taking COX-2-specific inhibitors simultaneously. Although investigations of this possible association have not been reported, many case studies provided evidence that the use of certain COX-2 inhibitors causes liver damage.13–17 Due to the emerging and broad usage of COX-2-specific inhibitors, it is important to thoroughly understand their effects on preexisting liver diseases. The current study provides scientific basis for the potential danger of exacerbated tissue damage in patients, who use these drugs under certain pathological conditions of the liver. Our data also provided strong evidence that the protective effects of COX-2 are mediated through the production of PGE2 and PGI2, which exert anti-inflammatory functions. These findings suggest that COX-2-derived PGs may have great therapeutic potentials in treating patients with inflammatory liver diseases.
The authors wish to thank Dr. Christopher Franklin (UCHSC) for his help with the caspase-3 activity assay.