Induction of the heme-degrading enzyme heme oxygenase-1 (HO-1) has been shown to be beneficial in terms of improvement of liver allograft survival and prevention of CD95-mediated apoptosis in the liver. In the present study, we investigated the effects of HO-1, and its products carbon monoxide (CO), biliverdin (BV), and iron/ferritin, in a mouse model of inflammatory liver damage inducible by lipopolysaccharide (LPS) in mice sensitized with the hepatocyte-specific transcription inhibitor D-galactosamine (GalN). Our results show that HO-1 induction by cobalt-protoporphyrin-IX (CoPP) reduced cytokine expression, protected mice from liver injury, and prolonged survival. While in contrast to ferritin overexpression, single administration of the CO donor methylene chloride (MC) or of BV also protected mice from liver damage, only coadministration of both HO products prolonged survival and reduced the expression of cytokines, e.g., tumor necrosis factor (TNF) and interferon γ (IFN-γ). In conclusion, HO-1–induced prolongation of survival, but not the protection from liver damage, seems to be dependent on down-regulation of cytokine synthesis. (HEPATOLOGY 2004;40:1128–1135.)
Degradation of heme by heme oxygenases (HOs) results in the production of carbon monoxide (CO), free iron, and biliverdin (BV). HO-1, in contrast to the second isoform HO-2, is inducible by various stimuli.1, 2 To investigate HO effects in vitro as well as in vivo, HO-1 can be induced by application of cobalt-protoporphyrin-IX (CoPP), for example, while tin- protoporphyrin-IX (SnPP) suppresses HO activity.3, 4 In the liver, administration of CoPP to mice results in HO-1 expression in hepatocytes as well as in Kupffer cells.5In vivo, HO-1 induction has been shown previously to protect mice from apoptotic liver damage5 and to protect rats from liver graft rejection as well as from ischemia/reperfusion injury.6–8
To investigate mechanisms of cytokine-dependent liver injury, mice can be sensitized with the hepatocyte-specific transcriptional inhibitor D-galactosamine (GalN) in combination with the macrophage activator lipopolysaccharide (LPS).9 In this GalN/LPS model, mice develop severe liver injury, which is dependent on tumor necrosis factor (TNF)10, 11 and interferon γ (IFN-γ)12 induction, while interleukin (IL) 10 prevents injury.13
In the present study we show that HO-1, induced by CoPP, protects mice from GalN/LPS-induced liver injury, prolongs survival, and reduces cytokine expression. Protection from liver damage could also be achieved by pretreatment with the HO products CO and BV, but not by ferritin. Prolongation of survival and cytokine reduction was only detectable in mice pretreated with a combination of both hepatoprotective products CO and BV.
HO, heme oxygenase; CO, carbon monoxide; BV, biliverdin; CoPP, cobalt-protoporphyrin-IX; SnPP, tin-protoporphyrin-IX; GalN,D-galactosamine; LPS, lipopolysaccharide; TNF, tumor necrosis factor; IFN, interferon; IL, interleukin; MC, methylene chloride; ppm, parts per million; Ad-Ferritin, adenovirus encoding the recombinant human heavy chain of ferritin; Ad-lacZ, adenovirus encoding β-galactosidase; RT-PCR, reverse-transcriptase polymerase chain reaction; ALT, alanine aminotransferase; ELISA, enzyme-linked immunosorbent assay.
Materials and Methods
BALB/c-mice (6-8 weeks old; weight range, 18-22 g) were obtained from the animal facilities of the Institute of Experimental and Clinical Pharmacology and Toxicology of the University of Erlangen-Nuremberg, Erlangen, Germany. All mice received human care according to the guidelines of the National Institutes of Health and the legal requirements in Germany. They were maintained under controlled conditions (22°C, 55% humidity and 12-hour day/night rhythm) and fed standard laboratory chow.
Dosage and Application Routes.
All substances were dissolved in pyrogen-free saline unless indicated otherwise. GalN (700 mg/kg; Roth, Karlsruhe, Germany) was administered intraperitoneally together with lipopolysaccharide (LPS) from Salmonella abortus equi (5 μg/kg; 7.5 μg/kg for survival experiments; Metalon, Ragow, Germany) or 30 minutes prior to intravenous administration of recombinant murine TNF-α (Innogenetics, Brussels, Belgium; 8 μg/kg dissolved in saline/0.1% human serum albumin). Methylene chloride (MC; 6.2 μmol/kg; Sigma Chemical Co., Taufkirchen, Germany) was administered orally in corn oil 6 hours before GalN/LPS treatment, resulting in about 9% carboxyhemoglobin concentration14 (data not shown). BV (25 mg/kg; Sigma Chemical Co.) was administered intraperitoneally 24 hours before GalN/LPS. In parallel, control animals were treated with previously boiled BV to rule out protective effects of contaminating LPS. CoPP (5 mg/kg; Alexis Deutschland GmbH, Grunberg, Germany) was administered intraperitoneally 24 hours before the induction of liver injury. SnPP (25 mg/kg; Biotrend GmbH, Cologne, Germany) was administered intraperitoneally 4 hours prior to the induction of liver injury. Anti-TNF antibody (immunoglobulin G fraction of plasma derived from a sheep immunized with recombinant murine TNF-α) and anti–IFN-γ antibody (plasma derived from a rabbit immunized with IFN-γ) were administered intravenously 15 minutes before GalN/LPS administration.
CO at a concentration of 1% (10,000 parts per million [ppm]) in compressed air was mixed with balanced air (21% oxygen) in a stainless-steel mixing cylinder before entering the exposure chamber. CO concentrations were controlled by varying the flow rates of CO in a mixing cylinder before delivery to the chamber. A CO analyzer (Interscan Corporation, Chatsworth, CA) was used to measure CO levels continuously in the chamber. Mice were placed in the CO exposure chamber 24 hours before induction of liver injury and were kept in the exposure chamber thereafter. CO concentration was maintained at 250 ppm at all times.
Adenoviral Gene Transfer.
Adenoviruses encoding the recombinant human heavy chain of ferritin (Ad-Ferritin) or β-galactosidase encoding gene LacZ (Ad-lacZ) were injected intravenously (2 × 109 plaque-forming units/mouse) 40 hours prior to challenge. Expression of human ferritin in mouse liver tissue was verified by reverse-transcriptase–polymerase chain reaction (RT-PCR).
Analysis of Liver Enzymes.
Hepatocyte damage was assessed 8 hours—and in survival experiments, 6 hours—after the induction of liver injury by measuring plasma enzyme activities of alanine aminotransferase (ALT), using an automated procedure.15
Detection of Messenger RNA by Real-Time RT-PCR.
Isolation of total RNA from liver tissue was carried out using the NucleoSpin RNA II Kit (Macherey-Nagel, Duren, Germany). To analyze altered gene expression, messenger RNA was transcribed into complementary DNA using SuperScript II RNase H− Reverse Transcriptase (Invitrogen GmbH, Karlsruhe, Germany). Oligonucleotides for subsequent polymerase chain reactions were also obtained from Invitrogen. The following oligonucleotide pairs were used: β-actin (729-752 and 1076-1053 in GenBank X03765); TNF (158-178 and 386-371 in GenBank X02611); IFN-γ (113-131 and 513-494 in GenBank M28621); and ferritin heavy chain (human; 199-219 and 597-578 in GenBank M97164). Real-time RT-PCR was performed using a LightCycler rapid thermal cycler system (Roche Diagnostics GmbH, Mannheim, Germany) and the LightCycler-FastStart DNA Master SYBR Green I mix (Roche Diagnostics GmbH). Reactions were performed in a 10-μL volume. To confirm amplification, specificity polymerase chain reaction products were subjected to a melting curve analysis.
Cytokine Determination by Enzyme-Linked Immunosorbent Assay (ELISA).
Sandwich ELISAs for murine plasma TNF, IFN-γ, IL-6, and IL-10 were performed using flat-bottom high-binding polysterene microtiter plates (Greiner, Nurtingen, Germany). Antibodies were purchased from BD PharMingen Transduction Laboratories (Heidelberg, Germany). Streptavidin-peroxidase (Jackson Immuno Research, West Grove, PA) and the peroxidase chromogen tetramethylbenzidine (Boehringer Mannheim, Mannheim, Germany) were used according to the manufacturers' instructions. For determination of intrahepatic TNF concentrations, livers were prepared as described previously.16 Liver lysates were adjusted to equal protein concentrations and analyzed for murine TNF with the Quantikine M Kit (BioRad, Munich, Germany).
Determination of Caspase-3 Activity.
To determine the activation of caspase 3 in the liver tissue of mice, liver homogenates (50% wt/wt) were prepared in lysis buffer containing 10 mmol/L HEPES (pH 7.4), 1 mmol/L CHAPS (3-[(3-cholamidopropyl)dimethylammonio]-1-propane sulfonate), and 1 mmol/L DTT (dithiothreitol) and analyzed using a colorimetric caspase-3 assay kit (Sigma Chemical Co.), according to the manufacturer's instructions.
For determination of tissue structure, liver tissue was fixed in 2% Zamboni solution17 for 24 hours and subsequently embedded in paraffin. Sections were stained for hematoxilin-eosin using a standard protocol and analyzed by light microscopy.
The results were analyzed using Student t test if 2 groups were compared or by ANOVA followed by the Dunnett test if more groups were tested against a control group. If variances were inhomogeneous in the Student t test, the results were analyzed using the Welsh test. All data in this study are expressed as mean ± SEM. A P value less than or equal to .05 was considered significant.
HO-1 Induction Protects Mice From Inflammatory Liver Injury and Prolongs Survival.
Induction of HO-1 has previously been shown to protect rats from transplantation-related complications by preventing cytokine responses.18 We investigated the effect of HO-1 induction by CoPP, or HO repression by SnPP, on liver damage and survival in the cytokine-dependent GalN/LPS model of apoptotic liver damage. Measurement of plasma ALT revealed that HO-1 induction protected mice from liver injury, while repression of HO increased liver injury (Fig. 1A). Similarly, induction of HO-1 prolonged survival compared to mock-treated animals, and repression of HO had no further detrimental effect on survival rates (Fig. 1B).
HO-1 Induction Reduces Cytokine Expression in the Mouse Liver.
To elucidate mechanisms leading to protection from GalN/LPS-induced liver damage, we measured plasma cytokine expression in CoPP versus mock-pretreated animals. We found that CoPP pretreatment significantly reduced TNF, IL-6, IFN-γ, and IL-10 production (Fig. 2A). Since TNF, and possibly IFN-γ, are key mediators of liver damage in the GalN/LPS model,10, 12 their expression was also measured in liver tissue. Real-time RT-PCR for TNF (Fig. 2B) and IFN-γ (Fig. 2C), revealed that both cytokine expression levels were significantly reduced in the CoPP-treated mice. In contrast, expression of IL-6 and IL-10 in liver tissue was not altered in response to the CoPP treatment (data not shown). Additionally, we measured TNF content in liver tissue using protein ELISA and found that it was also significantly reduced in the CoPP-treated group (data not shown). Taken together, these results indicate that HO-1–mediated TNF and/or IFN-γ reduction, but not IL-6 or IL-10 reduction, might contribute to protection from immune-mediated liver damage.
Protection From GalN/LPS-Induced Liver Injury Is Mediated by the HO-1 Products CO and BV.
To determine the HO-1 product(s) responsible for protection from GalN/LPS-induced liver damage, we treated mice with the CO-releasing agent MC (Fig. 3A-D), with BV (Fig. 3E-H), or Ad-Ferritin. We found that neither CO (Fig. 3A) nor BV pretreatment (Fig. 3E) prolonged survival, though measurement of ALT (Fig. 3B, 3F) and measurement of caspase-3 activation (Fig. 3C, 3G) revealed that both HO-1 products protected mice from GalN/LPS-induced liver damage. The effect of MC was verified by exposing mice to CO at 250 ppm for 24 hours in a CO chamber. These mice were also protected from GalN/LPS-induced liver damage (ALT: GalN/LPS, 4544 ± 1344 U/L; CO/GalN/LPS, 1468 ± 376 U/L; n = 4; P ≤ .05). Surprisingly, none of the HO-1 products reduced cytokine expression (Fig. 3D, 3H; data not shown for CO exposure), compared to the CoPP pretreatment (Fig. 2A). Thus, protection from liver damage by either CO or BV seems to be independent of cytokine reduction. Furthermore, neither prolongation of survival nor cytokine reduction are mediated by any of the HO-products alone.
Ferritin Overexpression Does Not Mediate Protection From GalN/LPS-Induced Liver Injury.
Next we investigated the effect of iron, the third HO product, on GalN/LPS-induced liver damage. We found that overexpression of the iron-inducible binding protein ferritin by adenoviral gene transfer did not protect mice from GalN/LPS-induced liver damage (Fig. 4A) nor did it reduce cytokine expression (Fig. 4B). Expression of the human ferritin heavy chain in the mouse liver was verified by RT-PCR (data not shown).
Cooperative Action of CO and BV Prolongs Survival and Reduces Cytokine Expression Following GalN/LPS Treatment.
Since none of the HO products alone prolonged survival following GalN/LPS treatment, or reduced cytokine production, we pretreated mice with MC plus BV. Similar to single pretreatment with MC or BV, copretreatment resulted in significantly reduced aminotransferase release (Fig. 5A). But in contrast to single pretreatment, copretreatment with BV plus MC significantly reduced the cytokine expression (Fig. 5B) and prolonged survival (Fig. 5C). Measurement of caspase-3 activity in single or copretreated mice revealed that while single pretreatment reduced apoptosis in the liver (see also Fig. 3), copretreatment further reduced caspase-3 activity (Fig. 5D). These results suggest that cooperative action of CO and BV is necessary to mimic the protective effects of HO-1 induction by CoPP in the present model.
Protection from GalN/LPS-induced liver damage by CO or BV or copretreatment was also investigated using hematoxylin-eosin–stained liver sections. Although livers of GalN/LPS-treated mice showed hemorrhage, leukocyte infiltrations, and severe signs of apoptotic damage, such as pyknotic nuclei and condensed chromatin (Fig. 6A, B), single pretreated livers were almost free of apoptotic signs (Fig. 6C-F), and copretreated livers appeared to be even more protected (Fig. 6G, 6H).
Prolongation of Survival and Protection From GalN/LPS-Induced Liver Damage Is Dependent on TNF but Not IFN-γ Reduction.
Next we investigated whether TNF or IFN-γ reduction would result in survival prolongation after GalN/LPS challenge. We found that anti-TNF pretreatment, but not anti–IFN-γ pretreatment, prolonged survival (Fig. 7A) and resulted in protection from liver damage (Fig. 7B). The bioactivity of both antibodies has been tested in the concanavalin A model of immune-mediated liver injury, which is strictly dependent on the induction of TNF as well as IFN-γ.16–20 Both antibodies were found to reduce concavalin A-induced liver damage significantly (data not shown). Therefore, CO plus BV-mediated reduction of TNF—but not of IFNγ—seems to be responsible for the survival prolongation effect following the GalN/LPS challenge.
Experimental liver injury in mice, such as that induced by GalN/LPS administration, is a widely used model to investigate the immunomodulatory potential of developed substances.9 In this model, neutralization of TNF, the lack of TNF-receptor 1, or deficiency in IFN-γ signaling results in protection from fulminant liver damage.9, 10, 12 It has been shown recently that HO-1 induction inhibits the production of proinflammatory cytokines in response to LPS in vivo.21 In the present study, we investigated the effects of the HO-1 inductor CoPP, as well as the effects of the HO products, on the outcome of liver damage and on the prolongation of survival using the GalN/LPS model.
CoPP has been shown to induce HO-1 in the liver, in hepatocytes as well as in Kupffer cells.5 We found that CoPP administration protected mice from GalN/LPS-induced liver damage and prolonged survival, while repression of HO by SnPP pretreatment significantly aggravated liver damage. This aggravation of liver damage might be due to the effect of SnPP on the second isoform of HO, HO-2, which is expressed constitutively in various cells.2 This may also represent an endogenous protection mechanism of the liver via HO-2.
The mechanism by which HO-1 protects from GalN/LPS-induced liver damage could be the reduction of cytokine expression, especially of TNF and/or IFN-γ, because these cytokines have been shown to be central mediators of damage in this model.10, 12 In fact, we measured significantly reduced cytokine expression in CoPP-treated mice following GalN/LPS administration. Expression levels of TNF and IFN-γ were also measured in liver tissue and were found to be significantly down-regulated by HO-1 induction. Therefore, HO-1–dependent cytokine regulation as a protection mechanism seemed to be very likely.
The protective effect of HO-1 induction should also be achieved by the administration of at least one of its products, i.e., CO, BV, or iron. Indeed, each of these products has been shown to exert cytoprotective effects. CO protects endothelial cells from apoptosis in vitro22 by inhibiting inducible NO-synthase activity,23 which is a key factor for GalN/LPS-induced liver injury,24 and prevents anti-CD95–induced apoptosis in the liver.5 Bilirubin, an intracellular reaction product of BV, represents a physiologically important defense against reactive oxygen species.25, 26 It inhibits peroxynitrite-mediated protein oxidation27 and protects mice against acetaminophen-induced hepatotoxicity.28 Free iron can be toxic to cells by generating free radicals. However, in turn, it induces ferritin, which is a Fe2+-sequestrating protein.29 Ferritin has been shown to prevent lipid peroxidation and serves as a long-term iron detoxification mechanism.30, 31 Also, a recent study shows that overexpression of ferritin protects endothelial cells and hepatocytes from undergoing apoptosis and ameliorates ischemia/reperfusion of the liver.32 In the present study, we found that CO and BV, but not overexpression of ferritin, protected mice from GalN/LPS-induced liver damage. Surprisingly, in contrast to CoPP-treated mice, none of the HO-1 products reduced plasma TNF and IFN-γ expression levels, and neither CO nor BV alone prolonged survival, although liver protection occurred. These results indicate that reduction of cytokine expression is no absolute requirement for HO-1–mediated protection from liver damage. This is in line with a previous report showing that HO-1 induction or CO administration protects mice from cytokine-independent liver damage evoked by the administration of agonistic anti-CD95 antibodies.5 Therefore, protection from liver damage seems to be dependent on an HO-1–induced effect directly in hepatocytes.
When both protective HO-1 products, i.e., CO and biliverdin, were administered in combination, mice were completely protected from GalN/LPS-induced liver damage. In contrast to single administration of CO or BV, coadministration reduced cytokine production and survival was prolonged. This implies that there might be a correlation between HO-1–dependent cytokine reduction and prolonged survival. Most prominently, TNF and IFN-γ expression were down-regulated. Both cytokines have previously been shown to play important roles in GalN/LPS-induced liver damage.10, 12 Therefore, we further asked whether repression of TNF or IFN-γ would be the cause of HO-1–induced prolongation of survival, using antagonistic, bioactive antibodies. Pretreatment with anti-TNF, but not with anti-IFN-γ, protected mice from liver damage and prolonged survival, showing that prolongation of survival is dependent on down-regulation of TNF rather than of IFN-γ. The observation that administration of anti–IFN-γ antibody did not protect mice from GalN/LPS-induced liver damage is in contrast to previous reports showing that IFN-receptor−/− mice are resistant to GalN/LPS-induced liver injury.12 However, as shown in our previous study, anti–IFN-γ treatment also failed to protect mice in other GalN-dependent models of liver damage, inducible either by staphylococcus enterotoxin B or agonistic anti-CD3 antibody.20
Taken together, it seems that there are 2 independent mechanisms by which HO-1 protects mice from liver damage. One is independent of cytokine reduction and seems to directly affect downstream events of apoptotic signaling.5 The other is dependent on TNF reduction and, besides protection from liver damage, results in prolonged survival following GalN/LPS challenge.
In conclusion, our results show that cytokine reduction, especially of TNF, is a major protective effect of HO-1 in vivo. If HO-1 products should become a liver therapeutic tool in future, their impact on TNF expression should be carefully observed.
The authors thank Sonja Heinlein, Andrea Agli, Anita Hecht, and Hedwig Symowski for their perfect technical assistance.