Article Text

This article has a correction. Please see:

Preconditioning with death ligands FasL and TNF-α protects the cirrhotic mouse liver against ischaemic injury
  1. J-H Jang,
  2. W Moritz,
  3. R Graf,
  4. P-A Clavien
  1. Swiss HPB (Hepato–Pancreato–Biliary) Center, Department of Surgery, University Hospital Zurich, Switzerland
  1. P A Clavien, Department of Surgery, University Hospital Zürich, Rämistrasse 100, 8091 Zürich, Switzerland; clavien{at}


Background: Ischaemic preconditioning is the pre-emptive proven strategy to reduce ischaemic injury in the liver, but it can be harmful in the elderly or in patients with liver diseases. Ischaemic preconditioning induces a protective effect via activation of oxidative stress. We hypothesised that Fas ligand and tumour necrosis factor α can induce a similar response. Therefore, we tested if death ligands could mimic ischaemic preconditioning.

Methods: Ischaemia was maintained for 60 min in cirrhotic mice. Death ligands were given 40 min before ischaemia. Ischaemic injury was assessed by histology and biochemical assays. To elucidate the mechanism, we used zinc protophorphyrin, an inhibitor of haem oxygenase-1 (HO-1), and gadolinium chloride, an inhibitor of Kupffer cells.

Results: Compared with the control group, death ligand preconditioning strongly reduced all markers of injury: serum transaminase levels, necrosis and apoptosis. Preconditioning caused an upregulation of HO-1, predominantly in macrophages. When zinc protophorphyrin or gadolinium chloride was applied prior to preconditioning, the beneficial effect of preconditioning was lost.

Conclusion: These results demonstrate that ischaemic preconditioning can be replaced by death ligand preconditioning in the cirrhotic liver to prevent ischaemic injury. The protective mechanism depends on HO-1 induction in macrophages. These results open doors for novel hepato-protective strategies in liver surgery and transplantation.

Statistics from

Hepatocellular carcinoma arising in patients with underlying cirrhosis is the fourth most common cause of death from cancer,1 with a rising incidence worldwide due to the increasing prevalence of viral hepatitis, as well as non-alcoholic steatohepatitis.2 3 Transplantation can be offered to very few patients and therefore liver resection remains the sole option for cure in many patients.4 The presence of cirrhosis is a major risk factor for liver surgery; even when liver resection is restricted to patients with well preserved liver function and in absence of portal hypertension, morbidity is high and perioperative mortality is five to ten times higher than in non-cirrhotic patients.4 5

Another contributing factor to poor outcome following major liver surgery is intra-operative blood loss typically occurring during parenchyma trans-section.58 One strategy to prevent blood loss and the need for transfusion is inflow occlusion (i.e. Pringle manoeuvre). However, the interruption of blood flow to the liver induces ischaemic injury, which is a common cause of postoperative liver failure.5 911 Two strategies are available to minimise the negative impact of inflow occlusion, namely ischaemic preconditioning and intermittent clamping.12 13 Intermittent clamping is the only strategy used in patients with liver cirrhosis, as no evidence is currently available showing protection from ischaemic preconditioning in this population.14 Because intermittent clamping is associated with prolonged resection time and bleeding during each period of reperfusion, novel protective approaches are required for livers with underlying diseases in particular.

An attractive protective approach is to subject the liver to a sub-lethal stress, which has been shown to trigger a variety of protective mechanisms.15 16 This is the principle of ischaemic preconditioning, where for example we found that a short period of ischaemia causes a sub-lethal oxidative stress, which induces potent antioxidative mechanisms.15 Exposure to a low-dose of tumour necrosis factor α (TNF-α), an established mediator of injury in the ischaemic liver,17 was found to confer significant protection against prolonged periods of ischaemia in a rat model. Fas and Fas ligand (FasL) are key mediators in the innate immune system,1820 mediating injury in patients with viral and alcoholic hepatitis, and graft versus host disease.21 22 Therefore, both cell death cytokines, TNF-α and FasL, may be involved in the mechanism of ischaemia and reperfusion (I/R) injury in the cirrhotic liver.

This study was based on the hypothesis that “preconditioning” with FasL and TNF-α confers protection in the cirrhotic liver subjected to prolonged periods of ischaemia. First, we developed a model of cirrhosis in mice mimicking the clinical situation, i.e. child A cirrhosis, characterised by clear histological changes consistent with cirrhosis, but well preserved liver function and absence of portal hypertension. The next step was to identify optimal doses of both death cytokines conferring protection, followed by an investigation of the protective mechanisms triggered by each respective cytokine.



Male wild-type mice (C57BL6; Harlan, Horst, the Netherlands) were used for all experiments (n = 5 for each animal group). Animals were fed a standard laboratory diet with water and food ad libitum and were kept under constant environmental conditions. All experimental procedures were approved by the Swiss animal welfare authorities and performed in accordance with the institutional animal care guidelines.

Induction of cirrhosis

Cirrhosis was induced by carbon tetrachloride (CCl4) feeding by gavage. Equal volumes of CCl4 and soyabean oil were mixed. A dose of 4 ml/kg body weight was given three times per week for 6 weeks. For pathological evaluation of cirrhosis Trichrome staining was conducted (fig. 1A). To evaluate severity of liver disease induced by CCl4, Child–Turcotte–Pugh scoring was used. Levels of total bilirubin and albumin were measured using the serum multiple biochemical analyser (Ektachem DTSCII; Johnson and Johnson, Rochester, NY, USA) for the scoring.

Figure 1 Comparison of liver injury in normal and cirrhotic animals after 4 h of reperfusion. Cirrhotic change is shown by Trichrome staining after 6 weeks of CCl4 feeding (A). Normal and cirrhotic mice were subjected to 1 h ischaemia following 4 h reperfusion; histology (B, C, F) and serum transaminases (D, E) were assessed. Necrosis determined by histomorphometry was expressed as % necrotic area relative to total area. Haematoxylin and eosin staining for histomorphometric evaluation showed pale hepatocytes, condensed nuclei and heavy congestion after 4 h of reperfusion in normal (B) and cirrhotic (C) liver (50× magnification). The differences of tissue injury markers between normal and cirrhotic animals were statistically significant. *p<0.05 (t test).

Partial hepatic inflow occlusion

A 70% segmental ischaemia model was used as previously described23 in normal or cirrhotic mice. Under isoflurane/O2 inhalation anaesthesia hepatic inflow to the median and left lobes was occluded by application of a micro-vascular clamp (Aesculap, San Francisco, CA, USA). During ischaemic periods, the abdomen was closed by a single running suture. Reperfusion was initiated by removing the clamp. The animal was allowed to wake up during reperfusion until killed either 1, 4 or 24 h later.

Death ligand preconditioning

Mouse recombinant death ligand proteins (FasL and TNF-α) were purchased from R&D Systems (Minneapolis, MN, USA). To stimulate a protective effect against on an ischaemic insult the proteins were injected into the peritoneum 40 min before hepatic blood flow occlusion. As preconditioners, we selected FasL, which is an ischaemic injury non-specific death ligand, and TNF-α, an ischaemic injury specific death ligand.17

Serum transaminase levels

Serum levels of aspartate aminotransferase (AST) and alanine aminotransferase (ALT) were used as markers of liver injury. Blood samples were obtained at each time point of reperfusion from the inferior vena cava. Blood cells were pelleted by centrifugation at 6000 rpm for 10 min at room temperature. Serum levels of enzyme activities were measured using the multiple biochemical analyser (Ektachem DTSCII; Johnson and Johnson).

Histological assessment

Liver specimens were fixed in formalin. The tissues embedded in paraffin were sectioned and stained with haematoxylin and eosin. The necrotic area was visually identified by manual outlining and quantified by using an image analysis program (Saisam V.4.2.2; Microvision Instruments, Evry Cedex, France). The necrotic areas from ischaemic lobes were expressed as percentages of the whole surface area of the stained sections.

Double immunostaining

Tissue sections of paraffin-embedded specimens were rehydrated and incubated with two primary antibodies: rat monoclonal anti-mouse F4/80 antibody (BMA Biomedicals, Augst Switzerland) and rabbit anti-mouse haem oxygenase-1 (HO-1) antibody (Stressgene, Victoria, BC, Canada). We followed the manufacturer’s instructions for blocking of non-specific binding and antigen retrieval (proteinase K for F4/80 and heating for HO-1). After colour development of diaminobenzidine using biotin–streptavidin for F4/80 (brown colour), HO-1 immunostaining was performed by anti-rabbit second antibody linked with alkaline phosphatase using fuchsin colour development (bright red colour).

TUNEL stain

The ApopTag kit (MP Biomedicals, Solon, OH, USA) was used to visualise nuclear DNA nicks by an immunoperoxidase detection of digoxigenin-labelled genomic DNA. The procedure was performed according to the manufacturer’s instruction. Briefly, liver sections were digested by protease K followed by inactivation of endogenous peroxidase with 3% H2O2 in methanol. The sections were incubated with residues of deoxigenin nucleotide and terminal deoxynucleotide transferase. The sections were then incubated with the antidigoxigenin antibody coupled with peroxidase. The cells with evidence of nuclear DNA fragmentation could be identified after incubating the sections with DAB and peroxidase.


Serum samples were used to detect circulating TNF-α levels by ELISA (Duo Set mouse TNF-α; R&D Systems) following the manufacturer’s instructions.

Western blotting

Mouse liver tissue was homogenised in Tris-based lysis buffer containing protease inhibitor cocktail (Roche Diagnostics, Mannheim, Germany). The protein concentration was determined using a Bradford protein assay (BioRad, Hercules, CA, USA). Reduced SDS-PAGE was performed and samples were blotted onto a PVDF membrane. Rabbit anti HO-1 (Stressgene), rabbit anti-phosphorylated JNK (Cell Signaling, Danvers, MA, USA), and rabbit anti-FLICE (caspase 8) inhibitory protein (FLIP) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (Abcam, Cambridge, UK) polyclonal antibody were employed. Secondary antibody binding and detection were performed according to standard protocols with the ECL detection reagent (GE Healthcare, Chalfont St Giles, UK). All results were measured by densitometry and expressed by relative expression to normal liver.

HO-1 blocking experiment

A specific HO-1 inhibitor, zinc protoporphyrin (ZnPP; 0.38 mg/kg body weight) (Frontier Scientific, Logan, UT, USA), was administered by oral gavage 3 h before hepatic inflow occlusion. The concentrated stock solution was protected from the light until used and was diluted by saline just before gavage. One hour ischaemia following 24 h reperfusion was performed for this experiment. All other procedures were conducted as previously described.

Kupffer cell inactivation

Gadolinium chloride (GdCl3; 10 mg/kg body weight) was injected into peritoneum of cirrhotic mice 24 h before ischaemia in presence or absence of death ligand preconditioning. Mice were killed 24 h after reperfusion.

Statistical analysis

Data are presented as means (SD). Groups were compared with the Student’s t test for unpaired samples using Prism 4.0 (GraphPad Software, San Diego, CA, USA). A two-sided p value <0.05 was considered to indicate as statistical significant.


Is the cirrhotic liver more susceptible to ischaemic injury than the normal liver?

We compared in a first set of experiments the effect of I/R in normal and cirrhotic mice. The cirrhotic mice exhibit normal serum bilirubin and albumin levels, and no ascites and no encephalopathy, representing Child A cirrhosis. Serum transaminase levels and histology were used as endpoints. While after 1 h of ischaemia followed by 4 h of reperfusion serum AST and ALT levels were increased in both normal and cirrhotic animals, serum enzyme levels were significantly higher in cirrhotic animals (fig. 1D, E). Consistent with serum transaminase levels, the extent of tissue injury determined by morphometry was also significantly higher in cirrhotic animals. Semiquantitatively assessed tissue injury showed significantly more severe injury in cirrhotic animals (fig. 1C) than normal animals (fig. 1B) 4 h after reperfusion (fig. 1F). These experiments show that cirrhosis adversely affects the response of the liver to I/R.

Is death ligand preconditioning protective against ischaemic injury in cirrhotic livers?

In this set of experiments we tested whether preconditioning with death ligands is protective against ischaemia/reperfusion injury in the cirrhotic liver. In pilot studies FasL and TNF-α were injected at three different doses prior to ischaemia. Four hours after reperfusion AST levels were significantly elevated in control animals, while animals injected with 1 ng/g body weight of TNF-α or FasL exhibited reduced markers of hepatic injury (fig. 2A, B). Subsequent preconditioning experiments were then performed with this dose. To evaluate the time course of injury, cirrhotic mice were pretreated with TNF-α, FasL or saline and studied 1, 4 or 24 h after reperfusion. AST levels (Fig. 2C) were dramatically lower in animals pretreated with either of the death ligands, particularly at 4 h after reperfusion. Histological assessment 24 h after reperfusion demonstrated a high degree of necrosis (48 (SD 12)%) in saline-treated animals (supplementary fig. A). The extent of necrosis was dramatically reduced (fig. 2D) when FasL (22 (SD 12%)) (supplementary fig. B) or TNF-α (16 (SD 8)%) (supplementary fig. C) preconditioning was applied before ischaemia. To further confirm whether TNF-α and FasL would protect the liver from ischaemic injury, a survival study was performed. While saline-injected control animals had a 20% survival rate, 60% survived after TNF-α treatment and 100% after FasL application, supporting our conclusion that these cytokines are protective at low doses (fig. 2E). Tissue damage was characterised by extensive necrosis and infiltration of inflammatory cells 24 h after reperfusion.

Figure 2 Dose–response curve of death ligand preconditioning in cirrhotic livers. The most protective concentrations of FasL and TNF-α were found by pilot studies in cirrhotic animals. Serum AST levels and necrotic area were taken to evaluate the effects of death ligand preconditioning. Death ligands applied at 1 ng/g body weight showed significantly lower serum AST levels than either the control or other groups tested with higher concentration of death ligands (A, B). Time course of AST levels after 1, 4, and 24 h of reperfusion after saline treatment or preconditioning with TNF-α or Fas (1 ng/g body weight) (C). Histomorphometric evaluation of necrotic area (D). *p<0.05 (t test). (E) Survival rate of cirrhotic mice after 60 min of 70% ischaemia. Mice were preconditioned by FasL, TNF-α or saline injection. After ischaemia, the non-ischaemic lobes were removed and the mice were allowed to wake up. The survival experiment was terminated after 7 days.

Does death ligand preconditioning protect the cirrhotic liver from apoptotic cell death?

Apoptosis is a common feature of liver injury caused by ischaemic insult. To investigate the protective mechanisms we assessed apoptotic markers, e.g. TUNEL staining, enzymatic assay of caspase 3, serum TNF-α level by ELISA, detection of phosphorylated Jun amino-terminal kinase (pJNK), FLIP by western blotting after 4 h of reperfusion. TUNEL staining of cirrhotic livers without ischaemic injury revealed about 2% of cells with apoptotic nuclei. In contrast 4 h after an ischaemic insult, almost 35% of the nuclei appeared to be apoptotic (fig. 3A, supplementary fig. D). Pretreatment with death ligands caused a significant reduction in TUNEL-positive nuclei, suggesting that both FasL and TNF-α preconditioning induced a defence against apoptosis (fig. 3A, supplementary fig. E, F). This conclusion was supported by the assessment of active caspase 3. In saline-treated animals, the activity of caspase 3 was significantly increased compared with either sham-operated or preconditioned animals (fig. 3B). To further study entry into apoptosis, the presence of serum TNF-α, pJNK and FLIP was determined. Activation of JNK by TNF-α and reactive oxygen species after ischaemia following reperfusion leads to hepatocyte apoptosis and degradation of the anti-apoptotic molecule FLIP.23 24 Circulating TNF-α was low in sham-operated animals. In saline-injected animals, TNF-α was increased almost tenfold 4 h after reperfusion. Again, both death ligands effectively reduced the secretion of TNF-α 4 h after pre-treatment (fig. 3C). Activation of JNK was associated with turnover of FLIP. Activated JNK was significantly increased in saline-treated animals compared with preconditioned groups (fig. 3D, F). After 4 h of reperfusion the amount of FLIP was significantly reduced in saline-treated group but unaffected by death ligand preconditioning (fig. 3E, F).

Figure 3 Effect of death ligand preconditioning on apoptosis. TUNEL staining was used to indicate areas of apoptosis (A). Activity of caspase 3, the executer caspase, assessed by enzymatic assay (B). Determination of TNF-α secretion in serum of mice 4 h after reperfusion (C). Phosphorylation of JNK (D, F) and amount of FLIP is shown by western blotting (E, F). Sham: cirrhotic mouse undergoing a sham operation. Control: cirrhotic mouse with ischaemia for 1 h/4 h reperfusion. FasL/TNF-α was injected 40 min prior to ischaemia. *p<0.05 (t test).

Does death ligand preconditioning induce protective molecules against ischaemic injury in the cirrhotic liver?

Preconditioning with TNF-α has been shown to reduce tissue injury subsequent to 90 min I/R in healthy female mice through MAPK and NFκB activation.26 To identify the mechanism by which death ligand preconditioning induces cytoprotection we evaluated a number of proteins that are known to be protective in an ischaemic insult, since FasL and TNF-α preconditioning in cirrhotic liver appear to mediate their protection through mechanisms different from those in the healthy liver. Proteins such as heat shock protein 70 (HSP-70), B cell lymphoma 2 (BCL-2) and HO-1 are reportedly induced during ischaemia and provide protection in subsequent reperfusion.27 28 To investigate the effect of preconditioning, animals were untreated (normal) or saline (sham), TNF-α and FasL injected. After 40 min, corresponding to the preconditioning time prior to ischaemia, liver tissue was directly harvested and processed for analysis.

While neither HSP-70 nor BCL2 levels changed (data not shown), HO-1 was induced in tissue homogenates by a factor of 2 (FasL) and 2.5 (TNF-α) (fig. 4A). Although the induction was not very strong, we further investigated the distribution of HO-1 in the preconditioned livers. In comparison with normal or sham animals, FasL-treated or TNF-α-treated animals exhibited an increased number of HO-1 expressing F4/80 positive cells, presumably Kupffer cells (fig. 4B). A histomorphometric analysis demonstrated a more than threefold increase in the number of HO-1 and F4/80 double positive cells (fig. 4C). Other leucocyte infiltrates, i.e. neutrophilic granulocytes assessed by myeloperoxidase staining, were similar in the preconditioned and the saline control livers (data not shown).

Figure 4 Death ligand preconditioning induces HO-1 in non-parenchymal cells before ischaemia. HO-1 expression was determined by western blotting and immunohistochemistry. Western blotting of tissue extract for HO-1 (A) was performed after 40 min of preconditioning in normal and cirrhotic mice. Co-expression of F4/80 and HO-1 in the non-parenchymal cells by death ligand preconditioning (B). Examples of individual cells which are either F4/80 positive (D, brown colour), HO-1 positive (E, bright red colour), F4/80 and HO-1 double positive (F) are shown separately. Morphometric determination of F4/80 and HO-1 double positive cells (C). The numbers of double positive cells were highly increased by death ligand preconditioning. *p<0.05 (t test).

Does HO-1 play a central role in death ligand preconditioning?

The previous experiment pointed out to a protective role of HO-1. To further substantiate this hypothesis, we treated the cirrhotic animals with a specific inhibitor of HO-1, ZnPP. Three hours later all animals were pretreated with saline (sham), FasL or TNF-α for 40 min. Following ischaemia for 1 h livers were reperfused for 24 h. Markers of injury, i.e. AST (fig. 5A) and necrosis (fig. 5B), demonstrated considerable liver damage in saline-injected cirrhotic animals. The inhibitor ZnPP did not have any effect in cirrhotic control animals. As shown before, FasL or TNF-α pre-treatment of cirrhotic animals strongly reduced tissue injury after reperfusion. However, the effect of FasL or TNF-α was completely abolished by HO-1 inhibitor treatment (fig. 5 A, B). Therefore, we conclude that the induction of HO-1 is a central mechanism of cytoprotection in the cirrhotic liver.

Figure 5 Inhibition of HO-1 reverses ischaemic injury in cirrhotic liver. The effect of HO-1 was inhibited by ZnPP administration by gavage. The inhibitory capacity was assessed by serum AST and histomorphometry of necrotic area after 24 h reperfusion. Serum AST levels in ZnPP + death ligand groups were significantly higher than in the death ligand preconditioned group (A). Consistent with AST data, the necrotic area was significantly more pronounced in Zn-PP + death ligand groups than in control group (B). *p<0.05 (t test).

Are Kupffer cells involved in the response of death ligand preconditioning?

In this experiment we investigated whether Kupffer cells are responsible for the protective effect by TNF-α preconditioning. Therefore we inactivated Kupffer cells by GdCl3 (10 mg/kg body weight in saline), given through intraperitoneal injection 24 h before ischaemia. After 60 min of ischaemia and 24 h of reperfusion, serum AST levels (fig. 6A) and tissue injury (fig. 6B) reached similar levels as saline-treated control animals. To demonstrate that TNF-α is acting through Kupffer cells, cirrhotic animals were injected with TNF-α alone or in combination with GdCl3. TNF-α preconditioning, as expected, protected liver from ischaemic insult. When Kupffer cells were silenced by GdCl3, TNF-α was ineffective. This suggest that TNF-α mediates its effect through a Kupffer cell-dependent pathway.

Figure 6 Protective effect of death ligand preconditioning is Kupffer cell-dependent. GdCl3, an inhibitor of Kupffer cells, injected 24 h prior to ischaemia reversed ischaemic injury in TNF-α-treated animals. Serum AST and necrotic area were measured to demonstrate the inhibitory effect of GdCl3 against Kupffer cell activity after 24 h of reperfusion. GdCl3 reversed serum AST level similar to the control group (A). Similarly, the necrotic area was comparable with the control groups when GdCl3 and TNF-α were applied. *p<0.05 (t test).


Preconditioning with both FasL and TNF-α given shortly prior to an ischaemic insult significantly ameliorates injury subsequent to reperfusion with dramatic reduction of apoptotic cell death and subsequent necrosis. We further identified upregulation of HD-1 as the protective mechanism downstream of death ligand preconditioning. As HO-1 was located mostly in the macrophages in the liver (Kupffer cells) and Kupffer cell inhibition led to a reversal of the protective effects of FasL and TNF-α, we concluded that death ligand preconditioning confers dramatic protection in the cirrhotic liver though a HO-1–Kupffer cell-dependent pathway.

Clinical relevance and model

To reduce the postoperative morbidity and mortality, several procedures have been proposed that protect the liver. While both procedures of ischaemic preconditioning and intermittent clamping12 are successful in many patients, recent evidence has shown that both procedures are not effective enough in a subset of patients with complex disease and the elderly.29 Therefore, alternative procedures are required to induce protection, with pharmacological preconditioning being the most attractive, because it is not invasive and can be easily given prior to surgery.27 The novel finding in this study is the protective effects of FasL and TNF-α given prior to a prolonged ischaemic insult. Both mediators induce cell death at high doses, but at low doses they can activate a number of cell types involved in liver injury, including Kupffer and endothelial cells.

An additional formidable challenge to improve surgery in patients with cirrhosis is the availability of suitable experimental models where mechanism of injury can be investigated. The mouse model of cirrhosis is an accepted model of extensive parenchymal destruction with bridging fibrosis.30 Parenchymal destruction is induced by chronic administration of a toxic substance, i.e. CCl4.31 This toxic model is widely used, yet it has its limits and does not completely reiterate the pathophysiological changes seen in human cirrhosis. The mouse liver appears to constantly regenerate, and it has been observed that the mouse liver is able to completely reverse “cirrhosis” when CCl4-feeding is terminated. Thus, in the strictest sense of the definition of cirrhosis, i.e. irreversible fibrosis, the mouse liver is imperfect. Nevertheless, the maintenance of extensive fibrosis in the liver is paramount to the current efforts to understand and improve human liver cirrhosis.

Preconditioning effects in the mouse liver

In the current experiments we used 40 min of preconditioning prior to surgery. Pilot experiments with a dose–response curve for both death ligands identified a dose of 1 ng/g body weight. A significant improvement was observed for all parameters tested. The injury, as exemplified by transaminase levels, was transient; however, the ensuing necrosis was extensive 24 h after reperfusion. At 4 h after reperfusion apoptosis was significantly improved. From the rapid conversion of apoptosis to necrosis in this model, we assume that entry into apoptosis at 4 h after reperfusion might not result in a full execution of programmed cell death in these animals. Presumably, a significant number of cells will turn necrotic, particularly in those animals not protected by preconditioning. This process, termed “apo-necrosis” leads to inflammation and, when considering the extent of necrosis observed in the cirrhotic liver (almost 50%), potentially to a complete loss of organ function.32

FasL and TNF-α induce a cell death in many cell types. However, they might stimulate cellular pathways without detectable organ injury in a certain dose. We could not find injury markers assessed by TUNEL, haematoxylin and eosin, or caspase 3 and 8 western blotting (data not shown). While we had expected that death ligand preconditioning induces MAPK activation, including p38, Erk and JNK, as shown by Teoh et al in normal mice,26 it was not reproducible in the cirrhotic mice. It is not clear if HO-1-positive macrophages are recruited or derived from the liver.

The devastating impact of I/R on postoperative injury with extensive necrosis was almost completely prevented with a single dose of death ligand. This observation supports the concept of pharmacological preconditioning. While the application of TNF-α or FasL might be difficult in clinical practice at this moment, the potential to apply a single dose 40 min prior to surgery paves the way to safe drugs applied in a similar fashion. In our opinion, the crucial point will be to target a protective reaction, e.g. in Kupffer cells. Whether this will be with modified “blunted” death ligands or other activators of Kupffer cells is not clear. The beneficial effects, combined with ease of handling, might provide novel attractive strategies in the cirrhotic population, who are particularly prone to develop liver failure after surgery.

We identified HO-1 and Kupffer cells as a downstream protective target of both death ligands in cirrhotic livers. HO-1 is a well established hepatoprotective molecule which has an anti-oxidative, anti-inflammatory, vaso-relaxing and anti-apoptotic effect.3336 Accumulating data about HO-1 production in the liver indicate that Kupffer cells are a major source of HO-1.37 Several ways to utilise the benefits of this protein have been suggested. Yutamura et al showed induction of HO-1 by giving hemin, a potent HO-1 inducer in normal rat liver graft.38 But it is unclear whether it will induce Kupffer cell-specific HO-1 in cirrhotic liver.

In summary, we have shown that a single dose of TNF-α or FasL strongly protects the cirrhotic liver against I/R injury. We provided further strong evidence that these protective effects are mediated by HO-1 located in Kupffer cells. A single injection of drugs targeting HO-1 in Kupffer cells might provide an attractive alternative therapy to prevent loss of tissue integrity and improve morbidity in cirrhotic patients undergoing liver surgery.


The authors thank M Bain-Stucki, A Morger-Gartenmann and U Ungethüm for excellent technical help and K-J Kang for mentorship.

View Abstract


  • Funding: This study was supported by Korea Science & Engineering Foundation (KOSEF; M01-2004-000-20431-0), Seoul, Korea (to J-H Jang), Swiss National Foundation (SNF; 3200B0-109906), Bern, Switzerland (to P-A Clavien).

  • Competing interests: None.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.