Hepatocytes as cytotoxic effector cells can induce cell death by CD95 ligand-mediated pathway


  • Clifford S. Guy,

    1. Molecular Virology and Hepatology Research, Division of Basic Medical Sciences, Memorial University, St. John's, Newfoundland A1B 3V6, Canada
    Search for more papers by this author
  • Jinguo Wang,

    1. Molecular Virology and Hepatology Research, Division of Basic Medical Sciences, Memorial University, St. John's, Newfoundland A1B 3V6, Canada
    Search for more papers by this author
  • Tomasz I. Michalak

    Corresponding author
    1. Molecular Virology and Hepatology Research, Division of Basic Medical Sciences, Memorial University, St. John's, Newfoundland A1B 3V6, Canada
    2. Discipline of Laboratory Medicine, Faculty of Medicine, Health Sciences Centre, Memorial University, St. John's, Newfoundland A1B 3V6, Canada
    • Molecular Virology and Hepatology Research, Faculty of Medicine, Health Sciences Centre, Memorial University of Newfoundland, St. John's, NL, Canada A1B 3V6
    Search for more papers by this author
    • fax: 709-777-8279.

  • Potential conflict of interest: Nothing to report.


The liver plays an increasingly recognized role in the host's immune responses. The direct contribution of hepatocytes as effector cells to local immunity, pathogen containment, and liver disease is not determined. This in vitro study examined whether hepatocytes can eliminate other cells via a CD95 ligand (CD95L or FasL)/CD95 (Fas)–mediated mechanism and whether this cytotoxic activity can be modulated by cytokines such as interferon gamma (IFN-γ) or tumor necrosis factor alpha (TNF-α). We have found that normal woodchuck and human hepatocytes, both cultured and primary freshly isolated, as well as human HepG2 cells, intrinsically transcribe not only CD95 but also CD95L when examined by reverse transcription-polymerase chain reaction (RT-PCR) assays. The functional competence of CD95L, which was detectable in hepatocytes and HepG2 cells by Western blotting, was confirmed in bioassays by induction of apoptosis of CD95-bearing P815 and LS102.9 cell targets and validated by inhibition of the cell killing with CD95 antagonistic antibody or with a general caspase inhibitor. Furthermore, exposure of cultured hepatocytes to IFN-γ or their stable transfection with IFN-γ cDNA or TNF-α cDNA increased hepatocyte CD95L/CD95–mediated cell killing. In conclusion, hepatocytes express both CD95L and CD95 and they can induce death of other cells by a CD95L-dependent mechanism. IFN-γ and, to a lesser extent, TNF-α can enhance hepatocyte CD95L-mediated cytotoxicity. This suggests that the local cytokine environment may modulate the hepatocyte contribution to liver immunity. (HEPATOLOGY 2006;43:1231–1240.)

The removal of antigens and aberrant cells from circulation is a physiological function of the liver, making this organ an important contributor to the first line of host immune defense. After birth, the liver is the site of production of immune competent cells, including natural killer (NK), natural killer T (NKT), and extrathymic T cells,1 elimination of activated cytotoxic T cells (CTL),2–5 and synthesis of complement and acute phase proteins.6 The liver is considered an immunocompetent organ whose contributions toward induction of peripheral immunotolerance and surveillance against pathogens are increasingly recognized. The role of Kupffer cells and sinusoidal endothelial cells in clearance of antigens and cells has been well established.7, 8 However, whether hepatocytes can directly eliminate cells either passing through, trapped, or normally residing in hepatic parenchyma remains unknown. In this regard, the endowment of hepatocytes in asialoglycoprotein receptor not only may facilitate removal of desialylated glycoproteins from plasma9, 10 but also may provide a contact between hepatocyte and other cells through recognition of their surface proteins depleted of terminal sialic acid residues. This may initiate a cascade of events leading to elimination of the cells contacted by hepatocytes. In the current study, we examined whether hepatocytes are equipped in the molecular machinery to act as cytotoxic effectors and, if so, whether this activity can be modified by the local cytokine milieu.

The cytotoxic cells can induce cell death through interaction of their membrane-associated CD95 ligand (CD95L or FasL) with CD95 (Fas) receptor on cell targets11, 12 or by release of cytolytic proteins, perforin and granzyme B.13 CD95L-induced cell death is a main mechanism by which activated CTL eliminate targeted cells.14 In this regard, it has been shown that interferon gamma (IFN-γ) can enhance the CTL CD95L-mediated cell killing.14 The CD95L–CD95 interaction also is crucial for removal of autoreactive T cells during thymic maturation15 and in homeostatic protraction of T and B lymphocytes.16, 17 Lymphoid cells can display both CD95L and CD95 and act depending on circumstances either as cytotoxic effectors or as targets for the cytotoxic reactions. Furthermore, expression of CD95L is not restricted to activated T cells, NK, and NKT cells18, 19 but is also evident in cells at the immune privileged sites.20

Considering normal liver tissue, hepatocytes were found to be CD95L nonreactive by immunohistochemical methods.21, 22 However, the ligand expression was detected by immunostainings or in situ hybridization in hepatocytes in hepatocellular carcinoma,22 alcoholic hepatitis,23 liver allograft rejection,21 and Wilson's disease.24 This may suggest that CD95L is in fact expressed in normal hepatocytes but at levels not detectable by conventional methods, whereas hepatocytes in some pathological conditions carry the ligand at quantities more readily identifiable by the same approaches.

In contrast to CD95L, hepatocytes are CD95-bearing cells, and their apoptosis can be swiftly induced by anti-CD95 antibody25 or by binding of CD95L.26 Therefore, certain liver diseases or their distinctive morphological stages could be a consequence of interaction between a hepatocyte's CD95L and its cognate receptor in either a suicidal or fratricidal manner.21–24

In the current work, we investigated whether normal cultured hepatocytes, primary hepatocytes, and HepG2 cells manifest a phenotype of cytotoxic effector cells and, if so, whether they can kill other cells via a CD95L-activated pathway. In most experiments we used hepatocytes from healthy woodchucks known to be susceptible to woodchuck hepatitis virus. These animals represent an excellent natural model of hepatitis B virus infection.27, 28 We used woodchuck hepatocytes as cytotoxic effectors to facilitate future investigations on their potential involvements in the liver innate immunity and the development of different forms and outcomes of viral hepatitis. In addition, because of the critical role of intrahepatic IFN-γ and tumor necrosis factor alpha (TNF-α) in the recovery from or progression to chronic viral hepatitis,29, 30 the effect of these cytokines on hepatocyte-mediated cell killing was investigated.


NK, natural killer cells; NKT, natural killer T cells; CTL, cytotoxic T lymphocytes; IFN-γ, interferon gamma; TNF-α, tumor necrosis factor alpha; PBMC, peripheral blood mononuclear cells; rwIFN-γ, recombinant woodchuck IFN-γ; rwTNF-α, recombinant woodchuck TNF-α; RT-PCR, reverse transcription polymerase chain reaction; MHC, major histocompatibility complex; MAb, monoclonal antibody.

Materials and Methods

Cell Lines.

Woodchuck WCM-260 hepatocyte line was established from the liver of a healthy woodchuck, as reported.31, 32 The cells were maintained in Hepato-STIM culture medium (Becton Dickenson, Bedford, MA) and demonstrated stable growth, consistent morphology, and transcribed albumin and asialoglycoprotein receptor when passaged weekly.31, 32 Human HepG2 cells (ATCC HB-8065; American Type Culture Collection, Rockville, MD) and murine mastocytoma P815 (ATCC TIB-64) and lymphoma LS102.9 cells (ATCC HB-97), both constitutively bearing CD95,33, 34 were cultured as reported.31, 32, 34

Preparation of Hepatocytes and Lymphoid Cells.

Hepatocytes were isolated from livers of healthy woodchucks by two-step collagenase microperfusion.35 Preparations were at least 98% pure by phase-contrast microscopy. Peripheral blood mononuclear cells (PBMC) were prepared from the same animals and from a healthy human by Ficoll-Hypaque gradient centrifugation (Pharmacia Biotech, Quebec, Canada).32 Pure primary human hepatocytes (DPK-HCWP-H) isolated from a healthy fragment of the liver of a 58-year-old white man were purchased from Dominion Pharmakine (Derio-Bizkaia, Spain). Animal experimental protocols were approved by the Institutional Presidents' Committee on Animal Bioethics and Care.

Treatments of Hepatocytes With IFN-γ or TNF-α.

WCM-260 hepatocytes were treated with woodchuck IFN-γ (wIFN-γ) or TNF-α (wTNF-α), using two approaches. First, the cells were exposed to bioactive recombinant wIFN-γ or wTNF-α (rwIFN-γ and rwTNF-α, respectively) produced in the baculovirus expression system, as recently reported.36 After 18-hour exposure to 10-fold dilutions of the cytokines, cytotoxic activity of WCM-260 was measured (see below) and compared with naive hepatocytes or those exposed to culture supernatant from insect cells infected with wild-type baculovirus.36 In addition, mRNAs from WCM-260 hepatocytes treated with rwIFN-γ (150 U/mL) or rwTNF-α (35 U/mL) were analyzed (see below). Secondly, WCM-260 were stably transfected with the complete wIFN-γ or wTNF-α cDNA (GenBank accession number AF232728 for wIFN-γ37 and AF333967 for wTNF-α29) using pcDNA3.1 vector-based expression constructs (Invitrogen, Carlsbad, CA) and Lipofectamine 2000 (Invitrogen). Cells transfected with empty pcDNA3.1 were used as controls. Transcriptional activity of wIFN-γ and wTNF-α in WCM-260 hepatocytes was ascertained by reverse transcription polymerase chain reaction (RT-PCR) using primers and conditions reported.36 To verify biological activity of the cytokines expressed, WCM-260 hepatocytes exposed to or transfected with wIFN-γ or wTNF-α cDNA were examined for class I major histocompatiblity complex (MHC) antigen display by flow cytometry.36

RNA Extraction and Reverse Transcription.

Total RNA was extracted from 1 × 106 cultured or primary woodchuck hepatocytes, HepG2 cells, woodchuck or human PBMC, from 2 × 105 primary human hepatocytes, and from 100 mg normal woodchuck spleen using Trizol reagent (Invitrogen). RNA (2 μg) was reverse transcribed to cDNA in a 20-μL reaction volume with 200 U Moloney murine leukemia virus reverse transcriptase (Invitrogen), as described previously.38

Cloning of Woodchuck Gene Sequences.

For cloning of woodchuck CD95L and CD95, cDNA transcribed from woodchuck spleen or PBMC RNA was amplified by PCR using degenerate primers deduced through interspecies comparison of sequences available in GenBank. PCR amplicons were cloned into the dual promoter vector PCRII using the TOPO TA cloning system (Invitrogen) and sequenced. Woodchuck-specific primers spanning introns were designed, the gene fragments amplified, cloned, their sequence determined, and submitted to GenBank (accession numbers AF152368 and AY993960 for woodchuck CD95L and CD95, respectively).

RT-PCR and Southern Blot Hybridization.

Transcription of CD95L and CD95 genes in test cDNA samples was assessed by PCR. Thus, the CD95L 508-bp fragment was amplified with sense primer 5′-CAGCTCTTCCACCTGCAGAAGG and antisense primer 5′-AGATTCCTCAAAATTGATCAGAGAGAG during 32 cycles, each cycle at 95°C for 15 seconds, 55°C for 30 seconds, and 72°C for 60 seconds. Detection of the CD95 500-bp fragment was facilitated with sense primer 5′-GATGGAGGGCATGGTTTAGAAGTG and antisense primer 5′-AGCAGCTGGAGTTTCTGCTCAGC during 34 cycles, each cycle at 95°C for 15 seconds, 60°C for 30 seconds, and 72°C for 60 seconds. Woodchuck β-actin was amplified as a loading control.29 Specificity of the PCR products was routinely confirmed by Southern blotting with appropriate probes and autoradiography.29, 39

Real-time RT-PCR.

To quantify expression of CD95L, CD95, and β-actin, real-time RT-PCR was established using the Lightcycler Faststart Master SYBR I kit (Roche Diagnostics, Laval, Quebec, Canada) and the Roche LightCycler (Roche Diagnostics). Reactions were performed in 20-μL volumes, each containing 2 μL cDNA derived from 50 ng RNA using the following primer pairs: sense primer 5′-CCATTTAACAGGTAAGCCC and antisense primer 5′-TCATCATCTTGCCCTCC for CD95L, sense primer 5′-GTGCACCACGTGTGAACATGGAAT and antisense primer 5′-TAATCGGGAGTAGCAGTAGCAGGA for CD95, and primers 5′-CAACCGTGAGAAGATGACC and 5′-ATCTCCTGCTCGAAGTCC for β-actin.

CD95L Detection by Western Blotting.

Cell suspensions were treated with ice-cold RIPA buffer (1% NP-40, 0.5% DOC, 0.1% SDS, 150 mmol/L NaCl in 50 mmol/L Tris, pH 8.0) and cellular debris removed by centrifugation. Lysates were separated by SDS-PAGE at 10 to 20 μg protein/lane and then blotted onto a nitrocellulose membrane (Amersham Biosciences, Piscataway, NJ) by semi-dry transfer using the Bio-Rad SD cell system (Bio-Rad, Mississauga, Ontario, Canada).37 As positive controls, freshly isolated human and woodchuck PBMC were examined at 20 μg protein/lane. The blots were blocked with 5% skim milk in Tris-buffered saline, pH 7.4, overnight at 4°C. Both transmembrane (38 kd) and soluble (28 kd) isoforms of CD95L were detected using rabbit anti-murine CD95L IgG (Santa Cruz Biotech, Santa Cruz, CA) followed by horseradish peroxidase–conjugated goat anti-rabbit IgG F(ab ′)2 antibodies (Jackson ImmunoResearch, West Grove, PA). The signals were visualized using an ECL detection kit (Sigma, Oakville, Ontario, Canada).

Preparation of Target Cells for Cytotoxicity Assays.

P815 and LS102.9 cells were subcultured to a density of 5 × 105 cells/mL and labeled with 50 μCi 3H-adenine (Perkin Elmer, Wellesley, MA) for 18 hours before the JAM cytotoxicity assay. The cells were washed 3 times in HBSS (Invitrogen), resuspended in RPMI medium34 at 1 × 105 cells/mL, and immediately used.

JAM Cytotoxicity (DNA Fragmentation) Assay.

WCM-260 hepatocytes, HepG2 cells, or WCM–260 transfected with wIFN-γ or wTNF-α cDNA were grown to confluence (∼6 × 104 cells/well) in 96-well flat-bottom cell culture plates. In some experiments, WCM-260 were treated with rwIFN-γ or rwTNF-α at ∼95% confluence for 18 hours before the assay. In the case of primary hepatocytes, the cells were aliquotted at 6 × 104/well in 200-μL volumes. 3H-adenine-labeled P815 or LS102.9 cells were added at 20 × 103, 10 × 103, 5 × 103 and, occasionally, at 40 × 103 cells in quadruplicate to experimental wells yielding final effector:target (E:T) ratios of 6:1, 3:1, 1.5:1 and 12:1, respectively. Plates were centrifuged for 5 minutes at 45g and incubated at 37 °C in 5% CO2 for 18 hours. Well contents were harvested onto glass fibre mats (Perkin Elmer, Wellesley, MA) using a 96-well harvester (Tomtec, Hamden, CT). Counts per minute (cpm) were measured using a Top10 beta counter (Becton Dickinson, San Diego, CA) and percent lysis was determined by applying the formula: percent specific lysis = (control cpm − experimental cpm)/control cpm × 100, where control cpm for each target cell type were obtained in the absence of effector cells. Freshly isolated woodchuck PBMC were used in parallel experiments at E:T ratios of 50:1, 25:1 and 12.5:1, as described before.34 Specificity of CD95L-mediated lysis of P815 cells was confirmed using an antagonistic anti-CD95 monoclonal antibody (MAb) (Jo2; Becton Dickinson), as described.34 In the case of LS102.9 cells, caspase-dependent initiation of apoptosis was blocked by pre-treatment of the cells with the pan-caspase inhibitor z-VAD-fmk (Biomol, Plymouth, PA) at 150 μmol/L for 4 hours.

Statistical Analyses.

Results were analyzed by unpaired Student t test with Welch's correction using GraphPad Prism software (GraphPad Software Inc., San Diego, CA). Differences between experimental conditions were considered to be significant when two-sided P values were less than .05.


Hepatocytes Constitutively Transcribe CD95L.

Initial experiments were performed using a non-quantitative RT-PCR to determine whether hepatocytes express genes encoding for the CD95L and its receptor. As shown in Fig. 1A, CD95L mRNA was without difficulty detected in human and woodchuck lymphoid cells, as well as in HepG2 cells and in woodchuck WCM-260 hepatocytes. Importantly, pure preparations of woodchuck and human hepatocytes also displayed CD95L mRNA. In addition, transcription of CD95 mRNA was evident in woodchuck PBMC and hepatocytes, although CD95 mRNA was not detected in human PBMC and HepG2 cells (Fig. 1A) or primary hepatocytes (data not shown) due to the woodchuck-restricted specificity of the CD95 PCR primers used.

Figure 1.

Identification of CD95L mRNA and protein in hepatocytes. (A) Expression of CD95L and CD95 mRNA in human HepG2 cells, woodchuck WCM-260 hepatocytes, and in primary hepatocytes isolated from three healthy woodchucks (1-3) and a healthy fragment of human liver. Signals were detected by RT-PCR using 200 to 50 ng total RNA per reaction and visualized by Southern blot hybridization with appropriate 32P-labeled woodchuck gene probes. Woodchuck β-actin served as a housekeeping gene transcription control. RNAs from PBMC of a healthy human (hPBMC) and a healthy woodchuck (wPBMC), and plasmid DNAs carrying the appropriate woodchuck gene fragments were used as positive controls. Water instead of cDNA served as a negative PCR control. (B) Real-time RT-PCR quantification of CD95L and CD95 gene transcription in the livers, hepatocytes isolated from these livers and PBMC from three healthy woodchucks shown in A. Fifty nanograms total RNA was used for each RT-PCR, and each evaluation was done in triplicate. Gene transcription was normalized against woodchuck β-actin expression. Data represent mean copy numbers ± SEM. (C) CD95L protein detection by Western blotting in human and woodchuck PBMC, HepG2, and WCM260 cells, and in primary hepatocytes from six woodchucks (1-6). Proteins from PBMC and liver cell lines were loaded at 20 μg/lane and from primary hepatocytes at 10 μg/lane. The positions 38-kd and 28-kd isoforms of CD95L are indicated by arrows. Numbers below the panel represent relative densitometric values (DU) of the protein bands identified. PBMC, peripheral blood mononuclear cells.

When expression of CD95L was quantified by real-time RT-PCR in samples of whole liver tissue, in hepatocytes isolated from these livers, and in autologous PBMC, no statistically significant differences in the mean levels of CD95L mRNA were found (Fig. 1B). Similarly, the mean CD95 mRNA levels were not statistically different between the liver, primary hepatocytes, and peripheral lymphoid cells (Fig. 1B). A possibility that primary hepatocytes might be contaminated with lymphoid cells was excluded based on the absence of CD3 cDNA signals (data not shown), as analyzed by RT-PCR using conditions previously reported.29

Hepatocytes Display CD95L Protein.

To determine whether CD95L mRNA transcription was accompanied by CD95L protein, hepatocyte lysates were probed with anti-CD95L antibody by Western blotting. Fig. 1C shows that WCM-260 hepatocytes and HepG2 cells, as well as hepatocytes isolated from six different woodchucks, showed the 38-kd membrane-bound isoform of CD95L, which was also evident in control woodchuck and human PBMC. In addition, WCM-260 exhibited the 28-kd isoform representing pre-formed soluble CD95L, which were also detected in woodchuck PBMC (Fig. 1C). Densitometry analysis of the 38-kd bands suggested that the protein occurred in WCM-260 and HepG2 cells at approximately ninefold and twofold greater levels than in woodchuck and human PBMC, respectively. Primary hepatocytes contained approximately twice the protein of PBMC isolated from the same woodchucks when adjusted to the same total protein concentration (Fig. 1C and data not shown).

Hepatocyte CD95L-Mediated Cytotoxicity.

In preliminary experiments, it was established using woodchuck PBMC as effectors that the JAM DNA fragmentation assay can measure cytotoxic activity of woodchuck cells and that its sensitivity is comparable to or greater than that of the 51Cr-release assay, as was also observed by others.40 The ability of woodchuck PBMC to kill P815 cells was previously documented using the 51Cr-release cytotoxicity assay.34 Accordingly, the JAM assay was applied to examine hepatocyte cytotoxicity by using constant numbers of hepatocytes as effectors and increasing numbers of P815 or LS102.9 cells as targets. As shown in Fig. 2A-B, WCM-260 hepatocytes killed both cell types in a manner dependent on the target numbers. The killing was significantly reduced (P < .05 or P < .001) after pretreatment of P815 cells with an antagonistic CD95-blocking antibody, Jo2 (Fig. 2A), or when LS102.9 cell death was measured in the presence of z-VAD-fmk (Fig. 2B). HepG2 cells also killed P815 and LS102.9 targets, although a statistically significant inhibition by Jo2 MAb or z-VAD-fmk was usually seen at lower numbers of targets (Figs. 2C-D). Primary hepatocytes isolated from healthy woodchucks (Fig. 2E) and a healthy fragment of human liver (Fig. 2F) also eliminated CD95-sensitive P815 cells. Inhibition of the CD95 ligand–receptor interaction by Jo2 MAb markedly (P = .07) or significantly (P < .005 or P < .05) reduced the level of P815 cell killing.

Figure 2.

Hepatocyte killing of P815 and L102.9 cell targets mediated by CD95L. WCM-260 hepatocytes (A and B), HepG2 cells (C and D), and primary woodchuck (E) or human (F) hepatocytes were used as effectors against CD95-bearing P815 (A, C, E and F) or LS102.9 (B and D) cells in the JAM DNA fragmentation assay. Jo2 MAb was applied to block CD95L-dependent killing of P815 cells (A, C, E, and F), whereas lysis of LS102.9 cells was inhibited with pancaspase z-VAD-fmk inhibitor (B and D). Data bars shown in A to D are mean values ± SEM from three separate experiments with three to six experimental wells per condition tested. For E, hepatocytes isolated from three healthy woodchucks were used as effector cells, whereas F demonstrates target cell killing by primary human hepatocytes. Results are shown as mean values ± SEM, with each experiment performed with four to eight experimental wells per each E:T ratio. Data bars marked with ** are significant at P < .005 and with * at P < .05 when compared with the cells not treated with Jo2 MAb or z-VAD-fmk inhibitor. P value for data bars marked with + is .07. MAb, monoclonal antibody.

IFN-γ and TNF-α Upregulate CD95L Transcription in Hepatocytes.

In the first step, to learn whether exposure to IFN-γ or TNF-α would induce a measurable biological effect in hepatocytes, WCM-260 cells were examined for expression of class I MHC after treatment with rwIFN-γ (150 U/mL) or rwTNF-α (35 U/mL) or after transfection with the cytokine cDNA. Thus, class I MHC display was increased by 14.2-fold in hepatocytes exposed to rwIFN-γ (Fig. 3A) and by 12.9.-fold in those transfected with wIFN-γ cDNA (Fig. 3B). Exogenous rwTNF-α increased the expression by 2.4-fold (Fig. 3A), whereas transfection with wTNF-α cDNA induced a 2.8-fold increase (Fig. 3B). Therefore, both treatment approaches with IFN-γ or TNF-α enhanced expression of a functionally important molecule on hepatocytes.

Figure 3.

Evaluation of the effects of IFN-γ and TNF-α on class I MHC antigen display and CD95L gene expression in hepatocytes. The class I MHC heavy chain detection by flow cytometry in WCM-260 hepatocytes (A) exposed to or (B) stably transfected with wIFN-γ or wTNF-α and in control cells. For A, the cells were exposed to 150 U/mL rwIFN-γ, 35 U/mL rwTNF-α or to a supernatant from insect cells infected with wild-type baculoviral vector (wt super) for 18 hours before staining with B1b.B9 MAb. Intact WCM-260 cells not exposed to any of the above and stained with B1b.B9 MAb were used as an additional control. For B, WCM-260 hepatocytes were transfected with wIFN-γ (pcDNA3.1wIFN-γ), wTNF-α (pcDNA3.1wTNF-α) or with empty pcDNA3.1 vector and probed with B1b.B9 MAb. Non-transfected WCM-260 were used as an additional control. For C and D, CD95L mRNA was quantified by real-time RT-PCR in WCM-260 hepatocytes exposed to rwIFN-γ or rwTNF-α or transfected with these cytokine genes, respectively. Data are represented as mean percentage expression values ± SEM from three separate experiments after normalization to woodchuck β-actin and by taking expression of a given gene in unstimulated hepatocytes (C) or in hepatocytes transfected with empty pcDNA3.1 vector (D) as 100%. Data bars marked with ** are significant at P < .001 relative to respective controls. IFN-γ, interferon gamma; TNF-α, tumor necrosis factor alpha; wIFN-γ, woodchuck interferon gamma; wTNF-α, woodchuck tumor necrosis factor alpha.

The effect of rwIFN-γ or rwTNF-α on CD95L mRNA in hepatocytes was quantified by real-time RT-PCR. CD95L mRNA was upregulated (P < .001) by rwIFN-γ, but not by rwTNF-α (Fig. 3C). CD95L mRNA was also meaningfully (P < .001) elevated in hepatocytes transfected with wIFN-γ or wTNF-α cDNA, when compared with controls transfected with empty vector (Fig. 3D). These data clearly showed that IFN-γ is a potent inducer of CD95L transcription in hepatocytes and suggested that TNF-α may exert a similar effect under certain conditions.

IFN-γ But Not TNF-α Enhances Hepatocyte CD95L-Mediated Cytotoxicity.

As shown in Fig. 4, exposure of WCM-260 to increasing concentrations of rwIFN-γ enhanced hepatocyte cytotoxicity toward P815 and LS102.9 targets. Thus, rwIFN-γ significantly (P < .001) increased hepatocyte killing of P815 (Fig. 4A) and LS102.9 (Fig. 4B) under most of the conditions tested, whereas rwTNF-α had no measurable effect (data not shown). An MTT assay showed that neither rwIFN-γ nor rwTNF-α affected the viability of hepatocytes, P815 or LS102.9 cells at the concentrations used (data not shown), excluding the possibility that the cytokines may affect cell survival during the assay. Furthermore, P815 cell lysis was inhibited by pretreatment with Jo2 MAb (Fig. 4A), which was consistent with previous observations (see Fig. 2).

Figure 4.

Exposure to IFN-γ enhances hepatocyte CD95L-mediated cell killing. WCM-260 hepatocytes were stimulated with 15, 150, or 1500 U/mL rwIFN-γ and used as effectors at 6 × 104 cells/assay against indicated numbers of P815 (A) or LS102.9 (B) targets and their cytotoxicity measured by JAM assay. Data represent mean values ± SEM obtained from two separate experiments, each performed in duplicate. Data bars marked with ** are significant at P < .01, and those with * at P < .05 when compared with cells not exposed to the cytokine (unstimulated). Preincubation with Jo2 MAb reduced killing of P815 cells. Data bars marked with + are significant at P < .05 when compared with target cells not treated with the antibody. IFN-γ, interferon gamma; rwIFN-γ, recombinant woodchuck interferon gamma.

Transfection With IFN-γ or TNF-α cDNA Enhances Hepatocyte CD95L-Mediated Cell Killing.

As illustrated in Fig. 5, transfection of WCM-260 hepatocytes with cDNA of either wIFN-γ or wTNF-α enhanced CD95L-mediated killing of P815 and LS102.9 targets. Thus, transfection with wIFN-γ cDNA significantly (P < .01) increased elimination of P815 cells, which was consistent with the CD95L mRNA upregulation illustrated in Fig. 3C. This killing was blocked by 59% to 85% (P < .05) after pretreatment of P815 cells with Jo2 MAb (Fig. 5A). Elimination of LS102.9 targets was also significantly enhanced (P < .05 or P < .01) when compared with control cells transfected with the empty pcDNA3.1 vector. Also, transfection of WCM-260 hepatocytes with wTNF-α cDNA increased (P < .05 or P < .01) their cytotoxic activity toward P815 and LS102.9 targets (Fig. 5A and Fig. 5B, respectively), which was consistent with increased CD95L mRNA levels seen in Fig. 3D.

Figure 5.

Transfection with IFN-γ or TNF-α cDNA up-regulates hepatocyte CD95L-mediated cytotoxicity. WCM-260 hepatocytes stably transfected with wIFN-γ (pcDNA3.1wIFN-γ) or wTNF-α (pcDNA3.1wTNF-α) were used as effectors at 6 × 104 cells/assay against indicated numbers of P815 (A) or LS102.9 (B) target cells in the JAM assay. Preincubation of P815 cells with Jo2 MAb significantly decreased their killing by hepatocytes. Data bars marked with ** are significant at P < .01 and with * at P < .05 when compared with WCM-260 transfected with empty cDNA3.1 vector (pcDNA3.1). Data bars marked with + are significant at P < .05 when compared with P815 cells not treated with Jo2 MAb. IFN-γ, interferon gamma; TNF-α, tumor necrosis factor alpha; wIFN-γ, woodchuck interferon gamma; wTNF-α, woodchuck tumor necrosis factor alpha; MAb, monoclonal antibody.


Hepatocytes in normal livers have been considered to be CD95L nonreactive, and their ability to induce cell death had not been investigated. In this study, we documented by standard RT-PCR that both cultured hepatocytes and primary hepatocytes from livers not compromised by a disease process constitutively transcribe genes encoding molecules of the CD95L/CD95 pathway. To assess whether CD95L was displayed at functionally adequate levels in hepatocytes, previously established cytotoxic assays with heterologous cells as targets were adopted. The results showed that hepatocytes not only transcribe CD95L mRNA and synthesize the ligand, but can efficiently eliminate other cells via a CD95L-mediated mechanism. Furthermore, the CD95L-dependent cell killing was enhanced when hepatocytes were exposed to IFN-γ or transfected with IFN-γ or TNF-α cDNA, suggesting that hepatocyte cytotoxicity in vivo can be modulated by these cytokines.

Expression of the genes encoding effector molecules of the CD95L/CD95 pathway was until now an attribute of activated T cells, NK, and NKT cells.18, 19 As shown in Fig. 1, these genes are also transcribed in circulating lymphoid cells. Interestingly, the analysis showed that the levels of CD95L and CD95 transcription in hepatocytes and lymphoid cells were not meaningfully different. However, both cultured and freshly isolated hepatocytes displayed greater amounts of CD95L protein than lymphoid cells, suggesting that CD95L cDNA could be translated more efficiently in hepatocytes.

Previous attempts to identify CD95L protein in normal hepatocytes were not successful; however, the ligand has been detected in hepatocytes in some liver disorders and during allograph rejection21–24 and in HepG2 cells.24 In our study, a membrane-bound CD95L 38-kd protein was evident in woodchuck-cultured and primary hepatocytes, and in HepG2 cells. In addition, a 28-kd protein representing the soluble form of the ligand was seen in woodchuck WCM-260 hepatocytes and PBMC. Our success in identifying CD95L could be due to immediate isolation and preservation of RNA and proteins and using detection assays of superior sensitivity, but unlikely due to a species difference because both woodchuck and human primary hepatocytes carry seemingly comparable amounts of CD95L transcripts and eliminated test target cells to a similar extent.

Hepatocyte ability to eliminate cell targets via the CD95L/CD95 pathway was examined by the JAM cytotoxicity assay.40, 41 The validity of the finding that primary woodchuck and human hepatocytes, as well as WCM-260 and HepG2 cells, eliminated CD95-bearing cells was ascertained using P815 pre-incubated with Jo2 MAb. This antibody is uniquely noncytolytic to P815 cells, most likely due to a defect in the cell CD95-activated signal transduction, but blocks recognition of cell surface CD95 by CD95L.34, 42 Additional evidence was provided using LS102.9 cells treated with z-VAD-fmk, an inhibitor of caspase-dependent initiation of apoptosis.43 Thus, utilization of two different cell targets and two inhibitors blocking either CD95 or the intracellular cascade initiated by the CD95 ligand–receptor interaction clearly demonstrated the ability of hepatocytes to induce cell death via the CD95L-CD95 mechanism.

Constitutive expression of both CD95L and CD95 in hepatocytes could be considered a paradox; however, this finding follows the blueprint seen in different lymphoid cell subtypes in which both molecules are displayed.44 Although at this stage we cannot dissect whether a single hepatocyte expresses both of the molecules simultaneously, the display of functional CD95L depends, as in lymphoid cells, on cell activation.14, 44 Because CD95L protein in lymphoid cells is sequestered intracellularly and directed into secretory vesicles,45 this may prevent or postpone the ligand exposure on the effector surface and initiation of cell death via CD95L–CD95 interaction. The same could be true for hepatocytes.

Because of our interest in the mechanisms of liver injury in viral hepatitis and the fact that early and strong intrahepatic IFN-γ and TNF-α response is paramount to the resolution of hepadnaviral infection,29, 30 we wanted to determine whether IFN-γ and TNF-α may influence hepatocyte CD95L-mediated cytotoxicity. As we uncovered, wIFN-γ and, under certain conditions, wTNF-α, augment hepatocyte CD95L-dependent cell killing. Because IFN-γ has been known to enhance CD95L expression in both lymphoid and nonlymphoid cells,14, 46 the increases in hepatocyte CD95L mRNA and CD95L-dependent cell killing on exposure to wIFN-γ were consistent with the predicted outcomes. In addition, these data provided an independent confirmation that CD95L synthesized by hepatocytes is functionally competent and its expression can be regulated by extracellular factors. Furthermore, CD95L expression also can be enhanced by TNF-α through activation of NF-κB,47 which subsequently binds to the CD95L promoter.48 Although CD95L mRNA upregulation and an increase in CD95L-dependent cell killing were observed after transfection of hepatocytes with TNF-α cDNA, exogenous TNF-α did not produce such an effect. The reason behind this is not clear. The data on class I MHC expression suggested that TNF-α was recognized and activated downstream signaling in hepatocytes under both treatment conditions. Nonetheless, considering the results as a whole, IFN-γ was consistently more potent than TNF-α in enhancing hepatocyte CD95L-mediated cell killing.

The identification in this study of the constitutive expression of functionally competent CD95L in hepatocytes and demonstration that IFN-γ and, to some degree, TNF-α can modulate hepatocyte CD95L-dependent cell killing provide new insights into the intrinsic properties of liver parenchyma. Our results imply that hepatocytes are not just passive objects of actions exerted by other cells, but they can be active players determining fate of the cells brought in contact with their surface. Their cytotoxic activity may contribute to physiological and post-hepatectomic regeneration of liver parenchyma,49 removal of aberrant hepatocytes,23, 24 and elimination of activated T lymphocytes,3, 7 including those specific for viral pathogens. Hepatocyte cytotoxic activity, which is not readily apparent in normal liver, but can be heightened during inflammatory processes, may contribute to the pathogenesis of different forms of a given liver disease, including viral hepatitis, which is normally accompanied by up-regulated intrahepatic expression of IFN-γ and TNF-α.29 These issues will require further examination. Overall, the results of our study imply that investigations on immunological processes engaging the liver need to consider hepatocytes as active contributors.


The authors thank Norma D. Churchill and Colleen L. Trelegan for expert technical assistance, and Dr. Michael D. Grant from Faculty of Medicine, Memorial University, St. John's, Canada for valuable discussions and for providing LS102.9 cells.