Mechanism of action of the antifibrogenic compound gliotoxin in rat liver cells



Gliotoxin has been shown to promote a reversal of liver fibrosis in an animal model of the disease although its mechanism of action in the liver is poorly defined. The effects of gliotoxin on activated hepatic stellate cells (HSCs) and hepatocytes have therefore been examined. Addition of gliotoxin (1.5 μM) to culture-activated HSCs resulted in its rapid accumulation, resulting in increased levels of glutathione and apoptosis without any evidence of oxidative stress. In contrast, although hepatocytes also rapidly sequestered gliotoxin, cell death only occurred at high (50-μM) concentrations of gliotoxin and by necrosis. At high concentrations, gliotoxin was metabolized by hepatocytes to a reduced (dithiol) metabolite and glutathione was rapidly oxidized. Fluorescent dye loading experiments showed that gliotoxin caused oxidative stress in hepatocytes. Antioxidants—but not thiol redox active compounds—inhibited both oxidative stress and necrosis in hepatocytes. In contrast, HSC apoptosis was not affected by antioxidants but was potently abrogated by thiol redox active compounds. The adenine nucleotide transporter (ANT) is implicated in mitochondrial-dependent apoptosis. HSCs expressed predominantly nonliver ANT isoform 1, and gliotoxin treatment resulted in a thiol redox-dependent alteration in ANT mobility in HSC extracts, but not hepatocyte extracts. In conclusion, these data suggest that gliotoxin stimulates the apoptosis of HSCs through a specific thiol redox-dependent interaction with the ANT. Further understanding of this mechanism of cell death will aid in finding therapeutics that specifically stimulate HSC apoptosis in the liver, a promising approach to antifibrotic therapy. Supplementary material for this article can be found on the HEPATOLOGY website ( (HEPATOLOGY 2004;40:232–242.)

A common outcome in patients with chronic liver damage is the development of scarring fibrosis, a wound-healing response that is primarily associated with the production of extracellular matrix protein by hepatic stellate cells (HSCs).1, 2 HSCs in normal liver are in relative low abundance, reside in the space of Disse, and function to store vitamin A in the form of fatty acid esters.3 Chronic liver damage results in HSC transdifferentiation from the “quiescent” vitamin A–storing to an “activated” myofibroblast-like (α-smooth muscle actin-expressing) phenotype.1, 2 HSC activation results in their proliferation and up-regulation in the expression of genes contributing to matrix accumulation such as collagen type I and tissue inhibitors of metalloproteinases.4

Increasing evidence suggests that rats recover from liver fibrosis through up-regulation of apoptosis of activated HSC.5–8 The epipolythiodioxopiperazine fungal metabolite gliotoxin kills rat and human HSCs by stimulating their apoptosis; administration of gliotoxin to rats (treated with either carbon tetrachloride or thioacetamide to cause liver fibrosis) significantly reduced the number of activated HSCs in the liver and the extent of scarring fibrosis.5, 9 An up-regulation of the apoptosis of activated HSC may be an effective therapeutic approach to treat liver fibrosis.

Nuclear factor κB (NF-κB) is a transcription factor that regulates cell proliferation and apoptosis in addition to immune and inflammatory responses.10 In most cells, including “unstimulated” hepatocytes, NF-κB is inactive (under normal conditions it is bound to inhibitory I-κB proteins that mask a nuclear localization signal) and retained in the cytosol.10, 11 Cytokines such as tumor necrosis factor α (TNF-α)—which are known to play a role in liver fibrogenesis1, 2—activate NF-κB in hepatocytes by stimulating a degradation of the I-κB protein.10 Through the same receptor, TNF-α also activates an apoptotic pathway that involves the opening of a pore in mitochondria (termed the mitochondrial permeability transition [MPT]) resulting in cytochrome c release, activation of caspases and apoptosis.12 The parallel activation of NF-κB by TNF-α appears to prevent hepatocyte apoptosis through induction of antiapoptotic gene expression because an inhibition of NF-κB-dependent gene expression (using either a general inhibitor of gene expression or by specifically inhibiting NF-κB activation) results in hepatocyte apoptosis.12–15 Accordingly, in vivo, TNF-α stimulates hepatocyte proliferation/liver regeneration16—an essential response to liver insult. Gliotoxin, as an inhibitor of NF-κB activation through its inhibition of I-κB degradation, therefore diverts hepatocytes to undergo apoptosis when also treated with TNF-α.13 In culture-activated (i.e., α-smooth muscle actin-expressing) HSCs, NF-κB is constitutively active, and gliotoxin stimulates the apoptosis of HSCs without the need for the addition of cytokines such as TNF-α.5, 17 However, we have shown that gliotoxin only inhibits TNF-α–inducible NF-κB DNA binding activity in activated HSCs and not constitutively-activated NF-κB.5 Indeed, other inhibitors of NF-κB activation, such as calpain inhibitor I, did not stimulate activated HSC apoptosis.5

The hypothesis underpinning these experiments is that NF-κB is not involved in gliotoxin-dependent activated HSC killing (in the absence of TNF-α), and that there may be an NF-κB-independent pathway of gliotoxin-stimulated HSC apoptosis in addition to an NF-κB–dependent pathway. One potential target for gliotoxin may be the adenine nucleotide transporter (ANT), a protein located on the inner mitochondrial membrane that functions to carry adenosine diphosphate and adenosine triphosphate across the membrane. In combination with cyclophilin D, the voltage-depedent anion channel, and several kinases, the ANT regulates inner membrane permeability to low molecular weight compounds referred to as the MPT.18 The MPT is an important determinant in cell death that, under certain conditions, results in cytochrome c release from the mitochondrial membrane space, caspase activation, and apoptosis.18

A detailed examination of the effects of gliotoxin on liver cells (hepatocytes and HSCs) has been examined, because gliotoxin or a related compound could prove useful in the treatment of liver fibrosis.19


HSC, hepatic stellate cell; ANT, adenine nucleotide transporter; NF-κB, nuclear factor κB; TNF-α, tumor necrosis factor α; MPT, mitochondrial permeability transition; GSH, reduced glutathione; GSSG, oxidized glutathione; HPLC, high performance liquid chromatography; H2DCFDA, 2′,7′-dichlorodihydrofluorescein diacetate; ROS, reactive oxygen species; DTT, dithiothreitol; PDTC, pyrrolidine dithiocarbamate; SOD, superoxide dismutase.

Materials and Methods

Animals, Cell Isolation, and Culture.

Male Sprague-Dawley rats (250–300 g) were bred in house at the Biological Services Unit, Foresterhill, University of Aberdeen, UK. Rat hepatocytes were prepared by collagenase perfusion essentially as described previously.20 Hepatocytes were cultured in William's Medium E supplemented with 1 μg/mL of bovine insulin, 10% fetal calf serum, 80 U/mL of penicillin, and 80 μg/mL of streptomycin on collagen type I–coated plates (BD Biosciences, Oxford, UK). Rat HSCs were isolated through pronase/collagenase perfusion of the liver followed by density gradient centrifugation as described previously.5 HSCs were cultured in Dulbecco's Modified Eagle Medium containing 4.5 g/L of glucose, supplemented with 16% (v/v) fetal calf serum, 80 U/mL of penicillin, 80 μg/mL of streptomycin, and 32 μg/mL of gentamicin. Cells were seeded onto plastic culture dishes, under which conditions the cells became activated as evidenced by the expression of α-smooth muscle actin.1, 2 HSCs were used within the first two passages.

All cells were incubated at 37°C in a humidified incubator gassed with 5% CO2 in air. Hepatocytes were maintained in William's Medium E optimally, whereas HSCs were widely cultured in Dulbecco's Modified Eagle Medium. Experiments were therefore performed with cells in their respective optimal media. The effects of gliotoxin were not significantly affected by the type of culture medium employed for the two cell types in that the effects of gliotoxin on each cell type was also observed if incubations were performed in the alternative medium (data not shown).

Assessments of Cell Viability and Apoptosis.

Cell attachment was assessed by visual examination of the cells. The ability of cells to exclude the dye trypan blue was examined by the addition of trypan blue to give a final concentration of 0.05% (w/v).21 Caspase 3 (DEVDase) activity was determined using a colorimetric CaspACE kit (Promega, Southampton, UK) according to the manufacturer's instructions. Total protein was determined by the method of Bradford.22 4′,6′-Diamidino-2-phenylindole was used to stain DNA in formalin-fixed cells for the analysis of nuclear changes in response to various treatments.

Determination of Cellular Glutathione Levels.

Cellular reduced glutathione (GSH) and oxidized glutathione (GSSG) levels were determined essentially as previously described.23 The concentrations of GSH and GSSG were determined via peak integration using authentic standards.

Metabolism of Gliotoxin.

The medium was removed and cells were washed with HEPES/HBSS buffer (0.14 M NaCl, 5.4 mM KCl, 0.34 mM Na2HPO4, 0.44 mM KH2PO4, 5.6 mM glucose, 1 mM CaCl2, 6 mM HEPES, 4 mM NaHCO3 pH 7.4) and incubated with gliotoxin in 1.5 mL of HEPES/HBSS buffer per well of a six-well plate. The levels of gliotoxin were determined using high performance liquid chromatography (HPLC; Sollentuna, Sweden) essentially as outlined previously24 through direct injection of medium samples onto a Gilson HPLC machine fitted with a 10-cm Hichrom C18 reverse phase column (Theale, UK) (internal diameter: 4.6 mm).

2′,7′-Dichlorodihydrofluorescein Diacetate Detection of Reactive Oxygen Species.

Hepatocytes and HSCs were loaded with 2′,7′-dichlorodihydrofluorescein diacetate (H2DCFDA) and viewed under a confocal microscope to visualize reactive oxygen species (ROS) using a variation of a method described previously.25 Cells in a six-well plate were dosed with 10 μM H2DCFDA and incubated at 37°C for 20 minutes. Medium was removed and replaced with 3 mL of HEPES/HBSS buffer to which the required treatment was added. Cells were incubated for a further 20 minutes at 37°C. Cells were then examined using a confocal microscope using a ×20 objective at an excitation wavelength of 488 nm. Emission between 530–560 nm was detected and analyzed using Biorad Lasersharp software (Hemel Hempstead, UK).

Western blotting was performed essentially as outlined previously,26 except that in some cases (indicated in legends) the reducing agent (dithiothreitol) was omitted during SDS detergent denaturation of samples. A rabbit polyclonal antiserum to rat ANT was prepared as previously described.27 Detection was achieved using an ECL kit (Amersham, UK).


Student's t test (two-tailed) was used to test for replicate statistical significance from one experiment and was typical of the number of separate experiments reported in the figure legends. Results were consistently observed to be statistically significant in all other experiments.


The Toxicity of Gliotoxin in Liver Cells In Vitro.

Hepatocytes and HSCs were prepared from rat liver and cultured separately as outlined in Materials and Methods. Fig. 1A demonstrates that concentrations in excess of 10 μM of gliotoxin resulted in hepatocyte detachment (typically a concentration of between 30 and 50 μM of gliotoxin was required for extensive detachment of hepatocytes), whereas 1.5 μM of gliotoxin (a concentration without any apparent effect in hepatocytes) resulted in almost complete detachment of HSCs within 3 hours. Both hepatocytes and HSCs were equally sensitive to chlorpromazine (a Ca2+/calmodulin antagonist that causes mitochondrial dysfunction at concentrations above 150 μM),28 whereas microcystin—an agent known to cause hepatocyte apoptosis29, 30—specifically stimulated the detachment of hepatocytes (see Fig. 1).

Figure 1.

Effect of gliotoxin, microcystin, and chlorpromazine on the viability of (A) hepatocytes and (B) activated HSCs—substratum attachment. Hepatocytes and culture-activated HSCs were prepared as outlined in Materials and Methods. Cuture medium was changed and cells were treated with either 1.5 μM of gliotoxin (○), 10 μM gliotoxin (▵), 50 μM of gliotoxin (□), 200 μM of chlorpromazine (•), or 100 nM of microcystin (▴). Attachment was detemined by random selection of five fields (visual microscopic examination at ×10 objective magnification). Data are expressed as the mean and standard deviation percentage attachment compared with control vehicle-treated cells from one experiment. Data are typical of at least six separate experiments. HSC, hepatic stellate cell.

The mechanism of exclusion of trypan blue dye from cells is not fully understood, but we do know that it relies on an energy-dependent process.21, 31 Because apoptosis is an energy-dependent (often programmed) event, cells that undergo apoptosis exclude trypan blue for a period, despite the morphological appearance of death.31 Fig. 2A and 2C demonstrate that toxic (50 μM) concentrations of gliotoxin (i.e., a concentration that resulted in detachment from the substratum) gave rise to hepatocytes that did not exclude trypan blue. This is similar to the effects seen with 200 μM chlorpromazine but not with agents that cause hepatocyte apoptosis (i.e., 100 nM microcystin [see Fig. 2A and 2C]) or TNF-α together with the protein synthesis inhibitor cycloheximide12 (see Fig. 2C). In contrast, concentrations of gliotoxin above 1.5 μM that caused activated HSC detachment from the substratum resulted in HSCs that retained the ability to exclude trypan blue, whereas chlorpromazine treatment resulted in HSCs that were unable to exclude trypan blue (Fig. 2B and 2C). Fig. 2B also shows that gliotoxin treatment of HSCs resulted in the cleavage of DNA to a nucleosomal ladder as previously reported.5

Figure 2.

Effect of gliotoxin and other treatments on trypan blue exclusion by hepatocytes and activated HSCs. (A) Hepatocytes (bar = 100 μm) or (B) activated HSCs (bar = 50 μm) were treated as indicated for 3 hours followed by trypan blue staining as outlined in Materials and Methods. (B) The bottom right panel shows agarose gel electrophoresis of DNA that was isolated from control and gliotoxin-treated activated HSCs after 4 hours of treatment, using a 1-kB ladder from Promega. (C) Cells were treated with the indicated concentration of gliotoxin or with 20 ng/mL of TNF-α + 10 μM of cycloheximide, 100 nM of microcystin, 200 μM of chlorpromazine, or 1 mM of N-ethylmaleimide (all for 3 hours except TNF-α + cycloheximide, which was applied for 16 hours). Percentage of cells excluding trypan blue was determined from five randomly selected fields (visual microscopic examination at ×10 objective magnification). Data are expressed as mean and standard deviation percentage of total viability (excludes dye). Results are typical of at least four separate experiments. *Significantly different percentage viability versus control cells; P > 95% using Student t test (two-tailed). Abbreviations: HSCs, hepatic stellate cells; GT, gliotoxin; TNF-α, tumor necrosis factor α; CHX, cycloheximide; MC, microcystin; CLPRZ, chlorpromazine; NEM, N-ethylmaleimide.

Caspase 3 is a protease known to function as a distal executioner of many apoptotic programs.32 Fig. 3 indicates that hepatocytes treated with gliotoxin—including concentrations that caused substratum detachment—did not contain increased levels of caspase 3 activity similar to cells treated with chlorpromazine. In contrast, treatment with TNF-α/cycloheximide or microcystin alone—treatments known to stimulate hepatocyte apoptosis12, 29—resulted in significant increases in caspase 3 activity (see Fig. 3). In contrast to hepatocytes, HSCs treated with gliotoxin showed a dose-dependent significant increase in caspase 3 activity (see Fig. 3). Compounds that either had no effect in HSCs (microcystin) or caused HSC necrosis (chlorpromazine) did not result in an increase in caspase 3 activity (see Fig 3).

Figure 3.

Effect of gliotoxin and other treatments on caspase 3 activity levels in hepatocytes and activated HSCs. Hepatocytes and culture-activated HSCs were prepared as outlined in Materials and Methods. Culture medium was changed and cells were treated with the indicated concentration of gliotoxin or with 20 ng/mL of TNF-α + 10 μM of cycloheximide, 100 nM of microcystin, 200 μM of chlorpromazine, and 1 mM of N-ethylmaleimide (all for 3 hours except TNF-α + cycloheximide, which was applied for 16 hours). Data are the mean and and standard deviation of three separate determinations from the same experiment, which are typical of at least three separate experiments. *Significantly different fold caspase 3 activity versus control cells; P > 95% using Student t test (two-tailed). Abbreviations: GT, gliotoxin; TNF-α, tumor necrosis factor α; CHX, cycloheximide; MC, microcystin; CLPRZ, chlorpromazine; NEM, N-ethylmaleimide; HSCs, hepatic stellate cells.

Nuclear changes are diagnostic of regulated apoptotic mechanisms of cell death.32 Supplementary Fig. 1 shows that toxic levels of gliotoxin did not stimulate hepatocyte nuclear condensation as observed when hepatocytes were treated with known proapoptotic TNF-α/cycloheximide agents in contrast to both gliotoxin and TNF-α/cycloheximide treatments in activated HSCs.

These data show that activated HSCs were significantly more sensitive to the toxic effects of gliotoxin than hepatocytes and that gliotoxin stimulated an apoptotic mechanism of cell death in HSCs. In contrast, concentrations of gliotoxin toxic to hepatocytes resulted in unambiguous necrosis. It should be noted, however, that a proportion of hepatocytes treated with gliotoxin soon after their isolation (0–2 hours) underwent apoptosis (data not shown) likely associated with an inhibition by gliotoxin of NF-κB activity (which is briefly activated in response to hepatocyte isolation33).

The Uptake and Metabolism of Gliotoxin by Hepatocytes and HSCs.

To determine the underlying causes of the differential effects of gliotoxin in hepatocytes and HSCs, gliotoxin uptake and metabolism were examined in pure hepatocyte and activated HSCs cultures. Supplementary Fig. 2A indicates that both hepatocytes and activated HSCs rapidly removed gliotoxin from the culture medium in a time-dependent fashion. HSCs sequestered approximately 1.5 times as much gliotoxin per unit of protein than hepatocytes; however, the increased levels sequestered by HSCs could not account for the difference in the mode of cell death between hepatocytes and HSCs.

Supplementary Fig. 2B and 2C indicate that gliotoxin was metabolized to several different metabolites. Metabolite A had a retention time identical with dithiothreitol (DTT)-reduced or GSH-reduced gliotoxin (Supplementary Fig. 3B), suggesting that this metabolite is reduced dithiol gliotoxin. Metabolites B–D have been shown to be gliotoxin-glutathione conjugates.34 Metabolite E is unidentified. These metabolites were not produced by activated HSCs at 10 μM gliotoxin concentrations or below. At high (50 μM) concentrations of gliotoxin and after 120 minutes of incubation, reduced dithiol gliotoxin (metabolite A) and very low levels of a glutathione conjugate (metabolite B) were detected (see Supplementary Fig. 2B). These data suggest that metabolism of gliotoxin by HSC is not required for the stimulation of HSC apoptosis since apoptosis occurs in the absence of any apparent metabolite production. In contrast, the more rapid and extensive metabolism of gliotoxin by hepatocytes may be associated with either the avoidance of apoptosis and/or results in their necrosis.

The Effect of Gliotoxin on the Levels of Glutathione in Hepatocytes and HSCs.

Because gliotoxin contains a dithiol bridge that is neccesary for the proapoptotic effects of this compound in HSCs,5, 35–37 the effect of gliotoxin on the levels of glutathione in hepatocytes and activated HSCs was examined. Supplementary Fig. 3A indicates that the addition of gliotoxin to GSH in (a cell-free) HEPES/HBSS buffer system resulted in direct oxidation of GSH to GSSG, suggesting that there is potential for a direct reaction of gliotoxin with GSH (and other thiols) in cells. Supplementary Fig. 3B confirms that there was a conversion of gliotoxin to reduced dithiol gliotoxin by GSH or by DTT. The decrease in detectable (derivatized) GSH in this cell-free system could not be accounted for through oxidation alone, suggesting that GSH directly reacts with gliotoxin to form conjugates. However, although a gliotoxin-GSH conjugate would still be expected to be derivatizable, conjugates (such as metabolites B, C, or D seen with hepatocytes [see Supplementary Fig. 2B]) were not detected by this HPLC assay (not shown) or the gliotoxin HPLC assay (see Supplementary Fig 3B). Supplementary Fig. 3C indicates that the addition of 50 μM of gliotoxin (levels that cause hepatocyte necrosis) to rat hepatocytes resulted in a rapid and sustained depletion of GSH to approximately 30%–40% of control levels by 3 hours (Supplementary Fig. 3E) with detectable increases in GSSG levels—at least during earlier timepoints (see Supplementary Fig. 3C). This depletion of glutathione in hepatocytes was also observed when cells were treated with 200 μM of chlorpromazine or 1 mM N-ethylmaleimide (Supplementary Fig. 3E), compounds which also caused hepatocyte necrosis (see Figs. 1 and 2). In contrast, treatments that caused hepatocyte apoptosis (TNF-α plus cycloheximide or microcystin) did not result in a depletion of cellular glutathione (see Supplementary Fig. 3E). Subhepatocyte toxic 1.5 μM gliotoxin resulted in an increase in glutathione levels (see Supplementary Fig. 3E).

Supplementary Figure 3D and E indicate that a concentration of 1.5 μM of gliotoxin—a concentration that caused the complete apoptosis of activated HSCs—resulted in a significant increase in activated HSC GSH levels as also observed with TNF-α/cycloheximide treatment. Chlorpromazine results in a significant reduction in GSH levels and, along with the sulfhydryl reagent N-ethylmaleimide, caused cell death by necrosis (see Supplementary Fig. 3). In all treatments, however, there was no detectable generation of GSSG in HSCs.

ROS Generation as a Mechanism of Necrotic Cell Death in Hepatocytes Exposed to Gliotoxin.

GSH oxidation in cell-free systems and reductions in hepatocyte GSH levels in hepatocytes (with a transient increase in GSSG levels) suggests that gliotoxin may alter the redox status of cells and promote oxidative stress. Reduction of gliotoxin as observed in a cell-free system and in hepatocytes in these studies (see Supplementary Figs. 2B and 2C and 3B) could initiate a cycle of gliotoxin reduction and oxidation (redox cycling) that may generate superoxide and other ROS. Hepatocytes and activated HSCs were therefore loaded with a cell-permeable dye (H2DCFDA) that, when de-esterified in cells, is fluorescent under oxidative conditions.38 Figure 4A indicates that hepatocytes loaded with H2DCFDA fluoresced significantly after treatment with gliotoxin, particularly at 7.5-μM and 50-μM concentrations. In contrast, activated HSCs under the same treatment and optical conditions did not fluoresce significantly (see Fig. 4A). To confirm that a lack of fluorescence from gliotoxin-treated HSCs is not associated with different optical settings or treatment conditions, fluorescence was examined in H2DCFDA-loaded hepatocyte/HSC cocultures. Fig. 4B indicates that fluorescence was almost exclusively limited to hepatocytes. Because the apparent lack of fluorescence in gliotoxin-treated HSC may be associated with either a lack of uptake and/or de-esterification of H2DCFDA, hepatocyte and HSC extracts were incubated with H2DCFDA, and the level of fluorescence was examined after the addition of pro-oxidant (H2O2 and FeCl2; gliotoxin and GSH). Fig. 4C shows that H2DCFDA was not fluorescent in the absence of cell extract addition, even in the presence of pro-oxidants. Addition of either hepatocyte or HSC extract led to the production of fluorescence upon the addition of pro-oxidants, suggesting that both hepatocytes and HSCs are able to de-esterify H2DCFDA to a product that is fluorescent in the presence of oxidative conditions. However, the levels of fluorescence from H2DCFDA de-esterified by HSC extracts were lower than the levels of fluorescence from H2DCFDA de-esterifed by hepatocyte extracts. HSCs may therefore have a lower capacity for de-esterifying H2DCFDA, which may modulate the levels of fluorescence emitted by HSCs. To test this possibility, H2DCFDA-loaded hepatocyte and HSCs were treated with H2O2 and FeCl2. Fig. 4D indicates that both hepatocytes and HSCs significantly fluoresced after treatment with H2O2 and FeCl2. HSCs are therefore able to sequester and de-esterify H2DCFDA. The lack of fluorescence in HSCs treated with gliotoxin is therefore likely to be due to a significantly lower production of ROS in HSCs compared with hepatocytes.

Figure 4.

Detection of ROS/oxidative stress in cells treated with gliotoxin. (A) Hepatocytes (panels A–D) or activated HSCs (panels E–H) were loaded with H2DCFDA as outlined in Materials and Methods. After loading, the cells were washed twice with 5 mL HEPES/HBSS buffer. Cells were then incubated in 3 mL HEPES/HBSS buffer and treated with either dimethyl sulfoxide vehicle control (panels A and E), 1.5 μM of gliotoxin (panels B and F), 10 μM of gliotoxin (panels C and G), or 50 μM of gliotoxin (panels D and H). After 20 minutes incubation at 37°C, the cells were examined using a confocal microscope, and pictures were taken under the same optical conditions. (B) Activated HSCs were subcultured and seeded into a dish containing hepatocytes seeded at approximately 50% of normal density. After 24 hours, cocultures were loaded with H2DCFDA and treated with 50 μM of gliotoxin and examined as outlined above. Dimethyl sulfoxide–treated cocultures did not fluoresce (data not shown), whereas gliotoxin-treated cocultures appeared as shown (right panel). A tungsten light micrograph (left panel) was used to identify HSCs morphologically (outlined) and their position (yellow) was superimposed onto the fluorescence image of the same field. The fluorescence (arrow) is of a detached cell that was not associated with the HSCs indicated, which remained attached to substratum at the time of image capture. (C) Hydrolysis of H2DCFDA in hepatocytes and activated HSC was examined in 40 μg of lysed cell extract in 50-mM Tris buffer (pH 7.2) mixed with 200 μM H2DCFDA in a 50-μL volume after incubation for 20 minutes at 37°C. Fluorescence was measured 5 minutes after the addition of either 100 μL of 50-mM Tris buffer (pH 7.2) (clear bars); 50 μL of 3-mM H2O2 plus 50 μL of 300-μM FeCl2 (black bars); or 50 μL of 3-mM gliotoxin + 50 μL of 30-mM GSH (grey bars) to the H2DCFDA/extract mixes after addition of 3 mL of 50-mM Tris buffer (pH 7.2) (excitation λ 510 nm, emission λ 550 nm). Data are the mean and standard deviation of three separate determinations from the same experiment and are typical of three separate experiments. (D) Cells were loaded with H2DCFDA and treated as outlined above with either gliotoxin or 1 mM of H2O2 + 100 μM of FeCl2 (H2O2/Fe2+) and fluorescence was examined as outlined above under identical optical conditions. N.D. = no dye loading but cells treated with gliotoxin, a result typical of all treatments in which no dye was loaded into cells. The results are typical of at least three separate experiments. HSCs, hepatic stellate cells; H2DCFDA, 2′,7′-dichlorodihydrofluorescein diacetate.

The Effect of Thiol Reducing Agents and Antioxidants on the Toxicity of Gliotoxin in Hepatocytes and Activated HSCs.

The structure of gliotoxin (Supplementary Fig. 4) includes a dithiol bridge, and Supplementary Fig. 3B indicates that this compound is susceptible to reduction by cellular-reduced thiols such as GSH, even in cell-free systems. Fig. 5A indicates that apoptosis of activated HSCs was potently blocked by the thiol-reducing agent DTT but also by the thiol-active agent pyrrolidine dithiocarbamate (PDTC). In contrast, hepatocyte cell death in response to gliotoxin was unaffected or potentiated by the addition of DTT and PDTC, respectively. Other compounds that block the nucleosomal ladder formation in HSCs after treatment with gliotoxin (Quin2-AM and caspase inhibitor 1)5 also had no effect on gliotoxin toxicity in hepatocytes. In contrast, the antioxidant catalase significantly reduced gliotoxin toxicity in hepatocytes, whereas addition to gliotoxin-treated HSCs had no effect (see Fig. 5A). Note that in these experiments (Fig. 5), the mode of cell death for HSCs (apoptosis) and hepatocytes (necrosis) was not altered by the addition of other compounds as judged by trypan blue exclusion (data not included). The addition of several components to reduce the levels of ROS in HSC cultures was without any detectable effect on gliotoxin-dependent HSC apoptosis, whereas the addition of catalase and superoxide dismutase (SOD) was the most potent abbrogator of hepatocyte necrosis (see Fig. 5A). The addition of catalase and SOD significantly reduced the levels of ROS detectable in H2DCFDA-loaded hepatocytes (Fig. 5B), supporting the suggestion that ROS generation is the mechanism of cell death in hepatocytes treated with gliotoxin. A major component of the complex of proteins that regulate the MPT is ANT. Figure 6A demonstrates that there is a different expression profile of the ANT isoforms between HSCs and hepatocytes as determined by Western blotting, with HSCs expressing predominantly ANT1 (cardiac/skeletal muscle isoform), whereas hepatocytes express the ANT2 isoform (although culture conditions appear to result in the expression of ANT1 isoform). To determine if the specific effects of gliotoxin in HSCs is associated with the ANT, HSCs and hepatocytes were treated with gliotoxin, PDTC and/or DTT and the migration of ANT under denaturing but nonreducing gel conditions examined. Figure 6B demonstrates that the ANT1 isoform in HSCs had variable redox status under nonreducing gel conditions, because at least two distinct bands are observed. Treatment with 1.5 μM of gliotoxin to cause HSC apoptosis resulted in an alteration in the migration of the slower migrating HSC ANT band that was prevented by addition to the culture medium of either PDTC or DTT. Treatment with 50 μM of gliotoxin resulted in an apparent loss of ANT—although under nonreducing gel conditions this could be caused by an oligomerization of proteins to high molecular weight complexes that did not enter the gel. Interestingly, PDTC promotes the formation of the ANT dimer in HSCs, but neither PDTC nor DTT caused cell death in HSCs in the short term. Under nonreducing gel conditions, hepatocyte ANT isoforms migrated as a single band, and gliotoxin treatment did not result in detectable alteration in ANT migration (Fig. 6C).

Figure 5.

Hepatocyte and HSC-dependent cell death is differentially inhibited by antioxidants and thiol redox–active compounds. (A) Cells were pretreated with the indicated concentration of potential inhibitor (black bars) 1 hour prior to the addition of a toxic concentration of gliotoxin (hepatocytes, 30 μM; HSCs, 1.5 μM). Cell attachment (viability) was determined by assaying attached cell protein after removal of culture medium and two washes with phosphate-buffered saline. Results are presented as percentage cells attached versus untreated (100%). Gliotoxin alone–treated cells (grey bars). Data are the mean and standard deviation of at least four separate experiments. *Significantly different attachment versus gliotoxin-treated cells P > 95% using Student's t test (two-tailed). (B) Hepatocytes were loaded with H2DCFDA and treated as outlined in (A) with either vehicle controls (dimethyl sulfoxide and phosphate-buffered saline; panel A); 30 μM of gliotoxin (panel B); 30 μM of gliotoxin after 1 hour preincubation with 40 μg/mL SOD + 400 μg/mL catalase (panel C); 1 hour preincubation with 40 μg/mL SOD + 400 μg/mL catalase followed by addition of dimethyl sulfoxide vehicle only (panel D). Panels A–D show tungsten light images; panels E–H show fluorescence in the same fields. DTT, dithiothreitol; PDTC, pyrrolidine dithiocarbamate; CsA, cyclosporin A; Z-VAD-FMK, caspase inhibitor 1; SOD, superoxide dismutasel; CS, catalase + SOD; DCS, deferoxamine + catalase + SOD.

Figure 6.

ANT expression in HSCs and hepatocytes; redox-sensitive alteration of HSC ANT by gliotoxin treatment. (A) Western blots for ANT, total actin, and albumin under reducing conditions. IC = freshly isolated rat hepatocytes. Hepatocytes were cultured for 12 and 24 hours. All lanes contain 10 μg total cell protein. (B) Western blot for HSC ANT (under nonreducing conditions) and α-smooth muscle actin (under reducing conditions) after treatment for 1 hour with the indicated concentration of gliotoxin and either dimethyl sulfoxide vehicle control (−), 300 μM of PDTC, or 100 μM of DTT. Each lane contains 10 μg of cell protein/lane. (C) Western blot for hepatocyte ANT (under nonreducing conditions), total actin, and albumin (under reducing conditions) after treatment for 1 hour with the indicated concentration of gliotoxin and either dimethyl sulfoxide vehicle control (−), 300 μM of PDTC, or 100 μM of DTT. Note, all cell extracts were washed twice with ice-cooled PBS (137 mM of NaCl, 2.7 mM of KCl, 10 mM of phosphate [pH 7.4]) prior to protein determination and processing for Westerns to ensure the presence of DTT in some samples did not affect the nonreducing nature of nonreducing gels. The antibody to ANT was raised in rabbits against ANT purified from rat heart mitochondria and cross-reacts with all rat and human isoforms. Each lane contains 10 μg cell protein/lane. These data are typical of at least three separate experiments. Abbreviations: HSC, hepatic stellate cell; ANT, adenine nucleotide transporter; PDTC, pyrrolidine dithiocarbamate; DTT, dithiothreitol.


Our data demonstrate markedly altered effects of gliotoxin on different cell types of the liver. The mechanism of hepatocyte death is likely to be dependent on gliotoxin metabolism because hepatocytes rapidly generated several metabolites of gliotoxin that were not generated in HSCs. In view of gliotoxin's interaction with the cellular thiol GSH, it is likely that gliotoxin is directly reduced by cellular thiols within cells. However, a significant reduction of cellular GSH was only seen with high concentrations of gliotoxin in both hepatocytes and HSCs. It is possible, therefore, that at high cellular concentrations of gliotoxin, GSH reduces gliotoxin (oxidizing GSH to GSSG), resulting in redox cycling of gliotoxin and generation of ROS as outlined in Supplementary Fig. 2. In hepatocytes—which additionally conjugate gliotoxin with GSH—there would be a depletion of total cellular GSH and greater pressure on the reduced GSH pool to protect cellular thiols. This may therefore be the cause of the high levels of ROS observed in hepatocytes treated with a toxic concentration of gliotoxin (but not seen in HSCs).

The MPT is constituted of cyclophylin D, ANT, and possibly the voltage-activated anion channel.39, 40 Three cysteine residues in the ANT have been shown to be susceptible to a variety of thiol-reactive agents and to modulate the activity of the MPT.39, 40 It is likely that gliotoxin directly reacts with one or more cysteine residues in the ANT in a similar fashion to diamide41 or phenylarsineoxide.39, 42 This is supported by data showing that HSC apoptosis and altered HSC ANT mobility both occurred after gliotoxin treatment and both were inhibited by DTT or PDTC. DTT could function to reverse the putative gliotoxin–ANT conjugate, because it readily reduces many protein thiols. PDTC is an unusual compound because it is able to both reduce and oxidize thiol groups.43, 44 We show that PDTC does not reduce gliotoxin but is more likely to convert any reduced gliotoxin back to the oxidized form. Interestingly, PDTC contains vicinal sulphur atoms in its structure, as does gliotoxin. PDTC may therefore compete with gliotoxin for binding to ANT and prevent gliotoxin from stimulating a MPT.

With respect to apoptosis, the mechanism of action of gliotoxin may be best considered as interfering with two pathways of TNF-α receptor signaling (Supplementary Fig. 5). TNF-α is known to stimulate an apoptotic death pathway that involves an MPT-regulated release of cytochrome c from mitochondria and activation of caspase cascades (pathway 1).12, 45 However, in many cell types—including both hepatocytes and HSC—TNF-α also activates NF-κB activation (pathway 2), which inhibits apoptosis.12, 46 Accordingly, TNF-α alone stimulates hepatocyte proliferation in vivo.16 Gliotoxin is an inhibitor of NF-κB activation through its inhibition of I-κB degradation.47 Hepatocytes treated with both TNF-α and gliotoxin therefore undergo apoptosis, likely because of an inhibition of NF-κB–dependent inhibition of apoptosis. However, although gliotoxin is able to inhibit TNF-α–inducible NF-κB activity, activated HSCs contain a constitutively activated level of NF-κB activity through a CBF-1–dependent down-regulation of I-κB levels, which presumably results in a stoichiometric excess of Rel: I-κB protein.5, 7, 48 Gliotoxin treatment of HSCs therefore does not inhibit constitutive levels of NF-κB in activated HSC (note the same situation may not be the case in quiescent HSCs). The stimulation of apoptosis by gliotoxin in activated HSCs must therefore be independent of an inhibition of I-κB degradation.

In the context of liver fibrosis, these observations are important for understanding the potential mechanisms for therapeutic intervention. The HSC is pivotal to the development of liver fibrosis, and stimulation of HSC apoptosis is a potential avenue for treatment. Because cytokines such as TNF-α are known to be elevated in the diseased and fibrotic liver, an inhibition of NF-κB by any potential therapeutic approach is likely to promote the death of hepatocytes with potentially catastrophic consequences.49 However, in view of the potent effect of gliotoxin in stimulating HSC apoptosis, an analogue that retained pathway 1/MPT promoting activity without inhibiting pathway 2/26S proteasomal I-κB degradation (see Supplementary Fig. 3) could be a valuable tool in finding a treatment for liver fibrosis.