Glycogen synthase kinase 3-mediated voltage-dependent anion channel phosphorylation controls outer mitochondrial membrane permeability during lipid accumulation§


  • Cecile Martel,

    1. INSERM U769, LabEx LERMIT, Université Paris-Sud, Faculté de Pharmacie, Châtenay-Malabry, France
    2. INSERM U1004, Université de Paris-Sud, PRES UniverSud Paris, Hôpital Paul Brousse, Villejuif, France
    Search for more papers by this author
  • Maya Allouche,

    1. Institut Charles Gerhardt, UMR 5253 CNRS, Equipe MACS, UFR des Sciences Pharmaceutiques et Biologiques, Montpellier, France
    Search for more papers by this author
  • Davide Degli Esposti,

    1. INSERM U1004, Université de Paris-Sud, PRES UniverSud Paris, Hôpital Paul Brousse, Villejuif, France
    Search for more papers by this author
  • Elena Fanelli,

    1. Dipartimento di Fisiologia Generale ed Ambientale, Università degli Studi di Bari, Aldo Moro, Bari, Italy
    Search for more papers by this author
  • Céline Boursier,

    1. INSERM IFR 141, Université Paris-Sud, Faculté de Pharmacie, Châtenay-Malabry, France
    Search for more papers by this author
  • Céline Henry,

    1. INRA, UMR 1319 Micalis, PAPPSO, Jouy-en-Josas, France
    Search for more papers by this author
  • Joel Chopineau,

    1. Institut Charles Gerhardt, UMR 5253 CNRS, Equipe MACS, UFR des Sciences Pharmaceutiques et Biologiques, Montpellier, France
    2. Université de Nîmes, Nîmes, France
    Search for more papers by this author
  • Giuseppe Calamita,

    1. Dipartimento di Fisiologia Generale ed Ambientale, Università degli Studi di Bari, Aldo Moro, Bari, Italy
    Search for more papers by this author
  • Guido Kroemer,

    1. INSERM, U848, Villejuif, France
    2. Metabolomics Platform, Institut Gustave Roussy, Villejuif, France
    3. Centre de Recherche des Cordeliers, Paris, France
    4. Pôle de Biologie, Hôpital Européen Georges Pompidou, AP-HP, Paris, France
    5. Université Paris Descartes, Faculté de Médecine, Paris, France
    Search for more papers by this author
  • Antoinette Lemoine,

    1. INSERM U1004, Université de Paris-Sud, PRES UniverSud Paris, Hôpital Paul Brousse, Villejuif, France
    Search for more papers by this author
  • Catherine Brenner

    Corresponding author
    1. INSERM U769, LabEx LERMIT, Université Paris-Sud, Faculté de Pharmacie, Châtenay-Malabry, France
    • INSERM U769, LabEx LERMIT, Université Paris-Sud, 5 Rue J.-B. Clément, 92290 Châtenay-Malabry Cedex, France
    Search for more papers by this author
    • fax: 33-1 46 83 54 75

  • Potential conflict of interest: Nothing to report.

  • C.B. is supported by the University of Paris Sud (grant attractivité), Institut National pour le Cancer (INCa, 2008-1-PL BIO-04-CNRS ON1), and LabEx LERMIT. C.B. and A.L. are supported by the PRES UniverSud Paris. C.M. received a fellowship from Association pour la Recherche sur le Cancer. C.B., M.A., and J.C. are supported by ANR (ANR-08PCVI-0008-01). G.C. received support from MIUR (PRIN20089SRS2X_003) and Fondazione Cassa di Risparmio di Puglia.

  • §

    Senior coauthors.


Nonalcoholic steatosis is a liver pathology characterized by fat accumulation and severe metabolic alterations involving early mitochondrial impairment and late hepatocyte cell death. However, mitochondrial dysfunction mechanisms remain elusive. Using four models of nonalcoholic steatosis, i.e., livers from patients with fatty liver disease, ob/ob mice, mice fed a high-fat diet, and in vitro models of lipotoxicity, we show that outer mitochondrial membrane permeability is altered and identified a posttranslational modification of voltage-dependent anion channel (VDAC), a membrane channel and NADH oxidase, as a cause of early mitochondrial dysfunction. Thus, in nonalcoholic steatosis VDAC exhibits reduced threonine phosphorylation, which increases the influx of water and calcium into mitochondria, sensitizes the organelle to matrix swelling, depolarization, and cytochrome c release without inducing cell death. This also amplifies VDAC enzymatic and channel activities regulation by calcium and modifies its interaction with proteic partners. Moreover, lipid accumulation triggers a rapid lack of VDAC phosphorylation by glycogen synthase kinase 3 (GSK3). Pharmacological and genetic manipulations proved GSK3 to be responsible for VDAC phosphorylation in normal cells. Notably, VDAC phosphorylation level correlated with steatosis severity in patients. Conclusion: VDAC acts as an early sensor of lipid toxicity and its GSK3-mediated phosphorylation status controls outer mitochondrial membrane permeabilization in hepatosteatosis. (HEPATOLOGY 2013)

Nonalcoholic fatty liver disease (NAFLD) is accompanied by hepatosteatosis, a clinical condition characterized by excessive accumulation of lipids within hepatocytes and complex metabolic alterations.1, 2 Although reversible in early stages, steatosis can lead to more aggressive forms of liver injury such as hepatitis, cirrhosis, and hepatocarcinoma.3 Investigation of patients with hepatosteatosis showed that mitochondria harbor prominent morphologic and functional abnormalities, suggesting a central role of these organelles in the pathogenesis.4 Mitochondria can influence cell fate at the levels of energy production, lipid metabolism, production, and detoxification of reactive oxygen species (ROS) and release of proapoptotic proteins.5 All these alterations favor an increase in apoptotic and necrotic hepatocyte cell death.

The voltage-dependent anion channel (VDAC) or porin is the most abundant protein expressed in the mitochondrial outer membrane (OM). In physiological conditions, VDAC allows the flux of ions and metabolites necessary to mitochondrial metabolism and cell growth.6, 7 Thus, VDAC channel closure by tubulin limits mitochondrial metabolism, thereby decreasing the mitochondrial inner membrane potential (ΔΨm).8 VDAC is also implicated in NADH oxidation and then plays a role in cellular redox metabolism.9 In conditions of lethal stress, VDAC can contribute to the proapoptotic mitochondrial membrane permeabilization (MMP) (reviewed5), either by way of homo-oligomerization, direct physical interactions with endogenous members of the Bcl-2 family (e.g., Bax), adenine nucleotide translocase (ANT), and virus-encoded Bcl-2-like proteins or by way of its impact on calcium (Ca2+) fluxes and ROS detoxification. Nonetheless, the exact role of VDAC in MMP and permeability transition (PT) is debated and the molecular mechanisms that determine VDAC transition from a normal to a lethal function are elusive.

Here we provide evidence that, in steatotic hepatocytes, the lack of VDAC phosphorylation sensitizes hepatocytes to Ca2+-induced MMP. Using cellular and molecular approaches, we demonstrate that VDAC lack of phosphorylation is accompanied by a decrease of interaction with the serine/threonine glycogen synthase kinase 3 (GSK3) and Bcl-XL in mitochondria and enhances the stimulation of VDAC functions by Ca2+.


AIF, apoptosis inducing factor; ANT, adenine nucleotide translocase; Ca2+, calcium; CsA, cyclosporine A; Cyt c, cytochrome c; DIDS, disodium 4,4′-diisothiocyanatostilbene-2,2′-disulfonate; ΔΨm, mitochondrial inner membrane potential; FA, fatty acids; GSK3, glycogen synthase kinase 3; HFD, high fat diet; MC, mitochondrial complex; MMP, mitochondrial membrane permeabilization; NAFLD, nonalcoholic fatty liver disease; ND, nondiet; OM, outer membrane; PI3K, phosphoinositide 3 kinase; PT, permeability transition; PTP, permeability transition pore; P-Thr, phosphorylated threonine, P-VDAC, phosphorylated voltage-dependent anion channel; ROS, reactive oxygen species; thr, threonine; TNF, tumor necrosis factor; VDAC, voltage-dependent anion channel; Wort, wortmannin.

Materials and Methods

Animal Models.

Female 6 to 12-week-old lean (C57BL/6J) and ob/ob (B6.V-Lepob/J) mice were purchased from Janvier (Le Genest Saint Isle, France). Male 4-week-old C57BL/6J mice (Harlan, Udine, Italy) were acclimatized to laboratory conditions for 1 week before being randomly assigned to either the high fat or standard chow diet (Altromin-Rieper, Vandoies, Italy).

Collection of Human Liver Biopsies.

Eight frozen biopsies were chosen among liver biopsies collected during graft harvesting in our institution. The study protocol follows the recommendations of the ethical guidelines of the 1975 Declaration of Helsinki and was approved by our ethical committee.

Cell Culture.

HHL-5, immortalized noncancerous primary hepatocytes (a generous gift from Dr. A. Patel, University of Glasgow, UK)10 were cultured in DMEM-F12 (Dulbecco's Modified Eagle's Medium) supplemented with 10% fetal bovine serum, 1% glutamax, and 1% penicillin/streptomycin, in 5% CO2 at 37°C supplemented with 2% bovine serum albumin (BSA) and 2 mM oleate/palmitate (2:1 ratio), as described in the literature,11 and the cells were treated with this medium for 3 to 30 hours.

All sample collections and protocols are described in the Supporting Information.


Steatosis Promotes Alteration of MMP.

Mitochondrial modifications were investigated in mice that become obese due to a genetic leptin-deficiency caused by the ob/ob mutation.13 Transmission electron microscopy revealed the presence of numerous mega-mitochondria with matrix swelling, loss of internal material, and OM ruptures in a population of isolated mitochondria from ob/ob mice (Fig. 1A). These structural alterations correlated with an increase in MMP and of mitochondrial volume. As compared to mitochondria from lean mice, ob/ob mitochondria exhibited accelerated Ca2+-induced matrix swelling (Fig. 1B) and ΔΨm dissipation (Fig. 1C). Permeabilized fatty acid (FA)-treated human immortalized hepatocytes (HHL-5 cells)10 (Supporting Fig. 1) also proved to be more sensitive to Ca2+-triggered ΔΨm loss than untreated cells (Fig. 1D). Moreover, mitochondria from ob/ob mice were more permeable to water, both in normal condition and upon Ca2+ stimulation of PT (Fig. 2A). In the presence of cyclosporin A (CsA), the prototypic inhibitor of PT, the permeability of control and ob/ob mitochondria was reduced to similarly low levels (Fig. 2A). As a consequence, the proapoptotic intermembrane space protein, cytochrome c (Cyt c), was found in the 100,000 x g-supernatants of isolated ob/ob mitochondria from obese mice (Fig. 2B). This was not the case for the apoptosis-inducing factor (AIF), another proapoptotic protein (Fig. 2B). Caspase 3/7 activities were not enhanced by FA accumulation in vitro or in vivo (Fig. 2C,D), suggesting that the apoptotic signaling cascade was not activated. The distribution of Cyt c in ob/ob and lean mouse livers was also analyzed by immunohistochemistry (Supporting Fig. 2). Cyt c was particularly expressed in portal tracts and in some centrolobular areas, whereas lobular hepatocytes presented a lesser staining (Supporting Fig. 2A,B). In livers obtained from lean mice, a punctate cytoplasmic staining was observed in hepatocytes, whereas in steatotic or also some nonsteatotic hepatocytes from ob/ob mice livers a more diffused cytoplasmic staining was observed. These results confirm that in steatotic liver, more hepatocytes present a cytoplasmic liberation of Cyt c from mitochondria probably due to an increased membrane permeabilization.

Figure 1.

Steatosis and FA accumulation alters mitochondrial morphology, mitochondrial membrane potential, and swelling sensitivity to calcium. (A) Representative morphology of isolated liver mitochondria from lean and ob/ob mice by transmission electron microscopy. Scale bar = 1 mm. (B) Kinetic measurement of matrix swelling (optical density at 540 nm) (n = 10) and (C) transmembrane potential (rhodamine 123 fluorescence arbitrary units) of isolated mitochondria from lean and ob/ob mice livers in response to 25 μM Ca2+ addition at t0 (n = 10). (D) Effect of 24 hours of treatment of immortalized hepatocytes (HHL-5) with 2 mM fatty acids (FA) on the sensitivity of permeabilized cells to 25 μM Ca2+-induced depolarization (rhodamine 123 fluorescence, arbitrary units) (n = 3). Values are mean of triplicate samples ± standard deviation (SD).

Figure 2.

Steatosis alters mitochondrial water permeability, cytochrome c location, but does not activate caspases 3/7. (A) Mitochondrial membrane water permeability expressed as a rate constant (Ki, s−1) of isolated mitochondria from lean (solid bars) and ob/ob (open bars) mice livers in response to Ca2+ and CsA (means ± SD, *P < 0.05 and **P < 0.01) (n = 4). (B) Subcellular localization of Cyt c and AIF determined by fractionation of total liver from lean and ob/ob mice in mitochondria and cytosol and followed by immunoblotting analysis. One representative immunoblot of three independent experiments is shown. (C,D) Caspases 3/7 activity ratio in HHL-5 cells treated (open bars) or not (solid bars) with 2 mM FA and in total liver lysates from lean (solid bars) and ob/ob (open bars) mice. Caspases activity was measured by a fluorimetric assay (n = 3).

Calcium Regulates VDAC Functions in Steatosis.

To assess the causative link between mitochondrial proteins modification and mitochondrial dysfunction, the pharmacological regulation of MMP was examined. Thus, Ca2+ induced the maximal swelling and depolarization in 30 minutes and CsA inhibited Ca2+-induced swelling and depolarization in both mitochondrion types nearly as efficiently as the Ca2+ chelator EGTA (Fig. 3A,B). The VDAC blocker disodium 4,4′-diisothiocyanatostilbene-2,2′-disulfonate (DIDS) and the permeability transition pore (PTP) inhibitor N-[(3,5-di-tert-butyl-4-hydroxy-1-thiophenyl)]-3-propyl-N′-(2,3,4-trimethoxybenzyl) piperazine difumarate (S15176)14, 15 were less efficient in preventing PT induction in ob/ob mitochondria than in control mitochondria (Fig. 3A,B). This points to a role for PTP and VDAC in the differential response to Ca2+. No differential effect of bongrekic acid, an adenine nucleotide translocase (ANT) inhibitor, was observed, suggesting that ANT is not involved in the difference of response of both types of mitochondria to Ca2+ (not shown).

Figure 3.

Steatosis impacts Ca2+-stimulated channel function of VDAC in ob/ob mice. (A) Effect of pharmacological inhibitors on Ca2+-induced mitochondrial matrix swelling. Triplicates of mitochondrial proteins were incubated with inhibitors (1 mM EGTA, 5 μM CsA, 40 μM DIDS, 6 μM S15176), then with 25 μM Ca2+ and optical density of the suspension was recorded for 30 minutes. (B) Effect of inhibitors on the transmembrane potential by the measure of the dequenching of Rhod123 fluorescence. *P < 0.05 versus lean mice (n = 5). (C) Expression level of VDAC revealed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting. ATPase β-subunit was used as a loading control (n = 12). (D) Ca2+ accumulation in the matrix of isolated mitochondria determined by the fluorescent probe Rhod2 (n = 3). (E) Ca2+ accumulation in VDAC reconstituted in proteoliposomes determined by the fluorescent probe Fluo4-dextran. VDAC was purified to homogeneity from lean and ob/ob mice liver and reconstituted into liposomes containing Fluo4-dextran. Twenty μM Ca2+ pulses were added onto liposomes and fluorescence was recorded until equilibration. The solid and open bars correspond to the lean mice and ob/ob mice, respectively. Results are means ± SD, *P < 0.05 versus lean mice, n = 3.

Similar levels of VDAC were detected in liver mitochondria extracts from lean and ob/ob mice (Fig. 3C). Nonetheless, isolated ob/ob mitochondria accumulated significantly more Ca2+ than control mitochondria (Fig. 3D). Moreover, control VDAC proteoliposomes accumulated less Ca2+ than proteoliposomes, which contained VDAC purified from ob/ob mice (Fig. 3E; Supporting Fig. 3). Furthermore, the NADH oxidase activity of VDAC was higher in VDAC purified from ob/ob mice and was enhanced in the presence of Ca2+ (Supporting Table 1). Both Ca2+ accumulation and NADH oxidase activity were inhibited by DIDS (Supporting Fig. 4).

Finally, we determined VDAC channel conductance following reconstitution of the pure native protein into planar lipid bilayer. In the absence of Ca2+, VDAC from lean mice exhibited classical hallmarks, i.e., alternation of open (o) and closed (c) states at low potentials with a symmetrical behavior (Fig. 4A). At ±20 mV, the main difference was that VDAC purified from ob/ob mice liver opened permanently (Fig. 4; Supporting Table 2). Moreover, in the presence of 0.5 mM Ca2+, VDAC from lean mice liver behavior remains symmetrical: the amplitude level of the open states and the opening duration increased significantly. In contrast, VDAC from ob/ob mice liver behaved asymmetrically in response to positive and negative potentials. Thus, at −20 mV the channel permanently closed (Fig. 4; Supporting Table 2). This suggests a remarkable change in the gating properties of the channel.

Figure 4.

Differential modulation of VDAC single channel by Ca2+. Single VDAC channels were reconstituted spontaneously into planar lipid bilayers upon addition of purified VDAC from lean (A) and ob/ob (B) mice livers to the cis-compartment containing 1 M KCl, 10 mM HEPES pH 7.4 with or without 0.5 mM Ca2+. The resultant channel activity was recorded for at least 2 minutes at ± 20 mV. For analysis, the data were filtered to 1 kHz, digitized at 4 kHz, and collected on a Pentium computer using Axoxcope10 and a digidata 1440A. Experiments were repeated three times giving similar results. c: closed channel; O: open channel.

VDAC Phosphorylation Status in Steatotic Patients and Experimental Models.

In mammals, VDAC is expressed as three homologous isoforms, VDAC1 to VDAC3, which possess multiple threonine (Thr) residues (Supporting Fig. 5).16 First, we analyzed the level of Thr phosphorylation of purified VDAC from liver of lean and ob/ob mice by immunoblotting with an antibody specific for phosphorylated Thr (P-Thr) and found a unique 34 kDa band comigrating with VDAC (Fig. 5A), consistent with a phosphorylation of VDAC on one or several Thr residues in lean mice and a lack of phosphorylation in ob/ob mice. Second, we analyzed total extracts of human liver biopsies with variable grades of hepatosteatosis and mitochondrial extracts from mice fed a high-fat diet (HFD) confirming the difference of P-Thr phosphorylation between steatotic and lean samples (Fig. 5B,C; Supporting Table 3). Moreover, VDAC phosphorylation status correlated with the severity of the liver pathology (Fig. 5B). Third, we investigated the lack of phosphorylation induced by FA treatment in HHL-5 cells and found that the loss of P-Thr phosphorylation is significant from 6 to 30 hours of FA treatment (Fig. 5D). To demonstrate the identity of VDAC as the 34 kDa band identified by P-Thr immunoblotting, two-dimensional polyacrylamide gel electrophoresis (2D-PAGE) followed by immunoblotting was performed. This led to the identification of eight spots reacting with VDAC-specific antibodies and, some of them, with the antibody for P-Thr (Fig. 5E,F). The identity of VDAC 1 to 3 in these double-labeled spots was confirmed by nanoliquid chromatography and mass spectrometry (Supporting Table 4). Thus, in normal conditions, VDAC is phosphorylated on one or several Thr residues. This phosphorylation is significantly reduced in steatotic samples from patients, in fat accumulating HHL-5 cells, as well as in obese mice. These results reveal the existence of a lipid-induced signaling pathway leading to the lack of phosphorylation of VDAC.

Figure 5.

Threonine phosphorylation of VDAC is reduced in steatosis. Phosphorylation of VDAC (P-Thr) in (A) purified VDAC from lean (three animals) and ob/ob mice (three animals); (B) total liver of steatotic patients classified in grade 0 (three patients), grade 1 (two patients), and grade 2 (three patients); (C) liver of nondiet mice (ND, three animals) and mice fed a high-fat diet (HFD, six animals) for 24 weeks; and (D) fatty acids (FA)-induced steatotic cells (results are representative of n = 3 experiments). Expression and phosphorylation of VDAC were determined by immunoblotting and densitometric analysis (30 μg proteins/lane). Indicated results are means ± SD with *P < 0.05 and **P < 0.01 versus lean or nondiet mice. (E) Identification of VDAC variants by 2D-SDS-PAGE of mitochondrial proteins isolated from lean mice liver, and (F) determination of their threonine phosphorylation status. One representative gel and one immunoblot of three independent experiments are shown. Circles indicate the position of VDAC as revealed by immunoblotting and spectrometric analyses.

VDAC Interactome Changes in Steatosis.

Next, blue native polyacrylamide gel electrophoresis (BN-PAGE) revealed the existence of numerous multiprotein complexes (MC) containing VDAC (Fig. 6A,B). Surprisingly, a complex of 175 kDa (MC175kDa), present in control mitochondria, was totally absent in ob/ob mitochondria (Fig. 6B). MC175kDa contains P-VDAC, the serine/threonine kinase GSK3, the antiapoptotic protein Bcl-XL (Fig. 6C). Glucokinase, ANT, Akt, P-Akt, Bax, Bak, and cyclophilin D, which are putative partners of VDAC,17 were not present in MC175kDa (not shown). Moreover, we observed that GSK3 was similarly associated with both types of mitochondria and mainly in the cytoplasm, whereas the amount of P-GSK3β increased in ob/ob mitochondria as well as in cytoplasm. Bcl-XL was found in the complex in lean mitochondria, whereas in ob/ob mice it was more abundant in the cytosol, suggesting a regulatory flux out of the mitochondria (Fig. 6D). Thus, MC175kDa might contribute to the relative stability of nonsteatotic mitochondrial membranes (Fig. 6E).

Figure 6.

Steatosis modulates VDAC interactome. (A) Mitochondrial proteins from lean mice were separated by 2D BN-PAGE and proteins were stained by silver nitrate. The horizontal insert indicates the position of VDAC-containing complexes and the vertical insert indicates the position of MC175kDa. (B) VDAC position in 2D BN-PAGE gels of proteins from lean and ob/ob mice liver was revealed by immunoblotting with an anti-VDAC mAb. A circle indicates the position of a VDAC-containing complex, named MC175kDa and shows its absence in obese mice liver. (C) Composition of MC175kDa, analyzed by immunoblotting of mitochondrial proteins from lean mice after 2D BN-PAGE. (D) Comparison of GSK3β, P-GSK3β, and Bcl-XL subcellular localization between lean and ob/ob mice after cellular fractionation followed by separation by SDS-PAGE. (E) Scheme of the 175 kDa complex centered on VDAC. In physiological conditions, P-VDAC interacts with BcL-XL and GSK3, protecting the mitochondria from Cyt c release.

GSK3 Phosphorylates VDAC.

Prompted by the fact that GSK3 can phosphorylate VDAC, we assessed the proportion of inactive, phosphorylated GSK3 among total GSK3 protein (P-GSK3/GSK3 ratio) in mitochondrial fraction by immunoblotting. In isolated functional mitochondria from lean and ob/ob livers, P-Thr phosphorylation of VDAC was inversely related to that of GSK3 (Fig. 7A). Moreover, upon addition of FA to HHL-5 cells, P-Thr of VDAC decreased (0.46 fold ± 0.1) and P-GSK3 increased (1.45 fold ± 0.2) (Fig. 7B) as the sensitivity of mitochondria to Ca2+ stimuli increased (Fig. 7C). These effects on VDAC phosphorylation (Fig. 7B) and inner membrane depolarization (Fig. 7C) could be reversed by exposure to wortmannin (Wort), a phosphoinositide 3-kinase (PI3K) inhibitor that stimulates GSK3 kinase activity and decreases GSK3 phosphorylation.18 Indeed, Wort rescues partially VDAC phosphorylation (0.74-fold ± 0.07) from FA treatment (Fig. 7B). In the absence of FA, SB216763, a specific GSK3 inhibitor,19 decreased P-Thr of VDAC (0.65-fold ± 0.1) (Fig. 7B) and increased the sensitivity to Ca2+ (Fig. 7C), mimicking FA accumulation and steatotic condition. To demonstrate that GSK3 is the kinase or, at least, one of the kinases that contribute to VDAC phosphorylation in lean condition, we reduced its expression of 80% by small interfering RNA (siRNA) and observed a 46% decrease in P-Thr VDAC in HHL-5 cells (Fig. 7D). Finally, using recombinant (i.e., VDAC1, GSK3, Bcl-XL) and native purified proteins (i.e., VDAC), we observed in vitro that VDAC and Bcl-XL can be direct substrates for phosphorylation by GSK3β and suggest that no other kinase is needed (Supporting Fig. 5).

Figure 7.

Pharmacological and genetic manipulation of GSK3 phosphorylation modulates VDAC phosphorylation and mitochondrial sensitivity to Ca2+. (A) Phosphorylation of VDAC and GSK3 in isolated mitochondria from lean and ob/ob mice liver. Proteins were immunoblotted for VDAC, GSK3, P-GSK3, and P-Thr and analyzed by densitometry. *P < 0.05 versus lean mice. (n = 4) (B) Effect of 24-hour treatment of HHL-5 cells with 2 mM FA ± 2-hour pretreatment with 20 mM SB216367 or 200 nM wortmannin (Wort) on phosphorylation of VDAC and GSK3. Total proteins were analyzed by immunoblotting and densitometry (means ± SD; *P < 0.05 and **P < 0.01) (n = 4). (C) Effect of kinase inhibitors on the transmembrane potential of HHL-5 (rhodamine 123 fluorescence arbitrary units) cells treated or not with 2 mM FA in response to 25 μM Ca2+ addition at t0 (n = 10) (*P < 0.05 and **P < 0.01) (n = 3). (D) Silencing of GSK3 by a pool of four siRNAs for 48 hours decreases VDAC phosphorylation in HH5 cells as shown by immunoblotting. Scramble siRNA is a negative control. Densitometry (arbitrary units) is given under each band. Results are representative of three independent experiments.


Unraveling the initial molecular mechanisms leading to fatty liver disease in humans is of major clinical importance. Because activation of signaling pathways may precede the clinical symptoms,1 a series of models of liver steatosis were used to study the early stage of NAFLD. This led us to show, for the first time, that liver steatosis in humans is associated with a lack of VDAC phosphorylation. This modification was also observed both in a genetic model of obesity, the ob/ob mice, which is devoid of inflammation and fibrosis, at least in young animals,13 and in the steatotic liver of HFD-fed mice. No change in other phosphorylable amino acids, such as tyrosine or serine, was observed (not shown). These observations suggest that VDAC lack of phosphorylation may represent a hallmark of steatosis in mammals.

In addition to their impaired metabolic function,4 we show for the first time that mitochondria from ob/ob mice are more prone to Ca2+-induced PT, water and small molecules entry, and Cyt c release, suggesting a sensitization of the mitochondrial pathway of apoptosis. Indeed, tumor necrosis factor (TNF)-induced apoptosis and Ca2+ release by endoplasmic reticulum have been shown to be responsible for the massive hepatocyte loss observed in the late stages of NAFLD.20, 21 The absence of apoptosis and caspase activation in ob/ob mice might be explained by the earliness of the model.22 The cellular experiments revealed that intracellular FA accumulation is rapidly followed by VDAC dephosphorylation and ΔΨm loss sensitization. Thus, our current data obtained in ob/ob mice and FA-treated cells offer novel insights into the early phases of NAFLD pathogenesis and early mitochondrial alterations.

Importantly, we found differences in VDAC functions that were exacerbated by Ca2+. Thus, the level of VDAC phosphorylation proved to modulate the ionic channel permeability of VDAC in response to Ca2+, which may influence mitochondrial OM permeability.23, 24 Moreover, the overstimulation of oxidase function of VDAC purified from ob/ob mice by Ca2+ may affect the cellular metabolism by increasing NAD+ levels available either for redox processes, ADP-ribosylation reactions, or sirtuin-mediated deacetylations.25 Thus, VDAC phosphorylation is a sensor of lipotoxicity controlling the balance between adaptive response to protect cells from lipotoxicity and stress response to favor hepatocyte cell death.

Lipid-induced lack of VDAC phosphorylation results in an increase of its hetero-oligomerization state and a change in its interactome. In normal mitochondrial membranes, VDAC was found in an hetero-oligomer of 175 kDa (MC175kDa) that was not detected in the mitochondria of steatotic hepatocytes, suggesting that this complex is involved in the stabilization of mitochondrial OM by way of an interaction with Bcl-XL, which exerts antiapoptotic function in mitochondrial membrane.26 MC175kDa also contains a fraction of GSK3, supporting the hypothesis that mitochondria-associated GSK3 is the kinase involved in the phosphorylation of VDAC in normal conditions, whereas we cannot exclude the association of other kinase or phosphatase. GSK3 activity is inhibited by phosphorylation on serine residues induced by the activation of upstream kinases depending on tissue and pathological contexts.27 In ob/ob mice, total GSK3 has been found to be less phosphorylated,28 but our data indicate that the mitochondrial fraction of GSK3 from ob/ob mice is more phosphorylated. Unexpectedly, Bcl-XL and GSK3 were also enriched in the cytoplasm of ob/ob mice. Based on the evidence that Bcl-XL could be a direct target for GSK3 in vitro, it is tempting to speculate a role of GSK3 phosphorylation of Bcl-XL in steatosis. All the pharmacological or genetic manipulations designed to directly or indirectly target GSK3 converged to propose that VDAC is phosphorylated on a Thr residue by, at least, this kinase. This is reminiscent of previous studies in ischemia/reperfusion of cardiomyocytes29 or cancer cells30 showing the modification of VDAC phosphorylation by GSK3 on Thr51, which would block VDAC interaction with hexokinase and favor tubulin interaction in vitro.31 Therefore, our functional data are clearly consistent with a tissue-specific role for GSK3 in protecting hepatocytes against lipotoxicity in steatosis. Moreover, GSK3 is a pleitropic kinase involved in metabolism by way of its glycogen synthase activity regulation, in insulin signaling pathway,32 and in insulin resistance mechanisms.33 Irrespective of the molecular mechanisms linking FA accumulation and GSK3 phosphorylation, mitochondria-associated GSK3 may link metabolism and MMP by way of its physical and functional interaction with VDAC.

Based on the hypothesis that VDAC is Thr phosphorylated by GSK3 and the solution structure of VDAC,16 we speculate that at least some phosphorylated Thr residues are exposed to the cytosol. However, because VDAC is encoded by three distinct isoforms with multiple threonines (from 26 to 31 Thr per isoform), the residues critical for the lipid-related regulation of VDAC remain still elusive.

In conclusion, we identified a novel phosphoregulated mitochondrial complex, MC175kDa, that is present in normal mitochondria, yet absent from steatotic mitochondria and involved in mitochondrial Ca2+ and water homeostasis of healthy hepatocytes. Thus, the activation of GSK3 (or the inhibition of the corresponding phosphatase(s)) and/or the pharmacological stabilization of VDAC phosphorylation might constitute a strategy to limit mitochondrial damage and tissue injury in obesity-linked liver pathologies.


We thank C. Longin from the microscopy and imagery platform of INRA, Dr. A. Patel, University of Glasgow, UK, for the generous gift of HHL-5 hepatocyte cell line, S. Campagna and N. Saint for the electrophysiological measurements, and C. Gallerne and E. Maillier for technical assistance. BCl-XL was a gift from Alexandre Chenal (Institut Pasteur, Paris) and Christine Almunia (CEA Marcoule, direction des sciences du vivant).