Introduction

Mitochondria and endoplasmic reticulum (ER) are organised as a network with specific contact points, referred to as mitochondrial-associated ER membranes (MAMs), which play a pivotal role in calcium (Ca2+) signalling and energy metabolism [1]. Efficient Ca2+ transmission from the ER to mitochondria is mediated through the interaction of the voltage-dependent anion channel (VDAC) of the outer mitochondrial membrane with the inositol-1,4,5-triphosphate receptor (IP3R) on the ER via the chaperone glucose-regulated protein 75 (GRP75) [2]. Mitochondrial Ca2+ uptake is essential for the regulation of both mitochondrial metabolism and ER homeostasis [3], and alteration of ER–mitochondria cross-talk may result in a disruption of inter-organelle Ca2+ transfer [4] and subsequent ER stress [5]. Under certain conditions excessive Ca2+ entry into the mitochondrial matrix may be detrimental, causing the opening of the permeability transition pore (PTP) and cell death [6]. Therefore, tight control of Ca2+ exchange between ER and mitochondria is required to regulate vital functions and metabolic homeostasis.

The molecular and functional characterisation of MAMs in physiological and pathological conditions has improved in the last few years, highlighting unexpected roles for MAMs in cellular signalling [7]. Particularly, we recently demonstrated that cyclophilin D (CYPD), a mitochondrial protein known to modulate the opening of the PTP [8], also interacts with the VDAC–GRP75–IP3R complex at the MAM interface in both heart [9] and liver [10]. In cardiomyocytes, we found that the loss of CYPD reduced mitochondrial Ca2+ overload by depressing ER–mitochondria interactions and protected cells against lethal reperfusion injury [9], suggesting that CYPD regulates Ca2+ transfer from ER to mitochondria. In liver, the loss of CYPD reduced organelle interactions and induced hepatic insulin resistance, pointing to a new role of MAM integrity in the control of insulin’s action [10]. Whereas other studies in mice also suggest a role for MAMs in the control of glucose homeostasis [11–14], the mechanisms by which disruption of MAMs alters insulin signalling are unknown.

Based on the role of MAMs in Ca2+ transfer from ER to mitochondria, we hypothesised that a disruption of Ca2+ transfer between both organelles could contribute to hepatic insulin resistance. To this aim, using pharmacological and genetic loss of function approaches both in vivo and in vitro, we investigated whether the disruption of Ca2+ transfer from ER to mitochondria could link MAM alterations to hepatic insulin resistance in CypD (also known as Ppif)-knockout (KO) mice.

Methods

Cell culture

HuH7 cells were a gift from G. Mithieux’s laboratory (Inserm U855, Lyon, France) and were free of mycoplasma. They were cultured as previously described [10] and stimulated either with CYPD (ciclosporin [cyclosporin A, CsA] or NIM811) or with IP3R (2-APB and Xestospongin C) pharmacological inhibitors. Inhibition of CYPD was also achieved by silencing of CYPD. See electronic supplementary material (ESM) Methods for further details. For measurement of insulin signalling, cells were depleted in serum for 3 h before incubation with insulin (10−7 mol/l, 15 min).

Animals

CYPD-KO mice on a C57Bl/6/SV129 genetic background were a gift from S. J. Korsmeyer’s laboratory (Boston, MA, USA) [15]. Both male wild-type (WT) and CYPD-KO mice were obtained by homozygous intercross in our laboratory. All experiments were performed on mice of 18–22 weeks of age, and were conducted in accordance with institutional guidelines for the care and use of laboratory animals, and a regional ethics committee has approved all procedures. No randomisation or blinding were performed. No data samples or animals were excluded from the study.

Ca2+ measurements

Using confocal fluorescence imaging, we evaluated variations in the time course of mitochondrial Ca2+ content simultaneously with intra-ER Ca2+ in a native cell environment in the presence of extracellular calcium (1.8 mmol/l CaCl2) [16]. To measure mitochondrial Ca2+, cells were loaded with Rhod-2 AM (3 μmol/l) whereas the measurement of ER Ca2+ was performed using a low-affinity Ca2+ indicator, Fluo-5N (5 μmol/l; ThermoFisher Scientific, MA, USA). For measurement of absolute cytosolic Ca2+ levels, cells were loaded with 1 μmol/l of Fura2-AM and pluronic acid (both from ThermoFisher Scientific). See ESM Methods for further details.

ER–mitochondria interactions

ER–mitochondria interactions were measured both by subcellular fractionation and by in situ proximity ligation assay (PLA), as previously described and thoroughly validated [10]. See ESM Methods for further details.

Real-time PCR

mRNA levels were measured by real-time RT-PCR. See ESM Methods for further details.

Western blotting

Protein expression was analysed by SDS-PAGE. See ESM Methods for further details.

Hyperinsulinaemic–euglycaemic clamp

Insulin sensitivity of mice was measured during a hyperinsulinaemic–euglycaemic clamp. See ESM Methods for further details.

Primary hepatocytes

Primary mouse hepatocytes were isolated via a modified collagenase perfusion method, as described previously [17]. See ESM Methods for further details.

Mitochondrial respiration

Mitochondrial respiration was measured in intact or permeabilised primary hepatocytes. See ESM Methods for further details.

Hepatic lipid content

Hepatic triacylglycerol (TG) and diacylglycerol (DAG) content were measured by spectrophotometry. See ESM Methods for further details.

PKCε activity

PKCε activity in liver was determined by western blot based on the membrane translocation. See ESM Methods for further details.

Statistical analysis

Results are expressed as mean ± SEM. Student’s t-test was used to analyse the difference between control and experimental groups. Statistically significant differences were assessed with a one-way ANOVA with a Newman–Keuls post hoc test when three or more groups were compared. A p value <0.05 was considered as statistically significant.

Results

Pharmacological and genetic inhibition of CYPD inhibits Ca2+ transfer from ER to mitochondria in HuH7 cells

We recently demonstrated by in situ PLA that CYPD interacted with the VDAC1–GRP75–IP3R1 calcium-channelling complex [9, 10]. Here, we challenged the interactions of CYPD with this complex using pharmacological and genetic loss of function studies, and investigated the repercussions on both ER–mitochondria interactions and Ca2+ transfer. As CYPD inhibitors we used CsA and NIM811 (a CsA derivative devoid of immunosuppressive activity), which are both known to detach CYPD from the inner mitochondrial membrane [18]. Treatment of HuH7 cells with CsA or NIM811 significantly inhibited CYPD–IP3R1 interactions (Fig. 1a). Reduction of CYPD expression using specific small interfering RNA (siRNA) (Fig. 1b) also significantly decreased the interactions between CYPD and IP3R1 (Fig. 1c).

Fig. 1
figure 1

Pharmacological and genetic inhibition of CYPD alters its interaction with the VDAC1–GRP75–IP3R1 complex. (a, c) Representative images (scale bar, 20 μm) and quantitative analysis of CYPD–IP3R1 interactions measured by in situ PLA in HuH7 cells, following treatment with either CsA or NIM811 (2 μmol/l, 16 h) (a), or silencing of CYPD (25 nmol/l, 48 h) (c). Nuclei appear in blue and PLA-specific signals in red. Magnification ×63. *p < 0.05 vs control, n = 3. Co, control. (b) Measurement of CYPD mRNA levels by real-time PCR in HuH7 cells silenced for CYPD. **p < 0.001, n = 3

Full size image

Using organelle-targeted fluorescent dye, we then measured Ca2+ flux in HuH7 cells under histamine stimulation (100 μmol/l). Histamine binding to its receptor induces inositol-1,4,5-triphosphate elevation and activates IP3R, causing Ca2+ release from ER stores. Specificity and non-overlapping of the Ca2+-sensitive dyes is illustrated in ESM Fig. 1. Histamine application rapidly reduced ER Ca2+ stores (Fig. 2a) and simultaneously increased mitochondrial Ca2+ content (Fig. 2b), illustrating transfer of Ca2+ from the ER to the mitochondria. Importantly, after NIM811 treatment or CYPD silencing, histamine was unable to induce Ca2+ transfer from ER to mitochondria, as illustrated by the absence of mitochondrial Ca2+ accumulation in both situations (Fig. 2a, b).

Fig. 2
figure 2

Pharmacological and genetic inhibition of CYPD alters IP3R-mediated Ca2+ transfer between ER and mitochondria in HuH7 cells. (a, b) Measurement by confocal imaging of Ca2+ flux into ER (a) and mitochondria (b) in histamine-stimulated HuH7 cells, in control situation (Co, black curve), after NIM811 treatment (2 μmol/l, 16 h, red curve) or after siRNA-mediated CYPD silencing (25 nmol/l, 48 h, blue curve). Curves represent the time course of Ca2+ exchange between ER and mitochondria for 500 s after histamine stimulation (100 μmol/l), in the presence of extracellular Ca2+. Mean Ca2+ flux was normalised to fluorescence values prior to histamine application. *p < 0.05 vs control, n = 18

Full size image

Inhibition of CYPD function induces ER stress and alters insulin response in HuH7 cells

Next, we measured the consequences of CYPD inhibition on both ER and mitochondrial homeostasis and on insulin signalling in HuH7 cells. NIM811 treatment significantly increased mRNA levels of GRP78, XBP1S and CHOP (also known as DDIT3) (Fig. 3a), as well as the protein level of GRP78 and the phosphorylation of eukaryotic translation initiation factor 2α (EIF2α) (Fig. 3b), indicative of ER stress. However, this treatment had no effect on the mRNA levels of HSP10 and HSP60 (Fig. 3a), two markers of mitochondrial stress that are increased during the mitochondrial unfolded protein response [19]. Inhibition of CYPD using CsA reproduced the same effects (ESM Fig. 2a, b). Finally, the partial invalidation of CYPD expression by specific siRNA also induced ER stress markers (Fig. 3c).

Fig. 3
figure 3

Pharmacological and genetic inhibition of CYPD induces ER stress and alters insulin signalling in HuH7 cells. (a–c) Analysis of both mRNA (a) and protein levels (b, c) of ER stress markers in HuH7 cells following 6 (grey bars) or 16 (black bars) h of NIM811 treatment (a, b) or silencing of CYPD (c, black bars). UPR, unfolded protein response; MT mitochondria. (d, e) Analysis of PKB phosphorylation in basal situation (white bars) or after insulin stimulation (black bars) in HuH7 cells following NIM811 treatment (d) or CYPD silencing (e). Co, control. *p < 0.05 and **p < 0.01 vs Co; † p < 0.05 vs vehicle, n = 3

Full size image

NIM811 treatment also significantly reduced insulin-stimulated protein kinase B (PKB) phosphorylation (Fig. 3d) and induced PEPCK mRNA expression (ESM Fig. 2c) in HuH7 cells. Inhibition of CYPD using CsA reproduced all these effects (ESM Fig. 2d, e). In addition, the partial invalidation of CYPD expression by siRNA also altered insulin signalling (Fig. 3e) and induced PEPCK expression (ESM Fig. 2f). To strengthen the link between IP3R-mediated Ca2+ signalling, ER homeostasis and insulin signalling, we investigated whether pharmacological inhibition of IP3R could alter insulin signalling independently of CYPD. Both 2-APB and Xestospongin C treatments, which antagonise the calcium-releasing action of inositol-1,4,5-trisphosphate at the receptor level, significantly induced ER stress (Fig. 4a, b) and reduced insulin-stimulated PKB phosphorylation (Fig. 4c, d) in HuH7 cells.

Fig. 4
figure 4

Pharmacological inhibition of IP3R induces ER stress and alters insulin signalling in HuH7 cells. (a–d) Representative western blots and quantitative analysis of both ER stress markers (a, b) and insulin-stimulated PKB phosphorylation (c, d) in HuH7 cells treated (black bars) or not (white bars) with 2-APB (a and c, 50 μmol/l, 18 h) or Xestospongin (b and d, 1 μmol/l, 18 h). Co, control. *p < 0.05, **p < 0.01 and ***p < 0.005, n = 3–6

Full size image

Loss of CYPD in mice alters Ca2+ transfer from ER to mitochondria in isolated hepatocytes

We previously found that ER–mitochondria interactions were reduced in liver of CypD-KO mice [10]. We confirmed in this study our initial observation using an independent group of mice (Fig. 5a) and further analysed the composition of MAM fractions of WT and CypD-KO mice. We found no significant modification of IP3R1, VDAC1, GRP75, mitofusin 2 (MFN2) or sarco(endo)plasmic reticulum Ca2+ ATPase 2 (SERCA2) protein level in MAM fractions of CypD-KO mice compared with WT mice (Fig. 5b). This suggests that loss of CYPD induces a dissociation of organelles rather than a change in MAM protein composition. We then examined inter-organelle Ca2+ flux in isolated hepatocytes of CypD-KO mice, following the same experimental protocol performed in HuH7 cells. In WT hepatocytes, histamine induced Ca2+ release from the ER, immediately followed by mitochondrial Ca2+ uptake (Fig. 5c, d). This histamine-stimulated Ca2+ transfer from the ER to mitochondria was completely abolished in CypD-KO hepatocytes (Fig. 5c, d). This occurs in the absence of a modification of Serca2b (also known as Atp2a2) and Mcu mRNA levels in liver of CypD-KO mice (ESM Fig. 3a). Together, these results confirm in a more physiological model that CYPD participates in IP3-mediated ER-to-mitochondria Ca2+ transfer in hepatocytes.

Fig. 5
figure 5

Loss of CYPD alters IP3R-mediated Ca2+ transfer from ER to mitochondria in mouse hepatocytes. (a) Levels of MAM in liver of WT and CypD-KO mice, estimated by subcellular fractionation. *p < 0.05 vs WT (n = 4). (b) Representative western blots and quantitative analysis of proteins in hepatic MAM fractions of WT and CypD-KO mice (n = 4). White bars, WT mice; black bars, CypD-KO mice. (c, d) Measurement of Ca2+ flux into ER (c) and mitochondria (d) in histamine-stimulated primary hepatocytes from WT (black curves) and CypD-KO (red curves) mice. Experiments and presentation of the results are as described in Fig. 2. *p < 0.05 for CypD-KO vs WT (n = 12). (e, f) Representative curves (e) and quantitative analysis (f) of cytosolic Ca2+ (Fura2), following depletion of ER Ca2+ storage by thapsigargin (Thapsi.; 10 μmol/l), in the presence (1.8 mmol/l CaCl2) or absence (5 mmol/l EGTA) of extracellular Ca2+ in WT (black curve and white bars) and CypD-KO (grey curve and grey bars) mice. *p < 0.01 and **p < 0.001 for indicated comparisons (n = 16–25)

Full size image

We also measured cytoplasmic Ca2+ in response to a discharge of ER store by thapsigargin, both in the presence and absence of extracellular Ca2+. In both conditions, thapsigargin-sensitive Ca2+ stores were significantly higher in CypD-KO mice compared with WT mice (Fig. 5e, f), indicating that the reduced Ca2+ transfer from ER to mitochondria in CypD-KO hepatocytes was not due to a defect of ER Ca2+ storage. In addition, resting cytosolic Ca2+ was significantly increased in CypD-KO hepatocytes in the presence of extracellular Ca2+, whereas this effect was absent in the absence of extracellular Ca2+ (ESM Fig. 3b, c). The cytosolic Ca2+ peak after histamine stimulation did not differ significantly between WT and CypD-KO mice, in either the presence or absence of extracellular Ca2+ (ESM Fig. 3d).

Loss of CYPD in mice induces ER stress and mitochondrial dysfunction in liver

We then investigated ER or mitochondrial stress markers in the liver of WT and CypD-KO mice. We found increased mRNA levels of Grp78 and Chop in the liver of CypD-KO mice, without any change in Hsp10 and Hsp60 expression (Fig. 6a). In addition, the phosphorylation of PERK, EIF2α and of JNK (a serine/threonine kinase involved in ER stress-induced hepatic insulin resistance [20]) was increased in liver of CypD-KO mice (Fig. 6b), confirming hepatic ER stress.

Fig. 6
figure 6

Loss of CYPD induces hepatic ER stress and mitochondrial dysfunction in mice. (a, b) mRNA (a) and protein (b) levels of ER and mitochondria stress markers in liver of WT and CypD-KO mice. *p < 0.05 and **p < 0.01 vs WT mice (n = 3–10). UPR, unfolded protein response; MT mitochondria; tub, tubulin. (c, d) Oxygen consumption measured in either permeabilised primary hepatocytes (c, in response to 5 mmol/l glutamate + 25 mmol/l malate stimulation in both states 3 [+1 mmol/l ADP] and 4 [+ oligomycin]) or in intact hepatocytes (d, in the presence of 20 mmol/l glucose and after addition of oligomycin) from WT and CypD-KO mice. White bars, WT mice; black bars, CypD-KO mice. *p < 0.05 vs WT (n = 4)

Full size image

As Ca2+ import into mitochondria affects mitochondrial bioenergetics [21, 22], we studied mitochondrial respiration in either permeabilised (glutamate/malate) or intact (20 mmol/l glucose) primary hepatocytes of both WT and CypD-KO mice, in order to maintain ER mitochondria cross-talk. As shown in Fig. 6c, d, oxygen consumption was significantly reduced in CypD-KO hepatocytes, compared with WT hepatocytes.

Loss of CYPD in mice increases lipid accumulation and PKC activity and induces hepatic insulin resistance

We previously reported that CypD-KO mice are glucose intolerant, insulin resistant and showed increased gluconeogenesis, based on tolerance tests [10]. Here, we further performed hyperinsulinaemic–euglycaemic clamp to confirm hepatic insulin resistance. The glucose infusion rate required to maintain euglycaemia was significantly lower in CypD-KO mice compared with that in WT mice (Fig. 7a). In addition, the suppression of hepatic glucose production was significantly reduced in CypD-KO mice (Fig. 7b), whereas peripheral glucose utilisation was unaltered (Fig. 7a), indicating a specific state of hepatic insulin resistance. In agreement, the expression levels of gluconeogenic enzymes were increased in liver of CypD-KO mice (Fig. 7c).

Fig. 7
figure 7

Loss of CYPD induces hepatic insulin resistance and alterations of lipid homeostasis in mice. (a, b) Glucose infusion rate (GIR) and glucose uptake (a), as well as suppression of hepatic glucose production (HGP) (b) was measured under a hyperinsulinaemic–euglycaemic clamp in WT and CypD-KO mice (n = 4). (c) Expression of gluconeogenic genes G6p, Pepck and Pgc1α (also known as Ppargc1a) in the liver of WT and CypD-KO mice (n = 10). (d) Liver sections were stained with Oil Red O. Original magnification, ×20. (e) Hepatic TG and DAG content in liver of WT and CypD-KO mice (n = 6). (f) Measurement of PKCε activation in cytosolic and membrane fractions of liver of WT and CypD-KO mice (n = 3). (g) mRNA levels of genes related to β-oxidation, lipogenesis and TG export in liver of WT and CypD-KO mice (n = 10). White bars, WT mice; black bars, CypD-KO mice. *p < 0.05, **p < 0.01 and ***p < 0.001 vs WT

Full size image

We also examined lipid metabolism in both WT and CypD-KO mice. Circulating NEFA were not modified in CypD-KO mice compared with WT mice (0.19 ± 0.05 vs 0.16 ± 0.05 mmol/l, respectively, n = 10), whereas blood TG levels were significantly increased (0.90 ± 0.03 vs 0.60 ± 0.05 g/l, respectively, p < 0.05, n = 10). Interestingly, the lipid staining (Fig. 7d), as well as TG and total DAG levels (Fig. 7e), were significantly increased in the liver of CypD-KO mice. As PKCε was shown to be involved in DAG-mediated hepatic insulin resistance [23], we analysed its activity by measuring its translocation from cytosol to membrane. Consistent with DAG accumulation, PKCε activity was induced in the livers of CypD-KO mice (Fig. 7f). Furthermore, the mRNA levels of genes related to β-oxidation (Cpt1a) were decreased in liver of CypD-KO mice, whereas the expression of both lipogenic genes (Srebp1c [also known as Srebf1], Srebp1a, Fasn, Acaca) and genes involved in lipid export (Apob, Mttp) were increased (Fig. 7g). The expression of neither Mlycd nor Dgat2 was modified in liver of CypD-KO mice compared with WT mice. Altogether, these results suggest that lipid accumulation in liver of CypD-KO mice is likely related to both a reduction in lipid oxidation and an increase in lipid storage, the latter being mainly associated with increased de novo lipogenesis.

Role of PKCε in CYPD-related alteration of insulin signalling

Both JNK and PKCε enzymes, which are increased in liver of CypD-KO mice, are potential mediators of hepatic insulin resistance [20, 23]. To discriminate between the two, we measured the effect of CYPD silencing on insulin-stimulated PKB phosphorylation in the presence or absence of JNK and PKCε inhibitors. Inhibition of JNK with SP600125 did not modify CYPD siRNA-induced alteration of insulin-stimulated PKB phosphorylation (Fig. 8a), whereas inhibition of PKCε activity, using a specific peptide inhibitor of PKCε, prevented the reduction of insulin-stimulated PKB phosphorylation induced by CYPD silencing in HuH7 cells (Fig. 8b). Together, these data suggest a predominant role for PKCε, rather than JNK, in CYPD-related alterations of insulin signalling, at least in vitro.

Fig. 8
figure 8

Involvement of PKCε, but not of JNK, in CYPD siRNA-mediated alteration of insulin signalling in HuH7 cells. (a, b) Representative western blot and quantitative analysis of basal (white bars) and insulin-stimulated (black bars) PKB phosphorylation in HuH7 cells silenced for CYPD, in the presence or absence of an inhibitor of JNK (SP600125, 10 μmol/l) (a) or PKCε (specific peptide inhibitor of PKCε, 10 μmol/l) (b). For western blots shown in (b), four parts of the same gel are shown. *p < 0.05 and **p < 0.01 vs basal; † p < 0.05 vs respective Co siRNA (n = 3). (c) Schematic representation of the effects of loss/inhibition of CYPD on hepatic insulin resistance. Red, in vivo and ex vivo observations; blue, in vitro observations in HuH7 cells

Full size image

Discussion

Hepatic insulin resistance is a principal component of type 2 diabetes, but the cellular and molecular mechanisms responsible for its pathogenesis are only partly known. We recently proposed a role for MAM integrity in the control of hepatic insulin action and demonstrated that disruption of MAM integrity in the liver of CypD-KO mice can induce hepatic insulin resistance [10]. In the present study, we further investigated the underlying mechanisms and showed that the disruption of Ca2+ transfer from ER to mitochondria in the liver of CypD-KO mice is likely the molecular consequence of MAM disruption, accounting for defective insulin action in the liver.

ER–mitochondria contact points are known to be enriched in Ca2+-handling proteins and chaperones and to generate microdomains with a high Ca2+ concentration [3]. We provide in vivo and in vitro evidence that pharmacological or genetic inhibition of the mitochondrial chaperone CYPD results in modifications of ER–mitochondria Ca2+ transfer: (1) via CYPD binding to the mitochondrial inner membrane since its actions are inhibited by CsA and NIM811 and (2) via an interaction with the VDAC1–GRP75–IP3R1 Ca2+-channelling complex. These data are consistent with our recent study demonstrating that inhibition of CYPD in cardiomyocytes decreased the Ca2+ transfer from ER to mitochondria through IP3R under normoxic conditions [9]. Furthermore, CYPD-mediated disruption of ER–mitochondria cross-talk seems related rather to a dissociation of organelle than to a modification of protein expression at MAM interface, and seems independent of a modification of both Mcu and Serca2b mRNA levels. This new physiological role of CYPD in ER–mitochondria communication should now be considered as clinically relevant because inhibitors of CYPD function are either already used in clinical practice (e.g. CsA in transplanted patients) or are being developed (e.g. for treatment of hepatitis C virus or HIV infections) [24].

Our findings also point to a fundamental mechanism by which reduced structural and functional ER–mitochondria interactions alter organelle function and subsequently inhibit insulin’s action, leading to hepatic insulin resistance. We thus propose a model by which the loss of CYPD can induce liver insulin resistance (Fig. 8c). Invalidation of CYPD alters ER–mitochondria interactions and Ca2+ exchange, leading to increased ER Ca2+ storage and ER stress. It is likely that the loss of CYPD-mediated uncoupling of the ER from the mitochondria is compensated by increased levels of the ER protein-folding machinery and Ca2+ to re-establish ER homeostasis, as previously reported [4, 5, 21]. At the same time, lack of Ca2+ transfer to mitochondria in CypD-KO mouse hepatocytes can reduce mitochondrial respiration, as also seen in a recent study demonstrating that constitutive low-level IP3R-mediated Ca2+ delivery to mitochondria is essential to maintain normal cellular bioenergetics [22]. The chronic disturbance of Ca2+ homeostasis likely maintains the activation of unfolded protein response in liver of CypD-KO mice, despite increased GRP78 expression. Subsequently, both ER stress and mitochondrial dysfunction may contribute to an increase in hepatic lipid levels. Indeed, fat accumulation in the liver of CypD-KO mice seems to be related to both a reduction in lipid oxidation (based on reduction in Cpt1 expression) and an induction of de novo lipogenesis (based on induction of lipogenic genes), whereas esterification of lipid (based on the absence of change in circulating NEFA levels) and export of lipids (based on induction of ApoB and Mttp genes) from liver should not be altered. The induction of de novo lipogenesis is in agreement with the activation of both PERK and IRE1 branches of the unfolded protein response in the liver of CypD-KO mice, as both pathways were shown to activate the lipogenic transcription factor SREBP-1c [25–27]. Particularly, accumulation of DAG likely contributes to PKCε activation in the liver of CypD-KO mice. Together, these metabolic stresses would consequently result in alterations of hepatic insulin signalling and in the deterioration of glucose homeostasis in CypD-KO mice. Indeed, both induction of ER stress and accumulation of intracellular lipids have been involved in hepatic insulin resistance. Activation of JNK has been shown in ER stress-mediated hepatic insulin resistance [20] and activation of PKCε has been related to DAG-induced hepatic insulin resistance [23]. In CypD-KO mice, hepatic insulin resistance appears to be mainly secondary to ER stress modulation of hepatic lipogenesis and subsequent DAG-mediated activation of PKCε rather than to ER stress-induced JNK activation, at least in vitro. Whereas an unspecific effect of inhibitors could not be excluded, this result fits well with the mechanisms recently proposed for another mouse model of insulin resistance [28], as well as with recent data in human liver [23]. Nevertheless, we cannot exclude the possibility that other players could be involved in the insulin-resistant phenotype of CypD-KO mice. Particularly, as absolute cytosolic Ca2+ levels were increased in CypD-KO mouse hepatocytes, we cannot exclude the participation of Ca2+-sensitive kinases and/or phosphatases in the alteration of insulin signalling, as suggested by others studies [29, 30]. As disruption of MAM integrity has been observed in liver of different mouse models of obesity and type 2 diabetes [10], it is likely that the unexpected role of inter-organelle Ca2+ exchange in triggering hepatic insulin resistance might be extrapolated to these models. In agreement, the direct inhibition of IP3R, independently of CYPD, also induced ER stress and altered hepatic insulin signalling, suggesting that the proposed mechanism is a more general model and not restricted to the CypD-KO mouse model. Nevertheless, as controversy exists in this topic [31], further studies are required to clarify the role of inter-organelle Ca2+ exchange in triggering hepatic insulin resistance in obesity.

In conclusion, our data demonstrate that CYPD is an important regulator of MAM integrity and subsequently of Ca2+ exchange at the MAM interface, and provide the first evidence that IP3R-mediated Ca2+ transfer from ER to mitochondria is an essential cellular process involved in the control of hepatic insulin action. Therefore, modulation of ER–mitochondria Ca2+ exchange may provide an exciting new avenue for treating hepatic insulin resistance.