Disruption of Mitochondria-Associated Endoplasmic Reticulum Membrane (MAM) Integrity Contributes to Muscle Insulin Resistance in Mice and Humans

Whereas mitochondria and endoplasmic reticulum (ER) dysfunction were largely shown to contribute independently to insulin resistance (1,2), it has been now demonstrated that alterations in their physical interactions are also involved (3). Indeed, mitochondria and ER interact at contact points, called mitochondria-associated ER membranes (MAMs), to exchange calcium (Ca²⁺) and lipids, thus regulating cell metabolism and fate (4,5). A few years ago, we indentified that MAM integrity controlled hepatic insulin signaling and that disruption of organelle coupling contributed to hepatic insulin resistance (6). This was recently confirmed by an independent group (7). Accordingly, mice models with liver-specific invalidation of proteins located at MAM interface, such as mitofusin 2 (MFN2) (8) and mammalian target of rapamycin complex 2 (mTORC2) (9), as well as mutation of inositol 1,4,5-triphosphate receptor 1 (IP3R1) (10), display hyperglycemia, glucose intolerance, and increased neoglucogenesis. Alterations of mitochondrial Ca²⁺ uptake from the ER can also disrupt insulin signaling in cardiomyocytes (11). The literature is still equivocal regarding the exact role of MAMs in hepatic insulin resistance, however, because altered hepatic insulin sensitivity can be associated with enrichment of ER-mitochondria contacts (12) and reduced IP3R levels can be associated with decreased glucose production by the liver (13).

Another tissue crucial for glucose homeostasis is the skeletal muscle. This tissue is indeed the primary site of insulin-stimulated glucose uptake and, therefore, the main target for alterations in insulin resistant states (14). Interestingly, interactions of mitochondria with the sarcoplasmic-endoplasmic reticulum (SR/ER) have been demonstrated in skeletal muscle, although the molecular components of the tethers remain to be characterized (15). Furthermore, a local transfer of Ca²⁺ from SR/ER to mitochondria was illustrated upon caffein stimulation in skeletal muscle (16), which seems important for the control of oxidative metabolism (17). Accordingly, insulin-dependent Ca²⁺ mobilization participates to the translocation of the glucose transporter GLUT4 to the cell surface and to glucose uptake (18), suggesting a potential role of organelle coupling in muscle insulin action.

However, the relevance of MAMs in skeletal muscle insulin resistance has not been investigated up to now; thus, the current study was conducted to gain more insight into this important issue. We show here that SR/ER-mitochondria interactions in skeletal muscle are altered in insulin-resistant states in both mice and humans and that the modulation of organelle coupling regulates insulin signaling and action in human myotubes.

Animal studies were performed in accordance with the French guidelines for the care and use of animals and were approved by an ethics committee of the Rhône-Alpes region (France). Male C57BL/6J and ob/ob mice (12 weeks old) were purchased from Harlan and adapted to the facility for 1 week before study. Male C57Bl/6J mice (5 weeks old) were fed a standard diet (SD) or a high-fat and high-sucrose diet (HFHSD), prepared by the Unit of Experimental Food Preparation in Jouy-en-Josas, France, for 16 weeks, as previously described (6,19). Metabolic characteristics of these mice were reported previously (6).

For this study, gastrocnemius muscles were removed and freshly used for MAM purification or ex vivo insulin-signaling studies (±10⁻⁷ mol/L insulin, 15 min). An independent group of mice (5 weeks old) was fed the SD or HFHSD for 1, 4, 8, 12, or 16 weeks (n = 3 mice/group), and their gastrocnemius muscles were fixed in formaldehyde for paraffin inclusion or frozen. Finally, 12-week-old male C57BL/6J mice were used to infect gastrocnemius muscles (intramuscular injection, 10⁹ infection forming units/muscle) with recombinant adenoviruses encoding for green fluorescent protein (Ad-GFP) or FATE1 (Ad-FATE1) proteins for 7 days.

Isolation of MAM fractions was performed by differential ultracentrifugation as previously described (6).

Fixation and posttreatment of muscle fractions or myotubes and the observation of ultrathin sections were performed as previously described (6). For analysis of SR/ER-mitochondria contacts in human myotubes, we delimitated both organelles using ImageJ, and the fraction of mitochondria in contact with SR/ER within a 50-nm range was calculated and normalized to the mitochondria perimeter.

Primary culture of human myotubes was initiated from satellite cells of vastus lateralis muscle biopsy tissue samples obtained from donors during surgical procedures at Edourd Herriot Hospital (Lyon, France). Myotubes from seven donors (one man and six women; age: 68.8 ± 5.3 years old; BMI: 25.3 ± 0.7 kg/m²) were used for palmitate/oligomycin treatments. Myotubes from 10 lean subjects, 15 obese subjects without diabetes, and 12 obese patients with type 2 diabetes (T2D) were used to investigate whether organelle interactions are altered during obesity and T2D (Table 1). Patients with T2D were being treated with an oral antidiabetic drug or with insulin, or both, and did not interrupt their treatment before the operation. All participants gave their written consent after being informed of the nature and purpose of the study. The experimental protocol (DIOMEDE, agreement number 2012-111/A13-06) was approved by the Ethical Committees Sud-EST IV and performed according to the French legislation.

The myoblasts were purified, and differentiated myotubes were prepared according to the procedure previously described (20). Human myotubes were incubated with palmitate (500 µmol/L) for 24 h to induce insulin resistance (21) or treated with oligomycin (1 and 10 µmol/L, 16 h) to alter mitochondria function. For the analysis of insulin signaling, myotubes were depleted for 3 h in serum and further incubated with or without 10⁻⁷ mol/L of insulin for 15 min in serum-free medium.

Human myotubes were transfected in six-well plates, for 48 h, with 2 μg expression plasmids or small interfering (si)RNA (25 or 50 nmol/L; Qiagen) for specific targets, using ExGen 500 transfection reagent (Roche Diagnostics) or High Perfect transfection reagent (Qiagen) respectively, as previously reported (6).

Expression vectors pcDNA4/TO and pcDNA4/FATE1 (fetal and adult testis expressed 1) were a gift from Enzo Lalli (CNRS-UMR7275, Valbonne, France) (22). FATE1 was subcloned into pcDNA-internal ribosome entry site (IRES)-GFP vector, obtained as previously described (23), to obtain pcDNA-FATE1-IRES-GFP vector, allowing the simultaneous expression of both FATE1 and GFP proteins into cells under the control of the same promoter. Recombinant Ad-GFP (as a control) or Ad-FATE1 were generated by homologous recombination in the VmAdcDNA3 plasmid and amplified and purified as previously described (24,25).

Protein expression was analyzed by SDS-PAGE, and mRNA levels were measured by real-time RT-PCR. To compare intergel analysis of insulin-stimulated protein kinase B (PKB) phosphorylation in human myotubes from lean subjects, obese subjects, and subjects with T2D, a pooled sample of human myotubes was used as an internal standard on each SDS-PAGE gels and used to normalize protein expression. For that, the intensity (in pixel) of each targeted protein (i.e., phosphorylated PKB or total PKB) is normalized by the intensity of the internal standard before determining the ratio of phosphorylated PKB to total PKB.

Voltage-dependant anion channel 1 (VDAC1) and IP3R1 proximity were measured by in situ proximity ligation assay (PLA) (Olink Bioscience) to detect and quantify ER-mitochondria interactions, as previously described and thoroughly validated (6,26). PLA was also used here to analyze PKB phosphorylation by using separate primary antibodies (Cell Signaling) against PKB protein and the S473 phosphorylation site of PKB. In muscle tissues, in situ PLAs were performed on paraffin-embedded gastrocnemius sections, after an antigen retrieval at pH 6, using a bright-field revelation, as previously described (27).

The concentration of cytosolic calcium ([Ca²⁺]_c) was measured using the ratiometric dye Fura2- acetoxymethyl ester (AM) (5 µmol/L) loaded in human myotubes at 37°C for 1 h, whereas that of mitochondria ([Ca²⁺]_m) was measured using ratiometric 4mtD3cpv biosensor expressed after 2 days of adenovirus infection of human myotubes at 37°C. After three washes with a Ca²⁺-free Tyrode solution (140 mmol/L NaCl, 5 mmol/L KCl, 10 mmol/L HEPES, 1 mmol/L MgCl₂, 10 mmol/L glucose, and 100 µmol/L EGTA at pH 7.4), dishes were set on a Leica DMI6000 B microscope equipped with a ×20 objective, an ORCA-Flash 4.0 Scientific CMOS camera (Hamamatsu), and a Lambda DG-4+ illumination system (Sutter Instrument). Myotubes that were elongated (crossing the field of vision), branched, and polynucleated were selected. Cells were treated with 200 μmol/L Na-ATP (28) and 5 μmol/L thapsigargin (Tg). Because not all myotubes responded to Tg by an increase in free Ca²⁺ in cytosol or in mitochondria, we selected only cells showing a significant difference in the average fluorescence between the 40 s before and after Tg treatment. These cells were considered as having a good probe load and sufficient Ca²⁺ stores to enable measurement of variations in free [Ca²⁺]. Four independent experiments were analyzed. The fluorescence ratio was normalized by the fluorescence at the origin (F/F₀), and all cells were gathered for statistical analysis.

Glucose transport was measured in human myotubes, as previously described (29).

Total DNA was extracted from gastrocnemius muscle of mice, and mitochondrial and nuclear DNAs were quantified by real-time PCR, as previously described (30).

Cytochrome C oxidase (COX) and citrate synthase (CS) activities were measured spectrophotometrically in total lysates from gastrocnemius muscle of mice, as previously described (31,32).

Data are expressed as the mean ± SEM, and statistical significance was defined as a value of P < 0.05. Normal distribution of the data was tested using Shapiro-Wilk. Comparisons between more than two groups were analyzed by one-way ANOVA, followed by Turkey post hoc tests. For all other analysis, data were compared by Student t test. For Ca²⁺ imaging, statistical error (α), statistical power (β), and effect size (d) were calculated in function of the number of cells, and statistical difference was defined as α < 0.05 and β > 0.8.

We performed subcellular fractionation of gastrocnemius muscle from genetically (ob/ob) and diet-induced insulin-resistant mice. The skeletal muscles were sampled on the mice that have been previously used to study MAM integrity in the liver, and these mice were glucose intolerant and insulin resistant (6). The ob/ob (Fig. 1A) and HFHSD (Fig. 1B) mice both showed altered muscle insulin signaling, as illustrated by the significant reduction of insulin-stimulated PKB and glycogen synthase kinase 3-β phosphorylations in gastrocnemius explants. Interestingly, we found in both models a marked reduction of MAM amount in gastrocnemius muscle assessed by subcellular fractionation (Fig. 1C and D, respectively). This reduction of MAMs in diabetic muscle was also observed when MAM amounts were normalized relative to mitochondria amounts. The purity of MAM fractions was validated by Western blotting and TEM (Supplementary Fig. 1A and B), and the expression levels of 75-kDa glucose-regulated protein (GRP75), VDAC1, MFN2, and cyclophilin D were not modified in muscle MAM fractions by obesity and T2D (Supplementary Fig. 2A and B). We then used in situ PLA as an additional strategy to quantify the number of MAM contact points in paraffin-embedded gastrocnemius muscle by measuring the proximity between VDAC1 and IP3R1, as previously described (6,26). The ob/ob (Fig. 1E) and HFHSD (Fig. 1F) mice both showed a significant reduction of VDAC1/IP3R1 interactions compared with respective control mice, illustrating a marked disruption of organelle coupling in skeletal muscles of insulin-resistant mice.

We next induced insulin resistance in primary cultures of human myotubes from control subjects using palmitate treatment (500 µmol/L, 24 h). As expected, palmitate-treated myotubes displayed a reduction of insulin-mediated PKB and GSK3β phosphorylations (Fig. 2A) and an inhibition of insulin-stimulated glucose transport (Fig. 2B). Importantly, they also showed a dramatic reduction of VDAC1-IP3R1 interactions measured by in situ PLA (Fig. 2C) and a decrease in the percentage of mitochondrial membrane in contact with SR/ER, analyzed by TEM (Fig. 2D), indicating a disruption of organelle coupling.

Then, we analyzed organelle Ca²⁺ exchange in palmitate-treated human myotubes using Ca²⁺ video-imaging microfluorometry. Cells were infected with a 4mtD3cpv-endocing adenovirus for 48 h to measure free [Ca²⁺] in the mitochondrial matrix ([Ca²⁺]_m) or incubated with 5 μmol/L Fura2-AM for 1 h to measure free [Ca²⁺] in cytosol ([Ca²⁺]_c). Then, 200 μmol/L ATP were added in the medium to induce the IP3-mediated mobilization of Ca²⁺ stores (28). Conversely, cells were treated with 5 µmol/L Tg to release Ca²⁺ from ER stores independently of MAMs, because Ca²⁺ leak from ER stores could be carried by several types of channels that have not been described at MAMs (33). As shown in Fig. 2E and F, ATP induced a transient increase in both [Ca²⁺]_c and [Ca²⁺]_m in BSA-treated cells but only induced increased [Ca²⁺]_c in palmitate-treated myotubes, suggesting that palmitate induced a decrease of ATP-mediated Ca²⁺ mobilization from ER stores or a decreased Ca²⁺ uptake by mitochondria. We thus compared both ATP- and Tg-mediated variations in [Ca²⁺]_m and [Ca²⁺]_c. Tg induced a similar increase in the free [Ca²⁺]_c in BSA- and palmitate-treated myotubes, suggesting no modification in the ER Ca²⁺ stores. A slight but nonsignificant (Fig. 2G and Supplementary Table 5) decrease in [Ca²⁺]_m was observed in palmitate-treated myotubes, suggesting that a part of Tg-mediated Ca²⁺ could be released in MAMs.

Because mitochondrial physiology is markedly altered in skeletal muscle during obesity and T2D, it is difficult to determine whether alterations of muscle insulin action result from organelle miscommunication or mitochondria dysfunction, or both. To get more insight into this issue, we examined the relationship between organelle communication and mitochondria physiology in insulin-resistant states in vivo and in vitro. Firstly, we analyzed insulin sensitivity (insulin tolerance test), SR/ER-mitochondria interactions (in situ PLA), and mitochondrial biology (mitochondrial [mt]DNA amount and COX/CS activity) in muscle of SD and HFHSD mice after 1, 4, 8, 12, or 16 weeks of the diet. Diet-induced insulin resistance appeared in HFHSD mice after 12 weeks of the diet (Supplementary Table 6). SR/ER-mitochondria interactions are decreased in skeletal muscle of HFHSD mice as soon as 1 week of feeding and remained diminished during the entire feeding period (Fig. 3A). Conversely, mtDNA amount and COX/CS activity are reduced only after 16 weeks of HFHSD feeding (Fig. 3B and C, respectively), whereas we found a transient increase of COX/CS activity during the first weeks of HFHSD feeding (Fig. 3C).

Secondly, we thought to alter mitochondria function by oligomycin (an ATPase inhibitor) treatment (1 or 10 µmol/L for 16 h) in human myotubes and to analyze the repercussions on organelle contacts and insulin signaling. Oligomycin treatment markedly induced mitochondria fragmentation (Fig. 3D), without altering cell viability (data not shown). Importantly, 1 and 10 µmol/L oligomycin treatments markedly induced VDAC1-IP3R1 interactions (Fig. 3E), whereas only 1 µmol/L oligomycin significantly increased insulin-stimulated PKB phosphorylation (Fig. 3F).

Next, we investigated whether the inhibitory action of palmitate on insulin signaling could be prevented by increasing organelle coupling through Grp75 or Mfn2 overexpression, as previously performed in hepatocytes (6). Using a fluorescent vector, we estimated that 46.7 ± 2.3% of myotubes (n = 20 images) were transfected in our conditions (data not shown) and sufficient to observe an increase of Grp75 and Mfn2 protein expression (Fig. 4B). As expected, transient Grp75 or Mfn2 overexpression increased VDAC1-IP3R1 interactions in human myotubes (Fig. 4A). Interestingly, increasing organelle coupling by overexpressing Grp75 or Mfn2 restored palmitate-induced alterations of insulin signaling, as illustrated by the increase of insulin-mediated PKB and GSK3β phosphorylations (Fig. 4B). To evaluate the effects of MAM reinforcement on other parameters related to insulin action, we analyzed the effect of the hormone on the regulation of GLUT4, hexokinase II (HKII), and SREBP1c mRNA, because the regulation of these genes by insulin is altered in muscle of patients with T2D (34) and on glucose transport. We found that insulin induced expression of GLUT4, HKII, and SREBP1c in BSA-treated myotubes transfected with an empty vector, whereas palmitate treatment completely abolished these regulations (Fig. 4C). Importantly, Grp75 and Mfn2 overexpression both restored insulin-mediated regulation of these three genes in palmitate-treated myotubes (Fig. 4C). Similarly, palmitate treatment altered insulin-mediated glucose uptake in human myotubes transfected with an empty vector, whereas insulin-mediated glucose uptake was restored in palmitate-treated myotubes overexpressing Grp75 or Mfn2 (Fig. 4D). Altogether, these data demonstrate that increasing ER-mitochondria tethering in muscle cells is sufficient to restore metabolic insulin signaling and metabolic action.

In hepatocytes, reduction of MAM integrity has been associated with ER stress (35), which may contribute to insulin resistance. To determine whether the same phenomenon also occurs in human myotubes, we evaluated the effects of different treatments on the unfolded protein response (UPR) markers. Palmitate treatment increased mRNA levels of the 78-kDa GRP (GRP78), the spliced X-box-binding protein 1 (Xbp1-s), and CHOP (Fig. 4E), confirming palmitate-mediated ER stress in human myotubes (21). Interestingly, overexpression of Grp75 or Mfn2 counteracted palmitate-induced ER stress, as illustrated by the significant reduction of all UPR markers (Fig. 4E), indicating that reinforcement of MAM can improve palmitate-induced ER stress in muscle cells.

We next explored whether the reduction of organelle coupling, by reducing GRP75 or MFN2 protein levels by specific siRNA, can alter insulin action in muscle cells as observed previously in hepatocytes (6). We validated the efficiency of siRNA targeting in human myotubes and found that 51 ± 4% of human myotubes (n = 20 images) were targeted by fluorescent siRNA (data not shown). Silencing of Grp75 and Mfn2 using specific siRNA (Fig. 5A) reduced VDAC1-IP3R1 interactions in human myotubes (Fig. 5B). Importantly, in both conditions of disrupted ER-mitochondria interactions, insulin action was altered in muscle cells, as illustrated 1) by the reduction of insulin-mediated PKB phosphorylation (Fig. 5C), 2) by a reduction of Glut4 mRNA expression (Fig. 5D), and 3) by an inhibition of insulin-mediated glucose uptake (Fig. 5E). Lastly, cotreatment of human myotubes with siRNA (Grp75 or Mfn2) and palmitate did not lead to additive effects on insulin-stimulated PKB phosphorylation (Supplementary Fig. 3), suggesting that the two treatments may affect the same regulatory process.

Even if genetic manipulation of MAM proteins is a good approach to modulate SR/ER mitochondria interactions, none of these proteins are exclusively expressed at MAMs, and their manipulation may result in alterations of cellular functions outside of MAMs. Therefore, we sought to overexpress FATE1, a cancer-testis antigen recently identified as an uncoupler of MAMs (22) and normally not expressed in skeletal muscle. For that, we constructed a pcDNA-FATE1-IRES-GFP vector, which simultaneously overexpresses FATE1 and GFP proteins and allows us to analyze organelle interactions by in situ PLA only in transfected cells (green cells), compared with untransfected cells (unfluorescent cells). Firstly, we confirmed that FATE1 is not expressed in human myotubes and that its overexpression increased its protein levels (Fig. 6A). More importantly, transient FATE1 overexpression reduced VDAC1-IP3R1 interactions in FATE1-transfected myotubes compared with cells transfected with an empty vector or with untransfected cells in FATE1-transfected myotubes (Fig. 6B), validating the uncloupling activity of FATE1 in nontestis cells. To analyze the repercussions on insulin signaling similarly in both transfected and untransfected myotubes, we analyzed insulin-stimulated PKB phosphorylation by in situ PLA by using separate primary antibodies against PKB protein and the S473 phosphorylation site of PKB. As expected, insulin increased pS473 phosphorylation of PKB in myotubes transfected with an empty vector (Fig. 6C), validating the use of PLA for the analysis of insulin signaling. Importantly, insulin-stimulated PKB phosphorylation is markedly and specifically reduced in FATE1-overexpressing myotubes (Fig. 6C), indicating that disrupting organelle interactions independently of endogenous MAM proteins also alters insulin action.

Next, we constructed an adenovirus starting from the pcDNA-FATE1-IRES-GFP vector and we used an Ad-GFP as a control. We validated that the use of Ad-FATE1 in human myotubes reduced ER-mitochondria interactions (Supplementary Fig. 4A) and altered insulin-stimulated PKB phosphorylation (Supplementary Fig. 4B), as in transient transfection experiments. Interestingly, the same experiment performed in the HuH7 hepatic cell line also showed a clear reduction of both organelle interactions (Supplementary Fig. 5A) and insulin-stimulated IRS2 and PKB phosphorylations (Supplementary Fig. 5B), validating that FATE1-mediated organelle uncoupling also alters insulin signaling in hepatocytes, in agreement with our previous observations when we modulated endogenous MAM proteins (6). Therefore, we tested in vivo the injection of Ad-FATE1 in gastrocnemius muscle of mice. We validated that infected muscle fibers of both Ad-GFP and Ad-FATE1 mice express the GFP protein in muscle fibers (Supplementary Fig. 6). Importantly, FATE1 overexpression reduced VDAC1-IP3R1 interactions (Fig. 6D) and altered the effect of insulin on PKB phosphorylation (Fig. 6E) in gastrocnemius muscle of infected mice.

Because myotubes in primary culture are known to maintain insulin sensitivity of the donor (36,37), we investigated whether ER-mitochondria interactions are altered in primary myotubes established from obese subjects and donors with T2D compared with cells from healthy lean control subjects. The patients with T2D were matched for BMI and age with the obese subjects and for age with the lean donors (Table 1). We observed a reduction of insulin-mediated PKB phosphorylation in myotubes from obese donors and patients with T2D compared with lean donors (Fig. 7A), whereas no significant difference was observed between obese donors without diabetes and obese patients with T2D, indicating a similar degree of insulin resistance. Importantly, VDAC1-IP3R1 interactions, measured by in situ PLA, were decreased in myotubes from obese subjects without diabetes and those with diabetes compared with healthy donors (Fig. 7B). No significant difference was found between myotubes from obese donors without and with diabetes. Lastly, VDAC1-IP3R1 interactions were significantly and positively correlated with the effect of insulin on PKB phosphorylation (Fig. 7C) (P = 0.33, R² = 0.11; P = 0.04) when all the subjects were analyzed.

We recently showed that MAM integrity contributes to insulin action in the liver and is altered in situations of hepatic insulin resistance (6). Here, we investigated whether this mechanism is restricted to the liver or could be generalized to other insulin-sensitive tissues, particularly skeletal muscles. This study in muscle cells is particularly important because the few reported studies on this issue in the hepatocytes are highly controversial (6,12). We demonstrate in vitro and in vivo, from mice to humans, that the disruption of organelle coupling may contribute to muscle insulin resistance and that reinforcing MAM increases insulin action, at least in human myotubes. Therefore, targeting MAM could be a novel strategy to improve insulin sensitivity and restore glucose homeostasis.

As previously observed in the liver (6,27), we found that ER-mitochondria interactions are reduced in skeletal muscle of genetically and diet-induced obese and diabetic mice, in palmitate-treated human myotubes, and in myotubes from obese patients and patients with T2D, indicating a close relationship between MAM integrity and muscle insulin sensitivity. These structural analyses were systematically performed by in situ PLA in vivo and in vitro, whereas subcellular fractionation or TEM analysis confirms PLA results in vivo and in vitro, respectively. We tried to analyze ER-mitochondria interactions in skeletal muscle by TEM but were not able to perform a reliable analysis due to the architecture of muscle fibers. We believe that TEM is not really able to reveal fine structural details because of the overlapping density/noisy background in muscle tissues. This issue notwithstanding, cumulative evidence supports the use of in situ PLA as a reliable method to study in situ ER-mitochondria interactions in skeletal muscle, as previously validated in the liver (6,26,27). Furthermore, the functionality of MAMs is also altered in insulin-resistant states, because Ca²⁺ transfer from ER to mitochondria is reduced in palmitate-treated myotubes, supporting a potential implication of organelle miscommunication in muscle insulin resistance.

Next, we found that MAM integrity is required for an efficient insulin action in skeletal muscle, because the experimental disruption of MAM dampens insulin signaling and action in human myotubes, whereas the experimental induction of organelle interactions prevents palmitate-induced insulin resistance. These genetic experiments are interesting but present two weaknesses. Firstly, all myotubes are not transfected, and therefore, we analyzed organelle tethering on a pool of untransfected and transfected cells, reducing the robustness of the measurements. Secondly, none of MAM proteins is exclusively expressed at MAMs; therefore, their modifications may result in alterations of cellular function outside of MAMs. To get around these two problems, we sought to overexpress FATE1, an organelle uncoupler not expressed in skeletal muscle (22), by using the pCDNA-FATE1-IRES-GFP vector allowing the simultaneous expression of both FATE1 and GFP proteins. Using this strategy, we are able to analyze by in situ PLA VDAC1-IP3R1 interactions and PKB phosphorylation in transfected and untransfected cells as a negative control. We validated that FATE1-mediated reduction in organelle coupling is specific to transfected myotubes, indicating that FATE1 overexpression is a good strategy to reduce ER-mitochondria interactions in nontestis cells. Furthermore, we found that overexpression of FATE1 specifically altered insulin-stimulated PKB phosphorylation in transfected or Ad-FATE1–infected myotubes and in gastrocnemius muscle of mice. Altogether, these data demonstrate that FATE1-mediated organelle uncoupling is sufficient to alter insulin signaling in vitro and in vivo.

Because mitochondria physiology was previously reported to be altered in skeletal muscle of obese and diabetic mice (30,38) and in palmitate-induced insulin-resistant myotubes (21), the reduction of ER-mitochondria interactions in all of these models could be related to mitochondrial alterations. Nevertheless, we firstly confirmed that the reduction of MAM amount persists after normalization by mitochondria amount in the subcellular fractionation of mouse muscle. Furthermore, we demonstrated that organelle miscommunication is an early event in diet-induced insulin resistance preceding mitochondrial dysfunction and insulin resistance. Lastly, oligomycin-induced mitochondrial dysfunction in human myotubes rather increased SR/ER-mitochondria interactions and potentiated insulin signaling, probably as an adaptive mechanism to improve mitochondrial function and cell homeostasis. Altogether, these results suggest that muscle organelle miscommunication in the diabetic state is not a consequence of mitochondrial alterations or of the insulin-resistance state, supporting its participation in the development of muscle insulin resistance. Nevertheless, future experiments are required to demonstrate that preventing MAM disruption could improve muscle insulin sensitivity in vivo.

The mechanisms linking organelle miscommunication to muscle insulin resistance are currently unknown. Modulation of Ca²⁺ signaling and exchange between organelles could participate, as previously found in the liver (35). Activation of the UPR into the ER could also play a role, because we found here that the reinforcement of MAMs through Mfn2 or Grp75 overexpression prevented the regulation of UPR markers by palmitate. Nevertheless, because Mfn2 was shown to directly modulate UPR (39), we cannot exclude an effect independent from MAMs. Furthermore, the involvement of ER stress in muscle insulin resistance is more controversial than in the liver (38), and we previously found that reducing ER stress by genetic or pharmacological approaches is not sufficient to improve palmitate-induced insulin resistance in myotubes (21), suggesting that ER stress is probably not the link between MAM disruption and muscle insulin resistance. Alternatively, intramyocellular lipid accumulation has been associated to reduced muscle insulin sensitivity (40). In this context, mitochondria dysfunction and the subsequent impaired ability to oxidize fatty acids could play an important role in muscle insulin resistance, although the causality of this association is still controversial (1). We previously demonstrated (30) and confirmed here that mitochondrial alterations were not an early event in the diet-induced development of insulin resistance, contrary to MAM disruption. Therefore, organelle miscommunication could participate to mitochondrial dysfunction in skeletal muscle, subsequently leading to insulin resistance. Finally, boosting mitochondrial function might be beneficial to muscle insulin sensitivity and patient health (41). Because organelle coupling is known to control mitochondrial bioenergetics (42), MAM-mediated improvement of insulin action in human myotubes may be related to improved mitochondria function. However, further investigations are required to clarify this hypothesis.

Whereas organelle miscommunication recently emerged as a new mechanism of insulin resistance, the few reported studies on this issue are highly controversial. Indeed, we and others found that insulin resistance is associated with reduced ER-mitochondria contacts in the liver (6,7) and in adipose tissue (43), whereas another group found excessive organelle coupling in the liver of obese and diabetic mice (12). Of course, MAMs are dynamic structures strongly dependent on nutritional and metabolic status (27) as well as on environmental conditions and stress factors (44). In addition, their precise quantitative measurement is particularly tricky, potentially depending on the methodology, and most importantly whether the measurement takes into account the number, length, or thickness of the organelle coupling (45). That several types of mitochondria-ER contact sites exist is likely, with potentially different protein complexes and functions, and it is plausible that these different tethers could be differentially regulated, as recently suggested (46). Consequently, all of these factors probably participate to the contrasting conclusions in this exciting topic. Lastly, there are currently no good in vivo models of MAM disruption or reinforcement and no optimal reporter to dynamically follow ER-mitochondria interactions, making studies more difficult. Further studies are thus required to solve this controversy, and the future use of the very recent split-GFP-based contact site sensor (SPLICS) tool to follow narrow and wide organelle coupling (46) could be a relevant strategy. In the meantime and as a first step in this quest, we confirmed here that FATE1-mediated organelle uncoupling reduced insulin-stimulated IRS2 and PKB phosphorylation in HuH7 cells, as we previously observed when we modulated endogenous MAM proteins (6). In this context, the current study in skeletal muscle confirms that disruption of organelle interactions is also associated with insulin resistance, in another insulin-sensitive tissue, and further extends the results to humans, an aspect not investigated until now.

In summary, we demonstrated in vivo and in vitro that defective ER-mitochondria coupling is closely associated with impaired muscle insulin sensitivity in mice and humans. Furthermore, organelle miscommunication is an early event in diet-induced insulin resistance preceding mitochondrial dysfunction. Lastly, disruption of MAM integrity alters insulin signaling in vitro and in vivo, whereas reinforcing organelle coupling improves palmitate-induced insulin resistance in human myotubes. Taking into account that similar observations were done previously in the liver (6), and although additional studies are required, especially to demonstrate that MAM reinforcement in muscle of insulin-resistant patients improves insulin action, the presented data pave the way for considering targeting the MAM interface in insulin-sensitive tissues as a potential attractive strategy to improve glucose homeostasis.

Acknowledgments. The authors thank Enzo Lalli (CNRS UMR7275, Valbonne, France) for the gift of pcDNA-T0 and pcDNA-FATE1 vectors, Elisabeth Errazuriz for her technical help at the CIQLE Imaging Center (Lyon, France), Kassem Makki (INSERM U1060, Lyon, France) for his help with statistical analyses, Elsa Hoibian (INSERM U1060, Lyon, France) for her help with mice, Aurélie Vieille Marchiset (INSERM U1060, Lyon, France) for her help with culture of human myotubes, and Mélanie Paillard (INSERM U1060, Lyon, France) for her helpful discussion on FATE1.

Funding. E.T. was supported by a research fellowship from French government of higher education and research and by a scholarship from Lund University. This work was supported by INSERM, the Agence Nationale de la Recherche (ANR-09-JCJC-0116 to J.R.), and the “Fondation pour la Recherche Médicale” (DRM20101220461).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. E.T. and J.R. designed the experiments, researched data, contributed to discussion, and wrote the manuscript. S.C., N.B., G.B., M.-A.C., J.J.-C., C.D., and D.G.-R. researched data. M.R. and E.L. constituted the DIOMEDE biobank. G.B., H.V., and E.L., contributed to discussion and reviewed and edited the manuscript. J.R. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the Congress of the French Society of Diabetes, Lyon, France, 22–25 March 2016; at the 2016 Nutrition Congress of the French Nutrition Society, Montpellier, France, 30 November–2 December 2016; and at the Research and Health Day of INSERM, Paris, France, 22 November 2016.