Differential Effect of Glucose on ER-Mitochondria Ca²⁺ Exchange Participates in Insulin Secretion and Glucotoxicity-Mediated Dysfunction of β-Cells

Pancreatic β-cells are glucose sensors adjusting insulin secretion to glycemia via triggering pathways, including glucose oxidation by glycolysis and the Krebs cycle, an increase in the cytosolic ATP/ADP ratio, closing K_ATP channels, membrane depolarization, opening of voltage-dependent calcium (Ca²⁺) channels, an increase in cytosolic Ca²⁺, and exocytosis of insulin (1,2). Importantly, mitochondria and the endoplasmic reticulum (ER) play a crucial role in this process, as glucose-derived mitochondrial metabolism regulates insulin release by increasing the ATP/ADP ratio (3), and the ER controls the synthesis, correct folding, and sorting of insulin in response to glucose (4). Furthermore, both the ER and mitochondria control Ca²⁺ dynamics by regulating Ca²⁺ uptake and release from/into intracellular organelles.

β-Cell dysfunction is a major feature of type 2 diabetes (T2D), characterized by the incapacity to produce sufficient amounts of insulin to control glycemia due to a progressive decrease of both β-cell function and mass. Repeated or sustained exposure of β-cells to supraphysiological glucose levels, referred to as glucotoxicity, contributes to β-cell dysfunction during T2D (5). Particularly, chronic exposure of β-cells or pancreatic islets to high glucose (HG) concentrations impairs their glucose-stimulated insulin secretion (GSIS) and decreases the processing of insulin (6). Interestingly, mitochondrial dysfunction (7) and ER stress (8), as well as perturbations of Ca²⁺ homeostasis (9), have been largely associated with glucotoxicity-mediated β-cell dysfunction, highlighting the crucial role of these two organelles in T2D-related β-cell failure. Particularly, changes in mitochondrial structure (10), decreased mitochondrial respiration (11), reduced mitochondrial ATP production (12), and altered mitochondrial dynamics (13) underlie glucotoxicity-mediated β-cell dysfunction. In addition, ER Ca²⁺ depletion (9) and activation of the unfolded protein response (UPR) (8) are associated with glucotoxicity-induced β-cell dysfunction.

Interestingly, both organelles interact structurally and functionally at specific contact sites, called mitochondria-associated membranes (MAMs), in order to exchange phospholipids and Ca²⁺ and to regulate mitochondrial bioenergetics (14). Different physio-pathological roles of MAMs start to emerge, especially in metabolic diseases. Indeed, MAMs appear as new hubs of insulin signaling and glucose sensing, playing a key role in the control of glucose homeostasis (15). We recently demonstrated that MAMs controlled insulin action in both the liver (16) and skeletal muscle (17), and that organelle miscommunication was associated with insulin resistance in various mouse models of obesity and T2D (16–18). Importantly, we also evidenced a disruption of organelle interactions in β-cells of patients with T2D and in palmitate-treated MIN6-B1 cells (19), suggesting a potential implication of MAMs in β-cell dysfunction. Nevertheless, the direct role of MAMs in glucotoxicity-associated β-cell dysfunction is unknown.

The aim of the current study was therefore to clarify the role of MAMs in the regulation of insulin secretion by β-cells and to investigate whether MAM alterations could participate in the mechanisms of glucotoxicity-mediated β-cell dysfunction. Here we found evidence, in both a pancreatic β-cell line and human pancreatic islets, for a differential regulation over time of the ER-mitochondria Ca²⁺ exchange in response to HG, representing a continuum from a positive response to stimulate insulin secretion to a deleterious outcome participating in the progressive glucotoxicity-associated β-cell failure.

Rat pancreatic INS-1E β-cells (RRID:CVCL_0351) were cultured as previously described (20). INS-1E cells were treated with either 11 or 22.5 mmol/L glucose for 30 min, 48 h, or 72 h. INS-1E cells were also treated with either 0.5 µg/mL tunicamycin (Sigma-Aldrich) for 24 h or 2 µmol/L oligomycin (Sigma-Aldrich) for 1 h. INS-1E cells were transfected for 72 h with a cocktail of three Grp75-specific siRNA (20 μmol/L) (Quiagen) using DharmaFeCT duo (Dharmacon). Cells were transfected for 72 h using DharmaFeCT duo, with 2 µg of mAKAP1-mRFP-yUBC6 (linker) and mAKAP1-mRFP (pcDNA Ctrl) plasmids, provided by Hajnoczky and colleagues (21). RFP fluorescence of plasmids at 558 nm was used to sort cells (FACSAria; BD Bioscience).

Human pancreatic islets from eight donors without diabetes (Supplementary Table 1) were obtained through both the Geneva European Consortium for Islet Transplantation (ECIT; n = 7) and the European Diabetes Study Center (CEED, Strasbourg, France; n = 1). Islets were processed and cultured as previously described (9). Islets were cultured in either 5.5 or 16.5 mmol/L glucose for 48 h. For imaging analysis, islets were mechanically dissociated in order to obtain single cells and specifically analyze insulin-positive cells.

After treatment, cells were preincubated for 2 h at 37°C in Krebs-Ringer bicarbonate Hepes buffer (KRBH) containing 0.1% BSA and 5.5 mmol/L glucose. For GSIS experiments, cells were incubated for 1 h with 5.5 mmol/L glucose (supernatants were collected) and then further challenged with 22.5 mmol/L glucose for 1 h (supernatants were collected). Insulin was measured on culture medium by ELISA (ALPCO) for INS-1E cells or by a highly specific immunoradiometric assay (Bi-insulin IRMA; Cisbio International) for human islets (9) and normalized either by protein levels or insulin content (independent experiments). Insulin content was measured on acid-ethanol–extracted cell lysates from 300,000 INS-1E cells or 200 equivalent islets. Results are expressed in absolute values (ng/mL of secreted insulin) or are normalized by cellular proteins or insulin content. The fold of insulin secretion is also calculated.

Cell fixation, posttreatments, and image acquisition were performed as previously described (9). Mitochondria and ER were delimitated using ImageJ. The fraction of mitochondrial membrane in contact with ER within a 50-nm range was calculated and normalized to the mitochondrial perimeter and expressed as the total percent of contact between mitochondria and ER. Another calculation of the total percent of contact was also made according to different ranges of distance between both organelles, as previously described (22).

ER-mitochondria interactions were assessed by in situ proximity ligation assay (PLA), as previously described and thoroughly validated (16,23), by targeting two organelle surface proteins, VDAC1 and IP3R2 (19), using the following antibodies: VDAC1 (catalog no. ab14734, RRID:AB_443084; Abcam) and IP3R2 (catalog no. sc-28614, RRID:AB_305125; Santa Cruz Biotechnology). For human islets, in situ PLAs were performed in parallel with insulin immunofluorescence labeling (catalog no. A056401-2, RRID:AB_2617169; Agilent), and blobs per nucleus were quantified only in labeled β-cells. In both cases, analyses of dot number per cell represent the average of all analyzed cells in triplicate in three independent experiments.

We used the ratiometric Ca²⁺ probe plasmids encoding the 4mtD3CPV (24) (RRID:Addgene_36324) or erGAP1 (25) biosensors to quantify mitochondrial or ER Ca²⁺ content, respectively. Quantification of cytosolic Ca²⁺ levels was performed using the ratiometric Fura2-AM dye (5 µmol/L) (Invitrogen). More information is available in Supplementary Fig. 1. All dynamic Ca²⁺ measurements were performed in KRBH in the absence of extracellular Ca²⁺, either after 48 or 72 h posttransfection or after 30 min Fura2-AM loading. Cells were treated (added in puff, v/v) either with 1 µmol/L Na-ATP (Sigma-Aldrich) in order to stimulate Ca²⁺ transfer from ER to mitochondria or with 1 μmol/L thapsigargin (TG; Sigma-Aldrich) to release ER Ca²⁺ stock. Basal Ca²⁺ levels were quantified on 10 fields per dish before adding stimulus, and ATP- or TG-stimulated Ca²⁺ levels were measured on 1 field per dish. Therefore, quantification of basal Ca²⁺ levels is systematically done on a greater number of cells than those illustrated on average cures. Measurements were performed at 37°C using a wide-field Leica DMI6000B microscope equipped with a 40× objective and an ORCA-Flash4.0 digital camera (HAMAMATSU). The fluorescence ratio was analyzed with MetaFluor 6.3 (Universal Imaging) after removing background fluorescence. Results (average curves and histograms representing basal and stimulated Ca²⁺ levels) represent the average of all analyzed cells in triplicate in three independent experiments. However, we confirmed that we found in each batch of cells the same effect (or at least the same tendency for one case) as when results were analyzed with the total number of cells (data not shown).

Cells were pelleted and resuspended in 100 μL KRBH supplemented with 0.5% BSA and 11 mmol/L glucose. Oxygen consumption was measured in intact cells at 37°C using a Clark-type electrode (Strathkelvin Instruments). Oligomycin (10 µg/mL) (Sigma-Aldrich) was added in order to analyze resting oxygen consumption linked to nonphosphorylated mitochondrial activities.

The content of mtDNA was calculated using real-time PCR by measuring the threshold cycle ratio of a mitochondrial gene COX1 (rat INS-1E) or COX2 (human islets) versus a nuclear gene (cyclophilin A), as previously described (26).

The mitochondrial network was analyzed as previously described (27), either using MitoTracker (500 nmol/L, 20 min) (Invitrogen) for INS-1E cells or by performing immunofluorescence labeling, using both OXPHOS (catalog no. ab110413, RRID:AB_2629281; Abcam) and insulin (catalog no. A056401-2, RRID:AB_2617169; Agilent) primary antibodies in human islets, to specifically assess β-cells. Fluorescent images were analyzed using ImageJ, and circularity (4 * π * area/perimeter²) and length of major and minor axes were measured in order to calculate form factor (FF) (opposite of circularity) and aspect ratio (AR) (major axis/minor axis).

Total RNAs were extracted with the TRI Reagent Solution (Sigma-Aldrich). Levels of target mRNAs were measured by reverse transcription (PrimeScript RT; Takara), followed by real-time PCR using a Rotor-Gene 6000 (Corbett Research). TATA-binding protein (TBP) mRNA was quantified as a reference gene. The list of primers is given in Supplementary Table 2.

INS-1E protein lysates were prepared in RIPA buffer, and protein expression was analyzed by SDS-PAGE (10%) using the following antibodies: MCU (catalog no. HPA016480, RRID:AB_2071893; Atlas Antibodies), MICU1 (catalog no. HPA037480, RRID:AB_10696934; Atlas Antibodies), tubulin (catalog no. T6074, RRID:AB_477582; Sigma-Aldrich), and Grp75 (catalog no. sc-133137, RRID:AB_2120468; Santa Cruz Biotechnology).

All data are presented as mean ± SEM. Normality was tested using the D’Agostino and Pearson test included in the GraphPad Prism software. If values are normally distributed, differences between two groups were tested with unpaired Student t test. Mann-Whitney U test was applied when values were not normally distributed. Comparisons between more than two groups were analyzed by two-way ANOVA after confirmation of the normality. Correlative analyses were performed using Pearson r test. Statistical significance was accepted when P < 0.05.

We investigated the role of MAMs in the acute regulation of GSIS in the rat pancreatic β-cell line INS-1E. For that, we analyzed ER-mitochondria interactions and Ca²⁺ exchange after 30 min of glucose challenge (22.5 mmol/L), known to stimulate GSIS (28). MAM integrity was analyzed using in situ PLA by targeting VDAC1-IP3R2 proximity (19) in INS-1E cells treated with both 11 mmol/L (Ctrl) and 22.5 mmol/L glucose (HG). As shown in Fig. 1A, a 30-min exposure of INS-1E cells to HG increased VDAC1-IP3R2 interactions compared with Ctrl cells. The effect of HG on ER-mitochondria interactions was also seen when using a lower concentration of glucose (3 mmol/L) in Ctrl conditions (Fig. 1A). We then investigated the effects of HG on MAM function, by analyzing both basal and ATP-stimulated ER-mitochondria Ca²⁺ transfer of INS-1E cells using the mitochondrial-selective calcium fluorescent probe 4mtD3cpv (Supplementary Fig. 1A) (24). As shown in Fig. 1B, HG significantly increased basal mitochondrial Ca²⁺ levels. Then, we stimulated cells with ATP to induce the IP3-mediated mobilization of Ca²⁺ stores (29), and Fig. 1C illustrates average curves of intramitochondrial Ca²⁺ after ATP stimulation in both Ctrl- and HG-treated INS-1E cells. Quantitative analysis showed that HG significantly increased ATP-stimulated mitochondrial Ca²⁺ accumulation (Fig. 1C and D). ATP stimulation also increased cytosolic Ca²⁺ levels in the absence of extracellular Ca²⁺ in INS-1E cells (Supplementary Fig. 1B). As ATP-stimulated mitochondrial Ca²⁺ accumulation could therefore come from a direct transfer from the ER to mitochondria or a release by the ER in the cytosol followed by a recapture of Ca²⁺ from the cytosol to the mitochondria, we investigated whether TG-induced ER Ca²⁺ release into cytosol increases mitochondrial Ca²⁺ levels. We found that TG treatment weakly increased mitochondrial Ca²⁺ levels (ΔTG 0.57 ± 0.03, compared with ΔATP between 4 and 6) and entirely prevented the ATP-stimulated mitochondrial Ca²⁺ accumulation (Supplementary Fig. 1C), supporting that Ca²⁺ influx into mitochondria likely comes mainly from the ER through MAMs. Altogether, these data suggest that insulin secretion upon HG stimulation requires dynamic ER-mitochondria Ca²⁺ exchanges.

To validate this assumption, we explored whether the reduction of ER-mitochondria interactions, by silencing Grp75 as previously performed (16,17), can alter GSIS in INS-1E cells. We validated that silencing of Grp75 using specific siRNA reduced GRP75 protein expression (Fig. 1E) and disrupted ER-mitochondria interactions, as illustrated by the reduction of VDAC1-IP3R2 interactions by in situ PLA (Fig. 1F). Furthermore, Grp75 silencing significantly reduced ATP-stimulated mitochondrial accumulation (Fig. 1H and I), without modification of basal mitochondrial Ca²⁺ levels (Fig. 1G). Importantly, Grp75-mediated disruption of MAMs significantly altered GSIS in INS-1E cells (Fig. 1J), without modifying cellular insulin content (Fig. 1K). Grp75-mediated alterations of GSIS are conserved after normalization by cellular proteins (Supplementary Fig. 2A) or insulin content (Supplementary Fig. 2B). Therefore, organelle coupling is a new component regulating GSIS in β-cells.

Next, we investigated the effects of chronic HG concentrations (glucotoxicity) on ER-mitochondria interactions and Ca²⁺ transfer both in INS-1E cells and human pancreatic islets, after 48- and/or 72-h incubation times. We first validated that exposure of INS-1E cells to HG during 48 h significantly reduced both GSIS (Fig. 2A) and cellular insulin content (Fig. 2B), compared with Ctrl cells. We also normalized insulin secretion by cellular proteins or insulin content and confirmed the reduction of the glucose effect on insulin secretion after glucotoxicity (Supplementary Fig. 2C and D). Similar results were obtained after 72 h of HG treatment with an alteration of GSIS (Supplementary Fig. 3A–C and E) and a reduction of cellular insulin content (Supplementary Fig. 3D). As shown in Fig. 2C, a 48-h HG treatment significantly increased ER-mitochondria contacts, as illustrated by the upregulation of VDAC1-IP3R2 interactions. This result was confirmed by transmission electronic microscopy (TEM) analysis, as we observed a significant increase of the percentage of ER adjacent to mitochondria in a 50-nm range in 48-h HG–treated INS-1E cells (Fig. 2D and E). Specifically, only the 10- to 20- and 20- to 30-nm ranges of distance between both organelles were significantly increased by HG (Fig. 2F), whereas the number of mitochondria per field was not modified (Fig. 2G). Similar effects were observed after longer HG treatment, using both in situ PLA (Supplementary Fig. 4A) and TEM (Supplementary Fig. 4B–E). Importantly, we confirmed in human pancreatic islets, a more physiological model, that glucotoxicity-mediated β-cell dysfunction (Fig. 2H and Supplementary Fig. 2E and F) and reduced insulin content (Fig. 2I) are also associated with increased ER-mitochondria interactions, measured by both in situ PLA (Fig. 2J) and TEM (Fig. 2K–N) in insulin-positive cells. Once again, only the smallest range of contacts between both organelles was significantly increased by HG in human islets (Fig. 2M), without modification of the mitochondria amount (Fig. 2N). Therefore, glucotoxicity-mediated β-cell dysfunction is associated with an increase of ER-mitochondria contact sites.

We then investigated in INS-1E cells the effects of glucotoxicity on the ER-mitochondria Ca²⁺ exchange. HG significantly increased basal mitochondrial Ca²⁺ content in INS-1E cells after 72 h (Fig. 3B), whereas the effect was not significant after 48 h (Fig. 3A). Figure 3C and D illustrates average curves of intramitochondrial Ca²⁺ after ATP stimulation in both Ctrl and HG-treated INS-1E cells after 48 and 72 h of treatment, respectively. Quantitative analysis showed that HG significantly and drastically reduced ATP-stimulated Ca²⁺ mitochondrial accumulation after both 48 h (Fig. 3E) and 72 h (Fig. 3F) of glucotoxicity. As Ca²⁺ entry into mitochondria occurs through the uniporter MCU and its associated regulatory subunit MICUI (30), we monitored both proteins in both Ctrl and HG-treated INS-1E cells and found that MICU1 protein expression was significantly increased by 48-h HG treatment, whereas MCU expression was not significantly modified despite a tendency (P = 0.064) (Fig. 3G). Altogether, these data indicate that glucotoxicity-mediated β-cell dysfunction is associated with altered IP3R-mediated Ca²⁺ transfer into the mitochondria, despite an increase of MAM contact points.

As glucose stimulus is an acute signal that increases organelle interactions and Ca²⁺ exchange, we reasoned that sustained glucose stimulation over time could alter Ca²⁺ transfer from the ER to mitochondria by depleting ER Ca²⁺ stock. Therefore, we measured ER Ca²⁺ content after transfection of INS-1E cells with the aequorin-based fluorescent erGAP1 vector (25) (Supplementary Fig. 1D), both at the basal state and after TG-induced emptying of reticular Ca²⁺ stock. HG significantly reduced basal ER Ca²⁺ content after both 48 h (Fig. 4A) and 72 h (Fig. 4B) of treatment. Then, we stimulated cells with TG (1 µmol/L), a noncompetitive inhibitor of the sarco/endoplasmic reticulum Ca²⁺ ATPase that depletes almost all ER Ca²⁺ stores. Figure 4C and D illustrates average curves of ER Ca²⁺ content after TG stimulation in INS-1E cells after both 48 h and 72 h of HG treatment, respectively. Quantitative analysis showed that HG significantly reduced TG-stimulated ER Ca²⁺ release after 72 h (Fig. 4F), whereas no significant difference was observed after 48 h (Fig. 4E). The slopes of ER Ca²⁺ release after TG stimulation (reflecting the emptying speed of ER Ca²⁺) were unchanged between Ctrl and HG-stimulated INS-1E cells, after both 48 h (Fig. 4G) and 72 h (Fig. 4H) of treatment, suggesting that the reduction of Ca²⁺ release after HG treatment is not related to a change of ER Ca²⁺ channels but should, rather, be related to a reduction of the reticular Ca²⁺ content. Last, we measured cytosolic Ca²⁺ levels using Fura2-AM (Supplementary Fig. 1E) after Ctrl and HG treatments in INS-1E cells. Cytosolic Ca²⁺ levels were not significantly changed by HG after 48 h (Fig. 4I) or 72 h (Fig. 4J) of treatment, compared with respective Ctrl INS-1E cells. Therefore, HG induces ER Ca²⁺ depletion in INS-1E cells, hence impacting the capacity of reticular Ca²⁺ release over time, and this defect may participate in HG-mediated disruption of organelle Ca²⁺ exchanges.

As modification of organelle Ca²⁺ stores may stress both the ER and mitochondria (31), we then analyzed consequences of glucotoxicity on both ER stress and mitochondrial function, both in rodent β-cells and human islets. ER stress was assessed by measuring mRNA levels of key markers of the UPR. We found that HG only significantly increased activating transcription factor 4 (ATF4) mRNA levels after both 48 h (Fig. 5A) and 72 h (Supplementary Fig. 5A) in INS-1E cells. Interestingly, ER stress was more marked in human pancreatic islets, as 48-h HG treatment significantly increased mRNA levels of 78-kDa glucose-regulated protein ATF4 and spliced X-box binding protein 1 (Fig. 5B). We also measured basal mitochondrial oxygen consumption, as an indicator of mitochondria function, both in INS-1E cells and human islets. As shown in Fig. 5C, 48-h HG treatment of INS-1E cells significantly reduced basal respiration, as well as coupled respiration, without a change in proton leak measured in the presence of oligomycin. More pronounced effects were observed in INS-1E cells treated by HG during 72 h (Supplementary Fig. 5B). In human pancreatic islets, HG significantly decreased basal mitochondrial oxygen consumption but did not affect both proton leak and coupled respiration (Fig. 5D).

To determine whether HG-mediated organelle stresses could participate in increased organelle tethering as a protective mechanism to restore organelle homeostasis, we induced either ER stress with 0.5 µg/mL tunicamycin for 24 h (inhibitor of the synthesis of N-linked glycoproteins that is classically used to induce ER stress) (Fig. 5E) or mitochondrial stress with 2 µmol/L oligomycin for 2 h (blocker of proton transfer during mitochondrial respiration) (Fig. 5G), and measured repercussions on ER-mitochondria interactions. Importantly, we found that both tunicamycin (Fig. 5F) and oligomycin (Fig. 5H) treatments increased VDAC1-IP3R2 interactions in INS-1E cells, suggesting that the glucotoxicity-associated increase of organelle tethering could be secondary to organelle stresses.

To characterize the effects of HG on the mitochondrial network, we measured AR and FF parameters of labeled mitochondria in both rat and human β-cells. Consistent with previous reports (32), HG induced mitochondrial fragmentation in INS-1E cells, with a significant decrease of both AR and FF after 48 h of treatment (Fig. 6A–C), without change in mtDNA amounts (Fig. 6D). We further categorized and quantified mitochondria into tubules, fragments, and intermediates, as previously described (28). In resting conditions, a majority of mitochondria were in a tubule and intermediate state (Fig. 6E), whereas mitochondria became short and fragmented upon 48 h of HG incubation, (Fig. 6E). Similar effects of HG were observed in INS-1E cells after 72 h of treatment (Supplementary Fig. 5C–G). Importantly, we found consistent results in human pancreatic β-cells, as 48-h HG reduced both AR and FF parameters (Fig. 6F–H) without modifications of mtDNA amounts (Fig. 6I) and increased the percentage of fragmented mitochondria (Fig. 6J). Overall, these data demonstrate that HG switches the mitochondrial network from a tubular shape to a fragmented pattern in both rat and human β-cells.

We then analyzed with R software whether HG-mediated mitochondrial fission was linked to intramitochondrial Ca²⁺ accumulation, by creating a correlogram between both basal and ATP-stimulated mitochondrial Ca²⁺ accumulation (ATP response) and AR and FF from Ctrl/HG-stimulated INS-1E cells (Fig. 6K). As expected, we found that AR was significantly and positively correlated with FF (Supplementary Fig. 6A), whereas basal mitochondrial Ca²⁺ content was negatively, but not significantly, correlated with ATP-stimulated mitochondrial Ca²⁺ accumulation (Supplementary Fig. 6B). Importantly, basal mitochondrial Ca²⁺ was significantly and negatively correlated with both AR and FF parameters (Supplementary Fig. 6C and D, respectively), indicating that the more mitochondria have high Ca²⁺ levels the more they are fragmented. Inversely, ATP-stimulated mitochondrial Ca²⁺ accumulation was significantly and positively correlated with both AR and FF parameters (Supplementary Fig. 6E and F), indicating that mitochondria with disrupted Ca²⁺ transfer from the ER were the most fragmented. Altogether, these data suggest that mitochondrial dynamics is closely regulated by mitochondrial Ca²⁺ dynamics under chronic HG challenge.

To mechanistically connect sustained organelle communication with dysregulation of mitochondrial dynamics and β-cell function over time, we used a recombinant construct encoding a red-fluorescent synthetic linker increasing ER-mitochondria tethering (21) (Supplementary Fig. 1F). TEM analysis confirmed that the linker expression significantly increased the surface of organelle contacts in fluorescence-sorted INS-1E cells (Fig. 7A and B). Interestingly, the enhancement of organelle contacts affects all distances of contacts (Fig. 7C) without changing mitochondrial number (Fig. 7D). However, the ER appeared dilated and mitochondria swollen in INS-1E cells overexpressing the linker (Fig. 7A). Importantly, ATP-stimulated organelle Ca²⁺ exchange was altered in INS-1E cells overexpressing the linker (Fig. 7F and G), whereas basal mitochondrial Ca²⁺ levels were not modified (Fig. 7E). We then measured 48 h later the repercussions on the mitochondrial network and GSIS. As shown in Fig. 7H, enhancement of organelle contacts induced mitochondrial fragmentation, as illustrated by the reduction of both AR and FF parameters in red INS-1E cells (Fig. 7I and J). Importantly, the reinforcement of organelle tethering by the overexpression of the linker significantly reduced GSIS in INS-1E cells (Fig. 7K), whereas insulin content was slightly increased (Fig. 7L). Linker-mediated alterations of GSIS were conserved after normalization by cellular proteins (Supplementary Fig. 2G) or insulin content (Supplementary Fig. 2H). Altogether, these data confirm the important role of MAMs in the control of mitochondria dynamics and β-cell function.

MAMs were involved in the control of insulin action (16–18,29), and disruption of MAMs was observed in pancreatic islets from donors with T2D (19). Therefore, MAMs could be an important intracellular marker reflecting the capacity to fulfill insulin needs, and playing an important role in regulating glucose homeostasis (33). However, the role of MAMs in both β-cell function and glucotoxicity-induced dysfunction has remained largely unknown. In the current study, we pointed out in rodent and human in vitro models a time-dependent regulation of ER-mitochondria communication by glucose, starting from an acute induction of MAM amount and function by glucose toward a chronic disruption of organelle Ca²⁺ exchange, linked to an emptying of the ER Ca²⁺ store, which likely progressively participates in both ER stress and mitochondrial fission and ultimately in altered GSIS.

Here, we newly identified that acute glucose stimulation of INS-1E cells stimulates ER-mitochondria interactions and Ca²⁺ exchange, in parallel with insulin secretion. Furthermore, we demonstrated that Grp75-mediated structural and functional disruption of MAMs alters GSIS in INS-1E cells, pointing out that organelle coupling is a new regulator of insulin secretion by β-cells. As Ca²⁺ signaling at MAMs was reported to control mitochondrial bioenergetics (34), and as Ca²⁺ uptake by mitochondria is essential for glucose-stimulated ATP increase in β-cells (35), it is likely that this regulation participates in the stimulation of mitochondrial ATP production, which is required to stimulate insulin release. This acute regulation of MAM integrity with HG in β-cells is different from what we previously observed in the liver (27). We think this difference may be related to the different functions of the two tissues. Indeed, the liver is a lipogenic organ where an increase of glucose flux redirects glucose through the storage pathways into triglycerides, in addition to glycogen. Under such conditions, glucose-induced reduction of MAMs (which requires 4 h) should contribute to slowing down mitochondrial lipid oxidation and favoring lipogenesis. In contrast, pancreatic β-cells are hormone-secreting cells, where glucose is a dynamic signal to increase insulin secretion. Therefore, the upregulation of MAMs by glucose is more rapid (30 min) and allows the boosting of mitochondrial ATP synthesis. It is likely that the regulation of MAMs by glucose in both tissues involves different mechanisms. Anyway, the dynamic increase of MAMs by transient exposure to HG is a new factor participating in GSIS and has a beneficial effect on pancreatic β-cells, supporting the previous notion that mitochondrial Ca²⁺ accumulation is important for sustained insulin secretion (36).

However, as often happens, critical processes in physiology can promote disease in chronic situations. We found that glucotoxicity-mediated alterations of GSIS involve a defect of both insulin content and secretion, as previously observed in T2D islets compared with control islets (10). We further demonstrated that glucotoxicity-associated β-cell dysfunction also increased organelle tethering but blunted ER-mitochondria Ca²⁺ transfers over time. Furthermore, glucotoxicity is also associated with an increase of basal Ca²⁺ mitochondrial content and a reduction of ER Ca²⁺ stores over time, without a significant change in cytosolic Ca²⁺ levels. Altogether, these data point to an importance of the ER in HG-associated mitochondrial Ca²⁺ changes. This point differs from the study of Tarasov et al. (37), showing that change in ER Ca²⁺ levels poorly affected the interplay between cytosolic and mitochondrial Ca²⁺ concentrations. Their measurements were performed on dispersed mouse-isolated β-cells in unstressed conditions and after electrical stimulation (involving an influx of Ca²⁺), compared with our study where we used insulinoma-derived rat cells and ATP as agonist in free-Ca²⁺ extracellular medium to stimulate mitochondrial Ca²⁺ accumulation. All these parameters could have an impact on the influence of ER Ca²⁺ on mitochondrial Ca²⁺ levels. Furthermore, HG-mediated induction in the expression of MICU1, and likely of MCU, may participate in increased basal mitochondrial Ca²⁺ levels in INS-1E cells. Therefore, we propose that chronic HG stimulation empties the ER, due to the prolonged organelle Ca²⁺ exchange, leading to increased basal mitochondrial Ca²⁺ accumulation and reduced Ca²⁺ transfer to mitochondria over time. Of course, we cannot exclude at this stage that other mechanisms participate in both glucotoxicity-mediated disruption of organelle Ca²⁺ exchange (i.e., alterations in IP3 signaling or in the structure/activity of the Ca²⁺ channeling complex) and mitochondrial Ca²⁺ accumulation (i.e., direct import from cytoplasm) in β-cells.

As with the majority of studies, our Ca²⁺ experiments are subject to limitations. 1) They were performed only in INS-1E cells, as these experiments required transfection of Ca²⁺ probes. 2) They were performed in the absence of extracellular Ca²⁺ in order to exclude an activation of P2X receptors by ATP and an influx of Ca²⁺ that was reported to participate in glucose-mediated mitochondrial Ca²⁺ rise (35). 3) We used ATP to stimulate ER-mitochondria Ca²⁺ exchange in INS-1E cells, as classically performed in numerous cell types (17,27,29,38–40). Moreover, ATP is a physiological extracellular signaling molecule of β-cells, arising from presynaptic terminals innervating the islets (as acetylcholine), from insulin secretion vesicles, or from connexin hemichannels and pannexin channels (41), and activation of P2YR was known to stimulate insulin secretion in a glucose-dependent manner (42). Altogether, these data validate the physiological relevance of ATP in β-cells to stimulate organelle Ca²⁺ exchange. 4) Acetylcholine treatment (another way to stimulate ER Ca²⁺ release) did not modify cytosolic or mitochondrial Ca²⁺ levels in our batch of INS-1E cells (data not shown), and it is unclear for us whether this absence of effect reflects a specificity of our batch of INS-1E cells or is specific to insulinoma-derived rat β-cells. 5) The conditions of Ca²⁺ measurement are therefore different from the conditions of GSIS where extracellular Ca²⁺ is necessary. However, we confirmed that the effects of HG on basal mitochondrial or ER Ca²⁺ levels were similar in the presence of extracellular Ca²⁺ to those found in its absence (data not shown). Therefore, in our condition, ATP only stimulates IP3R-mediated ER Ca²⁺ release, by activating membrane-bound P2YR purinergic receptors and increasing IP3 levels by the activation of phospholipase C. We are aware that in these conditions, changes in mitochondrial Ca²⁺ could be due to either a direct transfer from the ER to mitochondria or a release by the ER in the cytosol and a recapture of Ca²⁺ from the cytosol to the mitochondria. However, if this last scenario occurred, its involvement was minor in our conditions, as we only observed a weak rise in mitochondrial Ca²⁺ after TG-mediated ER Ca²⁺ release into cytosol (around 10 times less than in response to ATP) and as ATP-stimulated mitochondrial Ca²⁺ accumulation did not occur anymore when the ER Ca²⁺ store was depleted by TG treatment. In addition, ATP-stimulated mitochondrial Ca²⁺ accumulation was really rapid in our experiments, in favor of a direct transfer from the ER to mitochondria thanks to microdomains of high Ca²⁺, allowing for the bypass of the low affinity of MCU for Ca²⁺ (43). We think that an intermediary increase of cytosolic Ca²⁺ should delay ATP-mediated mitochondrial Ca²⁺ change, as diffusion of Ca²⁺ into the cytosol is low (44). Altogether, these data therefore suggest that at least part of ATP-stimulated mitochondrial Ca²⁺ accumulation comes directly from the ER through MAMs and is important for the control of insulin secretion by β-cells.

Moreover, we found that glucotoxicity-related β-cell dysfunction was also associated with ER stress and altered mitochondrial function and dynamics in both rodent and human β-islets, in agreement with structural alterations of both organelles in β-cells of T2D islets (45). Although it is unclear from our study whether these intracellular changes are interrelated and what the underlying molecular mechanisms are, our data suggest that glucotoxicity-associated mitochondrial fission could be linked to mitochondrial Ca²⁺ accumulation, as the more mitochondria increase their Ca²⁺ contents, the more they are fragmented, in agreement with other studies (46,47). Interestingly, acute and hyperactivation of both UPR signaling and mitochondrial fission can also have beneficial and detrimental effects, respectively, on β-cell function, as observed here for Ca²⁺ fluxes at the MAM interface. Indeed, acute HG treatment induced mild UPR signaling, activating IRE1α and leading to glucose-induced insulin biosynthesis, whereas chronic HG stimulation caused hyperactivation of IRE1α, leading to the suppression of insulin expression (48). Similarly, transient HG-induced mitochondrial fission is required for optimal GSIS (28), whereas we and others (32) found that glucotoxicity induced mitochondrial fragmentation.

Last, we demonstrated that ER-mitochondria tethering using a synthetic linker increased ER-mitochondria interactions but disturbed organelle Ca²⁺ exchange, likely through a lack of structural arrangement of Ca²⁺ channels due to reduced contact gap between organelles (21,49). However, we cannot exclude that the chronicity of overexpression participates in the reduction of organelle Ca²⁺ exchange. In these conditions, we observed mitochondria fission and altered GSIS in β-cells, suggesting that sustained organelle tethering is detrimental for β-cell function. It is likely that reinforcement of organelle interactions also empties the ER over time, leading to disrupted organelle Ca²⁺ exchange and to β-cell dysfunction. In agreement, the rupture of ER Ca²⁺ homeostasis, likely linked to a combination of Ca²⁺ store release and extracellular Ca²⁺ influx, was shown to play a major role in β-cell dysfunction under diabetic conditions (50,51).

We thus propose a model by which glucose differentially regulates functional ER-mitochondria interactions over time and subsequently regulates GSIS in β-cells. An acute increase in glucose concentrations increases organelle communication, likely to stimulate mitochondrial bioenergetics and ATP synthesis, and triggers insulin secretion. However, overstimulation of organelle interactions by chronic HG exposure induces a progressive depletion of ER Ca²⁺ stocks, an increase in mitochondrial Ca²⁺ content, and a disruption of Ca²⁺ transfer from the ER to mitochondria. Simultaneously, a time-dependent alteration of Ca²⁺ signaling participates in both ER stress and mitochondrial fragmentation and dysfunction. Organelle stresses might maintain increased organelle tethering as an adaptive mechanism to restore homeostasis. Interestingly, the experimental induction of ER-mitochondria interactions using a linker mimics the glucotoxic situation, as it fragments mitochondria and blunts GSIS. Altogether, these effects subsequently could create a vicious cycle that contributes to the progressive decrease of β-cell function and thereby to the worsening of the disease over time (Fig. 8). Therefore, disrupted Ca²⁺ signaling at the MAM interface might be a key feature of glucotoxicity-mediated β-cell dysfunction.

A striking observation is the constant increase of organelle interactions with both acute and longer HG exposure, suggesting that the glucose-mediated increase of MAMs could be at the same time a stimulatory and a protective mechanism to compensate for reduced organelle Ca²⁺ exchange, respectively. HG-mediated organelle stresses could participate in the increased organelle tethering over time in β-cells, as both tunicamycin-induced ER stress and oligomycin-induced mitochondrial stress are able to increase ER-mitochondria interactions. In agreement, tunicamycin-induced ER stress increased MAMs in HeLa cells (52) and in liver of lean mice (29), pointing to increased MAM formation as a generalized response to ER stress. However, we assume that the adaptive increase of MAMs in glucotoxic situations should be transitory, as we observed reduced organelle interactions in β-cells of patients with T2D as compared with control subjects (19). Furthermore, cells in culture and patients with diabetes differ in terms of chronicity, as cells are exposed to several days of HG against years for patients. Last, patients also present chronic dyslipidemia (lipotoxicity), and we previously showed that palmitate reduced organelle interactions in Min6B1 cells (19). Presently, the combined effect of glucolipotoxicity on MAMs is unknown and requires further investigations to understand the human pathology. Therefore, we propose that MAM integrity could be dynamically and differentially regulated in β-cells during the progression of T2D pathology, highlighting that MAM regulation plays an important role in the control of β-cell function.

Acknowledgments. The authors thank Pierre Maechler (University Medical Center, Geneva, Switzerland) for the generous gift of INS-1E cells, Gyorgi Hajnoczky (Thomas Jefferson Institute, Philadelphia, PA) for the generous gift of the ER-mitochondria linker and for sharing his macro to analyze ER-mitochondria interactions by TEM, Roger Tsien (University of California, San Diego, San Diego, CA) for the generous gift of the 4mtD3cpv vector, and Maria Teresa Alonso (University of Valladolid, Valladolid, Spain) for the generous gift of the erGAP1probe.

Funding. This work was supported by INSERM. F.D. was supported by a research fellowship from the French Ministry of Higher Education and Research. Human islets were mainly provided through the JDRF award 31-2008-416 (ECIT Islets for Basic Research Program).

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

Author Contributions. F.D., C.T., J.R., and A.-M.M. designed the experiments, researched data, contributed to discussions, and wrote the manuscript. B.P. analyzed TEM data, contributed to discussions, and reviewed the manuscript. S.D., Y.G., and F.V.C. helped with calcium experiments, contributed to discussions, and reviewed the manuscript. S.P., M.-A.C., K.C., and E.E.-C. researched data. A.-M.M. 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.

Data and Resource Availability. All data generated or analyzed during this study are included in the published article (and its Supplementary Data). No applicable resources were generated or analyzed during the current study.

Prior Presentation. Parts of this study were presented as a poster at the European Molecular Biology Organization Workshop on Membrane Contact Sites in Health and Diseases, Arosa, Switzerland, 21–25 September 2018, and as an oral presentation at both the 53rd Annual Meeting of the European Association for the Study of Diabetes, Lisbon, Portugal, 11–15 September 2017, and the 2018 Annual Meeting of the Francophone Society of Diabetes, Nantes, France, 20–23 March 2018.