Imeglimin Normalizes Glucose Tolerance and Insulin Sensitivity and Improves Mitochondrial Function in Liver of a High-Fat, High-Sucrose Diet Mice Model Free

Type 2 diabetes is a worldwide threat that has been labeled as a great challenge to human health in the 21st century. The total number of people with diabetes is estimated to rise from 382 million today to 582 million by 2035 (1), and the prevalence of type 2 diabetes is expected to rise in children and adolescents around the world in all ethnicities (2). In 2013, 5.1 million deaths were attributed to diabetes (1). These alarming statistics highlight that only a few effective treatment strategies exist to fight this multifactorial disease. Currently, pharmacological management of the disease consists of combination therapy to achieve glycemic control. Therefore, development of antidiabetic agents with improved safety and effectiveness is urgently needed. Imeglimin is the first in a new tetrahydrotriazine-containing class of oral glucose-lowering agents—the glimins—and is currently in phase 2b clinical development (U.S./European Union EudraCT 2012-004045-33). Several clinical trials have shown imeglimin to be well tolerated and exhibit benefits on HbA_1c as monotherapy and add-on therapy (3–5). Imeglimin acts on the liver, muscle, and the pancreas (6), three key organs involved in the pathophysiology of type 2 diabetes through suspected mechanisms targeting the mitochondria and reduced oxidative stress. Imeglimin decreases hepatic glucose production and increases muscle glucose uptake (6). Recently, imeglimin demonstrated increased insulin secretion in response to glucose in diabetic patients during a hyperglycemic clamp study (7). To further elucidate the mechanism of action of imeglimin and its capacity to improve insulin sensitivity, we treated diabetic mice fed a high-fat, high-sucrose diet (HFHSD). This model, characterized by insulin resistance, glucose intolerance, liver steatosis, and mitochondrial dysfunction, is commonly used to study the mechanisms of insulin resistance and the effects of drugs.

Male C57BL/6JOlaHsd mice purchased from Harlan (Gannat, France) at 4 weeks old were housed at 22°C with a 12-h light/dark cycle. Procedures were conducted in accordance with the institutional guidelines for the care and use of laboratory animals and were approved by a regional ethics committee. After 1 week of acclimatization, 5–6-week-old mice were divided into two groups: one with free access to a standard chow diet (SD) (Harlan) and the other with free access to a pelleted HFHSD diet (TD99249; Harlan) for 16 weeks as previously described (8). Animals received imeglimin 200 mg/kg b.i.d. by oral gavage during the last 6 weeks of HFHSD feeding. Control SD and HFHSD mice were treated by oral gavage with methylcellulose 0.5% as a vehicle for drug treatment (5 mL/kg). Food intake was measured every day during the first week and twice a week until the end of the experiment. Results are expressed as grams per day per mouse.

Blood glucose was taken weekly from tails in a fed state until week 11. From weeks 11 to 16, blood glucose was taken each week from tails in a fed state 2 h postgavage. Blood glucose levels were measured with a glucometer (Roche Diagnostics). Blood samples were collected by retro-orbital sampling (fed state and 2 h postgavage) at 65 and 112 days after starting the HFHSD and at study end.

Glucose and insulin tolerance tests were performed on 6-h–fasted mice and 2 h postgavage. Glucose 2 mg/g body weight or insulin 0.75 mU/g body weight were injected intraperitoneally; blood glucose levels were monitored with a glucometer at the indicated time points. During the intraperitoneal glucose tolerance test (ipGTT), plasma insulin and C-peptide were quantified before and 15 min after glucose injection. C-peptide and insulin levels were measured by ELISA methodology (80-INSMSU-E10 and 80-CPTMS-E01, respectively; ALPCO). Relative insulin clearance was estimated 15 min after glucose injection as the ratio (C-peptide − insulin)/C-peptide.

A subgroup of mice (SD, HFHSD, and HFHSD + imeglimin, n = 10) were fasted for 6 h. NaCl (0.9%) or insulin (10 mU/mice in NaCl 0.9%) were injected intraperitoneally, and animals were killed 15 min later by cervical dislocation. Liver and gastrocnemius muscle were rapidly removed and frozen until use for Western blot analysis of phosphokinase B (PKB) phosphorylation.

Total lipids were extracted from tissue with ethanol/chloroform (1:2 volume for volume [v/v]). Before extraction, internal standards (1,2-diheptadecanoyl-sn-gycero-3-phosphocholine; 1,2-diheptadecanoyl-sn-gycero-3-phosphoethanolamine; cholesterol ester 17:0; di-17:0 diglyceride; tri-17:0 triglyceride; and stigmasterol) were added. The organic phases were evaporated under N₂, and the different lipid classes separated by thin layer chromatography using the solvent hexane-diethylether-acetic acid (80:20:1 v/v/v) as eluent. Total phospholipids, diglycerides, triglycerides, and cholesterol esters were treated with 14% boron trifluoride in methanol (BF3/methanol). The resulting fatty acid methyl esters were analyzed by gas chromatography using a DELSI chromatograph model DI 200 equipped with an SP-2380 capillary column (60 m × 0.22 mm). Cholesterol was derivatized with bis(trimethylsilyl)trifluoroacetamine and analyzed by gas chromatography–mass spectrometry operated in positive chemical ionization mode. Acylcarnitine analysis was performed by flow injection tandem mass spectrometry (API 4500; AB SCIEX) as previously described except that butyl derivatives were used (9).

Mouse liver mitochondria were isolated using a standard differential centrifugation procedure in 250 mmol/L sucrose, 20 mmol/L Tris-HCl, and 1 mmol/L EGTA, pH 7.4 (10). Mitochondrial oxygen consumption rate was measured on freshly prepared mitochondria at 30°C using a Clark-type O₂ electrode (S1; Hansatech Instruments, Norfolk, U.K.) in a 1-mL chamber filled with respiration buffer (125 mmol ⋅ L⁻¹ KCl, 10 mmol ⋅ L⁻¹ Pi-Tris, 20 mmol ⋅ L⁻¹ Tris-HCl, and 0.1 mmol ⋅ L⁻¹ EGTA, pH 7.2) and using 1 mg mitochondrial proteins ⋅ mL⁻¹. Measurements were conducted in the presence of either glutamate (5 mmol ⋅ L⁻¹)/malate (2.5 mmol ⋅ L⁻¹) (GM) or succinate (5 mmol ⋅ L⁻¹) as substrates after the addition of 1 mmol ⋅ L⁻¹ ADP (state 3) followed by 1.5 μg ⋅ mL⁻¹ oligomycin (state 4) (Supplementary Fig. 3). Reactive oxygen species (ROS) production was estimated by measuring H₂O₂ release in a stirred 1-mL chamber containing 0.2 mg mitochondria and filled with a respiration buffer containing 6 IU horseradish peroxidase and 1 μmol ⋅ L⁻¹ Amplex Red (excitation 560 nm, emission 584 nm) and the same substrates as for respiration and using a fluorescence spectrophotometer (F-7000 FL; Hitachi U.S.A.). Measurements were conducted in basal conditions and after sequential additions of various substrates and 2 μmol ⋅ L⁻¹ rotenone (Supplementary Fig. 3). Results are expressed in pmol H₂O₂ ⋅ min⁻¹ ⋅ mg ⋅ protein⁻¹ using H₂O₂ standard solutions.

Liver and gastrocnemius muscle samples were lysed in PBS containing 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS supplemented with EDTA (5 mmol/L), Na₃VO₄ (1 mmol/L), NaF (20 mmol/L), dithiothreitol (1 mmol/L), and protease inhibitor cocktail (P2714; Sigma). Proteins were separated by SDS-10% PAGE, transferred to polyvinylidene difluoride membrane, and incubated overnight with primary antibodies. Primary antibodies used were total Akt/PKB (9272; Cell Signaling), phospho-Akt/PKB Ser473 (9271; Cell Signaling), OXPHOS (ab110413; Abcam), PGC1α (516557; Calbiochem), OPA1 (612606; BD Biosciences), DLP1 (611112; BD Biosciences), PARKIN (ab77924; Abcam), PINK1 (ab23707; Abcam), ANT2 (SC-9299; Santa Cruz), UCP2 (615902; BioLegend), FAT/CD36 (SC-9154; Santa Cruz), and CPT1A (ab128568; Abcam). The signal was detected with a horseradish peroxidase–conjugated secondary antibody (172-10-19; Bio-Rad) and revealed with an enhanced chemiluminescence system (Pierce). Each protein was normalized by tubulin expression in each sample, with an internal control on each gel to normalize the intergel variability. Results are expressed versus the SD group.

Total RNA was extracted with TRI Reagent Solution (Sigma). Target mRNA levels were measured by reverse transcription followed by real-time PCR using a Rotor-Gene 6000 (Corbett Research). A standard curve was systematically generated with six different amounts of purified target cDNA, and each assay was performed in duplicate. We measured TATA-binding protein (TBP) mRNA as a reference gene; results are expressed as a ratio that refers to TBP expression and is normalized to the SD group.

Total DNA from liver was extracted using phenol/chloroform/isoamyl alcohol (25:24:1) followed by ethanol precipitation. The difference in mitochondrial DNA (mtDNA) was estimated by measuring the level of a mitochondrial-encoded gene (Cox1) versus a nuclear-encoded gene (cyclophilin A [Ppia]) using real-time quantitative PCR as previously described (8).

Liver was dissected and fixed in 2% glutaraldehyde for 2 h at 4°C, postfixed in 1% osmium tetroxide for 1 h at 4°C, dehydrated, and embedded in epon. The tissue was sliced using an RMC/MTX ultramicrotome (Elexience), and ultrathin sections (60–80 nm) were mounted on copper grids, contrasted with 8% uranyl acetate and lead citrate, and observed with a JEOL 1200 EX transmission electron microscope (JEOL Ltd.) equipped with a MegaView II high-resolution TEM camera. Analysis was performed with the Soft Imaging System (ELOÏSE s.a.r.l.).

Liver was frozen in Tissue-Tek, and 10-μm sections were cut with a microtome (CMT950A; MICROM GmbH). Oil Red O and hematoxylin-eosin staining were performed, and sections were observed with an optical microscope (Axioskop 2; Zeiss) linked to a camera (Axiocam ERc 5s; Zeiss).

Lipids were extracted from mitochondrial preparations (1–2 mg protein) according to the method of Folch et al. (11). Samples were solubilized in 100 μL chloroform/methanol/water (60/30/4.5). Extracted lipids were further analyzed using a ZORBAX Eclipse Plus C18 column plugged on an LC 1200-MS² 6460 QqQ system equipped with an ESI source (Agilent Technologies).

Cardiolipins (2–4 μL) were separated at a flow rate of 0.4 mL/min at 30°C with a linear gradient of acetonitrile/water/ammonium hydroxide/acetic acid (90/10/0.2/0.5 v/v/v/v) (A) and of isopropanol/water/ammonium hydroxide/acetic acid (90/10/0.2/0.5 v/v/v/v) (B) as follows: 50% B for 5 min, up to 80% B in 10 min, up to 100% in 15 min, and 100% for 5 min. Acquisition was performed in negative selected reaction monitoring ion mode (source temperature 325°C, nebulizer gas flow rate 10 L/min, sheath gas flow 12 L/min, temperature 350°C, capillary 3,500 V, nozzle 1,000 V, fragmentor 280 V, collision energy 76 V). Transitions from molecular ion [M − 1] to linoleic acid (charge/mass ratio 279.2) or oleic acid (charge/mass ratio 281.2) were used for cardiolipin quantitation.

Phospholipids (2–4 μL) were separated at a flow rate of 0.25 mL/min at 50 °C with a linear gradient of water/methanol (60/40 v/v), 10 mmol/L ammonium acetate, 1 mmol/L acetic acid (A) and of isopropanol/methanol (90/10 v/v), 10 mmol/L ammonium acetate, 1 mmol/L acetic acid (B) as follows: 40% B for 1 min, up to 95% B in 15 min, and maintained at 100% for 1 min. Acquisition was performed in selected reaction monitoring ion mode (source temperature 325°C, nebulizer gas flow rate 10 L/min, sheath gas flow 10 L/min, temperature 400°C, capillary 3,500 V, nozzle 1,000 V). The transitions used were as follows: [M + 1] + → 184.1 (fragmentor 160 V, collision energy 20 V), [M + 1] + → [M − 140] (fragmentor 120 V, collision energy 17 V), and [M − 1] − → [M − 87] (fragmentor 150 V, collision energy 19 V) for phosphatidylcholines (PCs), phosphatidylethanolamines (PEs), and phosphatidylserines (PSs), respectively.

Rotenone-sensitive NADH-ubiquinone oxidoreductase (EC 1.6.5.3, complex I [CI]) was assayed using 100 μmol ⋅ L⁻¹ decylubiquinone as an electron acceptor and 200 μmol ⋅ L⁻¹ NADH as a donor in a 10 mmol ⋅ L⁻¹ KH₂PO₄/K₂HPO₄ buffer, pH 7.5, containing 3.75 mg ⋅ mL⁻¹ BSA, 2 mmol ⋅ L⁻¹ KCN, and 7.5 μmol ⋅ L⁻¹ antimycin A. NADH oxidation was measured at 340 nm before and after the addition of 4 μmol ⋅ L⁻¹ rotenone to allow the calculation of the rotenone-sensitive–specific activity, which is characteristic of CI.

Succinate-ubiquinone reductase (EC 1.3.5.1, complex II [CII]) activity was quantified by measuring the decrease in absorbance resulting from the reduction of 100 μmol ⋅ L⁻¹ dichlorophenolindophenol at 600 nm. The measurement was performed in 50 mmol ⋅ L⁻¹ KH₂PO₄/K₂HPO₄ buffer, pH 7.5, in the presence of 30 mmol ⋅ L⁻¹ succinate, 100 μmol ⋅ L⁻¹ decylubiquinone, 2 μmol ⋅ L⁻¹ rotenone, and 2 mmol ⋅ L⁻¹ KCN.

Coenzyme Q-cytochrome c-oxidoreductase activity (EC 1.10.2.2, complex III [CIII], sometimes called the cytochrome bc1 complex), was quantified by measuring the increase in absorbance resulting from the reduction of 100 μmol ⋅ L⁻¹ cytochrome c at 550 nm. The measurement was performed in 50 mmol ⋅ L⁻¹ KH₂PO₄/K₂HPO₄ buffer, pH 7.5, in the presence of 100 μmol ⋅ L⁻¹ decylubiquinone previously reduced by dithionite, 50 μmol ⋅ L⁻¹ EDTA, and 1 mmol ⋅ L⁻¹ KCN. The specific activity was calculated by subtracting the activity obtained before and after addition of 5 μg/mL antimycin A.

3-Hydroxyacyl-CoA dehydrogenase (HAD) (EC 1.1.1.35) activity was quantified by measuring the decrease in absorbance at 340 nm resulting from the oxidation of NADH (200 μmol ⋅ L⁻¹) and the reduction of S-acetoacetyl-CoA (50 μmol ⋅ L⁻¹). The measurement was performed in imidazole (40 mmol ⋅ L⁻¹) and EDTA (60 μmol ⋅ L⁻¹), pH 7.

All data are presented as mean ± SEM. One-way ANOVA was used to determine the global effects of treatment. When appropriate, differences between groups were tested with a protected least significant difference Fisher post hoc test. Statistical significance was accepted at P < 0.05 (*significantly different between SD and HFHSD, #significantly different between HFHSD and HFHSD + imeglimin). Mann-Whitney tests were applied when values were not normally distributed.

HFHSD-fed mice were used as a model of altered glucose homeostasis in rodents (8,12). Despite a decrease in food intake (g/day) (Supplementary Fig. 1A), HFHSD induced an increase in daily caloric intake compared with SD (SD = 11.45 kcal/day vs. HFHSD = 13.04 kcal/day). As expected, HFHSD mice were obese, hyperglycemic, glucose intolerant, and insulin resistant compared with SD mice (Fig. 1A–C and Supplementary Fig. 1B–C). The mass of liver, gastrocnemius, and epididymal fat is increased in HFHSD mice (Supplementary Fig. 1D–G). HFHSD mice also showed higher levels of both insulin and C-peptide levels (Fig. 1D) than SD mice, indicating a modification of insulin secretion and/or clearance. Whereas C-peptide concentration was fivefold higher than insulin in SD mice, it was only twofold higher in HFHSD mice (Fig. 1D), indicating a large decrease in insulin clearance with this diet.

Imeglimin was administered orally at 200 mg/kg b.i.d. during the last 6 weeks of the HFHSD feeding protocol. A slight decrease in body weight and food intake associated with some diarrhea was observed but only during the first few days of treatment (Fig. 1A and Supplementary Fig. 1A). These effects were transient and disappeared after a few days, remaining comparable to control HFHSD mice until the end of treatment. Furthermore, no effect was measured on fat pad weight or on liver and skeletal muscles mass with imeglimin (Supplementary Fig. 1D–G). However, imeglimin significantly decreased hyperglycemia (Fig. 1B), restored normal glucose tolerance (Fig. 1B and Supplementary Fig. 1B), and improved insulin sensitivity (Fig. 1C and Supplementary Fig. 1C). Imeglimin restored insulin secretion during ipGTT (+98% vs. HFHSD mice, P < 0.01) and improved C-peptide secretion during ipGTT (Fig. 1D) but did not modify insulin clearance under HFHSD (Fig. 1D). In fact, relative insulin clearance [(C-peptide − insulin)/C-peptide] was decreased by HFHSD (82% in SD group vs. 50% in HFHSD group, P < 0.001) but was not modified by imeglimin (41% in HFHSD + imeglimin group). Together, these data suggest that imeglimin improves glucose homeostasis in HFHSD mice independent of an effect on whole-body composition.

HFHSD feeding altered insulin-stimulated PKB phosphorylation in both liver (Fig. 2A) and skeletal muscle (Fig. 2B). Imeglimin improved insulin response in both tissues of HFHSD mice (+157% and +198%, respectively; P < 0.05 vs. HFHSD) (Fig. 2A and B), indicating an insulin-sensitizing effect of the molecule in both organs.

HFHSD feeding induced lipid accumulation in the liver compared with SD mice, as demonstrated by Oil Red O staining (Fig. 3A), electronic microscopy analysis (Fig. 3B), and total lipid quantification (Fig. 3C). HFHSD mice showed higher hepatic content of triglyceride (Fig. 3D), cholesterol (Fig. 3E), and diacylglycerol (DAG) (Fig. 3F) compared with SD mice. Imeglimin reduced all these lipid parameters in the livers of HFHSD mice (Fig. 3A–F), indicating that imeglimin decreased hepatic steatosis in HFHSD mice. Biochemical measurements showed that HFHSD significantly increased the mRNA levels of several genes of lipid metabolism (Cpt1a, Acaca, Srebp1c, Mlxipl, Acly) (Table 1). However, imeglimin was not associated with a significant improvement in the expression of these genes despite a tendency for decreased expression of the Srebp1c gene (P = 0.16). In addition, glycogen content was not modified by either the diet or the treatment (data not shown).

Next, we investigated whether the reduction in lipid accumulation could be related to an increase in the oxidation of fatty acids in the liver of imeglimin-treated HFHSD mice. Measurements from indirect calorimetry were not obtained in this study because a previous study in rats fed a standard and high-fat diet showed that the respiratory quotient was not modified by imeglimin (G.V., unpublished data). Nevertheless, the effect of imeglimin on fatty acid oxidation was supported by the measurement of HAD activity and the level of FAT/CD36 protein. Imeglimin increased both parameters (+14% and +15% vs. HFHSD group, respectively; P < 0.05) (Table 2), suggesting an increased lipid oxidation in the liver of HFHSD mice. Moreover, total and free carnitine levels as well as short-chain acylcarnitine levels, which are decreased by HFHSD (−34%, −41%, and −33% vs. SD mice; P < 0.01), are significantly restored by imeglimin (+45%, +52%, and +44% vs. HFHSD mice; P < 0.01) (Fig. 3G). However, imeglimin did not modify CPT1A protein levels in the liver of HFHSD mice (Table 2 and Supplementary Fig. 2).

Liver mitochondrial function is altered in insulin resistance states (13,14); therefore, we investigated whether imeglimin affects mitochondria, which are involved in both hepatic glucose and lipid homeostasis. For that, we isolated mitochondria from mouse liver and assessed both the respiration rate and the H₂O₂ production with either GM or succinate (Supplementary Fig. 3). Mitochondrial state 3 (phosphorylating) oxygen consumption was higher in HFHSD mice than in SD mice independent of the substrate used (Fig. 4A and B). In contrast, state 2 and 4 respirations were not modified by the diet. Imeglimin decreased state 3 respiration with GM (Fig. 4A), but increased respiration with succinate (Fig. 4B). Succinate state 2 and 4 rates were also significantly increased with imeglimin (Fig. 4B).

Liver mitochondria from HFHSD mice displayed an increase in H₂O₂ production with GM, which was not altered by imeglimin (Fig. 4C). However, HFHSD dramatically increased mitochondrial H₂O₂ production with succinate compared with mitochondria isolated from SD mice (+69.5% vs. SD mice, P < 0.001) (Fig. 4D). Rotenone addition, known to inhibit reverse electron flux from CII to CI (15), totally inhibited this production (Fig. 4D), suggesting an involvement of CI in mitochondrial ROS production in HFHSD liver. Imeglimin significantly reduced succinate-induced mitochondrial H₂O₂ production in HFHSD liver (−32% vs. HFHSD, P < 0.001) (Fig. 4D). HFHSD increased the mRNA levels of Gpx1 and Cybb (subunit P40 of NADPH oxidase), whereas imeglimin treatment tended to reduce the adaptive increases observed in HFHSD mouse liver (−9% and −14%, respectively, vs. HFHSD) (Table 1). Imeglimin significantly reduced Ncf4 (subunit GP91 of NADPH oxidase) (−19% vs. HFHSD, P < 0.05) (Table 1).

HFHSD feeding increased mtDNA content in liver as well as PGC1α protein levels, suggesting an increase in mitochondrial density/biogenesis. Imeglimin further increased mtDNA content without modifying PGC1α expression (Table 2). Measurements of the expression levels of mitochondrial dynamics markers (OPA1, DLP1) and mitophagy markers (PARKIN, PINK1) showed no differences between groups (Table 2). HFHSD feeding significantly reduced the expression and activity of CIII (−25% and −28%, respectively, vs. SD mice; P < 0.05) (Table 2) without affecting other complexes. Imeglimin totally restored CIII content and activity in HFHSD liver, reaching the level of the SD mice (+20% and +40%, respectively, vs. HFHSD; P < 0.05). In contrast, ATP synthase was overexpressed in HFHSD mice and more intensively in HFHSD mice treated with imeglimin. Finally, despite a similar CI content in all groups, imeglimin decreased CI activity in liver mitochondria of HFHSD mice (−25% vs. HFHSD, P < 0.05) (Table 2).

Because mitochondrial phospholipids protect and control the activity of OXPHOS complexes (16–19), liver mitochondrial membrane lipid composition was analyzed. HFHSD induced large changes in mitochondrial phospholipid composition compared with SD mice: Total cardiolipin levels increased, whereas PC, PE, phosphatidylinositol (PI), and PS levels were decreased (Table 3). Imeglimin amplified the effects of HFHSD on both cardiolipin and PS content, whereas it tended to restore PC, PE, and PI content to normal values in HFHSD mitochondria (Table 3).

Altered phospholipid composition of endoplasmic reticulum (ER) is known to be associated with hepatic ER stress and insulin resistance (20,21); therefore, the effect of imeglimin on hepatic ER stress in HFHSD mice was explored. There was a slight increase in the expression of ER stress markers in the livers of HFHSD mice compared with SD mice (Table 1). However, the treatment did not modulate their expression, suggesting that the mechanism of action of imeglimin does not modulate ER homeostasis.

Because type 2 diabetes is a progressive and multifactorial disease, the combination of several medicines targeting different elements of the pathology are frequently needed for optimal disease management. Recent clinical data demonstrated that in addition to its efficacy as monotherapy, imeglimin could complement the actions of metformin or sitagliptin by significantly improving HbA_1c and fasting plasma glucose in type 2 diabetic patients (3,4). In this study, we investigated the mechanism of action of imeglimin on glucose homeostasis in HFHSD-induced diabetic mice and demonstrated the efficacy of imeglimin on glucose homeostasis after 6 weeks of treatment in this model. Imeglimin demonstrated benefits on insulin sensitivity with an improvement in insulin signaling in both muscle and liver. We propose that the beneficial effects of imeglimin on liver are mediated at least in part through an action on mitochondria (Fig. 5). Indeed, imeglimin reduced lipid accumulation in the liver by improving mitochondrial density and function and increased mitochondrial density by an action independent of PGC1α protein levels and probably without modifying mitochondrial dynamics and mitophagy. Imeglimin also modulated the OXPHOS chain activity; it reduced CI activity and ROS produced from this complex when mitochondria oxidize succinate. Furthermore, imeglimin restored the expression of a subunit of CIII and totally restored CIII activity in HFHSD mice. Therefore, this dual effect of imeglimin (CI inhibition and CIII restoration) allows mitochondria to oxidize more CII substrates and, thus, potentially more lipids. This assumption is confirmed by the increase in both HAD activity and FAT/CD36 protein levels in addition to the restoration of total and free carnitine content and the elevated short-chain acylcarnitine in the livers of imeglimin-treated HFHSD mice. Moreover, imeglimin increased the oxidation capacity of liver mitochondria for succinate (or FADH₂) independently of the mitochondrial working state, suggesting an energy waste. The observed energy waste is unconventional because it is not a classical uncoupling. In fact, imeglimin increases mitochondrial respiration in state 2 and 4 with succinate but not with GM; a classical uncoupling agent or protein (DNP, CCCP, or UCP) acts on both substrates. Furthermore, uncoupling is not supported by the reduction in UCP2 protein levels by imeglimin. This energy waste is closer to a slipping of the mitochondrial pump resulting from the change in efficiency of the complex in pumping electrons when mitochondria used succinate rather than a change in membrane permeability. This subtle energy waste following imeglimin treatment could participate in the decrease in fatty acid accumulation in the liver. In agreement, liver-targeted mitochondrial uncoupler was recently shown to improve hepatic steatosis and insulin resistance in mice (22). Furthermore, reduction in both liver triglyceride and DAG levels and oxidative stress may contribute to improvements in hepatic insulin sensitivity induced by imeglimin. DAG-mediated activation of PKC and ROS-mediated activation of Jun NH₂-terminal kinase are well-described contributors to hepatic insulin resistance (22–27). In addition, we cannot exclude a role of imeglimin preventing other deleterious effects of oxidative stress, as previously reported (28,29).

Cardiolipin content increases in mitochondrial membranes could contribute to the increase in the respiration rate during overfeeding because cardiolipins are known to induce nonphosphorylating energy wasting in mitochondria (16,30). Increases in ATP synthase and ANT content could also amplify mitochondrial respiration. Although these mitochondrial adaptations in response to overfeeding should lead to increased OXPHOS capacity, this probably failed in HFHSD mice mitochondria because CIII content and activity decreased. Imeglimin reinforced HFHSD-induced adaptations by overincreasing cardiolipin, ATP synthase, and ANT mitochondrial content and restored CIII content and activity. These effects can explain imeglimin’s capacity to further increase succinate-driven respiration because it is associated with CI inhibition. Specific, tightly bound phospholipids, such as cardiolipins, are essential for the activity of the cytochrome bc1 complex (CIII), an integral membrane protein of the respiratory chain (31); therefore, the modulation of phospholipid composition by imeglimin through increasing cardiolipin content could contribute to improved mitochondrial function. Moreover, because cardiolipins modulate the interactions and activities of mitochondrial complexes (32,33), regulate OXPHOS chain efficacy, and reduce ROS production (34), we propose that the imeglimin-induced increase in cardiolipins could participate in inducing CIII activity and reducing ROS production by liver mitochondria. We demonstrate for the first time in our knowledge that an efficient antidiabetic treatment positively affects the liver mitochondrial phospholipid composition of HFHSD mice.

This study demonstrates important positive effects of 6 weeks of imeglimin on glycemia, glucose tolerance, and insulin sensitivity in HFHSD-induced diabetic mice. We propose that the beneficial effect of imeglimin on glucose homeostasis, particularly insulin sensitivity, involves improvements in hepatic mitochondrial function, leading to increased lipid oxidation and reduced ROS production. Imeglimin is the first antidiabetic compound that induces an increase in mitochondrial phospholipid composition, contributing to improvements in hepatic mitochondrial function. The mitochondrial effects of imeglimin, therefore, could participate in the imeglimin-mediated improvement of glucose homeostasis in patients with type 2 diabetes. Future studies are required to determine whether improvements in insulin sensitivity in skeletal muscle of HFHSD mice are related to the hepatic effects of imeglimin or whether a similar mechanism could occur directly in skeletal muscle.

Acknowledgments. The authors thank the PBES (Plateau de Biologie Expérimentale de la Souris, Lyon, France) for animal facilities; Elisabeth Errazuriz (CeCILE Imaging Center, Lyon, France) for technical help; Emily Tubbs (INSERM U1060, Faculté de Médecine Lyon-Sud) for technical help and English correction; Emmanuelle Loizon, Amelie Bravard, Pierre Theurey, and Abdallah Gharib (INSERM U1060, Faculté de Médecine Lyon-Sud) for technical help; and Christine Saban (Service Maladies Héréditaires du Métabolisme, Centre de Biologie et Pathologie Est, Centre Hospitalier Universitaire de Lyon et UMR) for fruitful contribution to the discussions.

Funding. This work was supported by INSERM.

Duality of Interest. This work was supported by Poxel SA. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. G.V. and J.R. contributed to the experimental design, researched data, contributed to the discussion, and wrote the manuscript. M.-A.C., N.B., A.D., E.M., A.-M.M., N.B.-H., J.-P.P.d.B., and C.A. researched data. É.F. and H.V. contributed to the discussion and review of the manuscript. S.H.-B. and S.B. contributed to the experimental design, discussion, and review of the manuscript. G.V. 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 as an oral presentation at the 74th Scientific Sessions of the American Diabetes Association, San Francisco, CA, 13–17 June 2014.