The Sodium–Glucose Cotransporter 2 Inhibitor Dapagliflozin Prevents Cardiomyopathy in a Diabetic Lipodystrophic Mouse Model

Type 2 diabetes mellitus (T2DM) is a well-recognized independent risk factor for heart failure (HF) (1,2). Whereas the prevalence of HF in the general population is 1–4%, it reaches ∼12% in patients with T2DM (3). In 1972, Rubler et al. (4) reported a specific diabetes-associated cardiac injury called diabetic cardiomyopathy. This cardiomyopathy is defined by ventricular dysfunction occurring without coronary disease or hypertension (1,5). Diabetic cardiomyopathy is also characterized by left ventricular (LV) hypertrophy, diastolic dysfunction, and myocardial fibrosis (1,5).

A large body of work indicates that diabetic cardiomyopathy is associated with altered cardiac energy metabolism (6). Indeed, in obese patients with T2DM, heart lipid uptake is increased (7). Several studies support that free fatty acid (FFA) accumulation leads to the increased production of diacylglycerol (DAG), ceramides, and reactive oxygen species (ROS), affecting cardiac insulin sensitivity (8,9) and cardiac contractility (6,10,11).

On the other hand, hyperglycemia and glucose overload have been involved in cardiac hypertrophy and dysfunction in the context of T2DM and obesity (12). The diabetic heart is simultaneously characterized by impaired insulin-stimulated glucose uptake (13) and obvious signs of glucose overload, such as ROS and advanced glycation end product (AGE) production as well as hexosamine pathway chronic activation (12,14). Interestingly, when comparing obese patients with and without diabetes, we previously demonstrated that hyperglycemia per se plays a central role in the impaired cardiac mitochondrial activity associated with myocardial contractile dysfunction (15).

The exact contribution of lipotoxicity and glucotoxicity remains unclear. Most rodent studies have been performed in obese animals, and it remains difficult, in this setting, to distinguish the respective contribution of obesity, insulin resistance, and hyperglycemia. To provide new insight into the mechanism driving the development of diabetic cardiomyopathy, we aimed to use a unique nonobese model of diabetes: the lipodystrophic Bscl2^−/− (seipin knockout [SKO]) mouse model.

Berardinelli-Seip congenital lipodystrophy (BSCL) is characterized by an almost complete lack of adipose tissue from birth or early infancy. In ∼50% of the cases, BSCL is the result of mutations in BSCL2 encoding the protein Seipin (16). BSCL is associated with major metabolic complications, such as insulin resistance progressing to overt diabetes, hypertriglyceridemia, and liver steatosis (17). In addition, cardiomyopathy frequently occurs in patients with BSCL (18). Most patients have concentric LV hypertrophy, equally split into mild, moderate, and severe, often associated with diastolic dysfunction but generally with preserved systolic function. About half of the patients (54%) display abnormal electrocardiogram (ECG) with prolonged QT intervals and nonspecific T-wave abnormalities (18). It is not clear yet whether this cardiomyopathy is associated with ectopic myocardial lipid accumulation (19–21). Thus, the phenotype of lipodystrophic cardiomyopathy is poorly characterized, and the underlying molecular mechanism remains unknown.

We and others (22–24) have previously shown that SKO mice are severely lipodystrophic and display insulin resistance, diabetes, and massive liver steatosis. In this study, we performed cardiac phenotyping of the SKO mice, and we used dapagliflozin treatment, a sodium–glucose cotransporter 2 (SGLT2), and the insulin sensitizer pioglitazone to reverse the hyperglycemia that appears to be central to the development of cardiomyopathy in these animals.

SKO mice were generated as previously described (24) and housed at 21°C with a 12:12-h light/dark cycle with free access to food and water. The experimental procedures were approved by the regional ethics committee (Comités d'éthique en expérimentation animale, Pays de la Loire, France) according to Directive 2010/63/EU of the European Union. For physical activity assessment, the mice were submitted to graduated aerobic exercise training, performed on a motor treadmill (slop 0°) over 4 weeks. Dapagliflozin, an SGLT-2 inhibitor, was obtained from Interchim (Montluçon, France) and added to the drinking water at a concentration of 1 mg/kg for 8 weeks. Pioglitazone (Interchim) was added to the diet at a concentration of 300 mg/kg (45 mg/kg/day) for 8 weeks. Mice started the treatment at 6 weeks of age. All animals were killed in a random-fed state.

ECGs were performed as previously described (25). Briefly, under isoflurane anesthesia, a six-lead ECG was recorded with 25-gauge subcutaneous electrodes on a computer through an analog-digital converter (IOX 1.585; EMKA Technologies, Paris, France) for monitoring and offline analysis (ECG Auto v3.2.0.2; EMKA Technologies).

Mice were anesthetized with isoflurane, and two-dimensional echocardiography was performed on mice using Vivid 7 Dimension ultrasonography (GE Healthcare) with a 14-MHz transducer. The Doppler-derived mitral deceleration time, the early diastolic (E), late diastolic (A), and E/A ratio were obtained to assess diastolic dysfunction.

Cardiac magnetic resonance (CMR) was performed using a 7-Tesla Bruker Pharmascan magnetic resonance imager (Bruker Biospin, Ettlingen, Germany) interfaced with a dedicated small-animal electrocardiographic and respiratory triggering system (SA Instruments). Ventricular function and wall deformation were assessed using a triggered cine FLASH sequence and tagged CMR frames, respectively, as previously described (26).

Mice were fasted for 2 h and then anesthetized with isoflurane (1.5–3%) in an oxygen/nitrous oxide mixture (0.6 L/min) via spontaneous breathing. A 10-min computed tomography scan was first acquired for subsequent attenuation correction and anatomical identification. Then, a 60-min positron emission tomography (PET) acquisition was initiated, along with 10-s tail vein injection of [¹⁸F]fludeoxyglucose (FDG) or [¹¹C]acetate (15–20 MBq in 50–100 µL for both radiolabeled markers). For myocardial glucose uptake (MGU; mmol/g/min) and myocardial VO₂ (µmol/100 g/min) analysis, [¹⁸F]FDG and [¹¹C]acetate PET acquisition data were sorted in 24 and 56 dynamic frames, respectively, secondarily analyzed, and quantified with Carimas software (Turku PET Center, Turku, Finland) using Patlak kinetic analysis and a one-tissue compartment model as described elsewhere (27).

[³H]VLDL was produced as previously described (24). Wild-type (WT) and SKO mice were anesthetized with a xylazine/ketamine/PBS (1:1:8) solution and received 100,000 cpm [³H]VLDL. The mice were killed, and the hearts were washed in cold PBS 5 min after injection. The tissues were digested with Solvable (PerkinElmer, Courtaboeuf, France), and the radioactivity was counted.

Twelve-week-old SKO and WT mice were fasted for 2 h, and a single dose of 0.75 IU/kg human recombinant insulin (Umuline rapide; Eli Lilly, Suresnes, France) was administered by intravenous injection. The mice were killed 5 min after injection, and the hearts were collected and washed in PBS.

Membrane isolation of heart lysates was performed by ultracentrifugation (75 min at 110,000 × g), and the pellets were resuspended in solubilization buffer (50 mmol/L Tris [pH 7.4], 100 mmol/L NaCl, 50 mmol/L LiCl, 5 mmol/L EDTA, 0.5% Triton X-100, 0.5% deoxycholic acid, and 0.05% SDS). For Western blot analyses, we used phospho–Ser⁴⁷³-AKT/AKT, phospho–Thr^172–AMP kinase (AMPk)/AMPk, FOXO1, and GLUT1 antibodies from Cell Signaling Technology; SERCA2, Ser¹⁶-P-phospholamban (PLN), and PLN antibodies were from Santa Cruz Biotechnology; GLUT4 and Tubulin antibodies from Sigma-Aldrich; and an O-GlcNAc Western blot kit from Thermo Fisher Scientific. For O-GlcNAcylated protein isolation, 750 µg lysates was incubated with wheat germ agglutinin (Sigma-Aldrich) agarose conjugate and then analyzed by Western blot. The AGEs were quantified using the oxiSelect AGE competitive ELISA (Cell Biolabs), and ROS production was measured by incubating heart lysates with dichlorodihydrofluorescein diacetate (Life Technologies) for 30 min at 37°C as previously described (28). RNA expression was analyzed as previously described (24).

Triglyceride (TG) measurement was performed using a colorimetric assay after cyclohexane/isopropanol (3:2, volume for volume) extraction. For ceramide measurement, samples were analyzed using a 1290 UPLC system coupled to a G6460 triple quadripole spectrometer (Agilent Technologies) using MassHunter software for data acquisition and analysis. For neutral lipids, 1 mL lipid extract was analyzed by gas-liquid chromatography on a FOCUS Thermo Electron system (Thermo Fisher Scientific) using Zebron-1 Phenomenex fused silica capillary columns (29).

Freshly excised mouse hearts were immediately mounted onto a Langendorff apparatus, as previously described (30). Briefly, hearts were perfused in 0.7, 1.4, or 2.5 mmol/L Ca²⁺ using Krebs-Henseleit bicarbonate buffer supplemented with two different energetic substrates (i.e., glucose [11 mmol/L] alone or glucose [11 mmol/L] in the presence of octanoate [1.5 mmol/L]). The oxygen partial pressure was measured in the coronary perfusate and effluent using specific electrodes to calculate the myocardial O₂ uptake (in µL · min⁻¹ · g⁻¹).

All data are reported as the mean ± SEM. Data sets were analyzed for statistical significance using a nonparametric Mann–Whitney U test, Kruskal Wallis, or two-way ANOVA analysis when specified.

As previously described (24), 10-week-old SKO mice are hyperglycemic, insulin-resistant, mildly hypotriglyceridemic, and normocholesterolemic (Supplementary Fig. 1). Therefore, we performed cardiac phenotyping in WT and SKO mice from 10–24 weeks of age. As differences between both genotypes were quite stable with aging (data not shown), we present in this study the data from 14-week-old animals. The ratio between heart weight and body weight (HW/BW) (Fig. 1A) as well as between heart weight and tibias length (Supplementary Fig. 1E) were both increased in SKO compared with WT mice. ECG showed significant prolonged QT interval duration in SKO versus WT animals (Table 1), whereas the heart rate was similar in both groups (Fig. 1B). On Doppler flow analysis, LV diastolic function parameters were markedly affected in SKO mice: the E/A ratio was increased, and both the deceleration time and the isovolumetric filling time were increased (Fig. 1C–F). Consistently, CMR imaging highlighted an increased diastolic wall thickness in SKO mice (Fig. 1G) despite a normal end diastolic volume (EDV) (Fig. 1H). Right ventricular volumes and function were normal (data not shown). In addition, SKO mice exhibited a mild ejection fraction (EF) decrease (52 vs. 64% in WT animals) (Fig. 1I) and a blunted longitudinal strain (Fig. 1J). Longitudinal strain correlated positively with EF (Fig. 1K) and negatively with end diastolic wall thickness (EDW) (Fig. 1L), highlighting the relationship among LV hypertrophy, wall deformation abnormalities, and altered EF. Finally, correlations were found among random-fed mean plasma glucose, EDW, and EF, with the most hyperglycemic animals displaying the thickest LV walls and the lowest EFs (Fig. 1M and N).

As ectopic lipid deposition and lipotoxicity have been extensively described in diabetic cardiomyopathy, we studied cardiac lipid metabolism in SKO mice. Previously, using radiolabeled TG-rich lipoproteins, we demonstrated elevated TG uptake and a massive ectopic lipid deposition in SKO liver in vivo (24). Using the same methodology, we showed in this study that cardiac TG uptake was similar in SKO and WT mice (Fig. 2A). Cardiac lipidomic profiles showed similar TG and ceramide levels in both SKO and WT animals (Fig. 2B and C), and DAG levels were only modestly reduced in the SKO heart (Fig. 2D). In addition, we did not observe any change in the expression levels of the procatabolic nuclear receptor peroxisome proliferator–activated receptor α or of its target genes involved in FFA oxidation in SKO mice (Supplementary Fig. 2A). These data suggest that the seipin-deficient heart unexpectedly exhibits neither a lipotoxic profile nor a major impairment in lipid metabolism.

As insulin resistance and chronic exposure to hyperglycemia are also involved in the pathophysiology of diabetic cardiomyopathy, we assessed the insulin sensitivity and glucose uptake of the SKO heart. The insulin-induced phosphorylation of AKT was blunted by 70% in the hearts of SKO compared with WT mice (Fig. 2E). Consistently, in the fed state, the total and membrane GLUT4 protein levels were 50 and 40% lower, respectively, in SKO animals (Fig. 2F). GLUT4 mRNA tended to be lower in SKO, but this trend was not significant (Supplementary Fig. 2B). The GLUT1 protein (Fig. 2F) and mRNA levels (Supplementary Fig. 2B) were comparable in both genotypes. Using [¹⁸F]FDG PET scan, a marked increase in in vivo MGU was found in SKO hearts (Fig. 2G). To evaluate the fate of the glucose, we used [¹¹C]acetate PET scan–monitored oxidation to model oxygen consumption (31–33). Our data revealed a 30% decrease in heart oxygen consumption, suggesting a decrease of all oxidative metabolism, including glucose (Fig. 2H). This latter result was further confirmed using an ex vivo–perfused heart setup, with a significant reduction of oxygen consumption in SKO hearts whether they were supplied with glucose or glucose plus octanoate (Fig. 2I). AMPk plays a central role in energetic substrate catabolism and oxygen consumption regulation. In accordance with a defect in energetic substrate utilization, a 60% decrease in phosphorylated AMPk was observed in SKO heart lysates (Fig. 2J). However, phosphorylation levels of the acetyl-coA carboxylase, one AMPk target, were not significantly decreased in SKO samples (data not shown). Together, these data demonstrate that in SKO hearts, despite marked insulin resistance, glucose uptake is elevated but oxygen consumption is reduced.

Because glucose uptake was increased in SKO hearts, we tested different pathways that are classically activated by glucose overload in the heart. First, we observed no change in ROS production in AGE or p47-phox plasma membrane levels in SKO mice compared with WT mice (Supplementary Fig. 3). Second, we assessed the hexosamine biosynthetic pathway (HBP), which leads to the covalent addition of O-GlcNAc residues on various cellular proteins. The o-GlcNAcylated protein levels measured by Western blot were increased by 2.5-fold in SKO mice compared with WT mice (Fig. 3A). As chronic HBP activation has been previously shown to alter insulin signaling (34), we isolated O-GlcNAcylated protein using wheat germ agglutinin and analyzed these protein samples by Western blot. Importantly, the O-GlcNAcylated level of the transcription factor FOXO1, a key player in insulin sensitivity, was twofold increased, whereas its total level remains stable (Fig. 3B). In contrast, O-GlcNAcylated level of AKT was unaltered in SKO heart. In addition, the phosphorylation level of PLN, a regulator of the endoplasmic reticulum calcium pump SERCA2a, was markedly blunted in SKO hearts, whereas SERCA2 levels tended to be marginally lower in SKO heart samples (Fig. 3C). Finally, in SKO hearts, the myosin heavy chain (MHC) β/α ratio was threefold increased (Fig. 3D), and the number of positive cells in TUNEL staining was increased by 50% (Fig. 3E). Notably, no sign of fibrosis was found in SKO mouse hearts (Supplementary Fig. 3D). Together, these data demonstrate that activation of the hexosamine pathway by glucose overload is associated with molecular signs of cardiac dysfunction in SKO mice.

As cardiac parameters were strongly correlated with plasma glucose levels, we hypothesized that targeting hyperglycemia might correct the cardiac dysfunction observed in SKO mice. To that purpose, we used the SGLT2 inhibitor dapagliflozin, which reduces hyperglycemia by increasing glycosuria. Six-week-old SKO mice were treated for 8 weeks with 1 mg/kg dapagliflozin. Dapagliflozin treatment massively increased glycosuria (Supplementary Fig. 4A), normalized plasma TG levels (Supplementary Fig. 4B), and limited hyperglycemia (Supplementary Fig. 4C), but had no effect on insulin sensitivity, liver steatosis, or body weight (Supplementary Fig. 4D–F). The HW/BW ratio displayed a nonsignificant trend to be reduced (Supplementary Fig. 4G). Echocardiography and CMR imaging analyses revealed that dapagliflozin improved the E/A ratio, decreased the isovolumetric relaxation time, reduced the deceleration time, and normalized the EDW as well as the EF (Fig. 4A–F). Finally, we tested whether the normalization of random-fed glycemia by dapagliflozin decreased the cardiac molecular consequences of glucose overload. Interestingly, dapagliflozin-treated SKO mice had significantly lower total levels of O-GlcNAcylated proteins and specifically the O-GlcNAcylated FOXO1 form as compared with untreated SKO mice (Fig. 4G and H). Moreover, the activation state of PLN was normalized, and the phosphorylation level of AMPk was also increased with dapagliflozin (Fig. 4I). These results suggest that dapagliflozin partly corrected cardiac dysfunction by decreasing the glucose overload and the subsequent hexosamine pathway activation.

To further investigate the link between glucose overload and cardiac dysfunction, we used the insulin sensitizer pioglitazone as an alternative hypoglycemic drug. We previously published that pioglitazone treatment increases adipose tissue mass, insulin sensitivity, and random-fed hyperglycemia in SKO mice (24) (Supplementary Fig. 5A and B). Regarding the effect on cardiac function, we first noticed that pioglitazone increased the HW/BW ratio (Supplementary Fig. 5D). Cardiac echography and magnetic resonance imaging exams showed that pioglitazone treatment improved E/A ratio, deceleration time, and EDW in SKO treated as compared with SKO control mice (Fig. 5A, C, and D). In contrast, isovolumetric relaxation time was unchanged, and EF tended to be reduced, even though this decrease did not reach significance (Fig. 5B and E). Consistent with its hypoglycemic action, pioglitazone reduced the HBP pathway activation and the O-GlcNAcylated FOXO1 level (Fig. 5G and H). However, the improvements of AMPk and PLN activation did not reach statistical significance, despite a strong trend (Fig. 5I). Altogether, these data confirm that glucose-lowering agents protect SKO mice from glucotoxicity, but with pioglitazone, the cardiac dysfunction is only partially corrected compared with dapagliflozin.

In this study, we performed for the first time a complete cardiac phenotyping of SKO mice. We showed that SKO mice exhibit LV hypertrophy, along with diastolic/systolic dysfunction and prolonged QT intervals. Whereas no signs of lipid accumulation or lipotoxicity were observed, we report an alteration in glucose metabolism in the SKO heart, with increased glucose uptake and chronic hexosamine pathway activation. The SGLT2 inhibitor dapagliflozin, and in a lesser extent the insulin sensitizer pioglitazone, alleviates the glucose overload burden, leading to a marked reduction in the O-GlcNAcylated protein levels and an improvement of the cardiac phenotype.

In our rodent model of lipodystrophy, cardiac phenotyping revealed a decreased E/A ratio and an increased isovolumetric filling time, which are both signs of decreased heart compliance, leading to diastolic dysfunction. Consistently, the LV was found to be hypertrophic and stiff, as reflected by increased EDW along with an increase in heart mass and by blunted longitudinal strain.

Regarding the mechanisms involved in the associated heart disease, we first asked the question of whether a cardiomyocyte-autonomous effect of seipin deficiency existed. Seipin is expressed at very low levels in the heart of WT animals, and we also report that the β/α MHC ratio, a marker of pathological cardiac remodeling, was comparable between 6-week-old SKO mice, which were normoglycemic at this time, and WT mice (data not shown). Conversely, at 10 weeks of age, when the hyperglycemic phenotype arose, the β/α MHC ratio was elevated, and the first cardiac imaging abnormalities (decrease in E/A ratio) began to appear in SKO mice. The positive correlation between random-fed glycemia and cardiac abnormalities reinforce our hypothesis that systemic energy metabolism has a deleterious impact on cardiac function. The improvement of the cardiac phenotype with an antihyperglycemic drug is also consistent with this insight.

Lipotoxicity has been largely involved in diabetic cardiomyopathy (9,11), and we have previously shown that SKO mice display massive hepatic ectopic lipid deposition and increased hepatic triacylglycerol-rich lipoprotein (TRL) uptake (24). Surprisingly, we did not see any changes in TRL uptake or intracardiac lipid levels (TG and ceramides) in the SKO heart. This finding is consistent with several autopsy reports that did not show ectopic lipid deposition in the hearts of patients with BSCL (19,20). Overall, the SKO mouse is not a model for heart steatosis and lipotoxicity-induced cardiac dysfunction.

As glucotoxicity has also been implicated in T2DM-associated cardiac dysfunction (14,15), we studied carbohydrate metabolism in SKO hearts. Using [¹⁸F]FDG PET imaging, we established that SKO hearts displayed increased MGU compared with WT hearts. Literature is quite puzzling and heterogeneous regarding heart glucose uptake in diabetic animals. In one study, Zucker diabetic fatty rats displayed lower glucose uptake under hyperinsulinemic-euglycemic clamp conditions compared with lean rats (35). In another study, Zucker diabetic fatty rats were fasted for 6 h, and glucose uptake PET monitored was found to be unaffected, with a trend toward an increase (33). Nutritional status obviously affects glucose uptake, and in our case, the glucose uptake assay was performed after a short 2-h fasting period to avoid major hyperglycemia that would alter PET analysis. We also avoided longer fasting, which has been previously shown to alter SKO mouse insulin sensitivity (36). In our conditions, compared with WT animals, SKO mice are massively hyperinsulinemic (23,24,37). This latter point, along with hyperglycemia, probably explains why MGU is increased in vivo despite strong insulin resistance. A similar result was obtained by Kaczmarczyk et al. (38) in GLUT4 partially deficient C57BL/6 mice. Despite a 50% decrease in GLUT4 levels compared with WT animals, these mice presented a higher baseline MGU, related to their hyperinsulinemic state. Thus, we consider that, in our setting, PET measurement of glucose uptake appropriately represents the physiological balance among cardiac insulin resistance, hyperinsulinemia, and hyperglycemia and realistically reflects the entry of glucose into cardiomyocytes. As noted, the discrepancy between the lower GLUT4 protein levels and the unchanged mRNA levels was unexpected, as, in most T2DM mice models, both are concomitants (39). Interestingly, using [¹¹C]acetate measurement, we demonstrated that oxygen consumption was decreased in SKO hearts. When entering the cell, [¹¹C]acetate is quickly metabolized into acetyl-CoA and then is usually catabolized in the Krebs cycle to produce ATP (40). Our data clearly indicate that the Krebs cycle activity is low in SKO mice but do not determine whether FFA or glucose oxidation is primarily affected. The ex vivo data, in isolated hearts in the presence of glucose only, confirmed that glucose oxidation is deficient in SKO hearts. This would be consistent with a large body of work demonstrating that glucose oxidation is impaired in the diabetic heart (6,33,41,42). In addition, the decreased AMPk activity in SKO hearts is consistent with altered aerobic catabolism, given that AMPk is known to promote FFA and glucose cardiac oxidation (43). Interestingly, the TG levels and the expression levels of the main genes involved in FA oxidation were not altered in SKO mice, which supports the hypothesis that the alteration occurs mainly in glucose oxidation.

We tried to highlight how glucose overload may contribute to cardiac dysfunction. Several publications have demonstrated that glucotoxicity induces oxidative stress (14,41), but using several methods, we found no oxidative stress in SKO hearts. We next investigated the involvement of the other well-known glucotoxic pathway, the HBP (44). Indeed, we found a marked increase in O-GlcNAcylated protein levels in the SKO heart, and we interpreted this finding as a clear sign of glucose overload. The HBP pathway is activated in vitro and in vivo in the context of hyperglycemia and diabetes (44). Several studies have sought to elucidate whether HBP activation has a direct causal impact on cardiac dysfunction. Indeed, pharmacological inhibition of the HBP in db/db cardiomyocytes attenuated hypertrophic signaling (45). Hyperglycemia has been shown to alter the β/α MHC ratio both ex vivo and in vivo. The increase of uridine diphosphate N-acetyl glucosamine production, the main metabolite of the HBP pathway, is suspected to be involved in the overexpression of the β form of the MHC (46), a phenomenon that could explain the increased ratio reported in SKO mice. Chronic activation of the HBP by glucose overload has also been shown to alter calcium signaling and PLN phosphorylation levels (47). Consistently, PLN Ser phosphorylation levels were decreased in SKO mice. In addition, dysregulation of O-GlcNAcylation alters mitochondrial function, oxygen consumption, and ATP synthesis (48). Thus, increased O-GlcNAcylated protein levels would alter mitochondrial function, contributing to the reduced oxygen consumption in SKO hearts.

We specifically highlighted in this study the increased O-GlcNAcylated levels of FOXO1. This is particularly interesting for two main reasons. Firstly, FOXO1 O-GlcNAcylation increases its activity (49), and this activation has been suggested to be a key mediator of glucotoxicity in different organs (50). Secondly, FOXO1 has been involved in metabolically induced cardiac dysfunction, especially insulin resistance, and FOXO1 knockdown appears to be protective in a model of diet-induced cardiomyopathy (51). Our work shows for the first time that FOXO1 O-GlcNAcylation is associated with heart insulin resistance and cardiac dysfunction. Together, these data lead us to hypothesize that in our model, glucose overload is the main trigger of cardiac dysfunction and that chronic HBP activation is a central mechanism in glucose-adverse effects.

To test this hypothesis, we used two hypoglycemic agents: the SGLT2 inhibitor dapagliflozin and the insulin sensitizer pioglitazone. In SKO mice, pioglitazone reduce glycemia through insulin-sensitivity enhancement (24), whereas dapagliflozin does not improve the insulin resistance. Both treatments lowered glycemia and reversed the HBP chronic activation, including the increase in O-GlcNAcylated FOXO1 levels, suggesting an antiglucotoxic effect. In term of cardiac function, SKO mice treated with dapagliflozin presented a decrease in ventricular wall hypertrophy and an improvement of both diastolic and systolic function. In contrast, pioglitazone corrected only partially the cardiac dysfunction in SKO mice, improving the E/A ratio and ventricular wall hypertrophy, but with no effect on the EF and the isovolumetric relaxation time. Of note, pioglitazone induced an increase of the HW/BW ratio and a trend to increase the EDV. Therefore, despite a similar antiglucose overload action, these treatments act differently because of either a deleterious side effect of pioglitazone or an additional beneficial effect of dapagliflozine and potentially by a combination of both. Thiazolidinediones have been described to induce cardiac hypertrophy through volume expansion (52). This is further supported by a large meta-analysis showing that the use of pioglitazone was associated with increased risk of nonfatal chronic HF requiring hospitalization (53). In striking contrast, SGLT2 inhibitors exert osmotic diuretic and natriuretic effects contributing to plasma volume contraction (54). In addition, a recent study suggests that empagliflozin, another SGLT2 inhibitor, increases the production of ketone bodies that are more easily used by the heart (55).

In absence of steatosis and fibrosis, the origin of cardiac hypertrophy remains poorly understood in our model. Chronic activation of the HBP in the context of diabetes is tightly correlated with cardiac hypertrophy with several transcription factors regulating prohypertrophic genes being activated (56). However, dapagliflozin only tends to lower the HW/BW ratio, suggesting that either the reduction of the O-GlcNAcylation was not sufficient or that other mechanisms are involved. Future work is needed to highlight the contribution of the HBP induction to the hypertrophy in our SKO heart.

Although the precise mechanism remains yet to be completely elucidated, our results support a beneficial cardiac effect of dapagliflozin and therefore are consistent with recent clinical studies. Indeed, the BI 10773 (Empagliflozin) Cardiovascular Outcome Event Trial in Type 2 Diabetes Mellitus Patients (EMPA-REG OUTCOME) trial has recently demonstrated that empagliflozin strongly decreased both cardiovascular and total mortality (30–40% relative risk reduction) (57) in a large population of patients with T2DM at high cardiovascular risk. This protective effect of the SGLT2 inhibitor occurred very quickly after a few months and was sustained over the 4-year study. In line with our findings, it is tempting to speculate that SGLT2 inhibitors improve cardiac abnormalities at least partly by limiting cardiac glucose overload, HBP activation, and the functional consequences on calcium signaling and aerobic metabolism. Previously, Hong et al. (58) showed that the hypoglycemic SGLT1/SGLT2 inhibitor phlorizin improves the ventricular thickness in nonobese T2DM Akita mice. However, to our knowledge, our study is the first to demonstrate that dapagliflozin might have beneficial cardiac effects by protecting the heart from chronic HBP activation, especially at least through FOXO1 O-GlcNAcylation.

In conclusion, using a unique model for metabolic cardiomyopathy with no lipotoxic features, we highlight that glucotoxicity by itself can trigger cardiac dysfunction and that agents reducing glucose exposure can improve this cardiomyopathy, with a superiority of dapagliflozine over pioglitazone in this specific model.

Acknowledgments. The authors thank M. Takahashi, J.-F. Deleuze, and M. Lathrop (Centre de Génotypage–Commissariat à l'Énergie Atomique, Evry, France) for donating the mouse model; C. Dumet, S. Suzanne, S. Lemarchand-Minde (animal facility, L'Institut du Thorax, Nantes, France), and B. Haelewyn (CURB, UNICAEN, Caen, France) for animal care; Dr. Benjamin Lauzier and Dr. Mickael Dergangeon (L'Institut du Thorax, Nantes, France) for fruitful scientific discussions; and Pia Tager (EA 4650, UNICAEN, GIP Cyceron) for assistance during CMR certification exams. The authors also thank the MicroPICell core facility (Structure Fédérative de Recherche François Bonamy, Nantes, France) for microscopy. Lipidomic analyses were performed at Toulouse INSERM MetaToul-Lipidomique Core Facility (MetaboHub ANR-11-INBS-0010).

Funding. This work was supported by grants from INSERM and French associations Aides aux Jeunes Diabétiques, Société Francophone du Diabète, and Fondation de France. X.P. was awarded the European Foundation for the Study of Diabetes/Lilly Research Fellowship Programme 2015.

Duality of Interest. M.J. conducted clinical trials as coinvestigator, provided advisory services, and/or attended conferences as contributor for Lilly, Novo Nordisk, Sanofi, Takeda, Bristol-Myers Squibb, Novartis, AstraZeneca, Boehringer Ingelheim, and Janssen-Cilag. B.C. is taking part in advisory boards for AstraZeneca, Lilly, Sanofi, and Takeda. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. M.J. researched data and wrote the manuscript. B.J., D.M., X.M., A.T., A.A., L.D., G.T., and A.M. researched data. D.M. and B.S. contributed to the scientific discussion and edited the manuscript. C.L.M. and F.C. contributed to scientific discussion. J.M. wrote the manuscript. B.C. contributed to scientific design and wrote the manuscript. X.P. researched data, designed the experiments, and wrote the manuscript. X.P. 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 76th Scientific Sessions of the American Diabetes Association, New Orleans, LA, 10–14 June 2016.