Sodium glucose transporter 2 (SGLT2) inhibition with empagliflozin improves cardiac diastolic function in a female rodent model of diabetes

The global epidemic of obesity is largely responsible for the increased incidence of type 2 diabetes mellitus and associated cardiovascular diseases (CVD). Affected individuals are likely to have one or more CVD complications, such as hypertension and cardiac diastolic dysfunction; indeed, diabetes is often considered a CVD equivalent. In this regard, young (premenopausal) obese and diabetic women are particularly vulnerable to CVD [1, 2]. Normally, young lean woman are at lower risk for development of CVD compared to men. However, this sex-related cardioprotection is lost in conditions of obesity or diabetes [3]. Diastolic dysfunction, which is defined by delayed diastolic relaxation [4], is associated with insulin resistance, cardiac fibrosis, hypertrophy, and inflammation [5, 6]. Diastolic heart failure, a CVD risk factor usually associated with aging, has dramatically increased in incidence in association with the increases in obesity and diabetes. Therefore, developing new strategies to improve both glycemia and CVD out comes in individuals afflicted with diabetes would be highly desirable, especially for women.

Diastolic dysfunction is typically associated with interstitial fibrosis and left ventricular (LV) hypertrophy that promote LV stiffness and impaired relaxation. In this regard, serum and glucocorticoid regulated kinase 1 (SGK1), which is highly expressed in the diabetic heart and is stimulated by excess circulating glucose, is emerging as a mediator of cardiac fibrosis/stiffening and impaired cardiac relaxation [7–10]. SGK1 stimulates a number of ion channels, including the epithelial sodium channel (ENaC), as well as transporters, transcription factors and enzymes [11]. Indeed, we recently reported that fibrosis and stiffening of the aorta of overweight female mice fed a high fat, high sugar diet for 4 months was associated with increased aortic expression of SGK1 and ENaC [12].

In diabetes, hyperglycemia promotes a state of glucotoxicity, inflammation and oxidative stress which are associated with hypertension and end organ injury, including injury to the heart. In this regard, pharmacologic inhibitors of the renal sodium–glucose cotransporter 2 (SGLT2) are emerging as a novel group of drugs that lower blood glucose and HbA1c levels and improve whole body insulin sensitivity in animals and humans with diabetes, largely by blocking renal proximal tubular reabsorption of glucose which increases urinary glucose excretion (glycosuria) [13, 14]. The mild osmotic effects of SGLT2 inhibitor treatment can lead to modest reduction in blood pressure (BP), an effect, along with the improvement in glucose control and weight reduction that could reduce the risk of CVD. Indeed, a recent meta-analysis demonstrated favorable cardiovascular outcomes in diabetes patients treated with empagliflozin (EMPA) [15]. Administration of SGLT2 inhibitors to animals or humans with diabetes may also reduce adiposity, oxidative stress and expression of advanced glycation end products (AGE) and receptors for AGE (RAGE) [16]. Herein, we tested whether SGLT2 inhibitor therapy would attenuate the development of the earliest manifestation of diabetic heart disease, diastolic dysfunction, in part, by reducing blood pressure (BP), cardiac oxidative stress and pro-fibrotic factors. Diastolic dysfunction is especially pronounced in obese, insulin resistant and diabetic females [1, 2, 17–19]. Specifically, we hypothesized that the SGLT2i, EMPA would blunt the development/progression of diastolic dysfunction and the associated abnormalities in cardiac remodeling in insulin resistant female diabetic db/db mice (Lepr^db/db). Previous reports demonstrate that female db/db mice develop diastolic dysfunction, cardiac fibrosis and left ventricular hypertrophy (LVH) [20–22]. The db/db model is clinically relevant in that hyperleptinemia and leptin resistance, obesity and associated heart disease are seen in the human obese population and leptin levels are elevated in conditions of chronic heart failure and chronic hypertension. The db/db mouse exhibits a non-dipping BP pattern, diastolic dysfunction and cardiac remodeling; these CVD features of metabolic disease are also observed in obese and insulin resistant humans [20, 23–27]. Herein, we examined whether the anticipated improvement in diastolic function and cardiac remodeling with EMPA treatment would be associated with reductions in myocardial interstitial fibrosis, profibrotic signaling proteins, oxidative stress and improvements in myocardial mitochondrial ultrastructure.

Collectively, the results of this investigation support the hypothesis that treatment with the SGLT2i, EMPA, improves cardiac diastolic function in female mice in the setting of obesity and diabetes, even in the absence of a reduction in BP. The improvement in diastolic function was associated, not only with improved glycemia, but with improvements in cardiac structure, including reductions in interstitial myocardial fibrosis and associated pro-fibrotic SGK1/ENaC protein expression levels, cardiomyocyte hypertrophy and cardiomyocyte mitochondrial ultrastructure. Although EMPA is reported to reduce BP in diabetic humans [41], we did not observe lower BP in treated mice, nor did we see an improvement in BP dipping status compared to diabetic mice not receiving EMPA. Thus, functional and structural improvements in the heart of EMPA-treated mice occurred in the absence of BP reduction.

Administration of SGLT2 inhibitors is reported to improve glycemia and promote weight loss and body fat mass reduction in humans and rodents with diabetes, in part, due to caloric loss associated with increased urinary glucose excretion and a metabolic shift from carbohydrate to fatty acid oxidation [42, 43]. In this regard, a 4 week course of EMPA led to a slight reduction in triglycerides in patients with type 2 diabetes [44]. By the end of this study, we observed that DbC mice exhibited moderate glycosuria indicated by the 54-fold increase in urinary glucose excretion compared to CkC. Importantly, at the end of the study period, EMPA-treated db/db mice (10 mg kg⁻¹ day⁻¹) exhibited twice the urine glucose excretion of untreated db/db mice, thus validating long-term SGLT2 inhibition. Pharmacologic inhibition of SGLT2 has been reported to induce increases in plasma insulin concentration in db/db mice [45] and Zucker diabetic fatty rats [46] in association with preservation of β-cell insulin secretion and consequent improved glycemia. Consistent with those previous animal studies, we observed increased plasma insulin concentration. Our results demonstrate that the low dose of EMPA used in this study blunted the progression of hyperglycemia by inducing glycosuria. Improved glycemia is associated with preservation of β-cell function [47, 48]. Thus, the higher concentration of plasma insulin we observed with EMPA administration reflects therapeutic preservation of β-cell function and a slower decline in plasma insulin concentration. We did not observe differences in body weight or composition (lean or fat mass) or plasma or liver triglycerides between DbC and DbE. Similarly, treatment of female db/db mice for 8 weeks of EMPA at the same dose as that used herein in our study had no effect on body weight [13].

In this study we have explored the mechanisms underlying diastolic dysfunction in female db/db mice, including myocardial fibrosis, hypertrophy and calcium handling. A major determinant of impairment of the passive properties of diastolic relaxation is cardiac fibrosis. In the setting of obesity and diabetes, an increase in myocardial interstitial fibrosis can occur in response to increased pressure or volume loading conditions and this is thought to impair normal rapid ventricular relaxation. Female diabetic db/db mice exhibit increased cardiac interstitial fibrosis as early as 8 weeks of age [49] and we observed increased fibrosis in 16 week old females (Fig. 4b). Although such a pathological remodeling response reduces LV wall stress and may confer short-term benefits, stiffening of the myocardium can lead to reduced myocardial wall velocity during early diastole and more reliance on atrial contraction to complete filling of the LV in late diastole. Impairment in ventricular wall relaxation can be evaluated during routine echocardiography using Tissue Doppler Imaging (TDI) which allows for determination of the peak velocities of the myocardial wall during early (E′) and late diastole (A′). Indeed, our TDI examination revealed a reduction in the E′/A′ ratio in DbC mice before treatment began and at the end of the study suggesting chronic impaired LV wall relaxation early in diastole and enhanced late filling. Although at baseline this impairment existed in the DbE group by the end of treatment period this impairment was resolved. Perhaps most importantly, were the differences in LV filling pressure or the E to E′ ratio, among groups of mice. Unlike some diastolic parameters, such as the E/A ratio, E/E′ is a load-independent parameter. DbC exhibited increased LV filling pressure at baseline, as well as at the end of the study period, whereas, DbE mice had elevated LV filling pressure at baseline and this impairment was normalized by the end of the study. The decrease in LV filling pressure and reduction in interstitial fibrosis are consistent with an improvement in diastolic function with EMPA treatment. Our findings of impairments in TDI derived diastolic function parameters are also consistent with a recent preliminary analysis from the EMPA-REG OUTCOME trial [50]. In that study baseline echocardiography was performed on diabetes patients followed by treatment with 10 mg day⁻¹ EMPA for 3 months and follow up echocardiography. EMPA resulted in an increase in E′ and a decrease in the LV mass index. These interesting findings suggesting that EMPA improves diastolic function will likely be explored more extensively in ongoing clinical trials.

Like cardiac fibrosis, left ventricular hypertrophy, a frequent correlate of diastolic dysfunction [51], occurs in response to increased pressure or volume loading conditions common in the setting of obesity and diabetes. We observed that standard hypertrophic parameters, including LV mass, LV mass adjusted for tibia length and cardiomyocyte cross sectional area, were increased or tended to be increased in both DbC and DbE compared to CkC, and these results are consistent with a hypertrophic phenotype. Moreover, others have reported increased LV mass and wall thickness in female diabetic db/db mice of similar age and that remodeling was associated with diastolic dysfunction [20]. In this study, relative wall thickness (RWT) did not differ among the three groups of mice. Prior to the start of treatment, anterior (LVAWTd) and posterior wall (LVPWTd) thicknesses at end diastole were moderately increased in DbC compared to CkC (P < 0.5). Wall thickness were not different between DbC and DbE and they tended to be increased in DbE compared to CkC (P > 0.05). Moreover, there were no differences in RWT between the groups at the end of the study. Therefore, the increase in LV mass, in concert with no change in RWT, suggests that the LV of DbC exhibited eccentric, rather than concentric, hypertrophy. We speculate that cardiomyocytes may be lengthening in response to volume overload associated with profound obesity, as evidenced by the increase in CO in DbC. Nonetheless, DbC cardiomyocytes were more hypertrophied than those of DbE and CkC and this may be the result of stress on the LV wall caused by an increase in LV filling pressure. DbE showed lesser cardiomyocyte hypertrophy likely in response to a reduction in LV filling pressure and the decrease in cardiomyocyte hypertrophy was only a partial effect compared to the normalizing effect of EMPA on interstitial fibrosis. This suggests cardiac fibrosis contributes mainly to diastolic dysfunction.

We explored several pathways that contribute to cardiac remodeling and diastolic dysfunction, some of which have not been examined in previous studies. Emerging evidence supports a role for SGK1 and ENaC in promoting fibrosis and adverse hypertrophy in human and murine heart disease [7, 8]. SGK1 is highly expressed in human and murine hearts and is upregulated in pathophysiological settings, including obesity, heart disease and diabetes [40]. SGK1 regulates the expression of a number of ion channels, including ENaC [7, 9] which is upregulated in tissues in the setting of obesity and diabetes [52]. Indeed, SGK1 is rapidly activated and induces adverse ventricular remodeling, including, fibrosis, an increase in cardiomyocyte cell size and LV hypertrophy in a murine model of transthoracic aortic constriction [7, 8]. On the other hand, SGK1 may protect cardiomyocytes from apoptosis [8]. Little is known about the role of SGK1 and ENaC in the diabetic db/db heart or the impact of EMPA on myocardial ENaC and SGK1 expression/activation. To our knowledge, this is the first study to examine the expression of SGK1 and the α-subunit of the epithelial sodium channel, ENaC, in the parenchyma and vasculature of the LV. In this study, we observed increases in myocardial expression of SGK1 and ENaC and normalization of these protein levels by EMPA. These results suggest that the improvement in cardiac function may be modulated, in part, through improvement in SGK1/ENaC activity in the heart. The changes in expression of these proteins may be due to significant improvement in hyperglycemia since high glucose has been shown to increase the expression of SGK1 and ENaC in distal renal tubular cells in vitro [53]. However, the effects of high glucose on SGK1 and ENaC in cardiomyocytes or coronary endothelial cells of db/db mice has not been examined. We also examined the potential contribution of the pro-hypertrophic signaling pathways, Akt and ERK (Additional file 1: Figure S2). Although compared to CkC, Akt tended to be activated in both DbC and DbE, there was no difference in the activation state between DbC and DbE. The activation state of ERK did not differ among groups of mice. Thus, the modest improvement in hypertrophy in DbE is not likely mediated by effects of EMPA on Akt or ERK. The absence of evidence for contributions from Akt and ERK to cardiac remodeling in female db/db mice examined in this study highlight the potential contributions of SGK1 and eNAC to cardiac fibrosis and hypertrophy.

In this study, changes in sarco/endoplasmic reticulum Ca²⁺-ATPase (SERCA), SERCA to phospholamban (PLB) ratio and phosphorylation of PLB may potentially affect diastolic function [54]. Although others have reported altered gene expression of SERCA2A and PLB in 18 week old female db/db mice, we observed no differences among groups in protein expression (not shown) and thus cannot ascribe the differences in diastolic function observed among these proteins. Thus, we conclude these pathways do not explain the improvements in cardiac function observed. On the other hand, we did observe improved cardiomyocyte mitochondrial ultrastructure and sarcomere organization that may, in part, contribute to improved diastolic function as we have observed in this and other rodent models of cardiac dysfunction and obesity [29, 31, 32, 55].

Signaling through RAGE has been implicated in numerous disease conditions, including diabetes and heart disease. RAGE can bind multiple ligand types, including, but not limited to a heterogenous group of advanced glycation end products (AGEs). AGEs accumulate as a consequence of long-term hyperglycemia and contribute to the pathogenesis of diabetic cardiomyopathy [56]. AGE accumulation can lead to generation of reactive oxygen species (ROS) and upregulation of RAGE. Moreover, AGEs have been shown to induce the expression of ENaC via activation of SGK1 [57]. Recent evidence in mice suggests there is RAGE upregulation in cardiomyocytes that contributes to cardiomyopathy [58]. Herein, we show that myocardial AGE expression is increased in DbC and DbE, compared to CkC, and observe that it is widely distributed throughout the LV wall, including in the vasculature. This occurred in concert with increased RAGE expression. In the DbC, the increases in both AGE and RAGE were associated with an increase in oxidative stress. Nonetheless, despite the modest improvement in HbA1c in DbE, EMPA did not reduce AGE accumulation in the heart. In this regard, a recent study utilizing a rodent model of type 1 (streptozotocin model) diabetes, reported that a low dose of EMPA (10 mg kg⁻¹ day⁻¹), the same dose used in this study, was not associated with suppression of AGE, RAGE or ROS formation in aortic tissue or serum methylglyoxal, an AGE precursor, although treatment did improve aortic remodeling and reactivity to acetylcholine [16]. In that study, a threefold higher dose of EMPA did reduce AGE and RAGE expression, as well as ROS formation in the aortic wall. It should be noted that the glycation process in vivo results in formation of early glycation products, such as HbA1c and glycated albumin, whereas advanced glycation products require longer times to accumulate relative to early glycation products. In this regard, a previous study reported that advanced glycated Hb (Hb-AGE) was significantly increased with high level of HbA1c in diabetic patients, but Hb-AGE did not correlate with diabetes duration and correlated poorly in a well-controlled sub-group [59].

Evidence from a recent clinical trial (EMPA-REG BP™) examining the effects of a 12 week course of EMPA on male and female patients with type 2 diabetes and hypertension demonstrates improvements in both sexes in SBP and DBP (~3.9 and 1.5 mmHg, respectively compared to placebo), as well as indirect markers of arterial stiffness and vascular resistance [41, 60]. In the untreated db/db mice examined here (DbC), we observed increased SBP and DBP, as well as impaired BP dipping, all of which were unaffected by EMPA. It is possible that the 5 week duration of our study was too brief to induce a detectable decrease in BP.

Coinciding with the preparation of this manuscript there was a more recent report from the EMPA-REG outcome trial demonstrating improvement in Tissue Doppler derived diastolic function in diabetes patients administered EMPA [50]. That EMPA resulted in similar improvement in myocardial wall relaxation in both diabetes patients and in diabetic db/db mice suggests that the present db/db model has potential clinical translational relevance.

The absence of data on the effects of EMPA on myocardial metabolism and insulin sensitivity is a limitation of this study. In this regard, a recent study reported increases in glucose disposal rate and liver, kidney and heart tissue glucose uptake, by euglycemic–hyperinsulinemic clamp, in female db/db mice treated with EMPA for 8 weeks at the same dose used in this study [13]. Thus it is possible that the improvement in diastolic dysfunction observed in this study could be due, in part, to increased myocardial glucose uptake and improvement in myocardial insulin sensitivity. Further study is needed to determine whether improvement in myocardial insulin sensitivity contributes to improved diastolic function in db/db mice treated with EMPA.

In conclusion, data presented herein support a newly described pleiotropic protective affect of EMPA on diastolic function and cardiac structure in the diabetic db/db mouse with established impaired diastolic relaxation, albuminuria and elevated BP. Specifically, we observed improvement in diastolic function, myocardial fibrosis, cardiomyocyte hypertrophy and ultrastructure of inter myofibrillar mitochondria that were associated with a significant improvement in glycemia, as well as improvement in myocardial expression of SGK1 and ENaC. Despite the improvement in glycemic control, HbA1c and fasting glucose were still elevated above normoglycemia values. Moreover, we did not see any marked improvements in body composition, lipid (plasma or liver triglycerides) or BP control or BP dipping, or reductions in myocardial accumulation of AGE/RAGE and protein nitrosylation with 5 weeks of EMPA treatment. It was recently highlighted that the unexplained aspects of the EMPA-REG OUTCOME results are that the cardiovascular and kidney benefits of EMPA occurred without dramatic improvements in glycemic, lipid, or BP control [61]. Therefore, it is possible that the pleiotropic effects of EMPA relate to factors other than improvements in glycemia and lipidemia. In this regard, the effect of EMPA on metabolic remodeling in tissues is emerging [61, 62]. Additional studies are needed to further elucidate the potential role for myocardial SGK1 and ENaC as mediators of the efficacy of EMPA. The findings of this investigation in a preclinical model suggest a potential clinical utility for EMPA in the treatment of diastolic dysfunction given the high incidence of diastolic dysfunction and the cardiovascular risks associated with this abnormality in the diabetic population and in women in particular.