SGLT2 inhibition with empagliflozin improves coronary microvascular function and cardiac contractility in prediabetic ob/ob^−/− mice

Transthoracic echocardiography was performed using non-invasive Doppler ultrasound at 9, 16 and 21 weeks of age in all mice using a high-frequency ultrasound imaging system (Vevo 2100 VisualSonics, Inc., Toronto, Ontario, Canada) with a 40-MHz central frequency transducer to characterize coronary vascular function according to previously established protocols [40]. All mice were anesthetized with Isoba^® vet isoflurane (Schering-Plough Ltd., England) in a closed chamber with 3% isoflurane in oxygen for 2 to 5 min until immobile and 1.0–1.5% of isoflurane was used during the examination. Each mouse was placed supine on a heated procedure board with isoflurane initially at 1.5% supplied by a nose cone connected to the anesthesia machine. Chest hair was removed with chemical cream (Veet, Reckitt Benckiser, UK). Isoflurane was reduced to 1% to lower coronary flow to a baseline level and velocity profile in the left coronary artery was monitored for 3 min to ensure stable signal was achieved, after which signals were collected and stored. Isoflurane level was then increased to 2.5% to induce hyperemia, and the velocity profile was monitored for up to 4 min during which time signals were stored for offline analysis of maximum hyperemic response. Isoflurane was reduced back to 1%, and mice allowed stabilize. Detailed protocol using isoflurane to induce hyperemia has been described elsewhere [40].

Measurements of left ventricle dimensions were performed in standard B-mode using images in short axis views at the papillary level, before the CFVR protocol.

FAC was calculated as 100 * ((LV diastolic area − LV systolic area)/LV diastolic area), where LV is the left ventricle.

Left ventricular end-diastolic volume (LVEDV) and left ventricular end-systolic volume (LVESV) was calculated using the Teichholz’s formula: where LVEDD is LV end-diastolic dimension. where LVESD is LV end-systolic dimension.

Stroke volume (SV) was calculated as end-diastolic volume (EDV) − end-systolic volume (ESV).

Cardiac output was calculated as (SV * HR)/1000, where HR is heart rate.

CFVR was calculated as the ratio of peak diastolic flow velocities at baseline obtained using 1% isoflurane, and during hyperemia obtained using 2.5% isoflurane (CFVR = hyperemic coronary flow velocity/basal coronary flow velocity).

Heart rate was calculated as average from three consecutive cardiac cycles in Doppler mode, adjacent to the CFVR bridging heart rate and CFVR in time.

HbA1c, urine glucose, albumin and creatinine were measured at about 7 to 8 (baseline), 15 to 16 (5 weeks after intervention) and 20 to 21 (10 weeks after intervention) weeks of age, under non-fasting conditions. Further biomarkers were measured at termination after 4 h fast.

The measurement of HbA1c was performed using blood from tail vein (A1CNow+, Bayer Healthcare, Austria). Urine albumin was measured using a commercial ELISA kit (Cat No E-90AL) from ICL, USA. Urine creatinine was measured using an enzymatic colorimetric method (Kit No A11A01933) from Horiba ABX, France. Urine glucose was analyzed using an enzymatic (hexokinase) colorimetric method (ABX Pentra 400, USA).

l-Arginine, ADMA and SDMA was analyzed using a modified liquid chromatography–tandem isotope dilution mass spectrometry analysis (LC–MS) methodology previously described (Sigma-Aldrich, St. Louis, MO, USA) [40].

Tissue samples were immersion fixed in buffered formaldehyde solution, dehydrated and embedded in paraffin wax. Histological staining was performed on 4 μm thick sections. Hematoxylin–eosin staining was used to evaluate general morphology and for steatosis scoring, Picro sirius red (PS) for the evaluation of fibrosis. Histopathological assessment of fibrosis and liver steatosis was carried out using semi-quantitative scoring where 0 indicates no change and 3 a severe change [47].

CD31 staining on the heart tissue to detect endothelial cells as markers for vessels was performed using anti-CD31 antibody (1:25, #ab28364, Abcam) on the Ventana semi-automated staining machine (Ventana Medical Systems, Inc., USA). Slides were deparaffinized at 69 °C and then underwent antigen retrieval in cell conditioning media at 95 °C. After a blocking step anti-CD31 antibody was applied unto the slides, and incubated for 60 min at 37 °C. The CD31 signal was detected using OmniMap Anti Rb HRP, followed by Discovery DAB CM reagent to give a red color. Finally, sections were counter stained and slides were removed from Ventana and mounted manually.

To enable analysis, stained tissue sections were digitized using a Zeiss Mirax slide scanner (3D Histech, Budapest, Hungary) to image 100% of each section. The resulting virtual slides were imported into BioPix iQ (version 3.3.4, BioPix AB, Gothenburg, Sweden) for CD31 analysis. Analytical tissue sections were selected using the scanned region containing tissue only and resampled at 3.51 mega pixels. The CD31 positive endothelial cells were identified with the computerized color marker which was carefully determined using hue saturation and brightness. The same setting was used for all sections. The outer regions of the tissue, artefacts and large vessels (≥ 200 µm²) were identified and excluded from overall vessel analysis. The results are reported as the mean area fraction of the identified vessels compared to the total tissue area examined.

Before analysis, we assumed an increased difference due to time in bodyweight, HbA1c, coronary hyperemic flow velocity reserve (CFVR) and FAC between strains and effect of treatment. Therefore, Anova with 5%-significance was used to compare changes between lean controls, treated and vehicle group at 5 and 10 weeks after intervention. The non-parametric Mann–Whitney U test was used to compare the cardiac vessel density and t-test was used for semi-quantitative values.

As expected, body weight was higher in the ob/ob^−/− groups compared with the lean controls at all three time points. The relative increase in body weight was reduced in empagliflozin treated animals compared to untreated ob/ob^−/− mice (Fig. 2b) as has been seen in clinical studies. There was no significant difference in wet heart weight; however, treated ob/ob^−/− group had a higher kidney wet weight when compared with lean controls and the untreated ob/ob^−/− group (Table 1). Upon onset of empagliflozin treatment, the animals increased their food intake at week 5 and 10 compared to baseline and compared to untreated ob/ob^−/− animals at the same time points (Fig. 2a). The ob/ob^−/− groups had a higher water intake at all time points compared with the lean controls (Table 1). Furthermore, the treated group had a higher water intake at 5 and 10 weeks after intervention compared with the untreated group.

HbA1c, an indicator for medium term glucose control was at baseline ~ 2 times higher in both ob/ob^−/− groups compared to control. There were no differences between the ob/ob^−/− groups at baseline. The empagliflozin treated animals experienced a highly significant reduction in HbA1c compared to untreated animals both at 5 and 10 weeks treatment (Fig. 3a). Taken together, these data indicate that empagliflozin reduced glycemia, promoted caloric loss via glucose excretion and reduced body weight gain.

Urine albumin (µg/ml) levels were mostly undetectable in the lean group at baseline mean ± SEM (n = 2/12, 9.5 ± 3.0), detectable in both ob/ob^−/− groups (untreated n = 16/22, 22.7 ± 5.1, treated n = 16/21, 33.0 ± 4.1). At 5 weeks, fewer lean mice had detectable albumin levels (n = 3, 7.1 ± 2.0) compared with the ob/ob^−/− untreated (n = 19, 15.1 ± 3.1) and fewer ob/ob^−/− treated compared with untreated (n = 9, 7.2 ± 1.1). At 10 weeks, half of the lean group had detectable albumin levels (n = 6, 7.6 ± 1.4), about half of the untreated group (n = 12, 11.0 ± 1.8) and fewer in the treated group (n = 5, 5.2 ± 0.3). Albumin levels were significantly higher in ob/ob^−/− untreated compared with treated group at 5 weeks (p = 0.0424) and at 10 weeks (0.0186) (Table 1).

Urine creatinine (g) was significantly higher in the lean controls compared with the treated and untreated group at all time points. At baseline, both treated and untreated ob/ob^−/− groups had similar creatinine levels; however, at 5 and 10 weeks after treatment start urine creatinine was significantly lower in the treated compared with the untreated group (Fig. 3c).

Urine glucose and urine glucose/creatinine ratio (Fig. 3b, d) were significantly higher in the ob/ob^−/− groups at baseline compared with the lean controls; however, there were no differences between the ob/ob^−/− groups at baseline. Already at 5 weeks after start of empagliflozin treatment, urine glucose and glucose/creatinine ratio were increased more than threefold in ob/ob^−/− and these increase were maintained at 10 weeks intervention.

To assess cardiac and coronary artery function during the progression of insulin resistance in non-atherosclerotic ob/ob^−/− mice and the effect of SGLT2 inhibition, we performed sequential measurement of CFVR using transthoracic Doppler echocardiography.

Cardiac function was measured by FAC, an ejection phase index of contractile function. 10 weeks of treatment with empagliflozin improved cardiac function in the ob/ob^−/− group compared to both vehicle and lean controls (Table 2). At 5 weeks post treatment, treated ob/ob^−/− group had a higher FAC compared with lean control, however there was no significant difference between treated and untreated groups of the ob/ob^−/−. There were no significant differences in FAC between the groups at baseline.

Coronary vascular function as measured by CFVR was different at baseline between groups, and this was due to the ob/ob^−/− groups having reduced hyperemic flow compared to lean mice. After 5 weeks intervention, there were no detectable differences between groups. At 10 weeks intervention, there was a significant difference between groups, and this was attributable to the untreated ob/ob^−/− group having a lower CFVR compared with lean controls and the treated ob/ob^−/− group (Fig. 4a). Hyperemic flow (Fig. 4b) after 10 weeks treatment was lower in the vehicle group compared with lean control, and a trend towards having lower values compared with treated ob/ob^−/− group. Basal flow velocity was not different between groups at any of the time points (Fig. 4c).

Heart rate was significantly different between groups at all time points, at baseline and 5 weeks post treatment, this was due to the treated ob/ob^−/− group having a higher heart rate compared with the lean control group. At 10 weeks intervention the treated ob/ob^−/− group had a higher heart rate compared with the untreated ob/ob^−/− group, but not when compared to the lean controls (Table 2).

The ob/ob^−/− mice heart has been shown to be metabolically inefficient [48], with reduced myocardial efficiency at an early age [49], impaired cardiac contractile function and left ventricular hypertrophy [50]. We detected a statistically significant increase in fractional area change (FAC), a systolic contractility index [42] and we observed significantly improved coronary microvascular function after 10 weeks of treatment in the ob/ob^−/− mice. Based on the knowledge from humans that coronary flow reserve and cardiac systolic function are powerful predictors of cardiovascular outcome [51, 52], we believe that the current findings are of clinical relevance and provide direct physiological evidences of SGLT2 inhibitors CV benefit [5].

In a none atherosclerotic model like the one  used in this study CFVR, which usually is a measure for both coronary macro- and microvascular function, becomes a direct indicator of microvascular function [40]. This has been further demonstrated by us recently in patients with HFpEF without epicardial coronary artery diseases, where a correlation between CFVR and peripheral endothelial function was observed [34]. Coronary microvascular function is determined in part by both endothelial dependent [53] and independent mechanisms [37, 53]; improved l-arginine/ADMA ratio is proposed to be involved in the improved CFVR. Consistent with our findings, a recent study in db/db mouse showed an improvement in cardiac diastolic function upon empagliflozin treatment [54]. Furthermore, in diabetic Zucker rats, SGLT2 inhibition using dapagliflozin showed improved cardiac function [55], albeit these are advanced models of diabetes. Furthermore, in a non-diabetic pressure overload-induced heart failure mouse model, empagliflozin blunts the decline in cardiac function independent of diabetes [56]. Diabetes, in particular insulin resistance is proposed to play a role in the development of heart failure and cardiac dysfunction [57]. Obese individuals without diabetes but with insulin resistance syndrome manifest similar increased risk for CV disease when compared with T2DM patients [58], and in individuals with heart failure, FAC has been shown to be reduced [59]. Even though coronary flow reserve and systolic function independently of each other and synergistically predict CV mortality [52], improved CFVR may partly contribute to the improved FAC, given that CFVR may reflect improved energy utilization by decrease baseline energy need and improved flow/energy reserve.

It has previously been shown that plasma levels of l-arginine in ob/ob^−/− mice is lower compared to lean mice, indicating dysfunctional nitric oxide pathway, while no difference was seen in ADMA levels [40]. We reproduced these findings in our study, and further showed that l-arginine was higher in empagliflozin treated ob/ob^−/− mice compared to untreated, resulting in higher l-arginine/ADMA ratio. l-arginine is the substrate for both nitric oxide synthase (NOS), which converts arginine to nitric oxide (NO) and citrulline, as well as for arginase that converts l-arginine to urea and ornithine. ADMA is an endogenously produced inhibitor of NOS, which has a similar structure to arginine and competes with arginine for NOS binding, hereby blocking the formation of NO from arginine by NOS directly [60]. Endogenous arginine synthesis is an inter-organ process that involves citrulline produced in the small intestine which is then transported in the blood and used for arginine synthesis in the kidney [61]. We speculate that the increase in l-arginine with SGLT2 inhibition could be because of increased synthesis in the kidney. The main site of action of SGLT2 inhibitors is the kidney, and the extent of the pleiotropic effects that this class of drugs have been still unknown. It is known that l-arginine is a physiological precursor to the formation of NO in endothelium-dependent vasodilation; and NO has several intracellular effects that lead to vasorelaxation, endothelial regeneration, inhibition of leukocyte chemotaxis, and platelet adhesion [62]. In the ob/ob^−/− mice, decreased cardiac NO production and increased oxidative stress have previously been demonstrated [63]. We therefore speculate that the observed improvement in l-arginine/ADMA ratio in our study could lead to an increase in NO bioavailability, thereby improving coronary and cardiac function. The lack of detectable difference in CD31 staining suggests that the mechanisms involved in the observed improved CV function does not involve a recovered structural capillary rarefaction. The improved l-arginine/ADMA ratio points to NO-dependent improvement of endothelial function as the relevant pathway for improved microvascular function.

Plasma alanine aminotransferase, hepatic triglyceride and steatosis scoring were lower in the empagliflozin treated ob/ob^−/− compared with the untreated mice. This is consistent with the finding in a mouse model of non-alcoholic steatohepatitis (NASH) and diabetes, where empagliflozin attenuated the development of NASH showing anti-steatotic effects [64]. Furthermore, our data is consistent with that published on the effect of ipragliflozin treatment in ob/ob mice [65]. Hepatic triglyceride content is closely associated with steatosis in NASH as well as with de novo lipogenesis [66], therefore, SGLT2 inhibitors may have potential therapeutic efficacy in T2DM-associated hepatic steatosis. Furthermore, hepatic triglyceride content has been shown to be associated with diastolic function independent of confounding factors including the metabolic syndrome in obese individuals [67]. SGLT2 inhibitors also mediates a metabolic switch from glucose to lipid utilization [68], and given the bi-directional relationship between fatty liver and CV disease [3], it is plausible that the reduction in hepatic triglyceride we observed in our study contributed in part to the improvement in cardiac contractile function. We observed reduced total-cholesterol and no changes in plasma triglyceride with empagliflozin treatment, which is partly different to observations from clinical trials, where no changes were observed in total-cholesterol and plasma triglyceride [69]. It is plausible that these differences are because of species differences in lipid profiles.

As expected, urine glucose excretion was elevated in the treated group in line with the mode of action of SGLT2 inhibitors. Food and water intake for the lean mice were comparable to published data [70]. Food intake for treated ob/ob^−/− mice was higher than in the untreated group and body weight was not different prior to treatment start. In line with clinical studies, we found that body weight increase was lower in the treated mice compared to the untreated. Furthermore, consistent with published data [71], we found that empagliflozin enhanced water intake, possibly to compensate for the urinary glucose, and fluid loss, thereby maintaining fluid homeostasis and stabilizing body weight. We showed that urine albumin and creatinine were reduced in the treated groups. It is well established that increased urine albumin is an independent risk factor in kidney and cardiovascular disease [72], and the reduction of albuminuria has been shown to reduce cardiovascular events [73]. The depletion of creatinine however could be caused by various factors such as changes in muscle mass, creatinine reabsorption, and cell leakage [74], therefore we are unable to ascertain the exact cause or importance of the reduced creatinine we observed.

Positive association between higher HbA1c levels and increased CV event rates has been shown in several studies, highlighting the need for glycemic control to prevent or delay the onset of CV morbidity and mortality [75]. In line with published studies using rodent models [15], and clinical studies [5], glycemic status as measured by HbA1c was significantly improved in the treated group. T2DM is associated with increased basal plasma glucagon levels, and it has been shown that in individuals with T2DM, SGLT2 inhibition further increases plasma glucagon. In addition it has been demonstrated that SGLT2 inhibition increases glucagon secretion from pancreatic islets [45, 76]. In our study, we were unable to detect differences in fasting plasma glucagon levels between treated and untreated ob/ob^−/− mice, however, the glucagon/insulin ratio was significantly higher in treated compared with untreated ob/ob^−/− mice. Consistent with findings in clinical studies [77], and animal study [78], empagliflozin treatment caused a significant increase in fasting plasma ketone concentration in ob/ob^−/− mice. The mechanisms involved in increased circulating ketone body levels with SGLT2 inhibition are not fully understood, however it is proposed that it may relate to a decrease in renal clearance of ketone bodies [79]. In addition, it is well established that elevated glucagon/insulin ratios favor ketone production and release from the liver. Furthermore, we speculate that under this condition of increased circulating ketone levels, the heart and other organs take up β-hydroxybutyrate and this is oxidized in preference to fatty acids, resulting in improved cardiac efficiency [80].

In the current study we demonstrated improved coronary flow reserve in an animal model without macrovascular complications. In humans, CFVR reflects both macro- and microvascular disease [37]. Given the anti-atherosclerotic effects also may translate into man [81], the potential magnitude of change of CFVR may be even higher in humans during longer term treatment than what is demonstrated in the current study. Further, while we showed that NO-pathway maybe involved in the improved CFVR, in humans with complex disease background, e.g. hypertension, dyslipidemia, as well as environmental risk factors, the translatability of this mechanistic finding needs be to confirmed. Finally, despite the human relevant CFVR protocol we have developed and applied in the current study, this important physiological measure is still assessed during anesthesia. Even though all animals were handled in a similar way, cautions still need to be taken when extrapolating the finding into humans.