Cardioprotective effects of empagliflozin after ischemia and reperfusion in rats

The Sodium Glucose Co-Transporter-2 inhibitor, empagliflozin (EMPA), reduces mortality and hospitalisation for heart failure following myocardial infarction irrespective of diabetes status. While the findings suggest an inherent cardioprotective capacity, the mechanism remains unknown. We studied infarct size (IS) ex-vivo in isolated hearts exposed to global IR injury and in-vivo in rats subjected to regional myocardial ischemia reperfusion (IR) injury, in whom we followed left ventricular dysfunction for 28 days. We compared rats that were given EMPA orally for 7 days before, EMPA 1.5 h before IR injury and at onset of reperfusion and continued orally during the follow-up period. We used echocardiography, high resolution respirometry, microdialysis and plasma levels of β-hydroxybutyrate to assess myocardial performance, mitochondrial respiration and intermediary metabolism, respectively. Pretreatment with EMPA for 7 days reduced IS in-vivo (65 ± 7% vs. 46 ± 8%, p < 0.0001 while administration 1.5 h before IR, at onset of reperfusion or ex-vivo did not. EMPA alleviated LV dysfunction irrespective of the reduction in IS. EMPA improved mitochondrial respiration and modulated myocardial interstitial metabolism while the concentration of β-hydroxybutyric acid was only transiently increased without any association with IS reduction. EMPA reduces infarct size and yields cardioprotection in non-diabetic rats with ischemic LV dysfunction by an indirect, delayed intrinsic mechanism that also improves systolic function beyond infarct size reduction. The mechanism involves enhanced mitochondrial respiratory capacity and modulated myocardial metabolism but not hyperketonemia.


Materials and methods
Animals and design. Male Sprague Dawley rats (250-300 g, Taconic, Ry, Denmark) were kept for acclimatization at a constant temperature of 23 °C, with a 12 h light-dark cycle and with unlimited access to food and water.
Our study included three experimental series: 1) Impact of EMPA (30 mg/kg) on infarct size in vivo and ex vivo was investigated to determine if the cardioprotection by EMPA, if any, was dependent on a whole-body system or could be induced directly on the heart. We compared rats receiving chronic oral EMPA treatment daily for 7 days before MI (EMPA-Chronic) to rats receiving an acute oral administration 1.5 h before MI (EMPA-Acute) or at the onset of reperfusion (EMPA-Post). Placebo animals were matched in dose and time of administration of vehicle ( Supplementary  Fig. S1a). 2) Impact of EMPA on post-MI LV dysfunction after 28 days evaluated by echocardiography and high resolution respirometry for mitochondrial respiratory capacity. In this series we compared EMPA-Chronic, EMPA-Post, a sham group receiving EMPA (SHAM) and corresponding placebo animals. In all groups in this experimental series, the administration of EMPA (30 mg/kg) continued consecutively for 28 days after myocardial infarction ( Supplementary Fig. S1b). 3) Impact of EMPA on mitochondrial respiratory capacity measured by high resolution respirometry and myocardial intermediary metabolism measured by microdialysis during early reperfusion (30 min) was investigated to delineate the metabolic changes caused by EMPA treatment. In this series we compared EMPA-Chronic, a matched placebo group (PLACEBO-Chronic), EMPA-Acute and a sham group (SHAM) ( Supplementary Fig. S1c).
Times of administration were chosen for several reasons. To test the most acute effect of EMPA we chose to administer EMPA 1.5 h before intervention, since the drug reaches maximum bioavailability after 1 h 14 . To test any effect requiring longer time and repeated administrations we chose to administer EMPA repeatedly for 7 days before intervention. Also, levels of β-OHB were measured at the different time points to determine the relation to any potential cardioprotection.
Exclusion criteria included death during procedure, death during follow up, complications to oral gavage, no sign of myocardial infarction evaluated during procedure; no paling or hypokinesia, no sign of living mitochondria during high resolution respirometry, increase of > 10% in the oxygen consumption rate after addition of cytochrome C during high resolution respirometry, coronary flow rate > 20 ml/min in the isolated perfused heart model indicating unspecific leakage, incorrect microdialysis probe placement and leaking probe membrane.
Experimental models. In vivo rat model. Rats were anaesthetized in an induction chamber with 8% sevoflurane (Sevorane, AbbVIE A/S, Copenhagen, Denmark) mixed with atmospheric air (flow: 0.8 L/min). Upon induction of anaesthesia, the rats were intubated and connected to a mechanical ventilator (Ugo Basile 7025 rodent ventilator, Comerio, Varese, Italy) with an adjusted flow of 0.6 L/min with 3.5% sevoflurane. A temperature probe (UNO, Zevenaar, Holland) was inserted, and body temperature was kept at a constant 37 °C ± 1 °C. Prior to the procedure the rats were injected subcutaneously with buprenorfin (Temgesic, Reckitt Benckiser Pharmaceuticals Limited, Slough, England) (0.045 mg/mL) to ensure peri-, and post-operational analgesia. A www.nature.com/scientificreports/ left sided thoracotomy, via the fourth and fifth rib, followed by pericardiotomy gave access to the myocardium. The left anterior descending coronary artery (LAD) was located and ligated approximately 2 mm distally to the junction between the pulmonary conus and the left atrial appendage using a 4-0 silk suture (Sofsilk, Covidien, Dublin, Ireland) 14 . Myocardial paling and hypokinesia distally to the ligature confirmed ceased blood flow and ischemia. After 30 min of ischemia, reperfusion of the myocardium was ensured by removing the ligature and confirmed by visualization of hyperaemia and enlargement of the LAD. In animals used for infarct size evaluation, the ligature was only loosened but remained in the chest during reperfusion, to ensure exact delineation of the area at risk (AAR). After reperfusion the chest was closed. Animals in the long-term survival group were given buprenorfin (7.4 μg/ml) in the drinking water for three days following MI.
Preparation of the isolated perfused heart. The hearts were isolated and perfused ex vivo as previously described 15 . In brief, rats were anaesthetised by a subcutaneous injection of Dormicum (midazolam (0.5 mg kg −1 body weight); Matrix Pharmaceuticals, Herlev, Denmark) mixed with Hypnorm (fentanyl citrate (0.158 mg kg −1 body weight) and fluanisone (0.5 mg Kg −1 body weight)). The rats were tracheotomized and connected to a mechanical ventilator and ventilated with atmospheric air (Ugo Basile 7025  Infarct size. In the in vivo model, the ligature around LAD was tightened again after 2 h of reperfusion, and a 4% Evans Blue (Sigma-Aldrich, St. Louis, MO, USA) solution was injected in vena cava inferior, to outline the myocardial area at risk (AAR), as previously described 15 .
In the isolated perfused heart model, the same protocol was followed without Evans Blue staining, since the hearts were subjected to global no flow ischemia. IS/AAR, AAR/LV and IS/LV were calculated and adjusted to the wet weight of the individual slices.
Echocardiography. Transthoracic echocardiography was performed at baseline, day 1 and day 27 after MI, as previously described 15 . Transthoracic echocardiography was performed with a Vevo 2100 high-frequency ultrasound system (Visual Sonic, Toronto, ON, Canada) with a 21 MHz rat probe. Animals were lightly sedated (3% sevoflurane, atmospheric air), fixated in a supine position to a heating pad and connected to ECG electrodes and a rectal thermal probe. Movement of the transducer was facilitated by a mechanical setup to eliminate movement disturbances. Two-dimensional and M-mode images were obtained.
Left ventricular volumes in end diastole (LV Vol d ) and end systole (LV Vol s ) were calculated using the bullet method (5/6 × LV Vol d /LV Vol s area x LV length). Left ventricular ejection fraction (EF) was calculated using LV Vol d and LV Vol s by the formula: All images were analyzed using the Vevo 2100 software.
Mitochondrial respiratory capacity. We analyzed the mitochondrial respiratory capacity with non-fatty acid and fatty acid substrates using high-resolution respirometry (Oxygrah-2 k; Oroboros Instruments, Innsbruck, Austria) as described previously 16 . A mid-papillary biopsy of LV myocardial tissue was prepared by manual dissection of fiber bundles (~ 1.5 m), 30 min after MI in series 3 and 28 days after MI in series 2. In series 3 fibers were dissected at mid papillary level central in the area at risk, to ensure evaluation of mitochondria exposed to IR. In series 2 fibers were dissected from the remote myocardium, since fibrotic areas are not suited for analysis.
To avoid any O 2 limitations to respiration the chambers were hyperoxygenated and all measurements were carried out in duplicate. The integrity of the outer mitochondrial membrane was tested by adding cytochrome c and an increase of > 10% in the oxygen consumption rate led to exclusion. The respiratory rates (O 2 consumption rates) are expressed as the O 2 flux normalised to the cardiac muscle mass of the permeabilized fibers (pmol s −1 kg −1 wet weight of permeabilized fibers).
GM: complex I respiration without ADP. GM3: Complex I respiration with ADP. GMS3: Complex I + II respiration with ADP (maximal coupled respiration). 4o: LEAK/non-phosphorylating resting respiration. ROX: residual oxygen consumption. Moc: fatty acid respiration without ADP. Moc3: fatty acid respiration with ADP. The respiratory control ratio (RCR) was calculated as state 3 respiration/state 2 respiration and expresses the respiratory coupling efficiency of the electron transport system (ETS) independent of muscle weight and mitochondrial density.
Mitochondrial enzymatic activity. We analyzed citrate synthase (CS) activity in the cardiac tissue by spectrophotometry as described previously 16 . The results are expressed as μmol min −1 g protein −1 .
Ketone body assay. β-hydroxybutyrate (β-OHB) was quantified in rat plasma using hydrophilic interaction liquid chromatography tandem mass spectrometry as previously described 17 . The blood sample was taken from the left femoral vein immediately prior to excision of the heart.

Myocardial interstitial concentrations of metabolites assessed by microdialysis in vivo.
Myocardial interstitial concentrations of metabolites were assessed by microdialysis as described previously 18 . After placement of the ligature around the LAD in the in vivo model but before inducing ischemia, a microdialysis probe (membrane length 4 mm, cut-off 6 Da; AgnTho's, Lidingoe, Sweden) was carefully guided by a 26 g needle into the free anterior wall of the LV in the estimated area of infarction. The probe was connected to a microdialysis pump (Univentor Limited, Zejtun, Malta) and perfused at a flow speed of 1μL min -1 with deoxygenated KHB (95%N 2 and 5% CO 2 ). Following insertion, a 20-min period of constant flow was allowed for stabilization, to ensure the concentrations of myocardial metabolites reach equilibrium. We measured concentrations of the TCA cycle intermediates; citrate, malate and succinate and ATP degradation products; hypoxanthine, xanthine, adenosine and inosine, during stabilization, ischemia and reperfusion at 10-min intervals and stored samples at − 80 °C. During collection the vials were cooled to approximately 4 °C. Samples were analysed by liquid chromatography and mass spectrometry. The absolute values were corrected to recovery rate as described previously 19 .
Myocyte cross sectional area. Hematoxylin/eosin (HE) staining was used for visualization of myocyte crosssectional area.
Briefly, cross sections of the left ventricle (3 µm) were deparaffinized and counterstained with hematoxylin II (Ventana Medical Systems, AZ, USA) 28 days after MI.
Myocyte cross-sectional area was evaluated at 400 × magnification in the HE stained sections. Cells were measured manually by outlining the cell contour. Only cells with a visible nucleus, a clear and intact cell membrane and cells located perpendicular to the plane were measured. Five to seven randomly selected fields in each section were selected for evaluation in the remote myocardium, which was characterized as the area most distally from the infarcted area.
Microscopy was performed on a light microscope (BX50F4, Olympus, Tokyo, Japan) and image analysis was performed using Image J software (NIH, Bethesda, MD, USA).
Drug preparation and administration. EMPA (30 mg/kg) (Empagliflozin, Merck, Darmstadt, Germany) 20 was dissolved in a 0.5% hydroxyethylcellulose (Sigma-Aldrich, St. Louis, MO, USA) solution and administered by oral gavage. Placebo animals received vehicle only. The relatively high dose of EMPA was based on pivotal nonclinical safety studies 20 , a pilot trial in (Supplementary Fig. S2) and since this study serves as a proof-of-concept study exploring the mechanism of action rather than the clinical potential of EMPA.
Statistical analysis. Based on own experience and reports by other groups, a sample size of n = 10 was considered adequate to identify a treatment effect 12,15,21 . All results are expressed as mean ± SD unless otherwise stated. Comparisons of means between three or more groups were analysed by one-way ANOVA with post-hoc Bonferroni test. Comparison between two groups was done using a student's t-test. Concentrations of myocardial interstitial metabolites were analysed using two-way ANOVA. All analyses were performed using GraphPad Prism 8.2.0 (Graph Pad Software, CA, USA). P < 0.05 was considered statistically significant.

Results
In the in vivo infarct size study, we included 66 rats. Numbers for the final analysis and those excluded are shown in the flow chart in supplementary figure S1a. In the ex vivo infarct size series, we included 20 animals. Two rats were excluded due to protocol violations. In the post MI LV-dysfunction series, we included 50 rats. The flow chart is shown in supplementary figure S1b. In the mitochondrial respiratory capacity and intermediary metabolism series, we included 50 rats. Supplementary Figure S1c depicts the flow chart and specifies reasons for exclusion of 21 rats. EMPA treatment for 7 days prior to MI significantly reduced myocardial in vivo IS by 20% points compared to placebo (65 ± 7% vs. 46 ± 8%p < 0.0001) (Fig. 1a). Administration of EMPA 1.5 h before MI yielded no IS reduction in vivo (p > 0.99) (Fig. 1a). AAR did not differ between groups (Fig. 1b). EMPA did not reduce IS ex vivo (63 ± 16% and 53 ± 13%, p = 0.14) or affect hemodynamic performance (Supplementary Fig. S3).
LV EF was similar in all groups at baseline. LV EF was reduced after MI at day 1 and day 27 compared to baseline in the placebo (p < 0.0001), EMPA-Chronic (p = 0.0005) and EMPA-Post (p < 0.0001) groups (Fig. 3a), whereas there was no reduction in the sham groups between baseline and day 27 (p = 0.34). EMPA (EMPA-Chronic and EMPA-Post) significantly improved LV EF at day 1 and at 27 after MI compared to placebo (50 ± 5 and 49 ± 11 vs. 41 ± 9%, p < 0.05) (Fig. 2a).
Neither LV Vol s nor LV Vol d differed between groups at baseline (Fig. 2b,c). Animals treated with EMPA (EMPA-Chronic and EMPA-Post) had significantly improved LV Vol s at day 27 after MI compared to placebo (368 ± 63 and 322 ± 59 vs. 427 ± 71 µL , p < 0.05) (Fig. 2b). We observed no difference in LV Vol d between the EMPA treated and placebo groups at day 27 after MI (Fig. 2c).
Enzymatic activities of CS did not differ between either of the experimental series (Figs. 3d, 4c). At the beginning of stabilization, the myocardial interstitial concentration of the tricarboxylic acid (TCA) cycle intermediates citrate, succinate and malate were similar in all groups (Fig. 5), but the concentration of glutamate was significantly higher in EMPA-Chronic compared to placebo and remained so during stabilization. EMPA-Acute increased the myocardial citrate level and EMPA-Chronic increased the succinate level before the induction of ischemia compared to EMPA-Chronic and placebo respectively. During ischemia, the interstitial citrate concentration was significantly higher in the EMPA-Acute group compared to EMPA-Chronic and remained continuously increased during reperfusion. The interstitial concentrations of succinate, malate and glutamate were significantly increased by EMPA-Chronic compared to placebo during ischemia and also in early reperfusion.
At the end of stabilization, the interstitial concentrations of purine metabolites were similar in all study groups. Concentrations of adenosine and inosine were significantly elevated in early ischemia compared with EMPA-Acute. During early reperfusion, the interstitial concentrations of adenosine, inosine, hypoxanthine and xanthine were significantly increased in the EMPA-Chronic group compared to EMPA-Acute.
The interstitial concentration of lactate was significantly elevated during ischemia in EMPA-Chronic compared to EMPA-Acute but not compared to placebo.
EMPA increased circulating β-OHB acid levels after 1.5 h of its administration. After 7 days of its administration, the level of β-OHB acid was normalized (Fig. 6).
At 28 days after MI, myocyte cross sectional area was slightly increased in the EMPA-Post group compared to placebo but otherwise similar across the study groups (ANOVA, p < 0.05) (Fig. 7).

Discussion
The present study demonstrates that continuous treatment with EMPA for 7 days prior to MI reduces IS in an in vivo rat heart model, while a single administration 1.5 h before MI does not. IS reduction is associated with a subsequent improvement of LV function at 28 days post MI. The most notable finding is that administration of EMPA in a post MI period of 28 days improves LV function regardless of time of therapy initiation and hence independently of IS reduction in the in vivo rat heart. The hemodynamic effect of SGLT2 inhibition was associated with an improvement in mitochondrial respiration that was documented in healthy, sham-operated rats. The cardioprotective effect and the effect on mitochondrial respiration seem to be independent of circulating β-OHB levels. Our results confirm that EMPA yields no cardioprotection in the isolated heart. EMPA did not affect mortality rate (Fig. S1).
The mechanisms underlying the cardioprotective effect of treatment with EMPA is of importance in the light of the recent demonstration of a beneficial cardiovascular effect of SGLT2 inhibitors. Notably, the beneficial effects were obtained not only by patients with 8,22 but also in patients without diabetes 8,9 . In accordance with previous studies, we found that EMPA administered for one week prior to MI reduced IS in non-diabetic rat hearts [23][24][25] . Administration of EMPA as a single dose shortly before MI did not offer the same effect 23 . Even though acute intravenous administration of dapagliflozin and canagliflozin is protective 26,27 , this observation may indicate that the main mechanism of EMPA does not work by a primary and direct modulation of the myocardium. This assumption is supported by the fact that administration of SGLT2 inhibitors directly into the perfusate in an ex vivo, Langendorff setting leaves cardioprotection unobtainable 10,12 . We confirmed these findings in a Langendorff setting that eliminated any effect of circulating substances (e.g. metabolic substrates such as β-OHB), since the initial washout with high coronary flow rates rapidly dilutes substances, which are carried over from the body. Hence, the mediating cardioprotective signal appears to be dependent on a wholebody system. EMPA improves contractility in hypoxic cardiomyocytes 25 . We extended these findings by demonstrating that EMPA enhanced EF irrespective of the time of treatment onset in rats with post infarction compromised LV function. Our findings demonstrated that IS reduction alone is not responsible for the effect. EMPA improved end systolic volume, while end diastolic volume remained unchanged, reflecting increased contractility. In accordance with our findings, canagliflozin increased stroke volume and myocardial efficiency without altering myocardial substrate utilization, i.e. uptake of glucose, fatty acids or ketones, in otherwise healthy swine subjected to IR injury 28 and alleviated post-ischemic systolic and diastolic function in non-diabetic male rats 27 . Together, these findings suggest that SGLT2 inhibitors have an inherent beneficial modulating class effect that reduces the cardiac derangements following MI.
Expectedly, the mitochondrial complex I and complex I + II linked oxidative phosphorylation (OXPHOS) capacity was impaired by IR injury during early reperfusion after MI. EMPA-Chronic significantly improved mitochondrial function. EMPA-Chronic also improved mitochondrial fatty acid oxidation. The main electron entry sites in the ETS after β-oxidization of fatty acids are complexes I and II via NADH and electron transfer flavoprotein via FADH 2 29 . Hence, the increased respiration by EMPA might reflect an overall improvement in www.nature.com/scientificreports/  www.nature.com/scientificreports/ the electron flow through complex I and II, as reflected in the respiratory coupling efficiency (RCR) and driven by an increase in state 3 respiration. Reduction of the oxidative capacity of fatty acids plays an important role in the development of HF after IR injury 30,31 . This reduction provides an early indication of deranged cardiac mitochondrial performance in HF 30 . EMPA-Chronic improved both complex I and fatty acid respiration. Because we measured mitochondrial respiration under unloaded resting conditions in the absence of EMPA or its potential mediator, the preserved mitochondrial function may favour a mechanism underlying the improved cardiac function rather than a consequence and that the modulation has already happened at the organ level  www.nature.com/scientificreports/ in the intact body. Supporting this, we observed a significantly improved mitochondrial respiratory capacity in sham animals treated with EMPA compared with sham animals treated with placebo. Thus, our findings seem to represent a specific effect of EMPA on mitochondrial respiratory capacity. We have previously demonstrated a similar effect in an in vitro model 12 . ATP degradation end products (i.e. adenosine, inosine, hypoxanthine and xanthine) were elevated during ischemia 19 . In accordance with an increased ATP-turnover from the elevated respiratory capacity, in particular during reperfusion when the respiratory chain is recovering 32 , we found an amplified increase in the ATP degradation products in the EMPA-Chronic group. The myocardial interstitial concentrations of TCA cycle intermediates succinate, malate and glutamate were significantly elevated in the EMPA-Chronic compared to placebo during ischemia.
In our experimental rat model, we found that levels of β-OHB increased only transiently after administration of EMPA. Conversely, in the EMPA-Chronic group, circulating β-OHB levels were normalized at the time of exposure to IR with no apparent association with the observed IS reduction. Since we were unable to associate elevated β-OHB levels in plasma with IS reduction and increased mitochondrial respiration, it seems unlikely that the cardioprotective effect should be coupled to increased β-OHB metabolism.
Some limitation must be acknowledged. We used healthy young male animals without comorbidities and without previous exposure to pharmacological treatment. Furthermore, we used a supra-therapeutic dose of EMPA to investigate the cardioprotective properties and mitochondrial modulatory effects on MI and post-MI HF. During several of the study procedures, the animals underwent anesthesia with volatile anaesthetics, which has been shown to be cardioprotective 33 . However, we found no differences in length of anesthesia or doses of sevoflurane between our study groups.
We found no differences in the number of mitochondria, measured by citrate synthase activity. A biomarker such as citrate synthase may not be an optimal marker for mitochondrial content across varying pathological conditions such as IR 34 . However, the same constraint relates to all markers that seem to have similar validity as citrate synthase 35 . To circumvent the limitation, we calculated the respiratory control ratio, as a reliable measure of respiratory coupling efficiency, independently from the number of mitochondria. Mitochondrial respiration was measured 30 min after reperfusion, whereas infarct size was evaluated after 2 h. The damaging process of reperfusion may extend beyond 30 min and thus mitochondrial respiration may still be significant as part of the cardioprotective mechanism, as shown in the chronic experimental series. However, the time discrepancy may explain that we found no IS reduction in EMPA-Acute animals, while mitochondrial respiratory capacity was similar to that observed in the EMPA-Chronic group. Similarly, it might explain that concentrations in intermediary metabolites were different in EMPA-Acute and EMPA-Chronic, while we found no difference in mitochondrial respiration.
The microdialysis samples were collected in 10-min spans to ensure sufficient material for analysis. This may limit the interpretation as a result of low temporal resolution compared to the rapid changes that occur during early reperfusion. Catheter implantation and surgical procedures may per se affect levels of myocardial interstitial metabolites 36 and may challenge the interpretation of the physiology in our specific experimental setup. The influence was observed initially and was similar in all study groups, whereas no increase was observed in the sham group during ischemia. Hence, we considered the differences between the study groups during ischemia and reperfusion valid. Furthermore, microdialysis is not the most accurate method to estimate exact concentrations in tissues, but we chose the continuous measurement approach from the same animal to assess dynamic changes.
We did not investigate causes of mortality during the experiments, specifically we did not monitor arrhythmicity, which might have provided information about differences in causes of mortality between groups. www.nature.com/scientificreports/

Conclusion
EMPA yields cardioprotection against acute IR injury in vivo but not ex vivo, indicating dependency of an intact body and an indirect effect by a delayed intrinsic cardioprotective mechanism. The protection persists in the failing rat heart by restoring systolic function and the effect is present irrespective of infarct size reduction. The cardioprotective effect is associated with enhanced mitochondrial respiratory capacity. Our data support a beneficial effect of EMPA in non-diabetic individuals with post infarction left ventricular dysfunction.