Increasing Pyruvate Dehydrogenase Flux as a Treatment for Diabetic Cardiomyopathy: A Combined ¹³C Hyperpolarized Magnetic Resonance and Echocardiography Study

It is now firmly established that type 2 diabetes contributes to an increased risk for the development of heart failure (1). Although some of this risk can be attributed to increased coronary artery disease and hypertension, it is becoming clear that patients with type 2 diabetes are also at risk for the development of “diabetic cardiomyopathy” (2–5), which manifests across a spectrum from subclinical left ventricular (LV) diastolic dysfunction (i.e., transmitral early diastolic peak velocity/early diastolic myocardial velocity ratio [E/E′]) to overt systolic failure (6). As the incidence of type 2 diabetes is rapidly increasing, understanding the pathophysiology behind diabetic cardiomyopathy and developing new treatment strategies is of increasing clinical importance.

Cardiac metabolism and altered substrate use are now emerging as candidate mechanisms underpinning diabetic cardiomyopathy and, as such, are targets for novel treatments (7,8). The cardiac metabolic changes in type 2 diabetes are linked to an increase in circulating fatty acid levels that results from insulin insensitivity and a failure to suppress adipose tissue hormone-sensitive lipase (9). This increase in fatty acid availability, and consequently increased cardiac usage, is thought to result in a loss of efficiency between substrate use and ATP production in the diabetic heart (10). Changes in cardiac substrate use precede functional changes in the diabetic heart (11). As a result, metabolic therapies aimed at restoring the balance of substrate use are an attractive target to improve, or even prevent, diabetic cardiomyopathy (12). Specifically, diastolic dysfunction could be a good initial indicator of the effect of metabolic therapy, given that dysfunction is seen in up to 60% of diabetic patients (13), precedes systolic dysfunction (14), and is more sensitive to changes in myocardial energetics than systolic function (15).

One of the potential targets for metabolic therapy is pyruvate dehydrogenase (PDH), a key enzyme that regulates the balance between carbohydrate and fat metabolism in the heart. In diabetes, PDH activity is decreased and pyruvate oxidation is impaired (16,17), resulting in a lack of metabolic flexibility in substrate selection, contributing to the overuse of fatty acids. Therefore, it follows that the restoration of PDH activity may re-establish a normal fuel balance, thereby restoring cardiac function.

PDH kinase (PDK) is responsible for the phosphorylation, and consequent deactivation, of PDH. Dichloroacetate (DCA) is a pyruvate mimetic that inhibits all isoforms of PDK, leading to an increase in PDH activity and glucose oxidation (18). The mechanism of action and isoform-specific inhibition kinetics of DCA have been extensively studied and reported (19,20). Previous studies (21,22) have shown that increasing glucose use via DCA, or via carnitine palmitoyltransferase-1 inhibition, improved LV function in the isolated perfused diabetic rat heart. As yet, there have not been any in vivo studies investigating whether DCA improves cardiac metabolism or function as a potential therapy for diabetic cardiomyopathy.

Until recently, we have been limited in our ability to measure metabolism in vivo. However, advances in hyperpolarized ¹³C magnetic resonance (MR) spectroscopy (23) have been made that now enable us to obtain a real-time, in vivo assessment of carbohydrate metabolism (24). In combination with an echocardiographic assessment of cardiac function, hyperpolarized ¹³C MR spectroscopy provides an ideal tool to investigate the effects of systemic treatments on diabetic myocardial metabolism and function in vivo.

In this study, we aimed 1) to use hyperpolarized [1-¹³C]pyruvate MR spectroscopy to assess cardiac PDH flux, and echocardiography to assess cardiac function, in a rodent model of type 2 diabetes (25); and 2) to investigate the effects of increasing PDH flux using DCA on cardiac substrate selection and function in the diabetic rodent myocardium.

Twenty-four male Wistar rats (initial body weight ∼200 g; Harlan U.K.) were housed on a 12:12-h light/dark cycle in animal facilities at the University of Oxford. All metabolic studies were performed between 7:00 a.m. and 1:00 p.m. with animals in the fed state. All procedures conformed to the Home Office Guidance on the Operation of the Animals (Scientific Procedures) Act of 1986 and to University of Oxford institutional guidelines. To generate a model of type 2 diabetes, two groups of eight rats were fed a high-fat diet for 3 weeks (60% fat, 35% protein, 5% carbohydrate; Special Diet Services, Essex, U.K.). After 12 days, these high-fat diet–fed rats were fasted overnight before receiving an intraperitoneal injection of 25 mg/kg streptozotocin (STZ; freshly prepared in citrate buffer, 0.655 g citric acid, 0.552 g sodium citrate in 100 mL double-distilled H₂O, pH 4) (25). These animals then continued on the high-fat diet for the remainder of the study. This model aimed to generate a moderate type 2 diabetic phenotype by using the high-fat feeding to induce peripheral insulin resistance, followed by a low dose of the pancreatic β-cell toxin STZ. STZ is traditionally used at high doses to induce type 1 diabetes, as it results in impaired insulin secretion from the β-cell. Reed et al. (26) proposed that if a low dose of STZ was used after high-fat feeding, the function of the β-cell mass would be modestly impaired without completely compromising insulin secretion, resulting in a moderate impairment in glucose tolerance (i.e., hyperglycemia in the absence of hypoinsulinemia). The final group of animals (n = 8) were fed a standard chow diet and received no injection of STZ so that they could act as controls.

After the initial 3-week period to establish the diabetic model, echocardiography was carried out in both the control and diabetic groups. Following this, one group of diabetic animals began treatment with DCA for a period of 28 days. Dichloroacetic acid (Sigma-Aldrich, Dorset, U.K.) was dissolved in the animals’ drinking water (to a final concentration of 1 mmol/L) and neutralized to pH 7.2 with NaOH. At the end of the 28-day treatment period, all animals underwent metabolic and functional analysis before being sacrificed, and plasma and tissue samples were collected. For all in vivo studies, animals were anesthetized using 2% isoflurane with 2 L/min oxygen.

Echocardiographic indices were obtained according to the recommendations of the British Society of Echocardiography. Transthoracic echocardiography was performed in control and diabetic animals, by blinded observers, with the use of a commercially available Vivid I echocardiography system (GE Healthcare) using an 11.5 MHz phased array 10S-RS pediatric echo probe. Wall thickness and LV dimensions were obtained from a short-axis view at the level of the papillary muscles. LV fractional shortening was calculated as (LVd − LVs)/LVd × 100, where LVs is LV end-systolic diameter and LVd is the LV end-diastolic diameter. Two-dimensional guided pulsed-wave Doppler recordings of LV inflow were obtained from the apical four-chamber view to measure maximal E and late peak velocity (A). The E parameter provides a measure of the peak early diastolic mitral inflow velocity, and is affected by left atrial pressure, LV relaxation, and LV systolic pressure. This results in it being preload sensitive. Tissue Doppler imaging was therefore also recorded from the medial mitral valve annulus to record E′ and active LV diastolic myocardial velocity (A′). E′ provides a measure of the velocity of the mitral medial annulus and is considered a noninvasive surrogate for LV relaxation. It is significantly less preload dependent, and the E/E′ ratio is therefore assumed to overcome the influence of ventricular relaxation on E and provide a preload independent reflection of LV filling pressure.

Rats were positioned in a 7T horizontal bore MR scanner interfaced to a Direct Drive console (Varian Medical Systems, U.K.), and a home-built ¹H/¹³C butterfly coil (loop diameter 2 cm) was placed over the chest, localizing the signal from the heart (27). Correct positioning was confirmed by the acquisition of an axial proton FLASH image (echo time 1.17 ms; repetition time 2.33 ms; matrix size 64 × 64; field of view 60 × 60 mm; slice thickness 2.5 mm; excitation flip angle 15°). An electrocardiogram-gated shim was used to reduce the proton linewidth to ∼120 Hz. Hyperpolarized [1-¹³C]pyruvate (Sigma-Aldrich) was prepared by 40 min of hyperpolarization at ∼1 K, as described by Ardenkjaer-Larsen et al. (23), before being rapidly dissolved in a pressurized and heated alkaline solution. This produced a solution of 80 mmol/L hyperpolarized sodium [1-¹³C]pyruvate at physiological temperature and pH, with a polarization of ∼30%. One milliliter of this solution was injected over 10 s via a tail vein cannula (dose ∼0.32 mmol/kg). Sixty individual electrocardiogram-gated ¹³C MR slice-selective, pulse-acquired cardiac spectra were acquired over 60 s after injection (repetition time 1 s; excitation flip angle 5°; slice thickness 10 mm; sweep width 13,593 Hz; acquired points 2,048; frequency centered on the C1 pyruvate resonance) (28). The combination of surface coil localization with slice selection ensured that the signals measured were accurately localized to the cardiac lumen (pyruvate) and the front wall of the myocardium (bicarbonate, lactate, alanine). Minimal contribution was provided from the skeletal muscle because of the large differential in metabolic rate between the contracting myocardium and the resting skeletal muscle.

Animals were sacrificed in the fasted state with an overdose of isoflurane (5% isoflurane with 2 L/min oxygen). The heart was removed, washed briefly in PBS, and frozen in liquid nitrogen. Epididymal fat pads were removed and weighed, with weights reported relative to body weight. Blood samples were taken and placed in heparinized tubes before centrifugation at 13,000 rpm for 10 min at 4°C. The plasma was then frozen in liquid nitrogen for later biochemical analyses.

Frozen tissue was crushed and lysis buffer was added before tissue was homogenized; a protein assay established the protein concentration of each lysate. The same concentration of protein from each sample was loaded onto 12.5% SDS-PAGE gels and separated by electrophoresis (29). Primary antibodies for PDK1 and PDK2 were purchased from New England Biolabs and Abgent, respectively; an antibody for PDK4 was donated by Professor Mary Sugden (Queen Mary University of London, London, U.K.); primary antibodies for uncoupling protein 3 (UCP3) and medium chain acyl-CoA dehydrogenase (MCAD) were purchased from Abcam. Even protein loading and transfer were confirmed by Ponceau staining (0.1% weight for volume in 5% volume for volume acetic acid; Sigma-Aldrich), and internal standards were used to ensure homogeneity between samples and gels. Bands were quantified using UN-SCAN-IT gel software (Silk Scientific), and all samples were run in duplicate on separate gels to confirm results.

Fasting blood glucose levels were assessed using an Optium Blood Glucose Monitor. Fasting insulin levels were assessed on postmortem plasma using an insulin ELISA kit (Mercodia, Uppsala, Sweden). Plasma metabolites (nonesterified fatty acids, LDLs, HDLs, triglycerides, and cholesterol) were assessed using an ABX Pentra 400 (Horiba ABX Diagnostics).

All cardiac ¹³C spectra were analyzed using the AMARES algorithm in the jMRUI software package (30). Spectra were direct current offset corrected based on the last half of acquired points. The peak areas of [1-¹³C]pyruvate, [1-¹³C]lactate, [1-¹³C]alanine, and [¹³C]bicarbonate at each time point were quantified and used as input data for a kinetic model based on that developed by Zierhut et al. (31) and Atherton et al. (24). PDH flux was quantified as the rate of ¹³C label transfer from pyruvate to bicarbonate (32). The rate of ¹³C label transfer from pyruvate to lactate and alanine was used as a marker of lactate dehydrogenase activity and alanine aminotransferase activity, respectively.

Values are reported as the mean ± SEM. Differences among the three groups were assessed using ANOVA with a Holm-Šídák post hoc correction. A Pearson correlation was used to assess the correlation between cardiac PDH flux and diastolic dysfunction (E/E′). Statistical significance was considered at P ≤ 0.05.

At the end of the 7-week study period, the diabetic group had significantly elevated fasting blood glucose levels compared with control animals (Fig. 1A). Fasting plasma insulin levels, epididymal fat pad weight, and body weights were not significantly different between the control and diabetic groups (Fig. 1B and C). Diabetic animals treated with DCA for 28 days demonstrated fasting plasma glucose concentrations that were significantly lower than those of untreated diabetic animals (Fig. 1A). Fasting plasma insulin levels were not different in treated diabetic animals compared with untreated (P = 0.30) and control (P = 0.13) animals (Fig. 1B). Epididymal fat pads were similarly unaffected in comparison with both untreated diabetic (P = 0.39) and control (P = 0.20) animals (Fig. 1C).

The plasma metabolites (Table 1) showed a significant reduction in the plasma triglyceride levels in the untreated diabetic and DCA-treated diabetic groups relative to controls. This finding potentially supports a greater rate of fatty acid oxidation in the diabetic animals that was unaffected by DCA treatment. The only other difference in the plasma lipid profile was a small decrease in the LDL levels measured in the DCA-treated diabetic animals relative to those of controls. Taken together, these results provide little evidence for a significant effect of the DCA treatment on fatty acid metabolism in the diabetic animals.

Cardiac PDK4 expression was elevated (Fig. 2A) in diabetic compared with control animals, but PDK1 and PDK2 expression was unaffected by diabetes (Fig. 2B and C). Cardiac expression of PDK1, PDK2, and PDK4 was significantly decreased in DCA-treated diabetic animals compared with untreated diabetic animals (P = 0.023, 0.009, and 0.002, respectively). PDK4 returned to a level not significantly different from that of controls (P = 0.2) (Fig. 2A). The protein expressions of the peroxisome proliferator–activated receptor-α target genes UCP3 and MCAD were elevated (Fig. 2D and E) in the untreated and DCA-treated diabetic animals relative to controls, indicating an increased rate of fatty acid oxidation in the diabetic animals that was unaffected by the DCA treatment.

The rate of ¹³C label transfer from pyruvate to bicarbonate in vivo was used as a real-time assessment of PDH flux (Fig. 3). Control animals showed a mean PDH flux of 0.017 ± 0.002 s^-1, compared with a significantly decreased cardiac PDH flux in diabetic animals of 0.005 ± 0.001 s^-1 (P = 0.0002; Fig. 4A). ¹³C label transfer to lactate in diabetic animals was not significantly different from that in controls; however, ¹³C label transfer to alanine was significantly increased (0.0080 ± 0.0003 s^-1) in subjects with diabetes compared with control subjects at 0.0049 ± 0.0005 s^-1 (P = 0.0001; Fig. 4B and C). Cardiac PDH flux was significantly increased by DCA treatment (0.018 ± 0.002 s^-1, P < 0.0001; Fig. 4A), compared with untreated diabetic animals, to the extent that the rate of flux was not different from that in control animals (P = 0.6). The rate of ¹³C label transfer to lactate was unchanged, but ¹³C label transfer to alanine was decreased in treated diabetic animal hearts compared with untreated subjects with diabetes (0.0038 ± 0.0004 vs. 0.0080 ± 0.0003 s^-1, P < 0.0001), to a level not significantly different from that in controls (P = 0.06; Fig. 4B and C).

Cardiac function, both systolic and diastolic, was assessed by echocardiography to investigate the differences between control and diabetic animals at both 3 and 7 weeks (3-week data not shown; see 7-week data in Fig. 5). At both time points, there was no difference in ejection fraction between groups or in E/A ventricular filling rates. A further measure of diastolic function, E/E′, was shown to be significantly increased in the diabetic animals at both 3 and 7 weeks, indicating diastolic dysfunction (Fig. 5C). At 7 weeks, the ejection fraction and E/A ventricular filling rates were not different in DCA-treated diabetic animals compared with either control or untreated diabetic animals (Fig. 5A and B). Treatment with DCA did, however, result in a decreased E/E′ ratio compared with diabetic animals (14 ± 1 vs. 20 ± 2, respectively; P = 0.019; Fig. 5C), restoring diastolic function back to that found in control rats (P = 0.4). Thus, cardiac PDH flux and diastolic dysfunction (as assessed by E/E′ ratio) were significantly negatively correlated (P = 0.02).

Overall, this suggests that, similar to the findings in humans, this diabetic model is characterized by hyperglycemia, reduced PDH flux, elevated PDK4 expression, and diastolic, but not systolic, dysfunction. Treatment with DCA reversed hyperglycemia, decreased PDK4 expression, improved PDH flux, and reversed diastolic dysfunction. This suggests that modulating substrate selection may be a potential therapeutic target for the treatment and/or prevention of diabetic cardiomyopathy.

Type 2 diabetes is a growing global health concern. It is now well established that type 2 diabetes markedly elevates the risk of the development of heart failure (1). Given the fact that the incidence of type 2 diabetes is rapidly increasing, it is likely that the rates of diabetic cardiomyopathy and heart failure will also continue to rise. As a result, further understanding of the etiology of diabetic cardiomyopathy and the development of novel therapeutic strategies aimed at preventing or treating diabetic cardiomyopathy is of great clinical importance.

Using the combination of hyperpolarized [1-¹³C]pyruvate spectroscopy and echocardiography, we have shown the following for the first time: 1) that the diastolic dysfunction associated with diabetes is linked with a reduction in PDH flux; and 2) that short-term treatment with the PDK inhibitor DCA can restore PDH flux, reverse diastolic dysfunction, and improve whole-body glycemic control in a rodent model of type 2 diabetes.

Substrate selection is a fundamental step in myocardial metabolism. In the normal resting heart, most (60–90%) of the acetyl CoA that enters the Krebs cycle comes from the β-oxidation of free fatty acids with 10–40% of acetyl CoA coming from the oxidation of pyruvate, which itself is derived from either glycolysis or lactate oxidation (8,33). However, the heart displays great flexibility in this choice of substrate, depending on the prevailing metabolic conditions (34). In the diabetic myocardium, because of the combined effects of insulin resistance, high circulating free fatty acid levels, and PDH inhibition, fatty acids are used almost exclusively to support ATP synthesis (22). This shift toward fatty acid metabolism and reduced carbohydrate oxidation appears to be due to a combination of several factors, including reduced insulin-induced GLUT4-mediated glucose uptake, suppressed glycolysis, and reduced mitochondrial PDH activity involving PDK4 (35,36). Further, the inhibition of glucose oxidation by fatty acids at the level of the PDH complex is universally reported (37–39).

The crucial importance of this increase in fatty acid metabolism lies in the fact that the mitochondrial redox state and, as a result, also the free energy of hydrolysis of ATP are negatively affected by a change in the balance of substrate use (40). In agreement with this, a decreased efficiency of substrate use to create ATP has been demonstrated in diabetes, and reduced myocardial energetics and diastolic dysfunction have been shown in multiple studies (41,42). This is in line with the concept that an impairment in high-energy phosphate metabolism initially affects the ability of the sarcoplasmic reticular Ca²⁺ ATPase, the most energetically demanding of all enzymes involved in contractile function (43), to lower cytosolic Ca²⁺, impairing diastolic function.

This is the first study to show a relationship between cardiac substrate selection and diastolic function in vivo. We have shown that, in the heart, the presence of diabetes is associated with reduced PDH flux (as a result of elevated PDK4 levels) and, importantly, that this reduction in PDH flux is related to impaired diastolic function. This establishes PDH as a potential therapeutic target for improving diastolic function in diabetes.

DCA inhibits PDK, which results in an increase in the proportion of active PDH (18). Although the PDK isoforms PDK1, PDK2, and PDK4 are present in the heart (19), it has been widely shown in diabetes that the inhibition of the PDH complex and an inability to metabolize glucose are mediated by the upregulation of PDK4 (44). In agreement with this, we have shown that diabetes was associated with increased PDK4 expression and with reduced PDH flux (Figs. 2A and 4A).

In addition, we have shown that treatment with DCA resulted in a restoration of PDH flux to a level seen in normal, nondiabetic animals, along with the downregulation of all isoforms of PDK in the heart (Figs. 2A and 4A). It is likely that the restoration of PDH flux was a result of both the downregulation of PDK expression and the direct effect of DCA on PDK activity, although PDK isoform activity was not directly assessed in this work (19). Furthermore, in association with this increase in PDH flux we have shown not only a reversal of diastolic dysfunction but also, crucially, that PDH flux remains related to diastolic function after DCA treatment. This suggests that restoration of the balance of cardiac substrate selection in diabetes by increasing flux through PDH, and increasing glucose oxidation, may be a central mechanism behind the observed restoration of diastolic function. This provides a therapeutic target in the form of cardiac substrate selection and, more specifically, the PDK/PDH interaction for the treatment of, or potentially the prevention of, diabetic cardiomyopathy.

Several other investigators have explored the physiological relevance and therapeutic potential of the PDK/PDH interaction using a PDK4-deficient mouse model in addition to pharmacological modulation with DCA. Jeoung et al. (45) showed that starvation lowered blood glucose levels more in mice lacking PDK4 than in wild-type mice. They further showed that the activity state of PDH was greater in the kidney, gastrocnemius muscle, diaphragm, and heart, but not in the liver of starved PDK4^−/− mice, indicating that the upregulation of PDK4 in tissues other than the liver was clearly important during starvation for the regulation of glucose homoeostasis. Jeoung and Harris (46) went on to show that fasting blood glucose levels were lower, glucose tolerance was slightly improved, and insulin sensitivity was slightly greater in PDK4^−/− mice compared with wild-type mice subjected to diet-induced obesity. Work by Ussher et al. (47) demonstrated that direct stimulation of PDH in mice with DCA significantly decreased infarct size after temporary ligation of the left anterior descending coronary artery. These results were then recapitulated in PDK4-deficient mice, which had enhanced myocardial PDH activity. Finally, Mori et al. (48) showed that the deletion of PDK4 prevented diastolic dysfunction and normalized blood glucose levels in a rodent model of cardiac hypertrophy.

While fatty acid oxidation rates were not specifically measured in this study, the protein expression levels of the peroxisome proliferator–activated receptor-α target genes UCP3 and MCAD were shown to be elevated in the diabetic animals irrespective of treatment with DCA. This would suggest an increase in fatty acid oxidation rates in diabetes that was unaffected by treatment with DCA. This finding is supported by the work of Ussher et al. (47), who have shown that palmitate oxidation rates are unaffected by DCA treatment in isolated perfused mouse hearts during ischemia/reperfusion experiments.

In addition to the cardiac glucose oxidation increase seen in this study as a result of DCA treatment, we observed a more systemic effect, in the reduction in circulating glucose levels. This is likely to be related to the direct effects of DCA on the liver (49). We have also shown a significant increase in cardiac ¹³C label transfer to alanine in diabetic animals (Fig. 4C), which is reduced to normal levels after DCA treatment, potentially demonstrating a reduction in the supply of this gluconeogenic precursor to the liver, facilitated by the glucose-alanine cycle. It is likely that this is at least partially involved in the restoration of normal blood glucose levels that is seen with DCA treatment. This also suggests that targeting the PDK family in diabetes is likely to result in positive effects in blood glucose management.

While our data support very strongly a direct effect of DCA on the heart (i.e., metabolic changes that correlate with the change in function), we cannot say that the whole-body effects of DCA are not having a contributory effect on the changes in cardiac function that we have observed. In support of the hypothesis that the direct cardiac effects of DCA are causing the changes in cardiac function is the study by Nicholl et al. (21), in which DCA was provided to an isolated working heart preparation from an STZ-induced diabetic rat. In this study, they demonstrated that isolated diabetic hearts provided with DCA revealed a marked improvement in function (both in terms of heart rate and rate pressure product). This work provides evidence of a direct effect of enhanced glucose oxidation on cardiac function but cannot exclude the potential effect of changes in whole-body metabolism from having made a contribution to the results observed in our study.

The incidence of diabetes is continually increasing, with >600 million people worldwide expected to have type 2 diabetes by 2030 (50). Further, subclinical diastolic dysfunction has been linked to an increased risk of heart failure (51). Therefore, a therapeutic target that both improves generalized glycemic control and reverses cardiac dysfunction is likely to be of great clinical importance. DCA has previously been the subject of clinical trials as a diabetic treatment (52). However, despite the benefits in glycemic control, it has not been widely used as a diabetic treatment because of side effects, which include peripheral neuropathy (53). Although other agents, such as phenylbutyrate (54), are now available to increase PDH activity, their use has, in general, been limited to the treatment of rare conditions of inborn errors in metabolism, including congenital PDH deficiency (55). The findings in our study support further work into the development of PDK and PDH modulators as targets for the treatment of type 2 diabetes. Finally, with clinical hyperpolarization studies imminent, studies that can identify reversible changes in cardiac substrate selection, which, if targeted, could improve cardiac dysfunction in diabetes, are likely to become a reality in the near future (56).

As with any study, the limitations of the techniques used need to be taken into consideration when interpreting the results. One limitation of the current study was that the use of hyperpolarized pyruvate only probes one side of the metabolic process, namely, carbohydrate metabolism, and is unable to report directly on fatty acid oxidation because of the low solubility and short hyperpolarized lifetimes of the physiologically relevant long-chain fatty acids.

Another limitation was that because of the echocardiography equipment used, we were able to assess only the diastolic parameters E/A and E/E′. The use of a higher-frequency echo probe operating at >12 MHz would have allowed a more in-depth study of the diastolic dysfunction observed in our diabetic animals (e.g., isovolumic relaxation time, myocardial performance index, propagation of mitral inflow, and flow propagation velocity). However, we feel that the use of the load-independent parameter E/E′ offered a sensitive and valuable assessment of diastolic dysfunction in our model and its normalization under DCA treatment.

No direct assessment of either glucose tolerance or cardiac insulin resistance was undertaken in this work, and so we cannot make any comment about the contribution that improved cardiac insulin sensitivity may have played in the restoration of PDH flux and the improvement in diastolic function that we observed. However, previous work by Lloyd et al. (57) has demonstrated distinct effects of insulin and DCA in the isolated perfused rat heart, with DCA specifically enhancing pyruvate oxidation, while insulin was shown to increase glucose uptake and glycolytic flux. It would, therefore, seem likely that the results observed in our study occur via the direct effect of DCA on PDH flux and pyruvate oxidation rather than via a secondary effect on cardiac insulin resistance.

Diabetes is a significant global health burden and is strongly linked to the development of widespread cardiovascular problems, including heart failure. We have shown here that by specifically targeting the PDK/PDH control of substrate selection in diabetes, benefits in both diastolic function and general glycemic control can be achieved.

Acknowledgments. The authors thank Lucia Giles, Lucy Ambrose, Latt Mansor, and Jack Miller (all from the University of Oxford) for their technical assistance and Professor Mary Sugden (Queen Mary University of London) for the donation of a primary antibody for pyruvate dehydrogenase kinase 4 (PDK4).

Funding. This study was funded by the British Heart Foundation (grants FS/10/002/28078 and FS/14/17/30634) and Diabetes UK (grant 11/0004175), and equipment support was provided by GE Healthcare.

Duality of Interest. L.M.L.P. was supported in the form of a partial contribution to her DPhil studies by AstraZeneca PLC, London, U.K. D.J.T. has previously received grant support from GE Healthcare. E.J. is an employee of AstraZeneca PLC, London, U.K. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. L.M.L.P. undertook the experiments, analyzed the data, and wrote the manuscript. O.J.R. undertook the experiments, analyzed the data, and reviewed and edited the manuscript. A.J.L., V.B., C.A.C., and L.C.H. undertook the experiments and reviewed and edited the manuscript. K.C. and E.J. reviewed and edited the manuscript. D.J.T. conceived the study, analyzed the data, wrote the manuscript, and reviewed and edited the manuscript. D.J.T. 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.