The Heart has Intrinsic Ketogenic Capacity that is Facultatively Upregulated in Heart Failure with Preserved Ejection Fraction

Heart Failure with preserved Ejection Fraction or ‘HFpEF’ is now the most common type of heart failure worldwide, with considerable morbidity and mortality, and with no effective pharmacotherapies. For the rst time, we show that the human and murine HFpEF heart have impaired uptake of ketone bodies, and reveal an intrinsic capacity of human and murine hearts to generate their own ketones via the canonical rate-limiting ketogenic enzyme, 3-hydroxy-3-methylglutaryl CoA synthase 2 (HMGCS2). We demonstrate that protein levels of HMGCS2 are dramatically elevated in HFpEF myocardium that does not overcome reduced specic activity of this enzyme. This appeared to be due to hyperacetylation caused by depletion of the de-acetylating enzyme sirtuin 3 and decreased oxidized/reduced nicotinamide adenine dinucleotide (NAD+/NADH) ratio, upon which the deacetylation activity of sirtuins is dependent. Also for the rst time, we show that the canonical stimuli that serve to turn on ketogenesis in the liver are present locally in the HFpEF heart, augmented by a profound energy deciency as determined by depleted phosphocreatine/ adenosine triphosphate (PCr/ATP) ratio. Thus, metabolically inexible HFpEF hearts try to auto-regulate their energy supply by upregulating ketogenesis as an apparent “rescue strategy”. These results serve to reframe our understanding of cardiac ketone metabolism and open opportunities for novel therapeutic strategies for HFpEF, urgently needed. the y-axis, signicance indicated with *. (c) Bar plot illustrating changes in substrates and products of the fatty acid transport, fatty acid oxidation, ketogenic, and ketone oxidation pathways in HFpEF vs control myocardium with log2 fold change on the y-axis, signicance indicated with *. *P ≤ 0.05 (murine experiments, n = 8 mice/group) (d) Schematic summarising the changes in ketogenic and lipid oxidation pathways with subcellular location indicated. HMGCS2 is seen as an inection point between increases and decreases in pathway intermediates; proteins indicated in yellow denote upregulation and blue downregulation in the HFpEF heart. The ratios of percentage cardiac uptake (Arterial - coronary sinus concentrations / Arterial concentrations * 100): PCWP with exercise of (e) β-Hydroxybutyrate and (f) acetoacetate were signicantly decreased in HFpEF hearts compared to controls. A positive number indicates net uptake by the heart, whereas a negative number indicates net secretion. (g) Percentage cardiac uptake assessment illustrated more acetoacetate (Y-axis) uptake than β-hydroxybutyrate (X-axis) in human HFpEF hearts (human transcardiac gradients, n= 13 controls, n= 19 HFpEF patients).


Introduction
Of all organs, the heart has the highest energy requirements and resting metabolic rate per unit mass (440 kcal/kg/day) 2 , and highest oxidative demand 3 . Therefore, it is critical that the heart can remain exible in its choice of substrate in the face of changing nutrient supply, metabolic stress, and pathological injury. Certain states, such as diabetes and heart failure, impair this exibility. In these states, alternative substrates such as ketone bodies become extremely important. It is thought that ketone bodies are a favoured substrate under conditions of stress and/or metabolic in exibility as they are a "thrifty" substrate 4 ; compared to fatty acids, ketones are more energetically e cient, yielding more energy per unit cost in oxygen consumption (P/O ratio) [5][6][7] . Recently, both clinical 8 and preclinical 9 data demonstrated bene cial effects of ketone administration in heart failure, and in HFpEF speci cally 10 .
Mechanistic study of HFpEF has been challenging due to the heterogeneity of clinical presentation, poor diagnostic criteria, and the absence of clinically-representative models. Recently, it was shown that invasive haemodynamic assessment during exercise provides superior diagnostic 11 and predictive 12 classi cation of HFpEF. Furthermore, a recent murine model was developed that is highly-representative of the human HFpEF syndrome 13 . Therefore, our goals were to: 1) use this murine model to address fundamental questions about metabolic exibility and the role of myocardial ketone utilisation in HFpEF; 2) examine ketone metabolism in human myocardium; and 3) determine cardiac ketone uptake in carefully-de ned human HFpEF that incorporates invasive exercise haemodynamic assessment.
To isolate the contribution of high-fat diet (without L-NAME) to changes in ketogenesis and ketolysis, we used high-fat diet only compared to chow. High-fat diets alone did not signi cantly change myocardial HMGCS2, HMGCL, BDH1, or OXCT1 levels (Extended Data Fig. 2a). We also observed no change in the βHb-to-AcAc ratio of the HFD hearts when compared to the control mice hearts (Extended Data Fig. 2b), while HFpEF mouse myocardium showed a signi cant increase in βHb-to-AcAc ratio (Extended Data Fig.   2c).
Transcardiac gradients in human HFpEF reveal decreased ketone uptake compared to healthy controls, with a predilection for AcAC uptake compared to βHB In a cohort of HFpEF patients and controls (Extended Data Table 1), we determined transcardiac ketone body gradients. All patients underwent invasive haemodynamic assessment before and during exercise. As in Fig. 2e and f, the ratio of percentage cardiac uptake (Arterial -coronary sinus concentration/ arterial concentration *100) of both βHB and AcAc to pulmonary capillary wedge pressure was signi cantly decreased in HFpEF patients compared to controls, indicating impaired ketone transport in HFpEF, consistent with that seen in the HFpEF murine model (Fig. 2b,c). Further, comparing percent uptake of βHB with AcAc revealed a predilection for AcAC uptake compared to βHB. In murine HFpEF myocardium, there was an increased βHB/AcAc ratio (Extended Data Fig. 2c), and this is consistent with increased AcAc compared to βHB uptake in human HFpEF hearts.

Murine and Human Hearts Have Intrinsic Ketogenic Capacity
As described above, comprehensive proteomic pro ling revealed that 3-hydroxy-3-methylglutaryl CoA synthase 2 (HMGCS2), the canonical rate-limiting ketogenic enzyme, was ~16-fold elevated in HFpEF myocardium (Fig. 2a). We wanted to establish de nitively whether the heart is capable of synthesizing ketone bodies via this canonical enzyme, which we assessed by condensation of β-oxidation- HMGCS2 speci c activity is signi cantly impaired in HFpEF myocardium, which is not overcome by upregulation of protein levels Western blotting con rmed (Fig. 4a) the dramatic upregulation of HMGCS2 protein levels seen by proteomics (Fig. 3a). The overall activity of the ketogenic pathway, indicated by the rate at which [1,2,3,4-13 C 4 ]acetoacetate formed over time, however, did not change (Fig. 4a). Immunoprecipitation followed by western blotting revealed increased acetylation of lysine residues on HMGCS2 (Fig. 4b). This was accompanied by decreased Sirtuin 3 (FC = 0.76, p = 9.3 × 10 -4 ) levels in HFpEF myocardium (Fig. 4c). As reported in the liver, under normal conditions su cient levels of Sirtuin 3 deacetylate HMGCS2 to maintain its active site in the correct conformation for binding substrate acetoacetyl-CoA, i.e. maintaining its ketogenic activity 19 . Sirtuin 3 is dependent on preserved NAD + /NADH ratio 20 , which we found to be signi cantly decreased (Fig. 4c). Therefore, our results suggest that depleted NAD + /NADH leads to reduced Sirtuin 3 activity, resulting in over-acetylated HMGCS2 that is dysfunctional (Fig. 4d). We believe that this, in conjunction with the metabolic in exibility reported above, served as a stimulus to upregulate HMGCS2 protein levels in the HFpEF myocardium.

Discussion
It was reported several decades ago that in the liver conditions of fatty acid over-supply along with reduced oxaloacetate (due to suppressed glucose metabolism) serve as a ketogenic stimulus, leading to acetyl-coA entering the ketogenic pathway 21 . Remarkably, we see the exact same conditions in HFpEF myocardium (Fig. 3b, c), rea rming the same canonical generative pathway in the heart. Further, depletion of PCr/ATP con rms a state of signi cant energy depletion in the heart 22 , augmenting the stimulus to seek other sources of energy, and ketones are a preferred alternative energy source of the heart 23 . The current paradigm suggests that robust HMGCS2 expression only occurs in hepatocytes and gut epithelial cells: extra-hepatic tissues don't contribute ketone bodies to the circulation and they internalise ketones down a concentration gradient via MCT1/2-dependent mechanisms 24,25 . This includes the heart oxidising ketone bodies in direct proportion to their delivery, and the myocardium is the highest ketone body consumer per unit mass 26-31 . However, we observed decreased myocardial ketone body uptake from the circulation, whilst concurrently the levels of the rate-limiting ketogenic enzyme were dramatically upregulated.
It has been suggested that "apparent" extra-hepatic ketogenesis can occur in the context of decreased ketone oxidation/ consumption 24 , whereby decreased ketone oxidation leads to ketone body accumulation and gives the appearance of de novo ketogenesis. However, we did not see increased ketone body accumulation. Several extra-hepatic non-HMGCS2 "pseudoketogenic" processes have been described that include reversible enzymatic activity of thiolase and OXCT1 32 . However, our data con rms bone de canonical HMGCS2 ketogenesis in the heart. Regardless, we are not proposing that cardiac ketogenesis is important outside the heart or contributes to circulating ketones. Rather, our data demonstrates an unknown capacity of the heart to generate its own ketones at a time of metabolic in exibility and critical fuel deprivation.
Previous studies of rodent Type 1 diabetes models reported upregulation of myocardial mRNA levels of HMGCS2 in mice 33 , and both its mRNA and protein levels in rats 34 . However, these studies did not determine enzymatic activity, post translational modi cation, or the relative changes in fatty acid oxidation, glucose metabolism, or citric acid cycle changes at the metabolite level, essential to pinpoint the transition to ketogenesis. Our assessment of HMGCS2 activity was critical for determining the actual state of myocardial ketogenesis. Measuring only mRNA or protein levels would have provided a profoundly misleading conclusion. Activity of HMGCS2 is dependent on continual deacetylation of HMGCS2 at several lysine residues. Sirtuin 3 is the major deacetylating enzyme, itself dependent on the NAD + /NADH ratio 20,35 . In addition to its role deacetylating HMGCS2, there are other roles of Sirtuin 3 that are important to consider in the context of heart failure and HFpEF, including maintaining mitochondrial function [36][37][38] and preventing cardiac hypertrophy 36 .
Impaired cardiac ketone uptake in the murine model was supported by our data in humans. Exercise pulmonary capillary wedge pressure (PCWP) has been shown to be a more accurate diagnostic criterion for HFpEF 11,12 . To our knowledge, we are the rst to compare PCWP to transcardiac ketone gradients in HFpEF and controls. Using transcardiac gradients, others have shown that cardiac ketone uptake is increased in HFrEF compared to HFpEF 39 and in HFrEF and aortic stenosis patients compared to controls 40 . However, our data demonstrates decreased cardiac extraction of ketone bodies with elevations in exercise PCWP, a more accurate measure of HFpEF. This is an important nding, and is consistent with impaired ketone uptake in our murine HFpEF myocardium.
HFpEF hearts favoured AcAc over βHB uptake. In our murine model, there was depletion of AcAc but not βHB in murine HFpEF myocardium. As ketone uptake is reported to occur along circulating vs internal gradients via the MCT1/2 transporters, these ndings are consistent. Indeed, our HFpEF murine myocardium showed a decreased expression of key ketolytic enzymes BDH1 and succinyl-coA:3-ketoacid CoA transferase (SCOT/OXCT1), accompanied by increased myocardial βHB-to-AcAc ratio. AcAc-βHB redox coupling is important for maintaining myocardial energy homeostasis, and our data suggests uncoupling via BDH1 suppression, leading to depleted AcAc. This may serve as a further stimulus for ketogenesis to generate su cient AcAc to supply acetyl-coA for the Krebs cycle.
Exogenous ketone bodies are cardioprotective in HFpEF, allowing the heart to extricate energy at a lower oxygen cost 42 , and suppressing myocardial in ammation 10 . As such, local generation of ketones by the heart would serve as a valuable therapeutic target for multiple cardiac diseases. Herein, we provide data that challenges the current paradigm of the heart as only a consumer of ketones provided by the liver. We de nitively demonstrate that both murine and human hearts are capable of ketogenesis via the canonical ketogenic enzyme. For the rst time, we reveal that human HFpEF hearts have impaired ketone uptake. We show that in HFpEF, myocardial metabolic in exibility leads to decreased glycolysis along with upregulated fatty acid oxidation in the presence of depleted oxaloacetate, serving as a stimulus to potently upregulate the canonical ketogenic enzyme in an apparent strategy to rescue energy deprivation. We believe these results reframe our understanding of cardiac substrate utilisation and offer a novel potential therapeutic strategy in HFpEF: targeting the heart's inherent ketogenic capacity.

Animals and husbandry
All procedures involving animals received ethics approval from the Institutional Animal Ethics Committee at the University of Sydney. Twenty male 5-week-old wild-type C57BL/6 mice were obtained from The Australian BioResources Pty Ltd facility in Moss Vale, New South Wales. Mice were housed and maintained on a 12-hour light-dark cycle with free access to water and food. Two weeks after arrival, mouse cages were randomly assigned as control or treatment group. While mice in the control group received a normal CHOW diet (SF14-162, Specialty Feeds Pty Ltd), mice in the treatment group were fed a combination of HFD (60% calories from lard, SF18-072, Specialty Feeds Pty Ltd) and L-NAME (0.5 g/L in drinking water). After ve weeks of different dietary regimens, all 20 mice were sacri ced using overdosed intraperitoneal injections of pentobarbital (Nembutal ® , Abbott Laboratories, North Chicago, IL, USA; 25 mg/mL, 75 mg/kg) as approved by the Institutional Animal Ethics Committee at the University of Sydney.
Blood was collected via cardiac puncture using EDTA-coated syringes and sample tubes, then centrifuged to isolate plasma. Aliquots of plasma were frozen on dry ice and stored at -80 °C until analysis. The hearts were obtained, weighed and then clamped in liquid nitrogen stored metal clamps, ash frozen in liquid nitrogen and stored at -80 °C until use.
Anesthesia for mice echocardiography Echocardiography was performed in anaesthetised mice. Anaesthesia was commenced using 2-3% iso urane in oxygen in an induction chamber. Anaesthesia was maintained during imaging using 1-2% iso urane in oxygen delivered via nose cone. During imaging, heart rate was maintained between 450-550 beats per minute (bpm) and body temperature was kept at 36-37°C (rectal probe). The mice were made to lie on a heated imaging platform on their ventral side. All animal studies were approved by the Institutional Animal Ethics Committee at the University of Sydney.
Echocardiography measurements A Vevo2100 system high-resolution ultrasound system with a 40-MHz linear probe (FUJIFILM VisualSonics Inc., Canada) was used to perform echocardiography on mice 5 weeks after commencement of diets. The echocardiographic measurements were performed according to a standard method published previously 43

Body composition
Longitudinal measures of body composition were assessed using an EchoMRI 900 (EchoMRI, TX, USA) prior to culling at each experimental endpoint.

Blood glucose and insulin measurement
Oral Glucose Tolerance Tests (oGTT) were performed via oral gavage (2 g/kg lean mass) after 6 hours fasting. Basal blood glucose was measured through tail vein bleeding using a clinical glucometer (Accu-Chek Performa, Roche Diagnostics Australia Pty Ltd). Blood glucose was measured 15, 30, 45, 60, and 90 mins after glucose administration. Blood insulin levels were determined by the Ultra Sensitive Mouse Insulin ELISA kit (Crystal Chem Inc). Area under the curve (AUC) was derived from an oral glucose tolerance test, which indicates the time taken to clear a bolus dose of glucose from the bloodstream and return to basal levels. The Homeostasis Model Assessment of Insulin Resistance (HOMA-IR) was used as an estimation for global IR for each group. It was calculated by multiplying the fasting plasma insulin (µU/L) and fasting plasma glucose (mM) and divided by constant 66 44,45 .
Human heart tissue Donor left ventricular (LV) myocardium was obtained from donors whose hearts did not proceed to transplantation due to transportation logistics, donor-recipient mismatch in size, and immune incompatibility, as previously described [46][47][48] . Histological analysis of the donor samples were demonstrated to be structurally normal as per formal pathological examination 46 . LV samples from the donors were collected and snap frozen (−196 °C) and were stored in the Sydney Heart Bank at −170 to −180 °C from the time of harvest 46 . Written consent was obtained to use tissue specimens for research.
The study was approved by the Ethics Committee of The University of Sydney (USYD #2016/923 & 2020/ ETH01161) and was conducted in accordance with the Declaration of Helsinki.

Invasive Exercise Haemodynamics and Transcardiac Gradients
HFpEF patients were referred to the Department of Cardiology, Alfred Hospital for investigation of symptoms consistent with a diagnosis of heart failure (NYHA II-III) in the presence of a left ventricular ejection fraction (LVEF) >50%. Exclusion criteria included signi cant coronary artery disease which had not been revascularized; moderate or greater aortic or mitral valve disease; in ltrative, restrictive or hypertrophic myocardial disease; pericardial constriction; or signi cant right ventricular disease. The diagnosis of HFpEF was con rmed by the presence of an end-expiratory pulmonary capillary wedge pressure (PCWP) ≥15mmHg at rest or ≥25mmHg during symptom-limited exercise. Healthy subjects were recruited from the general community and were free from known cardiac or respiratory disease and displayed a normal exercise hemodynamic pro le including a peak PCWP <25mmHg. The study was approved by the Institutional Ethics Review Committee, and all participants provided written informed consent.
Sample preparation for mass spectrometry-based proteomics Frozen powdered heart tissue (10-20 mg) was weighed into tubes and lysed in 100 µL of 4% sodium deoxycholate and 100 mM Tris HCl (pH adjusted to 8). Tissue lysates were then heated at 95°C and mixed at 1,000 rpm for 10 minutes using Thermomixer C (Eppendorf, Hamburg, Germany) before they were subjected to centrifugation at 18,000 x g for 10 minutes at room temperature. Protein concentration of tissue lysates was determined by Bicinchoninic Acid (BCA) Protein Assay Kit (Pierce, Thermo Scienti c, Waltham, MA, USA). 20 μg of protein was subjected to reduction with TCEP and alkylated with chloroacetamide for 10 minutes at 95°C and 1,000 rpm in the Thermomixer C. Samples were then processed for trypsin digestion and subjected to a peptide cleanup as described previously 49 . Peptides were resuspended in 5% formic acid and stored at 4°C until acquired by LC-MS.

Data-dependent mass spectrometry-based proteomics
Using a Thermo Fisher Dionex Ultimate 3000 UHPLC, peptides in 5% (vol/vol) formic acid (500 ng) were directly injected onto a 50 cm x 75 um fused silica analytical column with a ∼10 μm pulled tip containing C18 material (Dr. Maisch, Ammerbuch, Germany, 1.9 μm, 130A), coupled online to a nanospray ESI source. Peptides were resolved over a gradient 5-40% acetonitrile over 120 min with a ow rate of 300 nL min −1 . Peptides were ionized by electrospray ionization at 2.3 kV. Tandem mass spectrometry analysis was carried out on a Q-Exactive HFX mass spectrometer using HCD fragmentation (Thermo). The datadependent acquisition method acquired MS/MS spectra of the top 20 most abundant ions at any one point during the gradient. RAW MS data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identi er PXD026582 (Username: reviewer_pxd026582@ebi.ac.uk, Password: oPiW26Wz). RAW data were analyzed using the quantitative proteomics software MaxQuant 50 (version 1.6.3.4). Peptide and protein level identi cation were both set to a false discovery rate of 1% using a target-decoy based strategy. The database supplied to the search engine for peptide identi cations contained both the mouse UniProt database downloaded on the 8th Nov 2018 and the MaxQuant contaminants database. Mass tolerance was set to 4.5 ppm for precursor ions and MS/MS mass tolerance was 20 ppm. Enzyme speci city was set to trypsin (cleavage C-terminal to Lys and Arg) with a maximum of 2 missed cleavages permitted. Deamidation of Asn and Gln, oxidation of Met, pyro-Glu (with peptide N-term Gln) and protein N-terminal acetylation were set as variable modi cations. Carbamidomethyl on Cys was searched as a xed modi cation. We used the MaxLFQ algorithm for label-free quantitation, integrated into the MaxQuant environment 51 . Maxquant output was processed and statistical tests performed using the R software package (version 3.4.3) and processed data plotted using Tableau (version 10.0.2).

Sample preparation for targeted metabolomics and non-targeted lipidomics
Brie y, approximately 30-50 mg of powdered mice heart tissues were subjected to metabolite and lipid extraction prior to liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis. Targeted metabolite pro ling used in this study was established using reference standards for each individual metabolite to determine MS multiple reaction-monitoring transitions, declustering potentials, collision energies, and chromatographic retention time, as described previously 52 . A stepwise extraction method was performed as previously described, using methanol/chloroform/water to e ciently extract lipids and metabolites from the tissues 52,53 .

Targeted metabolomics
For targeted metabolomic analysis, an LC-MS/MS system composed of a Shimadzu Nexera LC-30AD UHPLC (Shimadzu Corporation, Kyoto, Japan) system coupled to a QTRAP6500 mass spectrometer (AB Sciex, Foster City, CA, USA) was used to measure hydrophilic metabolites in positive and negative ionization modes. The polar metabolites in both positive and negative ionization mode were separated in hydrophilic interaction liquid chromatography (HILIC) mode using an Atlantis ® HILIC column (Waters) and an XBridge TM Amide column (Waters), respectively, which allowed the separation of metabolites of different properties, both as previously described 52,53 . Sample analysis was conducted in a randomized sample order and each sample was run in duplicate. Data was acquired in the same batch on the same day.
The analysis software Sciex OS 2.0 (ABSciex) was used for MRM Q1/Q3 peak integration of the raw data les (Analyst software, v.1.6.2; ABSciex). The peak area corresponds to the abundance of that metabolite; the abundance values were then normalized to their bookended pooled tissue extracts in the subsequent analysis, which were included after every 4 study samples in the sample queue, to account for temporal drift in instrument performance.

Non-targeted lipidomics
For analysis of lipid extracts from mice hearts, isotopically-labeled lipid standard mixture (SPLASH™ Lipidomix® Mass Spec Standard, Avanti Polar Lipids) was spiked into each sample, and the concentrations of identi ed lipid species from 14 major lipid classes included in the SPLASH were calculated relative to the known concentration of internal standards included in the SPLASH standard mixture.
Lipidomic pro ling in this study used a Thermo Scienti c™ Vanquish™ UHPLC system and Thermo Scienti c™ Q Exactive™ HF hybrid quadrupole-Orbitrap™ mass spectrometer, with lipid separation performed on a Thermo Scienti c™ Acclaim™ C30 column (2.1 x 150 mm, 2.6 µm), operated at 45 °C, ow rate of 260 µL/min. Lipid extracts from the mice hearts were analyzed using the LC conditions and MS instrument conditions in both positive and negative ion modes, as described previously 54 . Lipid Search 4.1 software was used to align and process each acquired raw le's search results, as well as to identify the lipid species present in the heart extracts 55 .
Tissue homogenization and lysis for enzymatic activity assay For the determination of HMGCS2 activity in mice, a total of 2-3 mouse heart tissues were combined for each assay (n = 5-7 pooled heart tissues assayed for controls and n = 6-8 pooled heart tissues assayed for HFpEF mice). Heart extracts were prepared by lysis in 500 μL of ice-cold mammalian cell lysis extraction mixture containing protease inhibitor cocktail solutions (MCL1; Sigma-Aldrich). Tissues were homogenized by Qiagen TissueLyser LT (50 cycle/sec) for 2 minutes. Particulate matter was then separated by centrifugation for 10 minutes at 4 °C at 11,000 ×g. The resulting clear supernatant was collected for incubation. Human myocardial tissues were subjected to the same tissue homogenization procedures for extract preparation for mice.

HMGCS2 enzymatic activity
The protein content of the lysate mixture was determined using the Bicinchonic Acid (BCA) Protein Assay Kit (Pierce Biotechnology, Rockford, IL, USA) according to the manufacturer's instructions. HMGCS2 activity assay in mouse and human heart tissue was carried out with modi cation as described previously 74,75 . HMGCS2 activity was measured based on the incorporation of isotope-labeled [1,2-13 C 2 ]acetyl-CoA to isotope-labeled acetoacetate at pH 8 and temperature set to 37 °C. To 250 μL of reaction mixture containing 50 mM Tris, 0.01 mM of acetoacetyl-CoA, 1 mM of [1,2-13 C 2 ]acetyl-CoA (Sigma, 658650-5MG), 150 μL (0.225 mg/mL) of the tissue homogenate was added. Reaction proceeded at 37 °C for 5,10,20,30,45,60,90,120,150,180, and 210 minutes, with removal of 30 μL of aliquots of the reaction mixture, and the enzymatic reaction was stopped immediately by adding 5 μL of 5 M perchloric acid starting at time = 0 and the indicated time points. Measurement of 13 C-labeling in intermediates of the ketogenic pathway (acetoacetyl-CoA and HMG-CoA) was made using liquidchromatography tandem mass spectrometry.
To test for any non-enzymatic production of isotope-labeled acetoacetyl-CoA, HMG-CoA and acetoacetate, which would interfere with the HMGCS2 activity assay, samples comprising the lysis buffer plus the other chemicals, but no tissue, were incubated and analyzed exactly the way as normal tissue samples.
Calibration against analytical standards was carried out using labeled acetyl-CoA, unlabeled acetoacetyl-CoA, HMG-CoA and acetoacetate in the concentration ranges of 0.03-100 µmol/l. All standards were not exposed to the incubation but were added with the same amount of perchloric acid, followed by similar sample preparation steps to provide as similar as possible sample composition in comparison to the incubated tissue samples. Concentration values of 13 C 4 -AcAc were obtained in the unit of µmol/l. These values were divided by the incubation time (hrs), and the protein concentrations, respectively, in the unit mg/l. The resulting enzymatic activity values were expressed in total 13 C 4 -AcAc generated µmol/mg protein/hr, which were transformed to the nal unit nmol.mg p -1 .hr -1 for ease of interpretation.

Immunoblotting
Protein concentration of the resultant supernatant from the enzymatic assay was determined using the Bicinchonic Acid (BCA) Protein Assay Kit (Pierce), following manufacturer's instructions. 40 µg of total protein of heart tissue lysates were heat denatured with 4× Laemmli sample buffer (Bio-Rad, Hercules, CA) supplemented with 10% β-mercaptoethanol (Sigma) at 95 °C for 5 minutes, and the proteins were resolved on 4-15% Mini-PROTEAN ® TGX™ Precast Gel (Bio-Rad), and transferred to a PVDF membrane (Thermo Fisher Scienti c). Blocking of the membrane was performed using 5 % non-fat milk in Trisbuffered saline (TBS) containing 0.1 % Tween-20 for 1 hour. The membrane was then incubated with a primary antibody at 4°C overnight. Following three subsequent washings in TBST, the membrane was then probed with secondary horseradish peroxidase conjugated antibody [anti-rabbit (P0448, Dako, Carpinteria, CA, USA) or anti-mouse (P0447, Dako, Carpinteria, CA, USA)] for 45 minutes. The blots were then visualized using ImageQuant LAS 4010 digital imaging system (GE Healthcare). The protein levels were normalized to actin and compared to the control group. Quantitation was completed with Image J.
Each experiment was completed with n ≥ 2. The primary antibodies and dilutions probed for the target proteins are shown in the table below.  Figure 1 HFD and L-NAME administration led to obesity, glucose intolerance, insulin resistance and features of HFpEF, indicated by (a) fat mass, (b) increased glucose excursion, (c) increased HOMA-IR, (d) heart weight/tibia length, (e) increased transmittal doppler E/A wave ratio, (f) increased transmitral E to tissue doppler E' (E/E') ratio, (g) impaired global longitudinal strain; (h) pulmonary oedema indicated by increased wet/dry lung weight. (i) Heatmap showing decreased expression of glucose transporter and glycolytic enzymes in HFpEF myocardium, where red indicates high protein abundance and blue indicates low protein abundance. Each column represents an individual mouse. (j) Box-and-whisker plots with individual dot points illustrating signi cant reductions of several rate-limiting enzymes critical for the regulation of glycolysis and glucose oxidation, with upregulation of PDK4 that inhibits the pyruvate dehydrogenase complex leading to metabolic in exibility. Filled circles represent statistical signi cance    Stable-isotope tracing of myocardial ketogenesis. a) Corresponding 13C labelled ketogenesis intermediates from [1,2-13C2]-acetyl-CoA in murine (n = 3) and human (n = 7) myocardium. The circles above each metabolite name denote the number of labelled carbons considered for that metabolite. Isotopologues of each substrate are denoted by the number of 13C-labelled carbons: for instance, m0 has no labelled carbons (i.e. is unlabelled), m1 has one 13C-labelled carbon, m2 has two 13C-labelled carbons, etc. The data are presented as means ± SD (error bars) for individual isotopologues. (b) 13C4labelled acetoacetate measured at select time intervals in murine (n = 3) and human (n = 7) myocardium.