Integrated Systems Biology Identifies Disruptions in Mitochondrial Function and Metabolism as Key Contributors to HFpEF

Visual Abstract

Highlights

• •While hypertension is common, the necessity of both metabolic syndrome and obesity for the HFpEF phenotype underscores their pivotal role in disease development.
• •Transcriptional and metabolic remodeling in HFpEF is characterized by the up-regulation of inflammatory processes and the down-regulation of mitochondrial and energy metabolism pathways.
• •Mitochondrial ultrastructural and functional remodeling underlie the HFpEF phenotype and likely are an early contributor to cardiac dysfunction.
• •The HFpEF heart displays significant intramyocardial lipid accumulation characterized by large increases in lipid droplet size, number, and association with mitochondria.

Summary

Heart failure with preserved ejection fraction (HFpEF) accounts for ∼50% of HF cases. The ZSF1-obese rat model recapitulates clinical features of HFpEF including hypertension, obesity, metabolic syndrome, exercise intolerance, and diastolic dysfunction. We utilized a systems-biology approach to define the metabolic and transcriptional signatures to gain mechanistic insight into pathways contributing to HFpEF development. Male ZSF1-obese, ZSF1-lean hypertensive controls, and WKY (wild-type) controls were compared at 14 weeks of age for extensive physiological phenotyping and left ventricle (LV) tissue harvesting for unbiased-metabolomics, RNA-sequencing, and mitochondrial morphology and function. Utilizing ZSF1-lean and WKY controls enabled a distinction between hypertension-driven molecular changes driving HFpEF pathology, versus hypertension + metabolic syndrome. Comparison of ZSF1-lean vs WKY (ie, hypertension-exclusive effects) revealed metabolic remodeling suggesting increased aerobic glycolysis, decreased β-oxidation, and dysregulated purine and pyrimidine metabolism with few transcriptional changes. ZSF1-obese rats displayed worsened metabolic remodeling and robust transcriptional remodeling highlighted by upregulation of inflammatory genes and downregulation of the mitochondrial structure/function and metabolic processes. Integrated network analysis of metabolomic and RNAseq datasets revealed downregulation of most catabolic energy producing pathways, manifesting in a marked decrease in the energetic state (ie, reduced ATP/ADP, PCr/ATP). Cardiomyocyte ultrastructure analysis revealed decreased mitochondrial area, size, and cristae density, as well as increased lipid droplet content in HFpEF hearts. Impaired mitochondrial function was demonstrated by decreased substrate-mediated respiration and dysregulated calcium handling. Collectively, the integrated omics approach applied here provides a framework to uncover novel genes, metabolites, and pathways underlying HFpEF, with an emphasis on mitochondrial energy metabolism as a potential interventional target.

Heart failure (HF) is a growing epidemic. In the U.S. alone, more than 6.7 million people over the age of 20 years have HF, and this is projected to increase to more than 8.5 million people by 2030.^1,2 Nearly one-fourth of people develop HF in their lifetime,1, 2, 3 and current HF mortality rates are higher today than in 1999.⁴ Of those diagnosed with HF, ∼50% have heart failure with preserved ejection fraction (HFpEF).1, 2, 3^,5 HFpEF patients present with elevated left ventricular (LV) filling pressure despite normal LV ejection fraction (≥50%). At present there are very limited treatments for HFpEF, and the 5-year mortality rate is >50%.1, 2, 3 Clinical trials of drugs that are effective in HF with reduced ejection fraction (HFrEF) have uniformly failed in HFpEF.^1,3 Owing to our limited understanding of mechanisms that drive HFpEF and lack of therapeutic strategies to treat this devastating disease, the National Heart, Lung, and Blood Institute of the National Institutes of Health has issued a statement of emphasis detailing the research priority of HFpEF and identified HFpEF as the greatest unmet need in cardiovascular medicine.⁶

While clinicians struggle to treat HFpEF patients, research scientists grapple with preclinical models to study the pathobiology of HFpEF7, 8, 9, 10, 11, 12 to improve our understanding of this complex multi-organ disease. Clinically relevant models are required to fully elucidate molecular disease mechanisms and effectively translate new therapies. Toward this end, the ZSF1 rat has been proposed as an animal model for HFpEF.^13,14 This model was created by crossing rat strains with 2 separate leptin receptor mutations (fa and facp), the lean female Zucker diabetic fatty (ZDF) rat (+/fa) and the lean male spontaneously hypertensive heart failure (SHHF) rat. Offspring homozygous for both mutations (fa:facp) create a hybrid rat with central obesity and hypertension (ZSF1-obese rat) resulting in spontaneous cardiometabolic HFpEF, whereas the heterozygous lean offspring (ZSF1-lean rat) exhibit limited signs of obesity or diabetes.¹⁵ Previous studies have shown that ZSF1-obese rats develop significant diastolic dysfunction by 10-20 weeks of age with concentric LV remodeling and hypertrophy like that observed in HFpEF patients.¹⁶ In addition to LV diastolic dysfunction, studies have demonstrated skeletal muscle pathology, exercise intolerance, endothelial dysfunction, systemic inflammation, and renal and hepatic abnormalities that are consistent with cardiometabolic HFpEF.^13,16, 17, 18, 19

We have previously demonstrated that the ZSF1 rat is responsive to therapeutic interventions when delivered early during the progression of HFpEF.^18,19 The severity of HFpEF in terms of cardiometabolic pathology has been shown to be similar between male and female ZSF1-obese rats,²⁰ which is not the case for the popular “2-hit” mouse model of HFpEF, in which female mice are protected against the development of HFpEF.²¹ In summary, the ZSF1-obese rat represents a clinically relevant model for the elucidation of novel mechanisms responsible for the development and progression of HFpEF.

To uncover potentially novel and critical mechanisms in HFpEF, we provide an in-depth characterization of the ZSF1 obese rat model of HFpEF using several physiological, biochemical, molecular, and omics approaches. We evaluated male ZSF1-obese, ZSF1-lean hypertensive control, and Wistar Kyoto (WKY) (wild-type) lean normotensive control rats at an early stage in the development of HFpEF (14 weeks of age), performing extensive physiologic phenotyping in conjunction with unbiased metabolomics and transcriptomics. Our results reveal that the addition of obesity/metabolic syndrome in addition to hypertension and vascular dysfunction is a primary contributor to cardiac transcriptional and metabolic remodeling, driving the development of HFpEF. Most notably, mitochondrial energy metabolism pathways were highly disrupted, resulting in an energetic deficit that correlated with maladaptive mitochondrial ultrastructural remodeling and functional impairment. These findings support an integrated framework to identify metabolic and transcriptional pathways that are disrupted in, and contribute to, HFpEF progression that will optimally yield new therapeutic targets.

Methods

Experimental animals

WKY, ZSF1-lean, and ZSF1-obese male rats were purchased from Charles River Laboratories and used in all experiments in this study (n = 5-7 per group). Animals were purchased and held at the Lewis Katz School of Medicine at Temple University (TU) or Louisiana State University Health Sciences Center (LSUHSC) in a temperature-controlled and 12-hour light/dark cycle environment for the entirety of the study. All parts of the study were approved by TU and LSUHSC Institutional Animal Care and Use Committees and received animal care at TU and LSUHSC according to the Association for Assessment and Accreditation of Laboratory Animal International guidelines.

Study design

ZSF1-lean, ZSF1-obese, and WKY control rats were investigated at 14 weeks of age. Physiologic parameters of body weight, transthoracic echocardiography, and exercise capacity testing were as described below. Further investigation into pathophysiology of these separate animal models was performed using LV and systemic invasive hemodynamic measurements along with ex vivo assessments of mitochondrial ultrastructure and function. Isolated cardiac LV tissue samples were also submitted for RNA sequencing and unbiased metabolomics.

Echocardiography

Transthoracic echocardiography of all groups was performed with a Vevo 2100 echocardiography system (Fujifilm VisualSonics). LV diastolic measurements were performed with the use of an apical 4-chamber view of the heart. LV systolic measurements were performed with the use of a long-axis view. Animals were anesthetized with inhaled isoflurane at an induction dose of 3% and a maintenance dose of 1% for the longevity of the experiment. Heart rate was maintained at approximately 250-300 beats/min for the data collection period as previously described.²²

Exercise capacity testing

ZSF1-lean, ZSF1-obese, and WKY control rats were assessed for exercise intolerance with the use of a IITC Life Science 800 Series treadmill. Animals were first acclimated to the treadmill for a period of 5 minutes with no movement; they were then brought through a warm-up phase consisting of initially 6 m/min which was thereby increased to 12 m/min for a 4-minute ramp-up time, for a total warm-up phase of 5 minutes. For data collection as presented, the animals were run at a rate of 12 m/min with 0° incline until exhaustion, which was defined as animal placement on the shock pads for more than 3 seconds. Exercise capacity was then determined by the total distance run.

Terminal invasive hemodynamics and sacrifice

At 14 weeks of age, the animals were anesthetized with inhaled isoflurane at concentrations of 3% for induction and 1% for maintenance during the following procedure. The rodent neck and associated structures were dissected for exposure of the common carotid, which was cannulated with a 1.2-F high-fidelity pressure catheter, measuring the systemic pressures at systole and diastole accordingly for multiple cardiac cycles. The pressure catheter was then carefully introduced into the LV of the animal. LV and ventricular relaxation time constant were measured after multiple cardiac cycles to obtain an average measurement. The catheter was then removed, and the rat subsequently exsanguinated and killed, with tissues and plasma harvested for additional measurements as previously described.²³

Mitochondrial function

Heart mitochondria were isolated and subjected to respiratory function assays using the Seahorse XF96, similarly as described previously.^7,24 Briefly, ∼100 mg LV heart pieces were washed 5 times with cold buffer A (220 mmol/L mannitol, 70 mmol/L sucrose, 5 mmol/L MOPS, 1 mmol/L EDTA; pH 7.2 with KOH) followed by homogenization using a glass-col homogenizer in 2 mL buffer A containing 0.2% fatty acid–free bovine serum albumin (BSA). Homogenate was then subjected to centrifugation at 800g for 10 minutes followed by supernate collection and centrifugation at 10,000g for 10 minutes. The pellet containing mitochondria was then resuspended in 1 mL fresh buffer A (without BSA) and centrifuged at 10,000g, with this step repeated once. The washed mitochondrial pellet was then resuspended in 150 μL respiration buffer (120 mmol/L KCl, 25 mmol/L Sucrose, 10 mmol/L HEPES, 1 mmol/L MgCl₂, 5 mmol/L KH₂PO₄; pH 7.2 with KOH) and kept on ice.

To determine mitochondrial function, samples were diluted to a concentration of 2.5 μg (protein) in 50 μL respiration buffer per well and centrifuged onto XF96 microplates at 500g for 3 minutes at 4 °C. State 3 respiration in response to substrates were measured after injection of pyruvate + malate (final concentrations 5.0 mmol/L and 2.5 mmol/L, respectively) or succinate and rotenone (final concentrations 10 mmol/L and 1 μmol/L, respectively) to assess complex I and II rates, respectively. Fatty acid oxidation (FAO) was assessed in response to palmitoyl-l-carnitine and malate (final concentrations 50 μmol/L and 2.5 mmol/L, respectively). The oxygen consumption rates recorded after injection of oligomycin (1 μg/mL), an inhibitor of ATP synthase, served as a measure of state 4 respiration. After state 4 respiratory measurements, injection of FCCP, a mitochondrial uncoupler, provided electron transport chain complex maximal respiratory capacity. Respiratory control ratios, state 3/state 4, were calculated as a measure of the coupling of oxygen consumption to ATP production.

Mitochondrial calcium uptake assay

Isolated mitochondria were diluted in Isolated Mitochondria Assay Buffer (IMAB; 125 mmol/L KCl, 10 mmol/L NaCl, 20 mmol/L HEPES, 2 mmol/L MgCl₂, 2 mmol/L KH₂PO₄; pH 7.2 with KOH). Mitochondria were loaded into 96-well plates (final concentration 1 μg/μL), supplemented with 10 mmol/L succinate (Sigma-Aldrich; S3674), 10 mmol/L malate (Sigma-Aldrich; 240176), and 10 mmol/L pyruvate (Sigma-Aldrich; P5280), and 1 μmol/L calcium green-5N hexapotassium salt (Invitrogen; C-3737). Final volume at the start of the assay was 50 μL. Fluorescence was measured every 200 ms at 506 nm_ex/532 nm_em with the use of a TECAN Infinite M1000 Pro plate reader set at 37 °C. After 120 seconds of baseline measurements, successive injections of 2.5 μmol/L CaCl₂ (5 μL 25 μmol/L CaCl₂ stock prepared in IMAB) were administered every 120 seconds. To generate a standard curve of extramitochondrial Ca²⁺ (bath concentration), the same experimental setup was used without addition of mitochondria to the well. The standard curve was used to calculate the extramitochondrial calcium remaining post mitochondrial uptake (average of last 100 seconds per injection cycle) and to determine the percentage of mitochondrial calcium uptake after successive injections. All methods were as described previously.25, 26, 27

Mitochondrial swelling assays

Isolated mitochondria were diluted in IMAB (125 mmol/L KCl, 10 mmol/L NaCl, 20 mmol/L HEPES, 2 mmol/L MgCl₂, 2 mmol/L KH₂PO₄; pH 7.2 with KOH). Mitochondria were loaded into 96-well plates (final concentration 1 μg/μL), supplemented with 10 mmol/L succinate (Sigma-Aldrich; S3674), 10 mmol/L malate (Sigma-Aldrich; 240176) and 10 mmol/L pyruvate (Sigma-Aldrich; P5280). Final volume was 150 μL per well. Absorbance was measured every 5 seconds at 540 nm with the use of a TECAN Infinite M1000 Pro plate reader set at 37 °C with plate shaking between measurements. After 2 minutes of baseline measurements, a single Ca²⁺ bolus of 500 μmol/L CaCl₂ (7.5 μL 10 mmol/L CaCl₂ stock prepared in IMAB) was administered, with measurements recorded every 5 seconds for 10 minutes. All methods were as described previously.25, 26, 27

Transmission electron microscopy

Left ventricle tissues cut to ∼3 mm³ were fixed in 2% PFA + 2% glutaraldehyde in 0.1 mol/L sodium cacodylate buffer, pH 7.4, and stored at 4 °C for 48 hours. Tissues were washed 3 times for 15 minutes each time in 0.1 mol/L sodium cacodylate buffer, pH 7.4, and then post-fixed in freshly prepared 1.5% potassium ferrocyanide and 1% osmium tetroxide in 0.1 mol/L sodium cacodylate buffer, pH 7.4, for 2 hours. The samples were washed with water 4 times for 15 minutes each time followed by en-bloc staining overnight with 1% uranyl acetate. After washing 3 times with H₂O for 15 minutes each time, tissues were dehydrated in an ascending acetone series (25% acetone, 50% acetone, 75% acetone, 95% acetone, 100% anhydrous acetone, 100% anhydrous acetone), 15 minutes each step. Samples were infiltrated with Spurr’s resin (25% resin in acetone, 50% resin in acetone, 75% resin in acetone, 100% resin, 100% resin), 1 hours each step, followed by overnight incubation in 100% Spurr’s resin. The next day, one last exchange in 100% Spurr’s resin was performed before samples were placed in aluminum weigh dishes with fresh resin and polymerized at 60 °C overnight. After polymerization, tissues in proper orientation were excised from the resin with a jeweler’s saw and glued onto supports. Muscle tissues were sectioned with the use of a Leica UC7 ultramicrotome, and 60-nm-thick sections were collected onto 200-mesh copper grids with a formvar-carbon support film. Grids were post-stained with 2% uranyl acetate in 50% methanol and Reynolds lead citrate. Grids were examined and imaged in a FEI Tecnai 12 120 keV digital transmission electron microscopy (TEM), with images acquired at various magnifications (eg, ×1,100 to ×21,000).

Morphometric analysis of TEM images

Analysis of mitochondria, lipid droplets (LDs), LD-mitochondrion associations, and sarcomere lengths were performed with the use of ImageJ/FIJI (National Institutes of Health). After calibration for distance, shape descriptors and size measurements were obtained by manually tracing only discernable mitochondria or lipid droplets. Circularity was computed as [4π × (area/perimeter²)] and roundness as [4π × (surface area)/(π × major axis²); values of 1 indicate perfect spheres. Feret diameter represents the longest distance between any 2 points within a given mitochondrion.²⁸ A custom Phyton plugin (MitoCareTools) was adopted for quantification of lipid-mitochondria associations.^29,30 Areas where the LD was <100 nm from the outer mitochondrial membrane (OMM) were determined as an LD-mitochondrion interface. To obtain mean gap distance, the LD membrane was first traced, followed by tracing the mitochondrion OMM, with values obtained from the plugin.

Protein Immunoblotting

Remaining isolated mitochondria from our calcium and respiratory assays were pelleted and lysed in RIPA buffer supplemented with phosphatase inhibitors (Roche; 4906837001) and protease inhibitors (Sigma; S8830). Samples were kept on ice for 30 minutes with agitation via vortex every 10 minutes. Samples were then centrifuged at 13,000g for 20 minutes at 4 °C. The supernate was collected, and protein concentration was quantified with the use of the Pierce 660 nm Protein Assay Reagent (Thermo Fisher Scientific). Equal amounts of protein (5 μg) were run by gel electrophoresis on polyacrylamide Tris-glycine sodium dodecyl sulfate gels. Gels were transferred to polyvinylidene difluoride (EMD Milipore; IPFL00010), and membranes were blocked for 1 hour in Blocking Buffer (Rockland; MB-070), followed by incubation with primary antibody overnight at 4 °C on a rocker. Membranes were then washed in TBS-T 3 times for 5 minutes each time and incubated in a fluorescent secondary antibody for 1 hour at room temperature. Membranes were then washed in TBS-T 3 times for 5 minutes each time and imaged on a Licor Odyssey system. Antibodies in the study were used at a concentration of 1:1,000 and included VDAC1/3 (Abcam; ab14734), MCU (Cell Signal; 14997), MICU1 (Novus Bio; BP1-86663), MICU2 (Novus Bio; BP2-92063), MCUB (Sigma Aldrich; HPA024771), and Total OxPHOS Cocktail (Abcam; ab110413).

RNA sequencing

LV heart pieces were flash frozen in liquid N₂ immediately after excision and subjected to RNA sequencing (RNAseq) analysis. Total RNA was isolated with the use of a fibrous tissue RNA isolation kit (Qiagen). The TrueSeq stranded mRNA library prep kit was used to enrich polyA mRNAs via poly-T–based RNA purification beads, which were then amplified with the use of the HiSeq rapid SR cluster kit and multiplexed and run with the use of the HiSeq rapid SBS kit. Reading depth was ∼30 million reads per sample, and single-end 75 bp fragments were generated for bioinformatic analysis. All kits for sequencing were obtained from Illumina, and all sequencing was performed on the Illumina HiSeq2500 sequencer. RNA transcripts were aligned to the Rnor_6.0 assembly with the use of HISAT2 v2.1.0 and quantified with the use of HTSeq v0.11.2. Differential expression analysis was performed between groups with the use of DESeq2 v1.22.2. Genes were considered to be differentially expressed when they met a fold change (FC) of ≥2.0 with false discovery rate (FDR) of ≤0.05 (Benjamin-Hochberg FDR method). Gene Ontology (GO) analysis was accomplished using David GO analysis tools. Venn diagram for up-regulated, contra-regulated, and down-regulated genes was generated using VennPlex software (see Supplemental Table 4 for VennPlex comparison gene list).³¹ All RNAseq data has been submitted to the GEO (Gene Expression Omnibus) repository with the accession #GSE301671.

Metabolomic analysis

LV heart pieces were flash frozen in liquid N₂ immediately after excision to most accurately capture the in vivo cardiac metabolome. Samples were prepared by Metabolon using their automated MicroLab STAR system (Hamilton Co). First, tissue homogenates were made in water at a ratio of 5 μL per mg of tissue. For quality control, several recovery standards were added before the first step in the extraction process. To remove protein, dissociate small molecules bound to protein or trapped in the precipitated protein matrix, and recover chemically diverse metabolites, proteins were then precipitated with methanol (final concentration 80% v/v) under vigorous shaking for 2 minutes (Glen Mills GenoGrinder 2000) followed by centrifugation. For quality assurance and control, a pooled matrix sample was generated by taking a small volume of each experimental sample to serve as a technical replicate throughout the data set. Extracted water samples served as process blanks. A cocktail of standards known not to interfere with the measurement of endogenous compounds was spiked into every analyzed sample, allowing instrument performance monitoring and aiding chromatographic alignment.

The extract was divided into fractions for analysis by reverse-phase (RP)/ultraperformance liquid chromatography (UPLC)–tandem mass spectrometry (MS/MS) with positive ion mode electrospray ionization (ESI), by RP/UPLC-MS/MS with negative ion mode ESI, and by hydrophilic interaction liquid chromatography (HILIC)/UPLC-MS/MS with negative ion mode ESI. Samples were placed briefly on a TurboVap (Zymark) to remove the organic solvent. All methods used a Waters Acquity UPLC and a Thermo Scientific Q-Exactive high-resolution/accurate mass spectrometer interfaced with a heated ESI source and Orbitrap mass analyzer operated at 35,000 mass resolution. The sample extract was reconstituted in solvents compatible with each MS/MS method. Each reconstitution solvent contained a series of standards at fixed concentrations to ensure injection and chromatographic consistency. One aliquot was analyzed via acidic positive ion conditions, chromatographically optimized for hydrophilic compounds. In this method, the extract was gradient eluted from a C18 column (Waters UPLC BEH C18-2.1 × 100 mmol/L, 1.7 μm) with water and methanol containing 0.05% perfluoropentanoic acid (PFPA) and 0.1% formic acid (FA). For more hydrophobic compounds, the extract was gradient eluted from the C18 column with methanol, acetonitrile, water, 0.05% PFPA, and 0.01% FA. Aliquots analyzed via basic negative ion optimized conditions were gradient eluted from a separate column with methanol and water containing 6.5 mmol/L ammonium bicarbonate (pH 8). The last aliquot was analyzed via negative ionization after elution from a HILIC column (Waters UPLC BEH Amide 2.1 × 150 mmol/L, 1.7 μm) with a gradient consisting of water and acetonitrile with 10 mmol/L ammonium formate (pH 10.8). The MS analysis alternated between MS and data-dependent MSⁿ scans with the use of dynamic exclusion. The scan range covered 70-1,000 m/z.

Raw data were extracted, peak-identified, and processed with the use of Metabolon’s proprietary hardware and software. Compounds were identified by comparison to library entries of purified authenticated standards or recurrent unknown entities, with known retention times/indices (RIs), mass to charge ratios (m/z), and chromatographic signatures (including MS/MS spectral data). Biochemical identifications were based on 3 criteria: RI within a narrow window of the proposed identification, accurate mass match to the library ±10 ppm, and the MS/MS forward and reverse scores between experimental data and authentic standards. Proprietary visualization and interpretation software (Metabolon) was used to confirm the consistency of peak identification among the various samples. Library matches for each compound were checked for each sample and corrected, if necessary. Area under the receiver operating characteristic curve was used for peak quantification.

Original scale data (raw area counts) were analyzed with the use of Metaboanalyst 5.0 software (www.metaboanalyst.ca). Metabolites with more than 50% of the values missing were omitted from the analysis, and missing values were imputed by introducing values with one-fifth of the minimum positive value of each variable. An interquartile range filter was used to identify and remove variables unlikely to be of use when modeling the data. The data were log-transformed and autoscaled (mean-centered and divided by the standard deviation of each variable). Univariate (eg, volcano plots) and multivariate (eg, principal component analysis) analyses were then performed. For multiple comparison testing, q values based on FDR were calculated in R using a method embedded within the Metaboanalyst software (https://github.com/xia-lab/MetaboAnalystR),³² controlling for the FDR. Metabolites were considered to be significantly different when they met an FC of ≥1.25 and FDR of ≤0.05. Refer to Supplemental Tables 1-3 for the list of significantly different metabolites across our three comparisons.

Integrated pathway network analyses

Integrated network analyses using both the transcriptomic and metabolomic data sets were performed with the use of Metaboanalyst 5.0 software. Integrated pathway maps were generated with the use of BioRender. Refer to Supplemental Table 5 for the list of imputed metabolites and genes used for the integrated network analysis.

Statistical analysis

Statistical analyses was performed with the use of GraphPad Prism 10 (www.graphpad.com), Metaboanalyst (metaboanalyst.ca/home/xhtml), and the Metaboanalyst R program.³² Statistical parameters including the value of n (number of rats), the definition of center, dispersion and precision measures (mean ± SEM or SD), and statistical significance are reported in the figures and figure legends. A P value of ≤0.05 was considered to be statistically significant. For the metabolomics and transcriptomics data sets, an FDR value of ≤0.05 was considered to be statistically significant. For direct comparisons, statistical significance was calculated by means of unpaired (between 2 groups) or paired (within group) Student’s t-test. Comparisons among >2 groups utilized analysis of variance (ANOVA) with Sidak's post-hoc test for multiple pairwise comparisons. Shapiro-wilk normality test was employed to determine that the data were approximately normally distributed such that parametric methods could be used to present and compare data. Details on the statistical methods used for the metabolomics and RNAseq data sets are as described above in their respective methods sections.

Results

The clinical features of HFpEF are recapitulated in the ZSF1-obese rat

We investigated whether the ZSF1-obese rat, which is both hypertensive and obese, phenocopies the clinical characteristics of HFpEF, to identify potential molecular and metabolic mechanisms contributing to HFpEF (Figure 1A); ZSF1-lean (hypertensive lacking obesity/metabolic syndrome) and WKY rats were included as control groups. ZSF1-obese rats demonstrated a 60% and 40% increase in body weight compared with WKY and ZSF1-lean control rats, respectively (Figure 1B). Both lean and obese rats were hypertensive, with elevated systolic (∼155 mmol/L Hg) and diastolic (∼110 mmol/L Hg) blood pressures (Figure 1C). Distance run on a treadmill was 83% less in ZSF1-obese rats, indicating severe exercise intolerance, which was also observed in lean rats (Figure 1D). Echocardiography revealed a significant elevation in the E/e′ in ZSF1-obese rats with preserved ejection fraction (Figures 1E and 1F). Invasive hemodynamics (PV Loop) indicated a 6-fold increase in LV end-diastolic pressure) (Figure 1E), a hallmark feature distinguishing HFpEF from HFrEF.³³ LV, atrial, liver, and kidney weights normalized to tibia length were greatest in ZSF1-obese rats vs control rats (Figure 1G, Supplemental Figure 1), indicating tissue hypertrophy and/or edema. Collectively, the ZSF1-obese rat displays numerous features of clinical HFpEF, including obesity, hypertension, exercise intolerance, diastolic dysfunction with preserved ejection fraction, and cardiac hypertrophy.

Figure 1.

Clinical Manifestations of Heart Failure With Preserved Ejection Fraction (HFpEF) Are Observed in the ZSF1-Obese Rat

Physiologic characterization of HFpEF. (A) Schematic of study design. (B) Body weights of WKY (control), ZSF1-lean (hypertensive [HTN]), and ZSF1-obese (metabolic syndrome [MS] + HTN; HFpEF) rats. (C) Assessment of systolic (SBP) and diastolic (DBP) blood pressure obtained during invasive hemodynamics. (D) Distance run during a treadmill exercise capacity test. (E) Indices of cardiac left ventricular diastolic function assessed by means of echocardiography for the E/e′ ratio and by means of invasive hemodynamics for the left ventricular end-diastolic pressure (LVEDP). (F) Determination of cardiac systolic function assessed by echocardiography for the left ventricular ejection fraction (LVEF%). (G) Gravimetric assessment of left ventricle and left atria normalized to tibia length (TL). n = 4-7 male rats per group; mean ± SEM; 1-way analysis of variance with Holm-Sidak post hoc test: ∗P ≤ 0.05; ∗∗P ≤ 0.01; ∗∗∗P ≤ 0.001.

Lean hypertensive rats demonstrate significant metabolic remodeling with few transcriptional changes

To identify potential molecular and metabolic pathways contributing to disease development, RNAseq and quantification of the steady-state abundance of metabolites were performed in hearts from all 3 genotypes, initially assessing those changes mediated by hypertension alone by comparing ZSF1-lean and WKY control rats. We observed differential expression of 233 genes in ZSF1-lean hearts, with 149 increased and 84 decreased in expression (FC ≥2.0 and FDR ≤0.05) (Figure 2A). GO analysis of the differentially expressed transcripts surprisingly revealed no significant enrichment of biological or Kyoto Encyclopedia of Genes and Genomes pathways (Figures 2B to 2E), suggesting diffuse and nonspecific transcriptional remodeling. Metabolomics analysis identified 120 metabolites increased in abundance and 85 decreased in abundance (FC ≥1.25 and FDR ≤0.05) (Figure 2A). Unlike our transcriptomics data set, pathway enrichment analysis of the cardiac metabolome revealed significant changes (P < 0.05) in nucleotide metabolism, amino acid metabolism, and pathways critical for energy metabolism (eg, glycolysis, pyruvate, Krebs cycle) (Figure 2F and Supplemental Table 1). Collectively, these results suggest that chronic hypertension alone is sufficient to robustly remodel cardiac metabolism while minimally affecting the transcriptome.

Figure 2.

Hypertension Significantly Affects the Cardiac Metabolome With Minimal Effect on the Transcriptome

RNA sequencing (RNAseq) and metabolomic comparisons of hearts from ZSF1-lean vs WKY control rats. (A) Summary of the up-regulated and down-regulated transcriptional and metabolic changes. Gene Ontology analysis, revealing the top up-regulated (B) biological processes and (C) KEGG pathways of those genes found to be differentially expressed. Gene Ontology analysis, revealing the top down-regulated (D) biological processes and (E) KEGG pathways of those genes found to be differentially expressed. (F) Pathway enrichment analysis of the cardiac metabolome indicated those pathways found to be most significantly affected in ZSF1 rats due to hypertension. Fold change (FC) cutoffs of ≥2.0 (RNAseq) and ≥1.25 (metabolomics) were used with a false discovery rate (FDR) of ≤0.05. n = 6 male rats per group for RNAseq and n = 7 male rats per group for metabolomics. KEGG = Kyoto Encyclopedia of Genes and Genomes.

ZSF1-obese HFpEF hearts displays signatures of inflammation, mitochondrial dysfunction, and down-regulation of energy metabolism

Based on our physiologic phenotyping results, the 2 hits of obesity (i.e., metabolic syndrome) and hypertension are required for the robust development of HFpEF. Therefore, although we did examine the transcriptomic and metabolomic differences between ZSF1-obese and WKY rats (Supplemental Figure 2 and Supplemental Table 3), we were most interested in identifying potential transcriptional and metabolic alterations revealed with the addition of obesity/metabolic syndrome. A total of 5,691 genes were differentially expressed (3,123 up-regulated and 2,568 down-regulated; FC ≥2.0 and FDR ≤0.05) (Figure 3A). Interestingly, fibrosis and inflammation were the dominant signatures based on GO enrichment analyses, including pathways related to extracellular matrix assembly, immune cell activation, phagocytosis, B-cell activation and signaling, immune response, and nuclear factor κB signaling (Figures 3B and 3C).

Figure 3.

Transcriptional Cardiac Remodeling Is Dependent on the 2 Hits of Obesity and Hypertension in ZSF1-Obese HFpEF Rats

RNAseq and metabolomic comparisons of hearts from ZSF1-obese vs lean rats. (A) Summary of the upregulated and downregulated transcriptional and metabolic changes. Gene Ontology analysis revealing the top upregulated (B) biological processes and (C) KEGG pathways of those genes found to be differentially expressed. Gene Ontology analysis revealing the top downregulated (D) biological processes and (E) KEGG pathways of those genes found to be differentially expressed. (F) Pathway enrichment analysis of the cardiac metabolome indicated those pathways found to be most significantly affected in ZSF1 rats due to hypertension. Fold change (FC) cutoffs of ≥2.0 (RNAseq) and ≥1.25 (metabolomics) were used with an FDR of ≤0.05. n = 6 male rats per group for RNAseq and n = 7 male lean and 8 male ZSF1-obese rats for metabolomics. Abbreviations as in Figures 1 and 2.

GO enrichment analysis of significantly down-regulated transcripts revealed suppression of key metabolic and mitochondrial biological processes (Figure 3D). This included the down-regulation of ubiquinone biosynthesis, cristae formation, fusion, and protein import into the matrix (Figure 3D). Metabolic pathways that were down-regulated in ZSF1-obese hearts included the Krebs cycle, fatty acid metabolism, and pyruvate metabolism (Figure 3E). In agreement with the transcriptomic analyses, the metabolomic signature was affected to a greater degree than that observed with hypertension alone (ie, ZSF1-lean vs WKY [Figure 2]), with 148 metabolites that were increased and 130 metabolites that decreased in ZSF1-obese rats compared with lean control rats (FC ≥1.25 and FDR ≤0.05) (Figure 3A). Pathway enrichment analysis revealed nucleotide and amino acid metabolism as the most affected metabolic processes in HFpEF hearts (Figure 3F and Supplemental Table 2); although fewer total pathways were significantly affected, this was owing to the underlying metabolic remodeling invoked by hypertension alone. In fact, several metabolites associated with pathways significantly enriched by hypertension alone (eg, Krebs cycle) were further disrupted in the ZSF1-obese heart in the comparison of ZSF1-obese vs WKY (Supplemental Figure 2F), which shows enriched pathways similar to those in ZSF1-lean vs WKY.

Omics integration reveals transcriptional and metabolic coordination of the cardiac energetic deficit in HFpEF

To gain further mechanistic insight into HFpEF development, we next examined transcriptional changes dependent on the 2 hits of metabolic syndrome and hypertension vs those independent from hypertension. Changes independent from hypertension included 795 differentially expressed genes (Supplemental Figure 3A and Supplemental Table 4), with an enrichment in processes related to the cell cycle and proliferation. This transcriptional enrichment could be associated with the meta-inflammation known to occur in HFpEF and which appears evident in the ZSF1-obese hearts (Figures 3B, 3C and Supplemental Figure 3B). Transcriptional changes dependent on both hypertension and metabolic syndrome revealed 5,544 differentially expressed genes with a significant enrichment in energy metabolism pathways and additional signatures of inflammation (Supplemental Figure 3C).

Merger of our omics data sets provides a more comprehensive and integrated interpretation of the remodeling occurring in HFpEF. Using a multi-omics assimilation approach, we integrated the differentially expressed transcripts and metabolites that were significantly altered in abundance to reveal the most affected pathways that likely contribute to disease progression (Supplemental Table 5). The effects of hypertension alone (ZSF1-lean vs WKY) indicated glycolysis, purine and pyrimidine metabolism, and nicotinate and nicotinamide metabolism as pathways most affected (Figure 4A). HFpEF hearts (ZSF1-obese) had a greater impact on metabolic pathways related to ketone bodies, lipid metabolism, pyruvate metabolism, and the Krebs cycle, which was the most affected pathway (Figure 4B).

Figure 4.

Integrated Network Analysis of RNAseq and Metabolomic Data Set Reveals Unique Metabolic Pathways Affected in HFpEF

Integrated analysis of cardiac omics data sets to identify those pathways most affected by transcriptional and metabolic remodeling due to (A) hypertensive phenotype (lean vs WKY) or (B) the observed HFpEF phenotype (ZSF1-obese vs lean). FC cutoffs of ≥2.0 (RNAseq) and ≥1.25 (metabolomics) were used with an FDR ≤0.05. A greater pathway impact indicates a greater influence at the transcriptional and metabolic level to a given pathway. Labeled pathways had P values of ≤0.05. n = 6 male rats per group for RNAseq and n = 7 male lean and 8 male ZSF1-obese rats for metabolomics. Abbreviations as in Figures 1 and 2.

Because many of the identified pathways are central to cardiac energy metabolism (ie, glycolysis, pyruvate metabolism, Krebs cycle), we generated integrated metabolic pathway maps to better illustrate the transcriptional and metabolic changes in these pathways. The hypertensive effects (ie, ZSF1-lean vs WKY) on glycolysis revealed increased expression of Pfk (phosphofructokinase) and Pfkfb1 (6-phosphofructo-2-kinase:fructose-2,6-bisphosphatase), the latter which generates fructose-2,6-bisphosphate, a potent allosteric activator of PFK.²⁴ The downstream glycolytic intermediates 3-phosphoglycerate, 2-phosphoglycerate, phosphoenolpyruvate, and pyruvate were all increased in abundance, potentially suggesting increased glycolytic activity in ZSF1-lean hearts compared to WKY control hearts (Figure 5A). This is in stark contrast to the ZSF1-obese HFpEF heart, which showed an overall down-regulation of glycolytic enzymes. Obese hearts compared with lean had a higher phosphocreatine:ATP ratio and lower ATP:ADP ratio than lean vs WKY, indicating a lower cardiac energy state in HFpEF. These differences were largely driven by a reduction in ATP abundance in the HFpEF heart (Figure 5A). Interestingly, phosphocreatine levels were highest in the HFpEF heart, likely in part due to transcriptional down-regulation of creatine kinase isoforms (ie, Ckm, Ckmt2). Hypertensive and HFpEF hearts demonstrated increased abundance of acyl-carnitines, with greater increases in the 2-hit hearts (ZSF1-obese), suggestive of decreased utilization or increased synthesis (Figure 5B, Supplemental Figure 4). ZSF1-obese hearts also showed down-regulation of key β-oxidation enzymes and transporters (eg, Cact, Cpt1, Cpt2, Acat1) (Figure 5B). Transcriptional repression of all Krebs cycle enzymes, accompanied by increased abundance of the upstream metabolites citrate and isocitrate, suggests an overall decrease in Krebs cycle activity in obese hearts (Figure 5C). Because glycolysis and β-oxidation are central to cardiac oxidative metabolism, down-regulation of their enzymes along with additional pathways capable of input to the Krebs cycle (ie, branched-chain amino acids (BCAAs), ketones, amino acids) also likely contributes to the apparent overall decrease in Krebs cycle activity and the energetic deficit of the HFpEF heart (Figure 5, Supplemental Figure 5). These integrated analyses reveal a transcriptional and metabolic signature brought on by obesity in HFpEF, highlighting mitochondrial energy metabolism as a potential distinguishing and important feature.

Figure 5.

HFpEF Results in the Transcriptional and Metabolic Down-Regulation of Pathways Central to Energy Metabolism

Pathway maps of the transcriptional and metabolic alterations in ZSF1-lean vs WKY and ZSF1-obese vs ZSF1-lean rats in energy generating pathways, specifically (A) aerobic glycolysis, (B) fatty acid oxidation, (C) and the Krebs cycle. Additional pathways indicated in boxes provide an objective summary of the transcriptional and metabolic increase or decrease observed. Genes and metabolites significantly increased or decreased in expression or abundance (FCs ≥2.0 (RNAseq) and ≥1.25 (metabolomics); FDR ≤0.05) are as indicated. Abbreviations as in Figures 1 and 2.

Disrupted mitochondrial ultrastructure and impaired function are evident early in HFpEF development

Because of the strong mitochondrial signature unique to HFpEF, we looked deeper and examined mitochondrial ultrastructure by means of TEM. Gross qualitative assessment of electron micrographs revealed cardiomyocyte mitochondrial cristae disorganization, with decreased cristae density observed in the lean hearts and in the ZSF1-obese HFpEF hearts (Figure 6A). Quantitative mitochondrial morphologic analyses indicated no difference in mitochondrial number per cardiomyocyte area, but a decrease in the total mitochondrial area, indicating smaller mitochondria in ZSF1-obese hearts (Figure 6B). Damaged or fragmented mitochondria typically assume a smaller and more rounded morphology,^34,35 which was evident by a reduction in Feret diameter and an increase in the circularity index in ZSF1-obese HFpEF cardiomyocyte mitochondria (Figure 6C). Strikingly, obese hearts displayed a significant increase in LDs, which localized adjacent to interfibrillar mitochondria (Figure 6A). Quantification of LDs revealed a significant increase in number and size exclusively in obese hearts (Figures 6D and 6E). Because LDs strongly associated with interfibrillar mitochondria, we quantified mitochondria-LD interactions, which indicated an increase in the total number of mitochondria-LD contacts as well as the length of mitochondrial and LD membranes in close association with one another (Figure 6F). Finally, sarcomeric length was increased in ZSF1-obese cardiomyocytes, likely as a consequence of increased preload (ie, diastolic dysfunction) and LV dilation observed in the HFpEF heart (Figure 6G).

Figure 6.

Mitochondrial Ultrastructural Remodeling in the HFpEF Heart Is Characterized by a Decrease in Mitochondrial Content, Cristae Disorganization, and Lipid Droplet (LD) Association

Transmission electron micrographs of cardiomyocyte mitochondrial and LD ultrastructure. (A) Representative images of WKY, ZSF1-lean, and ZSF1-obese cardiomyocyte ultrastructure indicating cristae disorganization (yellow arrows) and lipid droplet accumulation and interaction with mitochondria (red arrows). (B) mitochondrial area (ie, content) quantified by the number of mitochondria per image area and the percentage area of mitochondria to total area. (C) Mitochondrial shape quantified by Feret diameter and circularity index. (D) Quantification of LD area (ie, content) quantified by the number of LDs per image area and the percent area of LDs to total area. (E) LD shape quantified by Feret diameter and circularity index. (F) Analysis of mitochondrial (mito)–LD interaction quantified by the number of mito-LD contacts per image area, the outer mitochondrial membrane (OMM) perimeter in contact with an LD, and the LD perimeter in association with the OMM. (G) Determination of sarcomeric length measured from z-line to z-line. n = 4 male rats per group; mean ± SEM; 1-way analysis of variance with Holm-Sidak post hoc test: ∗P ≤ 0.05; ∗∗P ≤ 0.01; ∗∗∗P ≤ 0.001. Abbreviations as in Figures 1 and 2.

After noting these mitochondrial ultrastructural changes, we examined mitochondrial function via respiratory activity and calcium handling assays. Determination of citrate synthase activity, a criterion standard for assessing mitochondrial abundance, was decreased in both lean and obese rats (Figure 7A). Both lean and obese cardiac mitochondria displayed lower overall respiratory rates compared with WKY (Figures 7B, 7D, and 7F). In the presence of pyruvate and malate (complex I) or succinate and rotenone (complex II), mitochondria from ZSF1-lean and -obese hearts showed a significant reduction in state 3 respiration (Figures 7C and 7E); fatty acid supported state 3 respiration (palmitoyl-l-carnitine) also trended lower but did not reach statistical significance (Figure 7G). Complex I respiratory control ratio (RCR) was reduced in the lean hearts, indicating reduced coupling of oxygen consumption to ATP production, and, surprisingly, this was improved in the obese hearts compared with the reduction in lean (Figure 7C). No differences were observed for complex II RCR, FAO RCR, or state 4 rates (Figures 7E and 7G, Supplemental Figure 6A).

Figure 7.

Mitochondrial Dysfunction Characterized by Impaired Respiratory Activity and Disrupted Calcium Handling Is a Key Feature of the HFpEF Heart

Functional assessment of the mitochondrial function in the HFpEF heart. (A) Determination of mitochondrial content assessed by citrate synthase (CS) activity. Interrogation of mitochondrial respiratory function by assessing oxygen consumption rates (OCR) of (B, C) complex I (pyruvate + malate)–specific substrates, (D, E) complex II (succinate)–specific substrate + rotenone (complex I inhibitor), and (F, G) fatty acid oxidation (FAO) capacity (palmitoyl-l-carnitine): state 3 (substrate-mediated) oxygen consumption and the respiratory control ratio (RCR) providing an index of oxygen consumption to ATP-production coupling. (H, I) Mitochondrial calcium uptake in response to repeated 2.5 μmol/L boluses. Mitochondrial (Mito) swelling in response to a 500 μmol/L bolus, displayed as both (J) uncorrected and (K) normalized before calcium addition. Quantification of mitochondrial swelling indicated by (L) percentage change from WKY baseline and (M) area above the receiver operating characteristic curve. Immunoblotting of (N) VDAC1/3 and components of the mitochondrial calcium uniporter (MCU and MCUIs) and (O) subunits of electron transport chain complexes. Proteins differentially expressed in protein abundance are indicated in red. n = 5 male rats per group (A to G), n = 4 male rats per group (I-O); mean ± SEM: 1-way analysis of variance with Holm-Sidak post hoc test: ∗P ≤ 0.05; ∗∗P ≤ 0.01; ∗∗∗P ≤ 0.001. Abbreviations as in Figures 1 and 2.

Mitochondrial calcium uptake is intricately linked to bioenergetics³⁶ and at high levels induces mitochondrial dysfunction. Indeed, HF is associated with mitochondrial calcium overload.^26,27,36,37 We isolated mitochondria from ZSF1-obese hearts and subjected them to repeated 2.5 μmol/L Ca²⁺ boluses. Interestingly, ZSF1-obese cardiac mitochondria failed to take up Ca²⁺, as demonstrated by the accumulation of Ca²⁺ in the bath (ie, extramitochondrial) (Figures 7H and 7I, Supplemental Figure 6B). This is suggestive of mitochondria that are either already calcium overloaded or that have down-regulated mitochondrial calcium uniporter (MCU) activity. Mitochondrial swelling, an indicator of susceptibility to mitochondrial permeability transition, was increased in lean hearts, whereas swelling of ZSF1-obese HFpEF mitochondria occurred faster and to a greater extent than both WKY and ZSF1-lean mitochondria (Figures 7J-7M, Supplemental Figure 6C). Protein expression according to immunoblotting of proteins involved in mitochondrial Ca²⁺ handling revealed a significant increase in both the 30 kD and 40 kD MCU isoforms and in the MCU gatekeeper, mitochondrial calcium uptake 1 (MICU1), exclusively in ZSF1-obese HFpEF cardiac mitochondria when normalized to a mitochondrial loading control, ATP synthase (ie, complex V) (Figures 7N and 7O, Supplemental Figure 7). VDAC1, which is also involved in mitochondrial calcium homeostasis,³⁶ was significantly decreased in ZSF1-obese mitochondria (Figure 7N). Collectively, our results indicate significant remodeling of the mitochondrial ultrastructure, accumulation of cardiomyocyte LDs, dysfunctional respiratory capacity, and dysregulated calcium handling, all of which underlie and likely contribute to the gross metabolic dysregulation and subsequent cardiac dysfunction observed in the HFpEF heart.

Discussion

In this study, we characterized and identified the underlying molecular changes associated with the HFpEF phenotype in a robust preclinical rat model. The following sections highlight those key findings, integrating them with the current literature in HFpEF.

Impact of the additive hit of metabolic syndrome and obesity

The ZSF1-obese rat model recapitulates the multifactorial clinical features that distinguish HFpEF. Importantly, our study validates the notion that “2 hits,” hypertension and metabolic syndrome/obesity, are necessary for the development of HFpEF, similarly to the “2-hit” L-NAME + HFD mouse model¹¹ and in agreement with human HFpEF populations, which are typically obese with vascular dyfunction.^38,39 A strength of our study is the inclusion of the WKY non-hypertensive control as this allowed us to examine transcription, metabolic, and functional changes that are dependent on and independent from these “2 hits.” Although hypertension alone resulted in cardiac metabolic remodeling, the addition of metabolic syndrome resulted in more drastic cardiac metabolism which was associated with an energetic deficit and mitochondrial abnormalities, both in ultrastructure and function. Similarly, only 152 transcription changes were exclusively dependent on hypertension, whereas 795 transcripts independent from hypertension and 5,544 transcript changes were dependent on both metabolic syndrome and hypertension. Other models of HFpEF have been proposed, namely, models of Western diet feeding,⁹ angiotensin II/phenylephrine (ANGII/PE) infusion,⁸ and senescence-accelerated aging (ie, SAMP/SAMPR mice),^10,12 and although pathology associated with these models may be multifactorial, those models do not contain 2 independent hits. Furthermore, long-term Western diet feeding progresses to HFrEF, which rarely occurs in humans,^40,41 and ANGII/PE and SAMP/SAMPR models lack metabolic syndrome or hypertension. Although these models are likely suitable for the study of diastolic dysfunction, this is distinct from the multifaceted, multiorgan nature of HFpEF. Thus, the ZSF1-obese rat model serves as an excellent preclinical model to study cardiometabolic HFpEF.

An interesting observation from our studies was that systolic function was preserved in the face of metabolic and mitochondrial dysfunction. This phenomenon is most likely due to the fact that cardiac contractility and energetics are preserved at all costs. For example, we have previously shown in genetic mouse models lacking mitochondrial calcium uptake, which is a critical second messenger for energetic matching of cardiac workload, that basal cardiac function is maintained. Indeed, only with additional stress, such as adrenergic overload or myocardial infarction, is any deficit in systolic function unmasked.^25,27 Thus, it is likely that with additional stress, the ZSF1 model would likewise show systolic dysfunction, but the value of the model is that there is no systolic dysfunction even with aging.

Identifying therapeutic targets for HFpEF has proven difficult owing to the combination of causes that contribute to the syndrome. Recently, sodium-glucose cotransporter 2 inhibitors (SGLT2i), which act to block glucose reabsorption in the kidney, have proven to be efficacious and safe in reducing cardiovascular events in nonhuman animal models^42,43 and in human HFpEF trials.^44,45 SGLT2 inhibitors have a minimal impact on hypertension yet result in profound weight loss and normalization toward glucose homeostasis.^42,46,47 These results are directly in line with our findings and overall conclusion that the primary driver of HFpEF is the metabolic syndrome component. Adjunctive therapy of SGLT2i and a hydrogen sulfide donor (H₂S, a well studied cardioprotective agent) in our ZSF1-obese HFpEF model was shown to be efficaciously superior to either treatment alone.¹⁹ This is interesting, because H₂S has been shown to modulate metabolism²³ and preserve mitochondrial integrity.⁴⁸ Although we observed numerous transcriptional changes independent from hypertension, most of the transcriptional remodeling was dependent on both hits, so treatments aimed at targeting metabolic syndrome alone will likely be insufficient for long-term efficacy.

This study provides a roadmap for the discovery of novel mechanisms driving HFpEF progression and provides a data set that can be correlated with the remodeling observed in human HFpEF.⁴⁹ Targeting of HDAC6 in an HFpEF mouse model was recently shown to be as efficacious as SGLT2i.⁴³ This is of note, because the mechanisms of SGLT2i cardioprotection remain unknown. Indeed, mice with global deletion of SGLT2 are protected from HF with SGLT2 blockade,^50,51 indicating that SGLT2i likely have an off-target mechanism of action.

Remodeling of energy metabolism

An important observation from our metabolomics data set is a drastic change in the energy state of the ZSF1-obese heart. Hearts with HFpEF displayed a greater phosphocreatine:ATP ratio (1.85 vs 0.884) and lower ATP:ADP ratio (0.37 vs 4.87) compared with hypertension alone. This was driven by a significant reduction in ATP in the HFpEF heart. This may, in part, be associated with an inability to liberate phosphocreatine stores, because all creatine kinase isoforms were reduced in ZSF1-obese rats. This indicates that the HFpEF heart is energy starved compared with the ZSF1-lean control (hypertension alone). The ZSF1-obese heart also displayed gross metabolic remodeling of pathways associated with energy metabolism, as discussed subsequently.

Glycolysis

We observed an increase in the expression of Pfkl and of Pfkfb1, activators of aerobic glycolysis, as well as increased downstream metabolites of glycolysis (ie, 3-PG, 2-PG, PEP, and pyruvate), suggesting increased glycolysis in the hypertensive heart. In contrast, nearly all glycolytic enzymes were decreased in abundance in the obese HFpEF heart. These results are in agreement with results from the Kass Lab, which also found a reduction in protein expression of these same glycolytic enzymes in human HFpEF endomyocardial biopsies; however, they detected only decreased abundance of the upstream glycolytic metabolites G6P and F-1,6-BP, whereas we observed a decrease in the downstream metabolite PEP.⁵² Glucose oxidation is also likely decreased, because we and others have shown changes in the abundance of pyruvate, PDH, MPC1, and PDK4,^52,53 with direct measurements of reduced glucose oxidation performed in the working heart.⁵⁴ In ZSF1-obese and human HFpEF hearts, changes in the pentose phosphate pathway (ie, purine and pyrimidine metabolism) were identified, indicating disruptions to ancillary biosynthetic pathways; these ancillary pathways are known to contribute to cardiac remodeling,^55,56 highlighting that their role in HFpEF is an area worthy of investigation. Collectively, these studies suggest a significant down-regulation of glycolytic metabolism and changes in ancillary pathways in the HFpEF heart.

Fatty acids

Fatty acids contribute the largest percentage to cardiac energy production, so loss of oxidative capacity in the failing heart would be detrimental to energy metabolism and cardiac function. Our results indicate a significant impairment in FAO, with associated fatty acid metabolic processes among the most down-regulated in the HFpEF heart, including key enzymes in transport and processing (eg, Acs, Cpt1, Cact, Acad, Acat1). Dysregulation of genes involved in fatty acid and oxidative metabolism seems a conserved signature, because similar findings to ours have been shown in other murine models^43,57 and human HFpEF populations⁵³; however, some studies have reported an increase in FAO transcripts.⁵⁸ Although there is a discrepancy in gene expression among different studies, a proteomic study of HFpEF samples, from the same group that reported an increase in FAO and oxidative phosphorylation transcripts, found an overall decrease in protein abundance,⁵⁹ a reminder that transcript and protein abundance often do not correlate in pathology.

In agreement with a decrease in FAO, our results indicate reduced abundance of short-chain acyl-carnitines and increased medium- and long-chain acylcarnitines, potentially suggesting inefficient oxidation. Medium- and long-chain acylcarnitines were decreased in human HFpEF, and the expression of FAO genes also were decreased⁵³; whether this discrepancy in findings is simply due to sampling of the right vs the left ventricle remains to be determined. Nonetheless, Krebs cycle intermediates are lower in human HFpEF,⁵³ and we demonstrate reduced utilization of fatty acids by mitochondria isolated from HFpEF hearts. While these collective results suggest impairments in fatty acid utilization, palmitate oxidation measured in the isolated working heart was increased in the mouse “2-hit” L-NAME + HFD mouse model⁵⁴; More work is needed to define how HFpEF remodels cardiac fatty acid metabolism.

Ketones, BCAAs, and amino acids

The potential for alternative fuel sources to contribute to cardiac energetics has become more appreciated. Our integrated network analysis approach identified the synthesis and degradation of ketone bodies as highly affected in HFpEF. We observed an increase in 3-hydroxybutyrate and a corresponding decrease in key ketone catabolic enzymes (ie, Bdh1, Oxtc1, Acat1), suggestive of reduced utilization. In a murine model of HFpEF, BDH1 protein abundance is reduced, with a corresponding trend of decreased oxidation rates.⁵⁴ In HFrEF, myocardial uptake, oxidation, and expression of BDH1 increases 2- to 3-fold,60, 61, 62 which is greater than predicted rates in HFpEF.⁶¹ This divergence in ketone body oxidation between HFrEF and HFpEF may provide insight into differential substrate/fuel treatment strategies. For example, increasing circulating ketones through the diet appears to provide beneficial effects in HFrEF,⁶³ but whether this would be beneficial in HFpEF has not been explored. Another potential target could be HMGCS, which we found up-regulated in the ZSF1-obese HFpEF heart and is generally known to be involved in ketone synthesis; thus, whether impaired ketone oxidation is due to competing synthesis mediated by HMGCS presents an interesting inquiry.

The BCAAs leucine, valine, and isoleucine have been proposed as an alternative fuel source for the heart, and suppression of BCAA oxidation has been implicated in HF.^64,65 Although we observed a global suppression of BCAA oxidation genes, the down-regulation of the nodal BCAA catabolic enzyme branched-chain α-keto acid dehydrogenase complex (Bckdh) agrees with data in human HFpEF.⁵³ Previous reports suggest that BCAAs accumulate in the human HFpEF heart, suggesting decreased oxidation.⁶⁶ However, contributions of BCAAs to energy production are likely minimal,^61,64,67,68 and the activation of cardiac BCAA oxidation does not provide energetic or functional benefit in models of HFrEF.⁶⁹ Thus, although BCAA oxidation seems down-regulated, targeting this pathway in HFpEF may not prove effective.

Our integrated pathway maps identified the down-regulation of numerous other amino acid pathways at the level of transcription and metabolite abundance. The observed changes in global amino acid metabolism could be related to the increased proteolysis that occurs in the failing heart.⁶¹ Many of these amino acids and represented pathways have yet to be explored, providing experimental opportunities to generate new hypotheses. For example, we observed a reduction in arginine metabolism, which when given as an oral supplement to HFrEF patients proved beneficial⁷⁰; whether similar benefits could be obtained in HFpEF patients is worth exploring.

Impact on mitochondrial and LD structure and function

Ultrastructural changes

Mitochondrial dysfunction is a hallmark of HF,71, 72, 73 and our transcriptomic and metabolic signatures implicates the derangement of several mitochondrial processes in the ZSF1-obese heart. Down-regulation of biological processes related to cristae formation and mitochondrial fusion were confirmed by ultrastructural remodeling characterized by the disruption and near disappearance of cristae and overall smaller and more rounded mitochondria, indicating that the fission-fusion balance is perturbed. Alterations in mitochondrial shape and cristae density can greatly affect the localization, structure, and function of the oxidative phosphorylation system, impairing cellular and mitochondrial metabolism.^34,35,74 Although we observed no change in total mitochondrial number, mitochondrial area was significantly reduced in HFpEF, likely because the smaller size of individual mitochondria. Interesting, mice treated with the SGLT2i empagliflozin demonstrated an increase in mitochondrial area per cardiomyocyte area.⁴² Recently, TEM of human HFpEF cardiomyocytes revealed no change in mitochondrial area but significant cristae derangement, which was most observable in patients presenting with obesity.⁵⁹

There exists a high correlation between myocardial adiposity and diastolic dysfunction.^75,76 Cardiac magnetic resonance imaging of HFpEF, HFrEF, and nonfailing patients revealed significant intramyocardial fat only in HFpEF.⁷⁷ We observed the accumulation of LDs in HFpEF hearts which was also recently seen in HFpEF patients via TEM imaging.⁵⁹ LDs act as an energy storage depot and are involved in transferring stored fatty acids to mitochondria for energy production. However, transcriptional down-regulation of FAO machinery and the structural remodeling likely limit utilization, thus promoting storage and LD accumulation. Although we observed greater mitochondria-LD interactions in HFpEF, the interpretation of this result is confounded by the fact that few, if any, LDs were observed in control hearts. To overcome this limitation, we examined the expression of known proteins that act as tethers to support mitochondria and LD approximation. Perilipin 5 (PLIN5), an LD protein reported to tether them to mitochondria,^78,79 was down-regulated 3-fold in our HFpEF hearts. Loss of PLIN5 decreases mitochondria-LD interactions and oxidative metabolism, whereas overexpression increases these interactions.⁸⁰ Similarly, we noted a down-regulation of Miga2, another mitochondria-LD tether involved in lipid metabolism and mitochondrial fusion.^81,82 Whether disruption of these tethers could play a role in HFpEF is unknown, but it is striking that these LD proteins were down-regulated in the context of massive LD biogenesis.

Mitochondrial dysfunction

Mitochondrial respiratory capacity was significantly impaired in both pre-HFpEF (hypertensive) and HFpEF (hypertension and metabolic syndrome) hearts, suggesting that although mitochondrial dysfunction is a key feature of HF, it is not necessarily unique to HFpEF. What is unique in the HFpEF heart is impaired mitochondrial calcium handling. Mitochondrial protein expression of MCU and MICU1, components of the MCU, were increased exclusively in the HFpEF heart. This could be a compensatory change to increase calcium-dependent activation of mitochondrial dehydrogenases to increase Krebs cycle flux and mitochondrial energetics. However, as previously reported by our group and others, though initially compensatory these expression changes in uniporter components turns maladaptive with chronic stress. In a mouse model of HFpEF, SGLT2i treatment improved HFpEF-mediated Ca²⁺ reuptake by the sarcoplasmic reticulum and rescued mitochondrial respiratory function⁴²; however, whether improved re-uptake was a consequence of improving mitochondrial Ca²⁺ buffering, improved energetics, and/or enhancing SERCA activity was not tested. Similarly, treating HF with a pan–histone deacetylase (HDAC) inhibitor decreased acetylation of proteins involved in oxidative metabolism, improving mitochondrial oxidative phosphorylation.⁸³ Whether HDAC inhibition plays a similarly protective role in HFpEF remains to be investigated.

Study limitations

There are several limitations to our study. Our results do not test a specific mechanism or hypothesis but provide an integrated systems biology approach to allow for the discovery of potentially important pathways and mechanisms contributing to HFpEF. We recognize that steady-state metabolomics, though powerful in its ability to capture a larger metabolite profile, provides only a snapshot of the cardiac metabolome and lacks the ability to determine anabolic and catabolic flux. Therefore, future studies should implement in vivo stable isotope metabolomics to resolve changes in metabolic pathways. The present study also used only male rats, partly owing to the high cost of acquiring a sufficient number of female rats to maintain power in our dual-control study. However, a recent study that exclusively used female ZSF1 rats reported similar findings related to the mitochondria,⁴² suggesting conserved mechanisms of action between sexes. Adjusting for sex in a human HFpEF RNA sequencing study, importantly, did not affect pathway enrichment.⁵⁸ Nonetheless, because the prevalence of HFpEF is slightly greater in women than in men, we understand and acknowledge the importance of potential for sex differences in molecular pathways. A benefit of the ZSF1 HFpEF model is that female rats display several of the same clinical characteristics seen in male rats (ie, hyperglycemia and obesity, cardiac hypertrophy, increased peripheral tissue weights, diastolic dysfunction, and preserved systolic function^20,42). This is an important feature because other models, namely, the “2-hit” L-NAME + HFD mouse model, were reported to display no cardiac hypertrophy and a smaller degree of diastolic dysfunction in females.²¹ This is of significance because clinical HFpEF displays a slightly higher prevalence in women vs men.⁸⁴ Although these aspects make the ZSF1 obese rat an excellent model for HFpEF, the “2-hit” model still holds value, because it is more amendable to genetic intervention and budget friendly. Current work in our laboratories is exploring whether similar functional, metabolic, transcriptional, and mitochondrial remodeling is found in female HFpEF or whether sex distinguishes between remodeling pathways and targets.

Also, aging is a critical risk factor for HFpEF and is not accounted for in our study. As we have shown in a large animal model of diastolic dysfunction, aging alone significantly alters the cardiac transcriptome and metabolome,⁷ making it difficult to tease out pathway changes due to disease progression, aging, or both.

Finally, it is evident that metabolic syndrome and obesity are primary drivers of HFpEF development, thus understanding the systemic changes at peripheral tissues is critical to complete our understanding of HFpEF pathophysiology. In a large-animal model that recapitulates several clinical features of HFpEF and diastolic dysfunction, we found skeletal muscle to have distinct transcriptional and metabolic signatures that were accompanied by mitochondrial dysfunction,⁷ and similar findings have been noted in human HFpEF skeletal muscle.⁸⁵ Similar approaches have also been performed to identify potential candidates for interorgan crosstalk between the liver and heart in HFpEF.⁸⁶ Investigating peripheral tissues and potential interorgan communication likely will yield novel and meaningful insights to understand HFpEF development and progression.

Conclusions

In summary, the results presented here demonstrate the power of applying integrated omics technologies to lead to the design of functional experiments to test specific hypotheses and discover novel therapeutic targets. The ZSF1-obese rat model recapitulates the clinical characteristics of human HFpEF and shares many of the same transcriptional, metabolic, and mitochondrial remodeling as seen in patients. Our findings provide a wealth of data that are likely to reveal novel metabolic pathways and molecular targets which will hopefully allow for the discovery of new therapeutics to treat HFpEF.

Perspectives.

COMPETENCY IN MEDICAL KNOWLEDGE: There is currently no known treatment for HFpEF. This systems-based study provides insight into the metabolic and transcriptional pathways in a preclinical animal model that recapitulates many clinical features of HFpEF.

TRANSLATIONAL OUTLOOK: We demonstrate that derangements in mitochondrial ultrastructure, function, and metabolism observed in this ZSF1-obese HFpEF model is similar to recent findings in HFpEF patients, indicating the strength of this preclinical animal model and a likely mechanism underlying disease development and progression. The characterization and validation of additional preclinical animal models and subsequent in vivo studies are needed to identify potentially targetable metabolic and transcriptional pathways. Future studies are needed in such animal models to advance our understanding of the disease to discover potentially novel therapeutic strategies for HFpEF patients.

Funding Support and Author Disclosures

This work was supported in part by grants from the National Institutes of Health (NIH; F32HL145914) and American Heart Association (AHA; Career Development Award 937591) to Dr Gibb, an AHA postdoctoral fellowship (20POST35200075) to Dr Li, a training fellowship to Dr Elrod from the NIH National CCTS awarded to the University of Alabama at Birmingham (TL1TR00316), NIH grant F30HL152564 to Dr Lazaropoulos, National Institute of Alcohol Abuse and Alcoholism grant AA029984 to Dr Sharp, NIH grant HL159428 to Dr Goodchild, NIH grants HL146098, HL146514, and HL151398 to Dr Lefer, and NIH grants R01NS121379, P01HL147841, 2P01HL134608, and T32HL091804 and AHA grant 20EIA35320226 to Dr Elrod. The authors have no relationships relevant to the contents of this paper to disclose.