Diabetes Mellitus Severity and a Switch From Using Lipoprotein Lipase to Adipose‐Derived Fatty Acid Results in a Cardiac Metabolic Signature That Embraces Cell Death

Abstract

Background

Fatty acid (FA) provision to the heart is from cardiomyocyte and adipose depots, plus lipoprotein lipase action. We tested how a graded reduction in insulin impacts the source of FA used by cardiomyocytes and the cardiac adaptations required to process these FA.

Methods and Results

Rats injected with 55 (D55) or 100 (D100) mg/kg streptozotocin were terminated after 4 days. Although D55 and D100 were equally hyperglycemic, D100 showed markedly lower pancreatic and plasma insulin and loss of lipoprotein lipase, which in D55 hearts had expanded. There was minimal change in plasma FA in D55. However, D100 exhibited a 2‐ to 3‐fold increase in various saturated, monounsaturated, and polyunsaturated FA in the plasma. D100 demonstrated dramatic cardiac transcriptomic changes with 1574 genes differentially expressed compared with only 49 in D55. Augmented mitochondrial and peroxisomal β‐oxidation in D100 was not matched by elevated tricarboxylic acid or oxidative phosphorylation. With increasing FA, although control myocytes responded by augmenting basal respiration, this was minimized in D55 and reversed in D100. Metabolomic profiling identified significant lipid accumulation in D100 hearts, which also exhibited sizeable change in genes related to apoptosis and terminal deoxynucleotidyl transferase dUTP nick‐end labeling–positive cells.

Conclusions

With increasing severity of diabetes mellitus, when the diabetic heart is unable to control its own FA supply using lipoprotein lipase, it undergoes dramatic reprogramming that is linked to handling of excess FA that arise from adipose tissue. This transition results in a cardiac metabolic signature that embraces mitochondrial FA overload, oxidative stress, triglyceride storage, and cell death.

Clinical Perspective

What Is New?

• Using dissimilar doses of streptozotocin (55 and 100 mg/kg), a β‐cell–specific toxin used to induce hypoinsulinemia, we established models of diabetes mellitus of varying intensities; despite the decline of insulin in D55, although these animals were hyperglycemic, they exhibited minimal change in average plasma fatty acid (FA) concentrations, and it was only in D100 that the predicted enlargement in FA emerged.

• With increasing severity of diabetes mellitus, when the heart is unable to control its own FA supply using lipoprotein lipase, it undergoes dramatic reprogramming like stimulation of β‐oxidation in peroxisomes and mitochondria, that is linked to handling of excess FA that arise from adipose tissue.

• As high β‐oxidation is not matched by elevated tricarboxylic acid cycle activity or enhanced oxidative phosphorylation, the subsequent increase of FA intermediates and their diversion to triglyceride, paired with increased reactive oxygen species formation secondary to excessive β‐oxidation, provoked a gene expression program supporting cell death (lipotoxicity).

What Are the Clinical Implications?

• In guiding the clinical management of type 1 diabetes mellitus, in addition to focusing on drugs that lower plasma glucose, attention should also be placed on dyslipidemia, which could be a major link between diabetes mellitus and cardiomyopathy.

Introduction

Coronary disease is a primary basis for the increased incidence of cardiovascular disease following diabetes mellitus.1, 2 However, patients with type 1 and type 2 diabetes mellitus have also been diagnosed with reduced or low‐normal diastolic function and left ventricular hypertrophy in the absence of vascular defects (labeled diabetic cardiomyopathy).3, 4 Animal models of diabetes mellitus also exhibit signs of cardiomyopathy.5, 6, 7 Factors contributing to the development of cardiomyopathy include connective tissue and insoluble collagen accumulation, impaired sensitivity to various ligands, mitochondrial dysfunction, endoplasmic reticulum stress, renin‐angiotensin‐aldosterone system activation and abnormalities in Ca²⁺ sensing and regulating proteins.5, 7 Abnormalities in pathways regulating cardiac energy metabolism also contribute toward the etiology of diabetic cardiomyopathy.8

The earliest metabolic change occurring in the diabetic heart is reduced glucose consumption,9 and a switch to exclusively utilize FA for energy.10, 11 This metabolic switching is likely initiated by a high‐glucose–induced release of endothelial cell heparanase‐1,12 an endoglycosidase exceptional in its ability to cleave side chains of heparan sulfate proteoglycan on the cardiomyocyte cell surface, thereby instigating release of bound ligands. These include lipoprotein lipase (LPL),13 which then traverses the interstitial space to the apical surface of the coronary lumen, where it breaks down the triglyceride core of circulating lipoproteins to release FA14 (Figure 1E). In this way, when glucose is being underutilized and a need for FA arises, the heart can tailor LPL to meet the ATP demands of the cardiomyocyte by providing the necessary amount of FA to replace ATP otherwise derived from glucose breakdown (LPL finely tunes FA delivery to cardiomyocytes). With increasing severity and disease progression, this cardiac heparanase‐1–LPL signaling axis is compromised in the diabetic heart, leading to a decline in LPL.15 A downside to the loss of this LPL rheostat renders the diabetic heart dependent on hydrolysis of adipose tissue (AT) triglyceride to generate FA. This could be a questionable adaptation, as the site of control of FA provision to the heart changes from a measured regulated delivery by coronary LPL to the unrestrained and unregulated provision by AT.

Figure 1.

The severity of streptozotocin (STZ) diabetes mellitus is uncovered by measuring insulin and not glucose. Diabetes mellitus was induced by injection of 2 different doses of STZ (55 [D55] or 100 [D100] mg/kg IV) and the animals terminated 4 days later (n=4–6). At termination, weight gain (change in body weight compared with initial value) was assessed (A). Glucose was determined throughout the study period from tail‐tip blood samples using glucose test strips and an Accu‐Chek glucose monitor (B). Following heart extraction, blood in the thoracic cavity was collected in tubes containing K2‐EDTA as anticoagulant. After centrifugation, plasma isolated from this blood was used for determination of insulin (C, left panel) and fatty acids (FA) and triglycerides (TG) (D). Pancreatic total insulin (ng) was determined following acid‐ethanol extraction and measured using a rat ultrasensitive insulin ELISA normalized to total protein (μg; D, right panel). Lipoprotein lipase (LPL) trafficking in the heart is summarized in E with the inset (stacked bar graph that characterizes LPL activity in the 1.0 mol/L fraction redrawn from original in Wang et al15) representing dimeric, catalytically active LPL determined in heart homogenates loaded onto a heparin‐sepharose column and eluted with increasing concentrations of NaCl. To measure LPL activity, we used in vitro hydrolysis of a [³H]triolein substrate. Data are presented as mean ±SEM. Significantly different from control, *P<0.05, **P<0.01, ***P<0.001. Significantly different from the D55 group, ^# P<0.05, ^## P<0.01, ^### P<0.001.

Streptozotocin (STZ) is an agent that produces pancreatic β‐cell necrosis and ensuing hypoinsulinemia. We used 2 doses (55 and 100 mg/kg IV) of STZ to differentially lower plasma insulin and produce diabetes mellitus of variable intensities. With D55, animals are moderately hypoinsulinemic, hyperglycemic, with little change in circulating FA or triglyceride and an increase in LPL. Interestingly, D100 displays striking hypoinsulinemia with augmentation of both circulating glucose and FA/ triglyceride15, 16, 17, 18 and a reduction in LPL. We tested how the heart adapts to this graded reduction in insulin and a switch from LPL‐derived FA to those obtained from AT.

Methods

The authors declare that all supporting data are available within the article and its online supplementary files.

Experimental Animals

This investigation conformed to the Guide for the Care and Use of Laboratory Animals published by the National Institutes of Health, the Canadian Council on Animal Care Guidelines, and institutional guidelines at the University of British Columbia (Certificate A17‐0072). Adult male Wistar rats (240–260 g; Charles River Laboratories, Wilmington, MA) were fed Laboratory Rodent Diet 5001 with a chemical composition of 5% total fat (ether extract) made up of 1.5% saturated, 1.6% monounsaturated, and 1.5% polyunsaturated FA. STZ is a β‐cell‐specific toxin used to induce hypoinsulinemia and diabetes mellitus. Under isoflurane anesthesia, animals were injected with a single intravenous dose of 55 (D55) or 100 (D100) mg/kg of STZ into the tail vein.15 After 24 hours, hyperglycemia (>13 mmol/L) was confirmed in tail‐tip blood samples using a glucometer and glucose test strips. Animals were euthanized after 4 days. Female rats were not used, as we have previously documented that female rodents are not as susceptible to the diabetogenic effects of D55‐STZ.19

Metabolic Assessments

At termination, hearts were exsanguinated and blood in the thoracic cavity collected in K2‐EDTA tubes and centrifuged immediately for separation of plasma that was used for determination of insulin (rat insulin ELISA; ALPCO, Salem, NH), FA (NEFA‐HR; Wako Diagnostics, Mountain View, CA) and triglyceride (Stanbio Triglycerides LiquiColor Mono; Thermo Fisher Scientific, Waltham, MA). After acid‐ethanol extraction, insulin content in the pancreas was assessed by ELISA.

RNA Sequencing and Analysis

Total RNA from 4 control, 6 D55, and 6 D100 rats was isolated using TRIzol (Thermo Fisher Scientific, Waltham, MA). Sequencing libraries were prepared from 400 ng total RNA using the TruSeq Stranded mRNA Sample Preparation kit (Illumina, San Diego, CA). Samples were checked for quality using a Bioanalyzer (Agilent Trchnologies, Sanra Clara, CA) and quantified using Qubit fluorometer (Thermo Fisher Scientific, Waltham, MA). Libraries were multiplexed and sequenced on the NextSeq 500 (Illumina). Following a previously published procedure,20 multiple analysis pipelines were applied and their results combined. The output for each pipeline was a list of genes ranked by the P value for differential expression after correction for multiple testing. A combined list was obtained by ranking the genes according to their median rank from the various analysis pipelines. One potential outlier was detected when clustering the samples and was therefore removed for the differential expression analysis. Network analysis and function categorization were conducted using STRING (set at the highest confidence with evidence from experiments, databases, coexpression, and co‐occurrence21). Gene ontology term enrichment and Kyoto Encyclopedia of Genes and Genomes pathway analysis (Enrichr software22) of differentially expressed genes were also applied to ascertain significantly enriched pathways.

Separation and Characterization of Plasma and Cardiac Lipids

Total plasma and cardiac lipids were extracted and solubilized in degassed chloroform:methanol:acetone:hexane (4:6:1:1v/v/v/v). Separation of triglyceride and FA were achieved using high‐performance liquid chromatography (HPLC; Waters 2690 Alliance HPLC, Milford, MA) equipped with an auto‐sampler and column heater, as previously described.23 Lipid classes were separated on a YMC DIOL column (4.6×250 mm, YMC, Asan, Korea) and the HPLC flow was split with ≈80% flow going to a FCII fraction collector (Waters, Milford, MA) and the remaining 20% flow going to a Waters evaporative light scattering detector. Phospholipid fractions were collected, solvent evaporated under a stream of nitrogen, then derivatized with boron trifluoride (14%) in methanol. Individual FA methyl esters were then separated using an 6850 GLC (Agilent Technologies) equipped with a flame ionization detector and an SP‐2330 capillary column (30 m×0.25 mm internal diameter) (Supelco, Bellefonte, PA) using hydrogen as a carrier gas. Peak areas were calculated using Chemstation software (Agilent Technologies) and FA quantified using heptadecaenoic acid (17:0) as the internal standard.

Isolation of Cardiomyocytes

Primary ventricular cardiomyocytes were prepared following previously described procedures.17, 24

Mitochondrial Stress Test

Oxygen consumption rate (OCR) was measured using the Seahorse XFe96 extracellular flux analyzer and the Seahorse XF Cell Mito Stress Test kit (Agilent Technologies). Mitochondrial inhibitor concentrations and cell density were optimized, and protocols were based on previously published data.25, 26 Cardiomyocytes were plated at a density of 5000 cells/well and incubated overnight in 150 μL of normal culture medium (replicates of 8–12 wells per condition and cell type). The next day, cells were washed twice with warm substrate‐limited assay media (XF Base Medium, 2.5 mmol/L GL, 0.1 mmol/L sodium pyruvate, 4 mmol/L l‐glutamine, 0.5 mmol/L l‐carnitine, 5 mmol/L HEPES, pH 7.4) and incubated with fresh assay media for 1 hour. Immediately before assay, either bovine serum albumin (BSA) 0.05 mmol/L palmitic acid (PA) or 0.1 mmol/L PA were added to the wells (PA:BSA, Agilent Technologies). The cartridge was loaded with 3 metabolic inhibitors, which were sequentially injected into the plate, and OCR measured after each addition subsequent to 3 initial baseline readings: oligomycin (2 μmol/L), carbonyl cyanide 4‐(trifluoromethoxy)phenylhydrazone (FCCP; 2 μmol/L), followed by the combination of rotenone and antimycin A (2 μmol/L/10 μmol/L). At the end of the assay, OCR was normalized to protein content.

Cardiac Apoptosis

Terminal deoxynucleotidyl transferase dUTP nick‐end labeling assay was carried out using a terminal deoxynucleotidyl transferase dUTP nick‐end labeling kit (Promega, Madison, WI) as described previously.23 Nuclei (n=2546, 2256, and 2382 for controls, D55, and D100, respectively) were counted from 3 different ventricular sections per animal, from 3 separate animals per group.

Metabolomic Profiling

Untargeted metabolomic analysis was performed using an Agilent 6530 quadrupole‐time of flight mass spectrometer, Agilent 1290 binary ultra performance liquid chromatography and MassHunter data acquisition software. The instrument was operated in positive ion, then negative ion for reverse phase and hydrophilic interaction (HILIC) chromatography involving a total of 4 separate runs for each sample. Negative ion tests used the same columns, but mobile phases were 5 mmol/L ammonium acetate (Fisher Optima LC/MS Grade, Thermo Fisher Scientific) in 98% water 2% acetonitrile, with pH adjusted to 9.2 using ammonium hydroxide. Reverse phase chromatography utilized a Waters BEH‐C18 column, 2.1×100 mm, 1.7 μm particle size, and similar buffered mobile phases to HILIC with the exception of using methanol rather than acetonitrile. Gradient chromatography was used for both HILIC and reverse phase with starting conditions of 90% organic solvent for HILIC and 90% aqueous for reverse phase and total run times of 35 to 45 minutes. The 6530 quadrupole‐time of flight mass spectrometer with Agilent Jet Spray was operated in electrospray ionization mode and used dual electrospray ionization nebulizers for simultaneous introduction of sample and reference mass solutions, data were collected at 2 scans/second with an m/z range of 65 to 1050 in all modes of operation. The quadrupole‐time of flight mass spectrometer was tuned specifically for low masses following the manufacturer's instructions, operated in high‐resolution mode using 4‐GHz data acquisition and used Agilent's standard reference mass solution providing corrected mass accuracy of about 2 to 3 ppm. Additional runs of pooled heart samples used tandem mass spectrometry mode and specifically targeted ions of interest with collision energies of 10, 20, or 40 volts to assist in identification of analytes. Data processing used multiple Agilent Technologies software packages. Profinder version 10.0 was used for molecular feature extraction; then, all features were exported to the chemometric software Mass Profiler Professional. Data imported to Mass Profiler Professional were normalized using a 75th percentile shift algorithm, baselined to the median of all samples, then filtered to remove entities that were highly variable within any one sample group. Identification of metabolites used a number of resources including Agilent MassHunter ID Browser B.08 software and Metlin accurate tandem mass spectrometry spectral library with matching mass tolerance set at 2 mDa±5 ppm, online mass spectral libraries from the Human Metabolome Database,27 and MassBank (https://massbank.eu/MassBank/), and our own mass spectral/retention time library created from about 200 compounds provided by British Columbia Children's Hospital Newborn Screening program.

Statistical Analysis

One‐way ANOVA followed by Bonferonni post hoc comparisons test was used to determine differences between group mean values. For some analyses, Kruskal–Wallis nonparametric test followed by Dunn post hoc comparisons test was applied for nonnormal distributions. Values are presented as means±SEM with individual data points. The minimum level of statistical significance was set at *P<0.05.

Results

Using dissimilar doses of STZ, we established models of diabetes mellitus of varying intensities. Accordingly, weight gain over 4 days was reduced in D55 animals compared with control such that these rats had lower body weights at the time of death. Induction of a more severe diabetes mellitus in D100 rats over the same time period resulted in a loss of body weight (as compared with body weight on day 0; Figure 1A). Both D55 and D100 were equally hyperglycemic compared with control over the duration of the study, with no measurable difference between the 2 diabetic groups (Figure 1B). Intriguingly, the severity of diabetes mellitus between D55 and D100 was uncovered only following determination of insulin; D100 animals showed markedly lower plasma and pancreatic insulin compared with D55 (Figure 1C). This dramatic pancreatic β‐cell destruction and hypoinsulinemia in D100 was accompanied by robust increases in plasma FA and triglyceride, effects that were essentially absent in D55 animals (Figure 1D). Our data advocate for D55 as a model of moderate type 1 diabetes mellitus with insufficient glycemic management, and an increase in cardiac LPL activity (Figure 1E, inset) to support FA provision to the heart when glucose use is curtailed. Additionally, D100 could be considered as a model of severe type 1 diabetes mellitus with poor glycemic control and dyslipidemia, resulting in a loss of LPL activity (Figure 1E, inset).

Subsequent to a reduction in circulating insulin, increased AT lipolysis attributable to lipase (adipose triglyceride lipase and hormone‐sensitive lipase) hyperactivity releases FA into the circulating plasma.28 Interestingly, despite the decline of insulin in D55, these animals exhibited minimal change in average plasma FA concentrations (Figure 2A through 2E). However, with increasing severity of diabetes mellitus and a further drop in insulin, the predicted enlargement in FA emerged. Accordingly, D100 animals exhibited close to a 2‐ to 3‐fold increase in various types of saturated (palmitic [16:0], stearic [18:0]), monounsaturated (oleic [18:1]), and polyunsaturated (linoleic (18:2), arachidonic (20:4)] FA that made up ≈80% of the total plasma pool (Table S1 and Figure 2A through 2E). It should be noted that despite the absolute increase in plasma saturated FA, monounsaturated FA, and polyunsaturated FA in D100, the percentage composition of these FAs remained unaltered (Figure 2F). Also of interest was the observation that circulating very‐long‐chain FAs (VLCFAs; with ≥22 carbons; eg, docosahexaenoic acid, C22:6n3), that made up a small percentage of the total identified FA and that must undergo initial peroxisomal β‐oxidation before entering the mitochondria for ATP generation, were also increased in D100 (Table S1). Our data suggest that the D100 heart could use the substantially increased plasma concentrations of these different FAs for generation of ATP instead of FA originating from LPL action.

Figure 2.

Plasma fatty acid (FA) composition is altered only when insulin reduction by streptozotocin (STZ) is substantial. Four days after injection of STZ, animals were terminated and plasma collected for determination of fatty acid (FA) composition (n=4–6). Plasma FAs were extracted with chloroform:methanol:acetone:hexane solvent. Separation of FAs was achieved using high‐performance liquid chromatography, followed by conversion to their respective methyl esters, and quantification by gas liquid chromatography. The unadjusted baseline saturated (SFA; palmitic acid [A]; stearic acid [B]), monosaturated (MUFA; oleic acid [C]), and polyunsaturated (PUFA; linoleic acid [D]; arachidonic acid [E]) FA composition is shown in control (Con) rats and animals with variable degrees of hypoinsulinemia (D55 and D100). Results are also expressed as the molar percentage of each FA over the total FA measured in the plasma (F). Data are presented as mean±SEM. Significantly different from control, **P<0.01, ***P<0.001. Significantly different from the D55 group, ^## P<0.01, ^### P<0.001.

To determine the capacity of the heart to respond to excess LPL or AT‐derived FA, we compared the ventricle transcriptome of D55 (Table S2) and D100 animals (Table S3). Figure 3A (lower panel) illustrates that in D55, there were 49 differentially regulated genes (padj <0.05 and significant in at least 5 out of the 10 analysis pipelines used), and when clustered according to function and ranked based on the false discovery rate, they were enriched largely for traditional glucose, protein, and lipid metabolic processes (Figure 3B). Strikingly, in D100, there were dramatic transcriptomic changes with 1574 genes differentially expressed (Figure 3A, upper panel). Like D55, the majority of genes were annotated as related to metabolic processes (with additional enrichment in genes related to cellular transport, mitochondrial and blood vessel organization, response to oxidative stress, and cell death [Figure 3A and 3B]). These results indicate that with the reduced glucose utilization, it is the superimposition of augmented AT‐derived FA that likely determines the extent of cardiac gene expression changes following diabetes mellitus.

Figure 3.

The largest magnitude of change in the ventricular transcriptome of rats with severe diabetes mellitus embraces metabolic pathway genes. Ventricle RNA from the different groups of rats was sequenced (n=4–6), and differentially expressed genes (padj <0.05 and significant in at least 5 of the 10 analysis pipelines used) were clustered according to function and ranked based on false discovery rate (FDR) (A). The insets describe the number of genes whose expression was up‐ or downregulated. The volcano plot (B) describes the profile of differentially expressed genes in D55 (inset) and D100 rats. The x axis represents Log₂ FC expression of genes vs Log₂ CPM on the y axis. Green circles highlight differentially expressed genes related to the metabolic process. CPM indicates counts per million.

Glucose and FA are the major sources from which the heart derives most of its energy. However, with the development of diabetes mellitus, the ability of this organ to utilize glucose is obstructed.5 Consistent with this finding, we observed a substantial decrease in the expression of genes controlling glucose transport (SLC2A4) and glycolysis (ENO3, HK2, PFKFB2) (Figure 4A). Furthermore, genes that modulate proteins (PDK1 and PDK4) that phosphorylate and inhibit pyruvate dehydrogenase to lower glucose oxidation, increased in expression. All changes related to genes controlling cardiac glucose utilization were more pronounced in the D100 compared with D55 (Figure 4A). Because glucose use is impaired in diabetes mellitus, the heart is obliged to primarily metabolize FA and does so by enabling its mitochondrial β‐oxidation followed by oxidative phosphorylation, to yield ATP. Related to this requirement, the genes controlling FA transport (CD36, FABP4) and mitochondrial β‐oxidation (CPT1A, HADHA, HADHB, ACADL) exhibited increased expression and more so for D100 (Figure 4B). The expression of genes encoding peroxisomal metabolism were also found to be universally elevated in the D100 heart (Figure 4C) and corresponds to the decline in VLCFA (eg, C22:5n3; C22:6n3) in this organ (Figure 4D and inset). Intriguingly, in D100, cardiac accumulation of the major polyunsaturated FA (18:2n6) and monounsaturated FA (18:1n9) exceeded that of saturated fats like palmitic (16:0) or stearic (18:0) acids (Figure 4D). This could indicate that saturated FA entering the hearts gets preferentially used, whereas monounsaturated FA and polyunsaturated FA (which require 2 additional steps to convert unsaturated to saturated bonds) get stored as triglyceride.29

Figure 4.

Metabolic gene expression reprogramming following diabetes mellitus of varying intensities emphasizes the increase in mitochondrial and peroxisomal β‐oxidation. Heat map of statistically significant genes related to cardiac metabolism of carbohydrates (A) and fatty acid (FA; B). (C) is a heat map pattern showing relative expression values of genes encoding peroxisomal metabolism. The red (high) and blue (low) colors reflect fold change (FC). Separation of heart FAs was achieved using high‐performance liquid chromatography, followed by their conversion to their respective methyl esters and quantification (μg/mg protein) by gas liquid chromatography. The FA depicted made up almost 80% of the eluted peaks and are expressed as a percentage of the total FA extracted (D). The inset in (D) describes the percentage composition of long‐chain FA (LCFA) and very‐long‐chain FA (VLCFA; chain‐length with 22 or more carbons) in hearts from the three groups (n=4–6). Data are presented as mean±SEM. Significantly different from control, *P<0.05, ***P<0.001. Significantly different from the D55 group, ^## P<0.01, ^### P<0.001.

Mitochondrial β‐oxidation of FA produces flavin adenine dinucleotide and nicotinamide adenine dinucleotide (consumed by the oxidative phosphorylation complex to produce ATP) as well as Acetyl‐CoA (that enters the tricarboxylic acid [TCA] cycle to also generate nicotinamide adenine dinucleotide and flavin adenine dinucleotide). We used Enrichr analyses of the transcriptomic data from D100 hearts and determined that of the biological processes that showed the highest enrichment, a disproportionate amount (the top 10 gene ontology terms for biological processes) was those related to mitochondrial functioning (Figure 5A), implicating a significant effect of excess FA on mitochondria. Furthermore, analysis of a protein‐protein interaction network in D100 identified functional networks enriched for components of mitochondrial ribosomal protein, respiratory complex 1, and ATP synthase (Figure 5B, pink squares). Closer inspection of the gene expression data revealed that core metabolic genes involved in making enzymes for the mitochondrial TCA cycle (Figure 5C) and oxidative phosphorylation (Figure 5D) were actually universally downregulated in D100 hearts. Unexpectedly, using a substrate limited medium (only low glucose; 2.5 mmol/L), measurement of basal OCR as a gauge of cardiomyocyte oxidative capacity indicated that myocytes from D100 hearts had the highest basal respiration (Figure 5E). This was possibly an outcome of metabolism of endogenous triglyceride‐derived FA as addition of etomoxir (carnitine palmitoyltransferase 1 inhibitor) lowered OCR in these myocytes (Figure 5E, inset). With addition of increasing concentrations of PA, although control myocytes responded by increasing their basal respiration, this effect was minimized in D55 and reversed in D100 myocytes (Figure 5E). Similarly, using FCCP to uncouple the proton gradient and maximize oxygen consumption to calculate spare respiratory capacity, control and D55 myocytes under basal conditions had the mitochondrial capacity to respond to this augmented energy demand, an outcome that was lacking in D100 (Figure 5F). FA decreased the response of myocytes from all 3 groups to FCCP‐stimulated OCR (Figure 5F).

Figure 5.

Mitochondrial oxidative phosphorylation is actively repressed with increasing severity of diabetes mellitus. Gene ontology enrichment analysis of biological processes using the Enrichr analysis tool (A). The gene signature identifying only the top 10 enriched gene ontology (GO) terms is depicted (A). Association network of genes that were significantly different between control (Con) and D100 hearts. The protein‐protein interaction network was assembled from RNAseq data. The pink squares illustrate the differentially regulated networks that are related to mitochondrial functioning. Lines represent associations based on differential expression evidence (B). Heat map pattern showing relative expression values of genes encoding tricarboxylic acid (TCA) cycle activity (C) and mitochondrial oxidative phosphorylation (D). The red (high) and blue (low) colors reflect fold change (FC). Oxygen consumption rates (OCRs) in cardiomyocytes isolated from Con, D55, and D100 hearts. Immediately before assay, either bovine serum albumin (BSA), 0.05 mmol/L of palmitic acid (PA), or 0.1 mmol/L of PA was added to the wells. Cells were exposed sequentially to the 3 metabolic inhibitors oligomycin, carbonyl cyanide 4‐(trifluoromethoxy)phenylhydrazone (FCCP) and rotenone plus antimycin A. OCR is expressed as pmole O₂/min per μg protein. Basal respiration (in BSA) was measured in the absence (E) or presence (E, inset) of etomoxir (ETO; 100 μmol/L), an inhibitor of carnitine palmitoyltransferase I (CPT1) (to assess the contribution of fatty acids arising from endogenous triglycerides. Spare respiratory capacity was calculated as OCR following FCCP minus baseline OCR (F). Three separate plates were assayed, each with cardiomyocytes isolated from 3 separate Con, D55, and D100 animals. Data are presented as mean±SEM. Significantly different compared with BSA treatment within each group, *P<0.05, **P<0.01, ***P<0.001. Significantly different compared with BSA‐Con, ^# P<0.0008.

The final products of an abnormal plasma FA profile, together with robust transcriptomic changes in the D100 heart to consume these FA, are metabolites. Using a nontargeted metabolomics approach by liquid chromatography–tandem mass spectrometry, we identified a broad set (363) of metabolites that were significantly differentially expressed in the D100 heart compared with control cardiac tissue (Table S4). We used principal component analysis (Figure 6A) and an orthogonal partial least squares discriminant analysis scores plot (Figure 6B) to identify the differences in metabolic patterns between control and D100 hearts and found clear separation between data for the 2 groups. Moreover, the corresponding S‐plot of orthogonal partial least squares discriminant analysis shows that the metabolite ions (the top 30) with the greatest influence on separation of control from D100 (ie, located furthest away from the center of the S‐plot and with a large variable important value ≥1) included increases in many types of diglycerides (ie, 18:2/18:1, 16:1/18:0) and phospholipids (ie, phosphatidylcholine [18:2/18:2, 16:0/18:2, 18:0/20:4]; phosphatidylethanolamine [18:2/21:0, 19:0/22:6, 22:6/21:0, 18:1/19:0]; phosphatidylserine [P‐20:0/18:0, P‐20:0/16:0, P‐20:0/18:1, P‐20:0/18:2, P‐20:0/22:4, O‐20:0/22:4, O‐16:0/20:0]), with the highest contribution identified as coming from triacid triglyceride (ie, 17:1/18:1/19:1, 14:0/20:2/20:2, 17:0/17:1/17:0, 12:0/20:0/22:5, 14:1/18:0/18:0, 16:1/18:0/18:3, 16:1/18:2/22:0, 18:2/19:1/19:1, 16:1/18:2/22:4, 12:0/12:0/12:0, 17:0/17:1/22:1) (Figure 6C and Table S4). These results, when added to the increased expression of (1) genes encoding acyl‐coenzyme A thioesterase (ACOT1, 2, and 4) that contributes to the conversion of acyl‐coenzyme A to FAs and coenzyme A (and thus triglyceride synthesis at the expense of FA oxidation), and (2) genes encoding glycerolipid synthesis (glycerol‐3‐phosphate acyltransferase, mitochondrial and glycerol‐3‐phosphate acyltransferase 3) (Figure 6D) implies that D100 animals with plasma lipid overload and mitochondrial dysfunction results in intracellular triglyceride accumulation (Figure 6E) and likely cell death.

Figure 6.

Substantial accumulation of lipid metabolites and triglycerides (TG) in heart tissue from D100 animals. Heart samples (5–10 mg) were powdered, transferred to a 1.7‐mL microcentrifuge tube, and 4 volumes of solvent added (acetonitrile for hydrophilic interaction or methanol for reversed phase). Following brief sonication and centrifugation at 20 000g for 10 min, the supernatant was removed to a high‐performance liquid chromatography autosampler vial. Hydrophilic interaction chromatography used a Waters BEH‐Amide column 2.1×100 mm, 1.7μm particle size, and mobile phases of 5 mmol/L ammonium formate (Fisher Optima LCMS Grade)+0.1% formic acid (Sigma) in 98% water/2% acetonitrile (EMD LCMS Grade) and acetonitrile containing 2% water and 5 mmol/L ammonium formate+ 0.1% formic acid for positive ion tests, both at pH 3.5. T‐tests were performed on the filtered data to determine statistical significance of P<0.05, and included Benjamini‐Hotchberg false discovery rate for multiple hypothesis testing correction. Principal component analysis (R2X=0.992, Q2 (cum)=0.974) (A) and orthogonal partial least squares discriminant analysis (R2X=0.635, R2Y=0.992) (B) score plots acquired by liquid chromatography–tandem mass spectrometry detailing cardiac metabolomics of control and D100 hearts. The S‐plot analysis (C, inset) and the variable important plot (C, top 30 metabolites) derived from the metabolite data illustrating the ions that contributed most to the separation of control from D100 hearts. Heat map of genes involved in glycerolipid synthesis in hearts from the different groups of rats (D). The red (high) and blue (low) colors reflect fold change. Total cardiac TG was extracted and solubilized in chloroform:methanol:acetone:hexane (4:6:1:1 v/v/v/v) and measured using high‐performance liquid chromatography (n=4–6). Data are presented as mean ±SEM. Significantly different from the D55 group, ^## P<0.01.

Electrons from nicotinamide adenine dinucleotide and flavin adenine dinucleotide enter the electron transport chain to cause buildup of the proton motive force for oxidative phosphorylation and production of ATP. With the availability of excess FA in D100, electrons are donated to molecular oxygen, with electron leakage resulting in abnormally large amounts of reactive oxygen species generation, leading to changes in gene expression related to oxidative stress (Figure 7A) and cell death (lipotoxicity). Indeed, of the genes that were differentially expressed in D100 hearts, a large number were those related to apoptosis, with a dramatic decrease in expression of antiapoptotic and an increase in proapoptotic genes (Figure 7B) together with a significant increase in terminal deoxynucleotidyl transferase dUTP nick‐end labeling–positive apoptotic cells (Figure 7C, Figure S1). Our data reveal that following severe diabetes mellitus, mitochondrial overload, incomplete FA oxidation, and oxidative stress pushes the heart toward potentially unrecoverable loss of cardiomyocytes (Figure 7D).

Figure 7.

Significant apoptotic cell death in D100 hearts. A Circos plot was used to display the association between differentially expressed genes in D100 (A) and D55 (A, inset) hearts (n=4–6). Expression level of genes involved in the pathways is indicated as a log₂ fold change (FC). The vast majority of genes are annotated as being directed towards modulation of metabolism (A) and cell survival mechanisms (B). Apoptotic cell death as identified by terminal deoxynucleotidyl transferase dUTP nick‐end labeling (TUNEL) assay (C). Isolated rat hearts were retrogradely perfused with phosphate‐buffered saline, then 10% formalin, embedded in paraffin, and 5‐μm sections prepared. Nuclei were counted from 3 different ventricular sections per animal, from 3 individual animals per group. Data are presented as mean±SEM. Kruskal–Wallis nonparametric test, Dunn post hoc. Significantly different from the control, ***P<0.001. The summary diagram (D) illustrates that following moderate hypoinsulinemia and hyperglycemia, when plasma fatty acids (FAs) have yet to increase, lipoprotein lipase (LPL) is “switched on,” and a robust expansion of coronary LPL follows. In animals with marked hypoinsulinemia and severe diabetes mellitus, there is a decline in vascular LPL. As these animals exhibit elevated plasma FA, we concluded that LPL‐mediated FA delivery would be redundant in these circumstances and is “turned off.” This is a questionable adaptation as the site of control of FA provision to the heart changes from a measured regulated delivery by coronary LPL to the unrestrained and unregulated provision by adipose tissue. Because of this metabolic reprogramming, FA supply exceeds the mitochondrial capacity of cardiomyocytes resulting in incomplete mitochondrial oxidation, accumulation of toxic FA metabolites, flux of excess unoxidized FA to triglycerides inducing a gene expression program supporting cell death (lipotoxicity).

Discussion

Cardiovascular disease is a leading cause of death in patients with diabetes mellitus.1 Although atherosclerosis is a primary cause for this cardiovascular disease, patients and animal models with type 1 and type 2 diabetes mellitus have also been diagnosed with heart failure in the absence of vascular defects (cardiomyopathy), with alterations in cardiac substrate metabolism contributing to this etiology.5 We used 2 doses (55 and 100 mg/kg) of STZ to differentially lower insulin and produce models of diabetes mellitus that imitate a continuum of glycemic management in patients, from insufficient (D55) to poor (D100) control. Our data suggest that with an increasing severity of diabetes mellitus, when the diabetic heart is unable to control its own FA supply using LPL, it undergoes dramatic reprogramming that is linked to handling of excess FA that arise from AT. This transition results in a cardiac metabolic signature that embraces mitochondrial FA overload, oxidative stress, triglyceride storage, and cell death.

Using a single injection of 55 mg/kg of STZ, we produced a model of moderate diabetes mellitus that we describe as imitating insufficient glycemic management in patients with type 1 diabetes mellitus where multiple finger pricks and daily insulin injections (3–4/day) may mean variable patient compliance and repeated exposure to bouts of hyperglycemia. With this dose of STZ, the insulin reduction was insufficient to increase circulating albumin‐bound FA or TG. As a result, production of energy in the diabetic heart is appropriated by LPL; there is augmented processing of this enzyme into an active dimeric form followed by its recruitment to the coronary vascular lumen.8 Out here, it breaks down the triglyceride core of circulating lipoproteins to release FA, representing an immediate compensatory response to guarantee FA supply when glucose cannot be used.15, 16 Intriguingly, 100 mg/kg of STZ produced animals displaying striking hypoinsulinemia, hyperglycemia, and an augmented pool of circulating plasma FA and triglyceride. We label these animals as being severely diabetic, representing patients with type 1 diabetes mellitus with poor glycemic control. In these animals, the unfettered hydrolysis of AT triglyceride to generate FA results in the diabetic heart using this pool of FA disproportionally. Consequently, to avoid lipid oversupply, there is conversion of LPL to inactive monomers, a reduction of enzyme at the coronary lumen, and inhibition of lipoprotein triglyceride hydrolysis.15 Our results imply that the greater the loss of glycemic control in type 1 diabetes mellitus, delivery of FA to the heart shifts from lipoprotein‐triglyceride hydrolysis by coronary LPL to AT‐derived FA. This is a questionable adaptation, as the site of control for FA provision to the heart changes from a measured regulated delivery by coronary LPL (which is then turned off to avoid further lipid overload) to unrestrained provision of this substrate from AT.

Therapeutic management of blood glucose and its monitoring are the foundational basis of diabetes mellitus treatment.30 Regrettably, recommended glycemic goals are infrequently achieved, and this strategy often overlooks other features that are commonly part of this complex disease such as hyperlipidemia. In this respect, insulin deficiency produces activation of lipolysis in AT that results in hydrolysis of stored triglyceride and release of large amounts of FA into the plasma. This is because the primary enzymes responsible for lipolysis in adipocytes, adipose triglyceride lipase and hormone‐sensitive lipase, are highly sensitive to inhibition by insulin.31 Unexpectedly, a reduction of plasma insulin by almost 50% was insufficient to stimulate lipolysis. Contrasting with these data, it was only when insulin fell by >80% that an increase in plasma FAs was observed. It is important to note that the percentage composition of the main FAs identified in D100 plasma remained unchanged and mirrored the composition of the dietary FA ingested. Following their release, FAs enter into the portal circulation and thus the liver (where they induce biogenesis of very‐low‐density lipoprotein), or are delivered to the heart as a complex with albumin10 and transferred across the plasma membrane to serve as an energy source. Given that the isolated heart exposed to equivalent concentrations of FA oxidize saturated FA, monounsaturated FA, and polyunsaturated FA at similar rates,32 our results indicate that the augmented FA in D100 would be nonproductive, as the supply of FA might exceed the flexibility of the heart to use this excess amount of substrate.

With uninterrupted contraction being a feature of the heart, cardiac muscle has a high demand for energy. As such, this organ demonstrates substrate promiscuity, enabling it to use multiple sources of energy, including FA, carbohydrates, amino acids, and ketones.32 Among these, carbohydrates and FA are the major participants from which the heart derives most of its energy, with FA producing almost 2.4‐fold more ATP than glucose. Accordingly, in a basal setting, glucose and lactate contribute to ≈30% of ATP generation, with FA oxidation accounting for the remaining 70%.32 Insulin plays an important role in glucose uptake and oxidation and suppression of FA oxidation in the heart.33 It was not surprising, then, that its reduction in D55 animals was characterized by cardiac gene expression changes that emphasized defects in glucose metabolism and an enrichment in FA utilization. However, in hearts from D100 animals, it was noticeable how much more pronounced this change in the metabolism gene profile was, related not only to altered substrate utilization but also to the wide spectrum of associated downstream consequences including mitochondria organization, oxidative stress, and cell death. Whether this effect is a consequence of the high amounts of plasma FA34 or the catastrophic loss of insulin in D100, or both, is currently unknown and can be assessed only if these results are compared with animals, where plasma FAs are increased without affecting insulin (using lipid‐ and heparin‐induced elevation of FA).34

The first degradative step in the cardiac utilization of FA involves its β‐oxidation in the mitochondria. However, another cytoplasmic organelle, peroxisomes, are also capable of β‐oxidation of FA (especially VLCFA), with 1 key difference. Within peroxisomes, VLCFA only undergo β‐oxidation for chain shortening and are incapable of being fully oxidized to CO₂ and H₂O to produce ATP35; the shorter chain moieties generated in the peroxisomes are in fact transferred to the mitochondria for additional β‐oxidation and subsequent ATP generation. When systemic hyperglycemia was superimposed with dyslipidemia, there were more intense gene expression changes in the D100 heart that direct this organ to use FA by activating both the peroxisomal and mitochondrial β‐oxidative pathways. One unappealing outcome of β‐oxidation of FA in these organelles is the production of chemically reactive ROS36, 37 and, when combined with the reduction of antioxidant enzyme genes, causes oxidative stress.38 Another drawback is that it results in the decrease of VLCFA like docosahexaenoic acid (22:6n3) and hence loss of the antioxidative and anti‐inflammatory effects of this omega‐3 FA.39 Thus, in attempting to handle excess FA through stimulation of β‐oxidation in peroxisomes and mitochondria, the D100 heart is exposed to oxidative damage and ultimately cellular demise.

The decrease in VLCFA may not be completely attributable to augmented oxidation but may also include a pathway to production of eicosanoids involved in an anti‐inflammatory response. However, an argument against this view is that D100 hearts had an enrichment of genes annotated as related to a response to oxidative stress, and decreased coenzyme Q9. Furthermore, linoleic (LA, C18:2n‐6) and oleic (OA, C18:1n‐9) acid are increased in both the heart and plasma FA in D100 rats. The accumulation of LA in metabolic tissues has specifically been proposed to induce oxidative stress through the generation of multiple oxidized LA compounds.40, 41, 42 Interestingly, some of them have cardiospecific toxicities, which can range from mishandling of calcium to inflammation.43 Such findings are also being increasingly reported in humans.44, 45, 46 Additionally, the upregulation of OA in cardiac tissues is known to generate triglyceride,47, 48 as we see in our D100 model. While this storage of triglyceride may be adaptive, over time, such accumulation of triglyceride would cause cardiomyopathy, as has also been shown clinically.49, 50, 51 We also emphasize that both OA and LA require additional steps by isomerases and reductases before they can be completely β‐oxidized compared with PA.52, 53 Thus, PA without double bonds can undergo β‐oxidation uninterrupted, is a better β‐oxidation substrate than LA and OA, and gets used faster (its levels are lower in both diabetic groups). Finally, reductases use nicotinamide adenine dinucleotide phosphate as the reducing equivalent donor.54 The preference of the D100 heart to underutilize OA or LA (hence leading to their accumulation) could be an adaptive response to preserve nicotinamide adenine dinucleotide phosphate levels within myocytes to prevent further oxidative stress.

Following β‐oxidation, FA catabolism resumes with sequential processing in the mitochondrial TCA cycle and oxidative phosphorylation system to yield ATP.55 Contrary to prediction, the augmented β‐oxidation was not matched by elevated TCA cycle activity or enhanced oxidative phosphorylation in D100 hearts. In fact, assessment of genes involved in mitochondrial metabolism revealed a critical decline in transcript levels of genes involved in these 2 pathways. In support of these observations, high‐fat feeding is also known to increase β‐oxidation but reduce TCA cycle intermediates in skeletal muscle of obese rodents.56 As a consequence, in our study, myocytes from D100 hearts were unable to increase their respiration in response to increasing concentrations of FA under both basal conditions or when energy demand is increased. This suggests that when FA supply exceeds the mitochondrial capacity for disposal of this substrate (mitochondrial overload), incomplete oxidation of FA and ATP deficiency is an expected outcome.56 The subsequent increase of FA intermediates and their diversion to triglyceride, paired with increased ROS formation secondary to excessive β‐oxidation, can provoke a gene expression program supporting cell death (lipotoxicity). Indeed, a large number of the differentially expressed genes in D100 hearts were those involved in apoptosis. Thus, even though a recent study has suggested that hyperglycemia/hypoinsulinemia alone, without dyslipidemia, is sufficient to impair cardiac function,57 data sets from this study indicate that with the increasing severity of diabetes mellitus, the sizeable increase in FA accounts for larger (compared with hyperglycemia alone) transcriptomic, metabolomic, and functional consequences and cardiomyocyte cell death.

In summary, our data establish the value of STZ models of diabetes mellitus with different degrees of hypoinsulinemia that imitate a continuum of glycemic management from insufficient (D55) to poor (D100) control. Globally, our data suggest that the differential reduction of insulin produced a dramatic change in cardiac gene expression largely related to substrate metabolism where glucose utilization is minimized and the heart adapts to use the robust increase in circulating FA. Unfortunately, this adaptation is nonproductive, as the supply of FA (especially monounsaturated FA and polyunsaturated FA) exceeds the flexibility of the heart to use these excess FAs resulting in triglyceride storage, oxidative stress, and cell death (lipotoxicity). These results may help guide the clinical management of type 1 diabetes mellitus, where in addition to focusing on drugs that lower plasma glucose, attention should also be placed on dyslipidemia, which could be a major link between diabetes mellitus and associated cardiomyopathy.

Limitations

The current study used a rodent model of STZ diabetes mellitus. It should be noted that most data from studies of rodent models of diabetes mellitus show minimal urinary glucose excretion when plasma glucose concentration is <22.2 mmol/L. This is a much higher capacity than what is typically observed in humans, where the renal threshold is 10 to 11 mmol/L.58, 59, 60 Our model of diabetes mellitus in the current study is one of acute loss of insulin and its associated metabolomic, transcriptomic, and lipidomic changes. However, in a separate ongoing study, D55 animals were followed for 6 weeks. The initial hyperglycemia causes a glucose‐induced pancreatic toxicity resulting in a further reduction in plasma insulin. This is accompanied by a modest increase in plasma FAs and triglycerides in D55 animals that is observed only in the long‐term. In these chronically diabetic D55 animals, the sustained increase in LPL16 together with the modest increase in adipose tissue lipolysis results in metabolomic changes that more closely resemble what was seen in 4‐day D100 hearts in the current study (Rodrigues Lab, unpublished data, 2019). It should be noted that the increase in cardiac LPL in acute D55 was unable to induce gene expression changes, oxidative stress, triglyceride accumulation, and cell death to the same magnitude as D100 hearts exposed to a robust enlargement in circulating plasma FA. This does not imply that augmentation of cardiac LPL, especially chronically, is without health risks. Cardiac‐specific LPL overexpression in mice (likely to a level much higher than that seen in D55 hearts) causes severe myopathy characterized by lipid oversupply and deposition, muscle fiber degeneration, excessive dilatation, and impaired ventricular function in the absence of vascular defects, a situation comparable to diabetic cardiomyopathy.61, 62 Additionally, in chronically diabetic D55 animals, the sustained reduction in insulin, modest increase in plasma FA, and expansion of cardiac LPL led to a depressed cardiac function.63

Sources of Funding

This work was supported by operating grants from the Canadian Institutes of Health Research (CIHR‐MOP‐133547) and the Heart and Stroke Foundation of Canada (G‐16‐00014536).

Disclosures

None.

Supporting information

Figure S1. Terminal deoxynucleotidyl transferase dUTP nick‐end labeling (TUNEL) assay to identify cardiomyocyte apoptosis. Apoptotic cell death was measured via TUNEL assay. Isolated rat hearts were retrogradely perfused with phosphate‐buffered saline, then 10% formalin, embedded in paraffin, and 5‐μm sections prepared. Nuclei were counted from 3 different ventricular sections per animal, from 3 individual animals per group and were quantified in Figure 7C. Sections were visualized by fluorescence microscopy where 4′,6‐diamidino‐2‐phenylindole–stained nuclei are blue, and fluorescein‐12‐dUTP incorporation resulted in localized green fluorescence only within the nuclei of apoptotic cells. CON indicates control; D55, 55 mg/kg streptozotocin‐induced diabetes mellitus; D100, 100 mg/kg streptozotocin‐induced diabetes mellitus.

Table S1. Micromolar FA in Plasma

Table S2. RNAseq D55

Table S3. RNAseq D100

Table S4. Metabolites D100

(J Am Heart Assoc. 2019;8:e014022 DOI: 10.1161/JAHA.119.014022.)