Mutation specific impairment of fatty acid and pyruvate oxidation in hypertrophic cardiomyopathy mouse models

Experimental animals:

Transgenic male C57Bl/6 mice expressing the R403Q mutation in Myh6 (obtained from Dr. Leinwand) [21], or the R92W mutation in Tnnt2 (obtained from Dr. Tardiff) [22] were bred with C57Bl/6N females [23–25] (Charles River). Mice were weaned and genotyped at 4 weeks of age by PCR, using tail DNA. All studies were performed in male mice at 14–16 weeks of age, which corresponds to the established disease stage [21, 22, 26]. Using heart weight after extraction and body weight, the heart-weight to body-weight ratio was calculated. Mice underwent echocardiography at UCSF, prior to transfer to the University of Florida for perfusion studies (Fig. 1A).

Echocardiography:

Echocardiographic imaging was performed using a Vevo 3100 platform (VisualSonics, Toronto, Canada) with a MX550D probe (40 MHz). Mutant (R403Q-MyHC, R92W-TnT) and littermate control mice were anesthetized using inhaled isoflurane (induction at 3 v/v% and maintenance at 1% v/v%). Parasternal long-axis (PLSAX) and short-axis (SAX) images were acquired in B- and M-mode. For assessment of diastolic function, pulse wave Doppler of mitral valve inflow, tissue Doppler imaging (TDI), and left atrial (LA) area were recorded in an apical 4-chamber view. Since isoflurane is known to depress LV systolic function [27], anesthesia was adjusted to ensure heart rates above 450 bpm during measurement of LV systolic function[28]. Image loops were analyzed using Vevo Image Lab software (version 5.6.0, VisualSonics, Toronto, Canada).

Heart perfusions:

All experiments involving mice were approved by the IACUC at the University of Florida and the University of California San Francisco. We studied two mouse models of hypertrophic cardiomyopathy, expressing the R403Q mutation in alpha myosin heavy chain (MyHC) [21]and R92W mutation in cardiac Troponin-T (TnT) [29]. Heart excision and perfusion was performed between 12 noon and 2 pm, using previously published methods [30, 31]. Briefly, mice were anesthetized using isoflurane and euthanized by cervical dislocation. Hearts were rapidly excised and transferred to ice-cold saline to arrest beating. Fat deposits and connective tissue was excised prior to cannulation of the aorta. Cannulated hearts were perfused in Langendorff mode under constant pressure (100 cm H₂O) for 30 minutes, followed by immediate freeze-clamping using liquid nitrogen. Heart rate was monitored using a pressure transducer connected to a Powerlab system (AD Instruments, Colorado Springs, CO, United States). Myocardial oxygen consumption was measured every fifteen minutes using an Oxygraph+ setup (Hansatech Instruments, United Kingdom). The O₂ consumptions were subsequently normalized to the heart wet weight. The perfusate contained Krebs-Henseleit electrolytes (25 mM NaHCO₃, 112 mM NaCl, 4.7 mM KCl, 1.2 mM each of MgSO₄ and KH₂PO₄, 0.5 mM sodium-EDTA, 2.5 mM CaCl₂), and 3 metabolic substrates (5.6 mM [1,6-¹³C₂] glucose, 0.17 mM [1, 3–¹³C₂] sodium salt of 3–hydroxybutyrate (β–hydroxybutyrate; BHB), 0.63 mM [U-¹³C] mixed fatty acids (Cambridge Isotope Labs) (FFA) suspended in 2% (w/v) bovine serum albumin). Insulin (1 μU/ml) was added to promote glucose utilization by the perfused heart.

Reagents:

Reagents, unless otherwise stated, were purchased from Thermo Fisher Scientific (Waltham, MA, United States). Isotopically enriched compounds, unless otherwise stated, were purchased from Cambridge Isotope Laboratories (MA, United States). The algol-derived [U-¹³C]fatty acids vary in composition by lot; they are primarily composed of palmitic acid (35–55%), oleic acid (20–35%), linoleic acid (10–20%), and palmitoleic acid (10–15%).

³¹P NMR spectroscopy:

The perfusion column was located in the bore of an NMR magnet to record NMR spectra. Spectra collected during perfusion were measured using a 10 mm QNP cryoprobe (Bruker Biospin, Billerca, MA, United States) at 310 K. Shimming was accomplished using the ²³Na signal until a linewidth of ~12 Hz or lower at half-height was achieved. ³¹P spectra were recorded using a 90° excitation pulse, an acquisition time of 0.84 s, a spectral width of 80 ppm (16384 points), 64 scans, and a pre-scan delay of 1 s.

Extraction of metabolites:

Frozen hearts were homogenized in a FastPrep 24 homogenizer (MP Biomedicals, Irvine, CA, United States) using 3:3:2 volume basis mixture of acetonitrile, isopropanol, water. Homogenization was performed using 30 s pulses followed by a 5-minute resting period on ice. This cycle was repeated five times. Homogenized tissue was centrifuged at 10,000 g at 4°C for 1 hour to remove tissue debris, and the supernatant was lyophilized overnight. Lyophilized powder was resuspended in a 1:1 mixture of acetonitrile and water, centrifuged at 10,000 g at 4°C for 1 hour; pellet was discarded and supernatant was lyophilized. The dry powder was stored at −20°C.

Samples for measuring ¹³C NMR spectra were prepared by dissolving the lyophilized powder in D₂O containing 50 mM sodium phosphate (pH 7.0) and 2 mM EDTA. A 10% (v/v) standard solution containing 5 mM 3-(Trimethylsilyl)-1-propanesulfonic acid-d6 (d6-DSS) and 0.2% (w/v) sodium azide was added (final concentrations: 0.5 mM DSS, 0.02% NaN₃). The solution was centrifuged at 10,000 g at 4°C for 15 min and the pellet was discarded. Supernatant (~90% of the initial volume) was transferred to a 1.5 mm NMR tube (New Era Enterprises, Newfield, NJ, United States). Total sample volume in the NMR tube was ~55 μl.

NMR spectra of extracts:

¹³C spectra were acquired using a 14.1 T magnet (600 MHz ¹H frequency) running an NMR spectrometer (Agilent Technologies, Santa Clara, CA, United states) equipped with a 1.5 mm home-built high temperature superconducting (HTS) probe [32]. Spectra were acquired using a 45° radiofrequency pulse with an acquisition time of 1.5 s, pre-scan delay of 1.5 s, spectral width of 242 ppm (55147 complex points), and ¹H decoupling (decoupling field strength of 4800 Hz; WALTZ-16). ¹H spectra were acquired on the same probe with a spectral width of 12 ppm (16384 complex points), and an acquisition time of 2.28 s.

Data Analysis:

All NMR spectra were processed using MestReNova (v14.1 or newer). ³¹P and ¹³C spectra were zero-filled to 131072 points and exponential line broadening of 15 Hz and 0.5 Hz respectively was applied. ¹H spectra were zero-filled to 32,768 points and 0.5 Hz exponential line broadening was applied. A polynomial baseline correction was applied as required. Peak areas were obtained by fitting mixed Gaussian-Lorentzian shapes to resonances of interest.

CAC flux was calculated by fitting NMR peak areas to a metabolic model using tcaCALC [33]. The model used to fit NMR spectra included parameters representing ¹³C enrichments of acetyl-CoA generated from fatty acids and ketones, pyruvate, and fluxes through pyruvate dehydrogenase (PDH), pyruvate carboxylase (Y_PC), and succinate anaplerosis (Y_S). Pyruvate cycling, i.e. malic enzyme flux from malate to pyruvate, was not included in the modeling since no evidence of lactate isotopomers other than those produced from the [1,6-¹³C₂]glucose was observed (Fig. S5,6). Alanine enrichment was estimated from the ¹³C satellites in ¹H NMR spectra and used as a fit parameter defining enrichment of substrate supplying PDH. Relative fluxes (fluxes normalized assuming CAC flux as unity) were converted to absolute values using median MVO₂ values and normalization to the individual wet heart weights [17, 34]:

where O₂ is myocardial oxygen consumption μmol/(g×min), FR represents the fractional contributions of glucose, FA, or BHB to acetyl-CoA production. R is the normalization value for each substrate based on the number of NADH/FADH₂ reducing equivalents produced by complete metabolism of each substrate [17]. Pyruvate carboxylation is assigned an R value of 1, and succinate anaplerosis is assigned an R-value of 0.5. The variables Y_PC and Y_S refer to the relative flux through pyruvate carboxylase and succinyl-CoA normalized to total CAC turnover. Y_PC and Y_S contribute to production of reducing equivalents, but flux through these pathways does not yield labeled acetyl-CoA, and their values can easily exceed CAC flux in general [33, 35]. FR_endo is oxidation of unlabeled substrates. The absolute CAC flux (V_CAC in equation 1) is calculated from the measured fractional values and the R-values. The output of TcaCALC is normalized to a citrate synthase turnover assigned as 1. With inclusion of the rates of O₂ consumption and the R values for each substrate, the absolute rates of CAC flux associated with each substrate are calculated. Analysis via tcaCALC yields the fraction of carbohydrate oxidation versus other sources of acetyl-CoA. With the estimate of labeled pyruvate determined from alanine by ¹H NMR, by difference the remainder of unlabeled pyruvate is calculated. The sum of these two quantities is the carbohydrate-derived fraction. The remaining sources of acetyl-CoA, the non-carbohydrate fraction, are then assessed from the isotopomers derived from the ketone and fatty acid labels, with an endogenous unlabeled fraction bringing the total CAC flux to 1 (Table S1). Goodness of fit is determined by the F-values calculated by tcaCALC after 100 Monte Carlo simulations which add noise to the data and perform independent fits. Typical F-values exceeded 200. The program also generates confidence intervals based on the Monte Carlo simulations. All estimates of uncertainty cited here reflect the biological replicates completed for the experimental groups; the confidence limits generated by tcaCALC for each individual heart were always significantly smaller than the biological uncertainties (Table S1). Therefore we did not perform propagation of errors. Examples of the fitted isotopomer data are included in the supporting information (Fig. S3,4). In general, tcaCALC utilizes the appearance of singlet signals as part of the modeling. The fractional enrichments achieved make natural abundance corrections unnecessary except in the case of the C5 of glutamate, where the singlet reports the natural abundance contribution and the oxidation of BHB. For the C5 of glutamate, the doublet (C5D) component reports on the fractional oxidation of the [U-¹³C]fatty acids. This same value is reported directly from the C4 quartet (Fig. 2). With this complimentary data, the relative enrichment from BHB oxidation and the natural abundance contribution to C5 can be determined.

Statistical Analysis:

Data from each mutant (MyHC-mut, TnT-mut) was compared to respective littermate control data (MyHC-CTRL, TnT-CTRL). A Welch’s t-test was used to analyze echocardiography data and the two-tailed Student’s T-test was used to analyze metabolic data; p<0.05 was considered statistically significant. Pie charts of fractional substrate oxidation were produced in Microsoft Excel. Analysis of absolute oxidative rates of the labeled substrates was completed in GraphPad Prism 11 (GraphPad Software) using an unpaired, two-tailed t-test. As the number of variables was small, no adjustments to the p-values were made to account for sample size. Due to the high variability in O₂ consumptions for the hearts, median values were used as normalization for the tcaCALC derived fluxes.

Mouse physiological phenotyping:

MyHC mutant hearts were larger (Fig. 1B) which was reflected by higher heart-weight to body-weight ratio compared to controls (Fig. 1C). Echocardiography confirmed higher LV mass in MyHC mutants and bi-atrial enlargement in TnT mutants, compared to respective controls (Fig. 1B, D–E). Fractional shortening, which reflects myocardial contractility was higher in TnT mutant (compared to controls), but similar in MyHC mutant and controls (Fig. 1F). Taken together, our data indicates mutation-specific differences in cardiac phenotype, with MyHC mutants demonstrating higher LV mass, and TnT mutants demonstrating diastolic dysfunction, LA enlargement and higher fractional shortening, when compared to respective controls.

Metabolic flux measurement:

Hearts were perfused with a mixture of three ¹³C-labeled substrates (glucose, FA, BHB). Heart rates were very consistent across all groups, but left ventricular developed pressures were lower in both the TnT and MyHC mutants compared to respective controls (TnT-CTRL and MyHC-CTRL, Fig. S1). Cardiac substrate selection was inferred by analyzing glutamate isotopomer distributions assessed by ¹³C NMR of the extracts of the freeze-clamped hearts (Fig. 2). The ratio of doublet to singlet resonances in glutamate C5 provides a direct, quantitative assessment of the fraction of ketone versus fatty acid oxidation. Acetyl-CoA derived from [1,3-¹³C₂]β-hydroxybutyrate produces an increase in intensity of singlet resonance (C5S) in glutamate C5, as opposed to the doublet (C5D), which can only arise from FA oxidation. Glutamate C3 and C4 resonances encode oxidation of both glucose and FA, as well as anaplerotic flux. TcaCALC fitting of the relative intensities of the C3, C4, and C5 multiplets produce estimates of the relative contribution of each of the three labeled substrates to the formation of acetyl-CoA (Fig. 2). Analysis using tcaCALC produces flux estimates at multiple levels, including carbohydrate versus non-carbohydrate metabolism, with subsequent partitioning into endogenous versus exogenous sources as well as fat versus ketone oxidation if the ¹³C labeling conditions provide sufficient information to deconvolute the various contributions. For the MyHC mice, total carbohydrate oxidation trended higher but was not significantly elevated (Fig. 3A). In control hearts, labeled glucose contributed to ~16% of the acetyl-CoA, with a smaller contribution from endogenous sources (Fig. 3B) which were probably composed of residual intracellular glucose and lactate not labeled by the exogenous labeled glucose. A large proportion of the non-carbohydrate contributions arose from unlabeled sources, with similar contributions from the labeled fats and BHB (Fig. 3B, left panel). For hearts from MyHC mutant mice, the increase in carbohydrate oxidation came primarily from unlabeled sources, with similar fractions of fatty acid and ketone oxidation to the control hearts (Fig. 3C). These proportional rates were normalized to O₂ consumption to produce quantitative estimates of the metabolic flux using equation (1). Total CAC turnover was dramatically depressed in the MyHC mutant hearts (1.24 ± 0.1 vs 0.39 ± 0.05 μmol/(g×min), being driven primarily by changes in endogenous fat oxidation, but also by significant changes for all the exogenous, labeled metabolites (Fig. 3D). Flux through pyruvate dehydrogenase (PDH) is a separate parameter in the tcaCALC model that is directly estimated by the intensity of the C4D34 doublet. PDH flux was ~4-fold higher in MyHC controls when compared to mutant hearts (0.21 μmol/(min*g) in MyHC vs 0.05 μmol/(min*g). Exogenous fatty acid, 0.17 μmol/(min*g) in MyHC-CTRL vs 0.04 μmol/(min*g) in MyHC-mut, and ketone metabolism, 0.15 μmol/(min*g) in MyHC-CTRL vs 0.04 μmol/(min*g) MyHC-mut, was similarly downregulated. The MyHC mutant phenotype suggests compromised metabolic output irrespective of the available fuel. Anaplerosis, or influx of substrate into the CAC via either pyruvate carboxylase or malic enzyme (both of which would comprise Y_PC), or Y_S was not changed between groups.

Fuel selection in the TnT mutant hearts was very different from the MyHC mutant. The fractional contribution of carbohydrate versus non-carbohydrate sources of acetyl-CoA were not different between the TnT-CTRL and mutant hearts (Fig. 4A). Subdivision of these sources showed relatively small changes in the distribution of substrate utilization (Fig. 4B–C). Myocardial oxygen consumption rate, was similar between TnT mutants and respective controls. CAC flux was significantly decreased in the TnT-mut, but this was driven exclusively by a loss of endogenous fat oxidation, 0.19 μmol/(min*g) in TnT-CTRL vs 0.12 μmol/(min*g) in TnT-mut (Fig. 4D). Exogenous glucose, fatty acids, and BHB showed no changes in utilization (Fig. 4D). In this case, Y_PC was not different between groups, but Y_S was significantly decreased, 0.27 μmol/(min*g) in TnT-CTRL vs 0.11 μmol/(min*g) in TnT-mut.

To investigate differences high energy phosphates in heart tissue, we measured ³¹P spectra of functioning hearts during perfusion (Fig. 5A). The PCr/ATP ratio calculated using peak areas from ³¹P spectra ([γ-P] of ATP; Fig. 5B) was ~1.5-fold higher in MyHC-CTRL when compared to the mutant (p = 0.04), but similar in TnT-CTRL versus the mutant. Decreases in PCr/ATP ratio reflect a loss of PCr, the short term reserve for replenishment of ATP, in MyHC mutants, which is consistent with results of ³¹P MRS imaging in HCM patients expressing the R403Q-MyHC mutation [12].

Substrate utilization:

In normal hearts, FA oxidation is dominant under baseline conditions [43]. Changes in substrate utilization away from FA oxidation to increased reliance on glycolysis is common at the onset of cardiac hypertrophy, with a subsequent decrease in oxidative capacity as the heart fails [2, 40, 44]. Increased myocardial glucose uptake has been demonstrated by ¹⁸FDG-PET imaging [40, 45] in HCM patients. In our study, total glucose utilization trended lower in the MyHC mutant heart, but absolute flux through PDH as measured using [1,6-¹³C₂]glucose was significantly decreased (Fig. 3D). TcaCALC infers glucose oxidation from both PDH flux and from the enrichment of the lactate/alanine precursor pools. By difference, the fraction of unlabeled carbohydrate oxidized by the heart is also estimated. Presumably, the unlabeled carbohydrate could originate from the limited glycogen stores in the heart or residual substrate remaining in the cytosol after excision. Lower PDH flux is indicative of impaired glucose oxidation but does not require changes in glycolysis. Endogenous (unlabeled) fatty acid oxidation contributed to the greatest drop as a fraction of total CAC flux. Based on the noted changes in the metabolic genes associated above, the overall decrease in CAC turnover is feasibly explained by the loss of the machinery (proteins) that facilitate the Krebs cycle. Note that O₂ consumptions are reported as per gram wet weight. As the heart is 80–85% water, normalization to the dry weight would increase rates to greater than 10 μmols/min∙gdw, in line with other examples of Langendorff perfused mouse hearts [20, 46].

Prior studies in mouse models have linked decreases in PDH activity and glucose oxidation with development of left ventricular hypertrophy (LVH) [47–49]. In our study, significant reduction in PDH flux and a trend towards decreased glucose oxidation without changes in anaplerotic flux was observed in MyHC mutant mice that have higher LV mass than the control, but not in TnT mutants that have similar LV mass as the respective control. Decreases in FA oxidation observed in MyHC mutant mouse hearts was also seen in human myectomy tissue [13–15]. It is unclear whether this is deleterious or represents an adaptive response, because while FAs produce the greatest ATP yield per 2 carbon unit of all energy substrates, they also have the highest O₂ consumption to produce ATP [50, 51].

Oxidative phosphorylation and phosphotransfer:

In the normal heart, the majority (~95%) of myocardial ATP requirements are provided by oxidative phosphorylation (OxPhos)[50, 51], with glycolysis providing the remaining 5%. Myocytes are equipped with phospho-transfer networks composed of creatine kinase (CK), adenylate kinase (AK) and glycolytic enyzmes that phosphorylate ADP to ATP at ATP-consuming sites, and deliver ADP to mitochondria[52] where it stimulates respiration. In our study, only MyHC mutants demonstrated a reduction in energy reserves, reflected by lower PCr/ATP ratios. Decreases in myocardial PCr/ATP ratio and CK flux [12] has also been reported in human ³¹P-MRS studies [8, 10, 11] and myectomy tissue. Additionally, abnormalities in mitochondrial structure, changes in expression of electron transport chain (ETC) genes and reduced respiratory capacity of ETC complexes have been described in human myectomy tissue [13–15] and HCM mouse hearts [6, 24, 25, 53]. Dysfunction of ETC complexes leads to a decrease in ATP production and increase in reactive oxygen species (ROS), which can damage mitochondria and worsen energetic deficits.

The MyHC mutant hearts in general show decreased MVO₂, CAC flux, FA oxidation, and flux through PDH. This ultimately manifests in a decreased PCr/ATP ratio, which is indicative of the canonical “engine out of fuel” that characterizes many heart failure phenotypes [43]. In contrast, the TnT hearts manifest no overt phenotype associated with energy metabolism. This is not entirely surprising, as the TnT mutation is associated with sudden death, not progressive heart failure.

Limitations:

A limitation of this study is a lack of estimate of mitochondrial “leakiness”, i.e., as estimate of the O₂ consumption not related to oxidative phosphorylation. The error bars for the MyHC mice are quite large. HCM is known to exhibit profound phenotypic heterogeneity[16]. It is likely that we are seeing phenotypic differences driven by factors more complicated than the single variable associated with the MyHC mutation. However, the significant differences shown in the MyHC mutants versus the homogeneity of the results for the TnT mutants convince us that our conclusions are robust. Phosphorous-31 saturation transfer experiments for quantitation of ATP synthase flux were not part of the study design. In the future, the ADP:O ratio could provide further insight into the nature of the MyHC failure.

This work represents mice that are maintained as a colony at UCSF, but were shipped to UF for the perfusions. Due to IACUC restrictions, the mice could only be maintained in temporary housing for a maximum of 7 days before sacrifice. Stress-related transport caused fighting among some animals, and overall, the measured O₂ consumptions were quite variable. To address the variability, we used the median values of the O₂ consumptions for each separate group to normalize the flux calculations using the relative peak ratios of the glutamate signals [17]. There is also the possibility that the O₂ electrode was not performing optimally in some instances. Several hearts recorded no O₂ consumptions, but nonetheless produced labeling from the exogenous substrates commensurate with hearts that did show O₂ consumption. All hearts used for the analysis performed at almost 400 bpm (Fig. S1). As the multiplet values for the hearts used were internally consistent, the use of median values of O₂ consumption for normalization allowed us to use more hearts for comparison. We chose to use alanine ¹³C enrichment as our proxy of pyruvate enrichment for acetyl-CoA production in accord with previous work that showed lactate provided a less suitable estimate [20].

From a modeling perspective, the use of the R-values for each substrate relies upon the assumption that mitochondrial (un)coupling does not change the relative efficiency of glucose, FFAs, or BHB for production of NADH/FADH₂ in the pathophysiology [17]. We acknowledge the weakness of the assumption, but without it a large body of metabolic research would be undermined.

The use of Y_PC and Y_S in the tcaCALC model produces notably better F-values for the fits, but both these parameters are less well-determined by the modeling than the fractional oxidation of the substrates producing acetyl-CoA. Pyruvate anaplerosis is tracked by the ¹³C label derived from the glucose, but flux via Y_S is inferred from CAC label washout. While Y_S is significantly decreased in the TnT hearts, without further experiments that provide labeling to this pathway, we do not choose to emphasize this result.

Conclusion:

In this study, we used the MyHC and TnT mouse models of HCM to study substrate selection and energy provision in the perfused heart. Only the MyHC mutant mice demonstrate a significant shift in energy-substrate metabolism characterized by decreases in FA oxidation, PDH flux and PCr/ATP ratio, all of which are associated with increase in LV mass. Using models that accurately recapitulate human disease, we identify dysregulated energy metabolism as a primary element of MyHC related cardiac remodeling. Our results suggest that treatment paradigms that actively modulate substrate metabolism in the heart should be tailored to knowledge of FA and carbohydrate utilization in each individual.