Modeling the Transition From Decompensated to Pathological Hypertrophy
Background Long‐chain acyl‐CoA synthetases (ACSL) catalyze the conversion of long‐chain fatty acids to fatty acyl‐CoAs. Cardiac‐specific ACSL1 temporal knockout at 2 months results in a shift from FA oxidation toward glycolysis that promotes mTORC1‐mediated ventricular hypertrophy. We used unbiased metabolomics and gene expression analyses to examine the early effects of genetic inactivation of fatty acid oxidation on cardiac metabolism, hypertrophy development, and function.
Methods and Results Global cardiac transcriptional analysis revealed differential expression of genes involved in cardiac metabolism, fibrosis, and hypertrophy development in Acsl1H−/− hearts 2 weeks after Acsl1 ablation. Comparison of the 2‐ and 10‐week transcriptional responses uncovered 137 genes whose expression was uniquely changed upon knockdown of cardiac ACSL1, including the distinct upregulation of fibrosis genes, a phenomenon not observed after complete ACSL1 knockout. Metabolomic analysis identified metabolites altered in hearts displaying partially reduced ACSL activity, and rapamycin treatment normalized the cardiac metabolomic fingerprint.
Conclusions Short‐term cardiac‐specific ACSL1 inactivation resulted in metabolic and transcriptional derangements distinct from those observed upon complete ACSL1 knockout, suggesting heart‐specific mTOR (mechanistic target of rapamycin) signaling that occurs during the early stages of substrate switching. The hypertrophy observed with partial Acsl1 ablation occurs in the context of normal cardiac function and is reminiscent of a physiological process, making this a useful model to study the transition from physiological to pathological hypertrophy.
What Is New?
We describe, for the first time, an intermediate phase of cardiac metabolism whereby alternative downstream uses of glucose are observed before the establishment of the strictly glycolytic state that results from long‐term inhibition of fatty acid activation.
What Are the Clinical Implications?
The hypertrophy observed in the early stages of a cardiac metabolic switch resulting from partial Acsl1 ablation occurs in the context of normal heart function and bears a transcriptional signature that is reminiscent of physiological hypertrophy, making this model particularly useful to study the precise mechanisms underlying the imbalances in cardiac remodeling that lead to pathological hypertrophy and heart failure.
Mitochondrial oxidation of long‐chain fatty acids (FAs) accounts for 60% to 90% of the energy used by the adult heart under normal physiological conditions, with glucose and lactate oxidation providing the remainder.1, 2, 3, 4 The healthy heart is metabolically flexible and can readily switch between energy substrates as dictated by substrate availability, hormonal status, and physiological environment.5 Pathological conditions can also be accompanied by alterations in cardiac substrate use; however, it remains unclear as to whether the shift in energy substrate is a cause or consequence of the pathological insult. Metabolic heart disease causes an increase in FA oxidation,6, 7, 8, 9 whereas increased glucose use is observed under conditions that induce pathological hypertrophy, myocardial ischemia, or heart failure.10 Analyses of some mouse models suggest that ventricular hypertrophy precedes the metabolic changes that result in reduced cardiac FA oxidation and increased glucose use.11, 12 Conversely, studies of ex vivo working hearts have shown that hemodynamic stress is sufficient to elicit mechanistic target of rapamycin complex‐1 (mTORC1)‐responsive increases in glucose use and cardiac remodeling, whereas perfusing working hearts with glucose results in mTORC1‐mediated hypertrophy and contractile dysfunction,13 compelling indicators that reliance on glucose may be detrimental to cardiac function.
Murine cardiac acyl‐CoA synthetase (ACSL) 1 deficiency results in decreased mitochondrial FA oxidation and suggests that a shift to glycolysis promotes the progressive development of mTORC1‐mediated hypertrophy and diastolic dysfunction.14, 15 The activation of FA via the formation of acyl‐CoA derivatives is required for their downstream utilization as energy substrates or complex lipid precursors. ACSL1 belongs to a family of 5 closely related ACSL isoforms and is highly expressed in oxidative tissues like brown adipose tissue, skeletal muscle, and heart, where it directs FA preferentially toward mitochondrial β‐oxidation.14, 16, 17 Hearts lacking ACSL1 exhibit a >90% reduction in mitochondrial FA oxidation and a 35% decrease in the uptake of the nonmetabolizable FA analog 2‐bromo‐[1‐14C]‐palmitate. In contrast, the uptake of non‐metabolizable 2‐deoxy[1‐14C]‐glucose increases 8‐fold, and pyruvate oxidation and tricarboxylic acid cycle flux also increase significantly, suggesting that ACSL1‐deficient hearts must rely almost exclusively on glucose for energy.15 These metabolic changes are accompanied by a marked increase in mTORC1 activation and the development of mTORC1‐mediated pathological hypertrophy, diastolic dysfunction, and increases in oxidative stress, suggesting that a forced substrate switch toward glucose oxidation is not entirely benign.15, 18
While cardiac hypertrophy is generally considered to be an initial, compensatory response to increased cardiac workload, chronic ventricular hypertrophy typically precedes the onset of heart failure, and the transition from physiological to maladaptive hypertrophy is still poorly understood. Previous studies have shown that cardiac function is not affected by partial ablation of Acsl1.14 Thus, the present work was designed to examine the molecular mechanisms that underlie the development of cardiac hypertrophy in the context of preserved cardiac function during the early stages of cardiac ACSL1 depletion and the resulting inhibition of FA use. We hypothesized that this incomplete ablation model would allow us to study the changes in cardiac metabolism that precede morphologic or functional changes; such changes would shed light on the role of glucose metabolism in mediating these processes. Thus, we asked whether partial Acsl1 knockdown would result in transcriptional and metabolic changes similar to those observed in ACSL1‐deficient hearts, and we performed unbiased gene array and metabolomics analyses to uncover unidentified players in the regulation of mTOR‐mediated cardiac hypertrophy.
Materials and Methods
The data that support the findings of this study are available to other researchers on request. All the materials used in this study are available commercially from the indicated vendors. Additional methods can be found in Data S1.
The University of North Carolina Institutional Animal Care and Use Committee approved all protocols. Mice were housed in a pathogen‐free barrier facility (12‐hour light/12‐hour dark cycle) with free access to water and food (Prolab RMH 3000 SP76 chow). Mice with LoxP sequences inserted on either side of exon 2 in the Acsl1 gene were backcrossed 6 times to C57BL/J6 mice and then interbred with mice in which Cre expression is driven by an α‐myosin heavy‐chain promoter induced by tamoxifen (B6.Cg‐Tg(Myh6‐cre/Esr1)1Jmk/J, Jackson Labs) to generate tamoxifen‐inducible, heart‐specific Acsl1 knockout (Acsl1H−/−) mice.14 At 6 to 8 weeks of age, Acsl1H−/− and littermate Acsl1flox/flox control male mice were injected intraperitoneally with tamoxifen (75 mg/kg BW, Sigma) dissolved in corn oil (20 mg/mL) for 4 consecutive days (3 mg/40 g of body weight). Subgroups of mice were also injected daily intraperitoneally for 2 weeks with rapamycin (Sigma, 1 mg/kg in PBS, 8% ethanol, 10% Tween 20, 10% PEG‐400) or with vehicle alone. Two weeks after tamoxifen induction, animals were anesthetized with 2,2,2‐tribromoethanol (Avertin) and tissues were removed and snap‐frozen in liquid nitrogen. To isolate total membranes, heart ventricles were homogenized with 10 up‐and‐down strokes using a motor‐driven Teflon pestle and glass mortar in ice‐cold medium I (MedI) buffer (10 mmol/L Tris [pH 7.5], 1 mmol/L EDTA, 250 mmol/L sucrose, 1 mmol/L dithiothreitol [DTT], plus Halt™ protease and phosphatase inhibitor cocktail [ThermoScientific]). Homogenates were centrifuged at 100 000g for 1 hour at 4°C; the membrane pellet was then resuspended in MedI buffer. Protein content was determined by the BCA assay (Pierce) with bovine serum albumin (BSA) as the standard. Plasma was collected from mice in 5% 0.5 mol/L EDTA. Plasma triacylglycerol, nonesterified free fatty acids, glucose (Wako), and free and total glycerol (Sigma) were measured with colorimetric assays.
Doppler analysis of the mitral valve to determine the inflow velocity was performed on lightly anesthetized mice (2% [vol/vol] isoflurane/100% oxygen) as previously described.14, 19, 20 Mitral valve flow Doppler was acquired by positioning the transducer angled cranially in a supine mouse at 45° in an epigastric position to achieve an apical 4‐chamber view. Peak E and A heights were determined on mitral valve sequential waveforms in at least 5 waveforms. The mean performance index was calculated as the isovolumetric contraction time (ICT) plus the isovolumetric relaxation time (IRT) divided by the ejection time ([ICT+IRT]/ET) to determine if either systolic or diastolic dysfunction was present.21, 22, 23, 24, 25 Doppler measurement data represent 3 to 6 averaged cardiac cycles from at least 2 scans per mouse.
ACSL‐specific activity was measured as described.14 Briefly, total membrane fractions were isolated by centrifuging homogenized tissues at 100 000g for 1 hour at 4°C. Protein (2–6 μg) was incubated with 50 μmol/L [1‐14C]‐palmitate, 10 mmol/L ATP, 250 μmol/L CoA, 5 mmol/L dithiothreitol, and 8 mmol/L MgCl2 in 175 mmol/L Tris, pH 7.4 at room temperature for 10 minutes to measure initial rates. The enzymatic reaction was stopped with 1 mL of Dole's solution (heptane:isopropanol:1 mol/L H2SO4; 80:20:1, v/v); heptane and water were added to separate the phases, and radioactivity of the acyl‐CoAs in the aqueous phase was measured in a liquid scintillation counter. No ACSL activity was measurable in the cytosolic fraction.
Global Gene Expression Analyses
Total RNA was isolated from heart ventricles (RNeasy Fibrous Tissues Kit, Qiagen). RNA integrity was verified on an Agilent BioAnalyzer 2100 (RIN ≥9). Gene expression microarrays were used to analyze changes in gene expression at either 2 or 10 weeks of Acsl1 knockdown in mouse hearts. For microarray analysis, cyanine‐5‐labeled cRNA was cohybridized to the G4122F mouse whole genome array (Agilent) with equimolar amounts of cyanine‐3‐labeled mouse reference RNA26 and scanned on a GenePix 4000B (Axon) with Feature Extraction (v220.127.116.11; Agilent). Total probes (9009) representing 7172 unique genes passed quality control (flagged as detected in ≥70% of the samples) (see Table S1). One‐way ANOVA analysis was performed to determine differentially expressed genes between genotypes, identifying 1244 probes (only 62 probes expected by chance) satisfying a corrected P value (Benjamini Hochberg false discovery rate) cutoff of 0.05. RNA sequencing (RNAseq) was used to analyze changes in gene expression after 2 weeks of Acsl1 knockdown in mouse hearts in the presence or absence of rapamycin. Library preparation of mRNA was performed using the RNA TruSeq Kit (Illumina), and 5 samples were randomly pooled together and sequenced via 50 base single‐end reads on a HiSeq 2000 sequencer (Illumina). Aligned reads were sorted and indexed using SAMtools27 and translated to transcriptome coordinates, then filtered for indels, large inserts, and zero mapping quality using UBU v1.0 (https://github.com/mozack/ubu). Transcript abundance estimates for each sample were performed using RSEM28 and raw RSEM read counts for all RNAseq samples were normalized to the overall upper quartile (see Table S2). 21 711 genes were mapped; the list was filtered to only include genes with a minimum count of 100 in all 20 samples, resulting in 9118 total genes for analysis. Three‐way ANOVA analysis using the categorical effects of genotype and drug, as well as the RNAseq lane as a random effect was performed with an absolute fold change minimum of 1.5 with a P<0.01 (Partek Genomics Suite v6.6). Complete, MIAME (Minimum Information About a Microarray Experiment)‐compliant microarray and RNAseq data sets were archived with the Gene Expression Omnibus of the National Center for Biotechnology Information and are accessible through the GEO SuperSeries accession No. GSE103375.
Total protein lysates were isolated in lysis buffer (20 mmol/L Tris base, 1% Triton X‐100, 50 mmol/L NaCl, 250 mmol/L sucrose, 50 mmol/L NaF, 5 mmol/L Na2P2O7, plus Halt™ protease and phosphatase inhibitor cocktail [Thermo Scientific]). Equal amounts of protein (100 μg) were loaded and resolved on 10% SDS polyacrylamide gels and transferred to nitrocellulose membranes. Blots were probed with antibodies against ACSL1 (catalog No. 4047), phosphorylated p70 (P‐p70) S6 kinase (S6K) (catalog No. 9234), phosphorylated 4E‐BP1 (catalog No. 2855), and phosphorylated AMP‐activated protein kinase (P‐AMPK) (catalog No. 2531), and were then stripped and reprobed with p70 S6K (catalog No. 9092), 4E‐BP1 (catalog No. 9644), or AMPKα (catalog No. 2532) antibodies, respectively (all antibodies from Cell Signaling). GAPDH (Abcam, catalog No. ab8245) was used as loading control.
Heart ventricles were fixed for 24 hours in PBS containing 4% paraformaldehyde, transferred to 70% ethanol, embedded in paraffin, serial sectioned, and stained with Masson's trichrome. Slides were scanned using an Aperio ScanScope (Aperio Technologies, Vista, CA) and fibrosis was quantified using Aperio ImageScope (v18.104.22.16813). The Positive Pixel Count v9 algorithm was used to measure staining of collagen (representing both fibrosis and collagen in extracellular matrix [ECM]), with hue value and width of 0.66 and 0.1, respectively.29 The N positive/N total value representing the percentage of collagen of the entire section was used to determine the weighted average for each slide, expressed as percentage of fibrosis.
2‐Deoxyglucose and 2‐Bromopalmitate Uptake
Anesthetized mice were injected intraperitoneally with 5 μCi 2‐deoxy[1‐14C]‐glucose in saline or retro‐orbitally with 1.5 μCi 2‐bromo‐[1‐14C]‐palmitate complexed to BSA in PBS (Moravek Biochemicals). Tissues were harvested and flash frozen in liquid nitrogen 30 minutes after injection. Radioactivity was measured in tissue homogenates by scintillation counting and normalized to the number of disintegrations per minute (DPM) present in 5 μL of serum obtained 5 minutes after injection ([dpm/mg tissue]/dpm in 5 μL serum).
Detailed methods for tissue sample preparation and analysis by gas or liquid chromatography/mass spectrometry as well as the raw, processed, and normalized data are available at the Carolina Digital Repository (https://doi.org/10.17615/c6794k). Metabolite concentrations were analyzed using Metaboanalyst (v3.0)30, 31, 32 run in the statistical package R (v3.3.2). Concentrations were log transformed and unit scaled, resulting in a Gaussian distribution of metabolite concentrations (Figure S1). The normalized data were used for principal component analysis and 2‐way ANOVA analyses to identify metabolites that associated with genotype or rapamycin treatment (see Table S3). Pearson correlation coefficients were calculated to measure the relationship between heart weight (normalized to body weight, resulting in a normal distribution) and each normalized metabolite concentration (Figure S1).
Values are expressed as the mean±standard error of the mean (SEM) as indicated. For studies that included rapamycin treatments, 2‐way ANOVA was used with the genotype and treatment identified as the sources of variation. Post‐hoc tests were performed with Tukey's multiple comparisons test. For all tests, P>0.05 was considered nonsignificant (ns). All variables used for correlation analyses (Pearson) were continuous and normally distributed, and linear relationships were confirmed via scatterplot analysis.
Partial ACSL1 Ablation in the Heart was Sufficient to Induce Cardiac Hypertrophy But Did Not Affect Cardiac Function
Analysis of Acsl1H−/− hearts 10 weeks after tamoxifen induction of cardiac‐specific ACSL1 deficiency results in >90% reduction in total ACS activity and FA oxidation.14 This markedly decreased ability to activate FA is accompanied by a compensatory increase in glucose uptake, consistent with a shift toward glycolytic oxidation, as well as the development of mTOR‐responsive hypertrophy with diastolic dysfunction.14, 15 Treatment with the mTOR inhibitor rapamycin for 10 weeks prevents the development of hypertrophy, whereas diastolic dysfunction and glucose uptake are not affected.15 To determine the initial effects of decreasing ACSL1 activity, we assessed the effects of a partial Acsl1 knockdown on the same parameters of cardiac metabolism and function 2 weeks after tamoxifen injection, when ventricular Acsl1 mRNA was reduced by 88% (Figure 1A). At this time, ACSL1 protein levels (Figure 1B) and total ACS activity (Figure 1C) had decreased by 52% and 70%, respectively, highlighting the relatively high stability of the ACSL1 protein as well as its role as the major FA‐activating enzyme in the heart. Lack of AMPK activation in Acsl1H−/− knockdown hearts (Figure 1D) suggested that the compensatory processes triggered in response to the impaired FA oxidation were sufficient to maintain adequate cellular ATP/AMP ratios. However, the partial knockdown of ACSL1 observed 2 weeks after tamoxifen induction was sufficient to elicit cardiac hypertrophy in vehicle‐treated Acsl1H−/− hearts (Figure 1E), whereas hypertrophy was not observed in hearts treated concomitantly with rapamycin, indicating that mTOR also mediates the compensatory hypertrophy observed at this early time point in the cardiac fuel switch. At this time point in the metabolic switch, phosphorylation of downstream mTORC1 targets S6K and 4E‐BP1 showed marked variation (Figure 1F), suggesting that mTORC1 activation, as measured by posttranslational modifications, occurs gradually. Doppler analysis of mitral and aortic valve blood flow did not identify differences in the mitral valve E/A ratio in Acsl1H−/− knockdown compared to control hearts 2 weeks after tamoxifen treatment, indicating that cardiac function was not affected by partial ACSL1 ablation (Figure 1G). Mean performance index calculations showed a mild rapamycin effect in both control and Acsl1 knockdown animals. Non–statistically significant functional issues were also identified but were considered to be of little relevance because no diastolic dysfunction is evident at this time point.
Knockdown of ACSL1 Resulted in Cardiac‐Specific Metabolic Derangements That Did Not Affect Systemic Metabolic Homeostasis
Acsl1 knockdown had no effect on blood glucose, whereas levels of circulating nonesterified FAs, free glycerol, and triacylglycerol showed mild, albeit non–statistically significant alterations, suggesting that cardiac‐specific knockdown of Acsl1 did not affect systemic metabolic homeostasis (Figure 2A). In contrast, the partial knockdown of ACSL1 resulted in dramatic changes in cardiac substrate uptake. The inability of Acsl1H−/− hearts to activate and therefore trap FA within cardiomyocytes resulted in a >50% reduction in 2‐bromo[1‐14C]‐palmitate uptake (Figure 2B), whereas 2‐deoxy[1‐14C]‐glucose uptake was significantly higher despite the high variability exhibited by Acsl1H−/− hearts (Figure 2C).
Short‐Term Inactivation of ACSL1 Caused Transcriptional Derangements That Were Distinct From Those Resulting From Total ACSL1 Knockout Observed 10 Weeks After Tamoxifen Induction
We analyzed global changes in cardiac gene expression after 2 and 10 weeks of Acsl1 knockdown in cardiomyocytes; control hearts included in this analysis were obtained from Acsl1flox/flox control mice injected in parallel and harvested at the same time points of 2 and 10 weeks. Correlation analyses (Figure 3A) identified 3 distinct sample groups: 10‐week Acsl1H−/− knockout, 2‐week Acsl1H−/− knockdown, and an interspersion of 10‐ and 2‐week Acsl1flox/flox control hearts. One‐way ANOVA analysis identified 1244 differentially expressed genes; all genes with a change ≥1.5‐fold in any pairwise comparison (Figure 3B) were clustered to identify patterns of gene expression among the 3 groups. The resulting heatmap shows a predominant signal for the 10‐week Acsl1H−/− knockout hearts, with distinct clusters of probes that were markedly altered after only 2 weeks of Acsl1 knockdown compared with either complete Acsl1 knockout at 10 weeks or to Acsl1flox/flox control hearts (Figure 3C).
We next looked at the expression of rapamycin‐sensitive genes previously identified in 10‐week Acsl1H−/− knockout hearts15, 18 to determine if these transcriptional responses were evident after short‐term Acsl1 inactivation (Figure 4A). Surprisingly, this collection of genes did not appear to represent an intermediate phenotype of mTOR activation. We analyzed the expression of pathological hypertrophy markers, amino acid responsive genes, metabolic genes and oxidative stress markers that had previously been shown to be regulated by mTOR15, 18 (Table 1). Expression of pathological hypertrophy markers in Acsl1H−/− knockdown hearts differed from the results observed after complete Acsl1 knockout, suggesting that the hypertrophy observed at this early time point in the metabolic switch was still compensatory. The expression of amino acid–responsive genes also remained unaffected in Acsl1 knockdown hearts, in contrast to what had been observed after complete Acsl1 ablation. Expression of genes involved in glucose and lipid metabolism showed intermediate effects compared with ACSL1 knockout hearts (Table 1), consistent with a transitional point in the cardiac substrate switch. In addition, the expression of oxidative stress markers differed markedly between the 2 models of Acsl1 deficiency.
We compared gene expression changes in the 2‐week Acsl1H−/− knockdown hearts to those of either control or 10‐week Acsl1H−/− knockout hearts to determine which genes were exclusively affected during the early stages of the metabolic fuel switch (Figure 4B). We observed a dramatic upregulation of genes involved in ECM remodeling and cell‐cell or cell‐matrix interactions, strong indicators of fibrosis. In addition, despite the observed increase in glucose uptake that results from partial ablation of Acsl1, these hearts did not exhibit a glycolytic signal (Figure 4C), suggesting that the residual ACSL activity present in Acsl1H−/− knockdown hearts may be sufficient for contractile energy production. Alternatively, the partial block in FA activation that results from incomplete Acsl1 knockdown may result in the initial use of alternative metabolic substrates before the switch to glucose is complete 10 weeks after the Acsl1 knockout is induced.14, 15
Rapamycin Treatment Identified Novel Cardiac‐Specific mTOR Targets Involved in the Initial Stages of the Cardiac Metabolic Switch
To further investigate the effects of partial Acsl1 ablation on cardiac metabolism, we used RNAseq to examine global changes in cardiac gene expression after 2 weeks of cardiomyocyte‐specific Acsl1 knockdown plus concomitant vehicle or rapamycin treatments. Of the 9118 total genes that passed quality control, 145 genes exhibited a change ≥1.5‐fold in Acsl1H−/− knockdown hearts compared with littermate controls (2‐way ANOVA analyses; see Table S2). Partial Acsl1 ablation resulted in the expected changes in lipid metabolism gene expression, with decreased FA uptake, intracellular FA transport, and triacylglycerol synthesis, as well as compensatory increases in peroxisomal and mitochondrial oxidation gene expression. Notably, no compensatory changes occurred in the expression of other activities that synthesize acyl‐CoAs (Figure 5A). Acsl1 knockdown hearts exhibited suppressed glucose transporter 1 and glucose transporter 4 expression despite activation of glucose transport regulators and the observed increase in 2‐deoxy[1‐14C]‐glucose uptake (Figure 5B), a phenomenon we also observed in 10‐week Acsl1 knockout hearts as well as in a skeletal‐muscle–specific model of ACSL1 deficiency.17, 18 RNAseq analyses once again showed shifts in glucose metabolism gene expression toward a downregulation of glycolysis and glycogen breakdown (Figure 5B). In contrast, genes involved in the pentose phosphate and hexosamine biosynthetic pathways were significantly increased (Figure 5C), suggesting that the excess glucose imported may, in fact, be diverted to these alternative routes rather than used to produce energy. Alterations in the expression of genes involved in the tricarboxylic acid cycle (Figure 5D) and in amino acid metabolism and ketogenesis (Figure 5E) also indicated a shift toward the use of alternate energy substrates during the early stages of the cardiac fuel switch. Rapamycin treatment resulted in differential effects between control and Acsl1H−/− knockdown hearts but reverted the observed changes in several genes (denoted by ▲), revealing novel mTOR targets involved in cardiac metabolism. Of note, Eno1 and Glud1 have been shown to be upstream of mTOR activation in cancer models,33, 34 whereas hepatic CD36 is known to be regulated by rapamycin but in an opposite manner to that observed in Acsl1H−/− knockdown hearts.35 Other novel cardiac targets of mTOR uncovered by this study include Pdha and Dbt, components of the pyruvate dehydrogenase and branched‐chain ketoacid dehydrogenase complexes, respectively.
Acsl1 Knockdown Results in a Transient Increase in Cardiac Fibrosis That is Unrelated to the Development of Hypertrophy
RNAseq analysis of Acsl1 knockdown hearts confirmed the upregulation of fibrosis genes observed in microarray analyses comparing 2‐ and 10‐week cardiac‐specific ACSL1 ablation (Figure 6A, Figure S3), which was further corroborated by Masson's trichrome staining (Figure 6B). Transcriptional alterations in the fibrotic program were observed in both the 10‐week cardiac‐specific Acsl1 knockout and 2‐week knockdown models; however, the induction of fibrosis genes was significant only in the short‐term model (Figure 6C). Moreover, partial ACSL1 ablation resulted in significant induction of genes involved in ECM breakdown, which could explain why complete cardiac ablation of ACSL1 does not result in increased fibrosis.14
We questioned whether the observed induction in ECM biogenesis and organization that is predominant in the short Acsl1 inactivation model is linked to the development of hypertrophy during the early stages of the cardiac substrate switch. Despite the drastic differences in the expression of pathological hypertrophy markers between the 2 ACSL1 deficiency models examined (Table 1), rapamycin treatment prevented the development of hypertrophy in this model; however, it did not reverse the Acsl1‐dependent alterations in the expression of fibrosis markers, indicating that the upregulation of fibrosis does not underlie the mTORC1‐mediated hypertrophy observed in Acsl1 knockdown hearts. In direct contrast with a total absence of cardiac ACSL1, partial Acsl1 ablation did not result in increased expression of pathological hypertrophy markers (Table 1, Figure 6D). Instead, we observed rapamycin‐sensitive changes in the expression of genes associated with physiological hypertrophy, and a shift toward increased expression of myosin heavy chain‐β.
Short‐Term Acsl1 Inactivation Caused a Clear and Unique Change in the Cardiac Metabolic Fingerprint
We analyzed changes in the cardiac metabolome during the early stages of the metabolic switch induced by FA oxidation inhibition. Metabolomic analysis identified 64 metabolites that were significantly altered in hearts exhibiting reduced ACSL1 activity (Figure 7A), with some overlapping as well as unique changes compared with the 10‐week model of cardiac ACSL1 absence.18 Moreover, concomitant treatment with the mTOR inhibitor rapamycin reverted the metabolic signature attributed to Acsl1 inactivation in pathways involved in amino acid and pentose shunt metabolism. Surprisingly, metabolic intermediates involved in FA synthesis and lysophospholipids were also normalized by rapamycin treatment. As observed in our RNAseq analyses, rapamycin affected the levels of some metabolites in Acsl1flox/flox control hearts; however, PCA suggested that treatment with rapamycin normalizes the changes observed in metabolites in Acsl1H−/− knockdown hearts but has little effect on cardiac metabolites in control mice (Figure 7B).
The variability in glucose uptake of Acsl1H−/− knockdown hearts prompted us to investigate the changes in the metabolic profile of hearts during the early stages of the substrate switch. We observed a negative correlation between glucose uptake and heart size (Table 2), suggesting that hearts do not become hypertrophied if they can compensate for their inability to use FA by rapidly switching to glucose use. Correlations of metabolites identified via metabolomic analysis and corresponding heart sizes revealed that levels of metabolites involved in glycogenolysis, glycolysis, and protein glycosylation decreased with heart size, whereas amino acid, ATP catabolism, and anaplerotic metabolites levels increased in hypertrophied hearts (Table 3).
We used unbiased metabolomics and gene expression analyses to examine the early cellular responses to altered cardiac substrate availability and to establish whether the underlying mechanisms that promote the development of cardiac hypertrophy are glucose‐mediated. Our results indicate that even partial deficiency of ACSL1 is sufficient to elicit cardiac hypertrophy; because rapamycin prevented the hypertrophy, we questioned whether enhanced glycolytic flux underlies the development of an mTOR‐mediated process. Changes in the uptake of nonmetabolizable FA and glucose analogs in Acsl1 knockdown hearts were comparable to those observed 10 weeks after temporal induction of ACSL1 knockout, suggesting that a residual ACS activity of 30% is not sufficient to sustain normal cardiac metabolism, forcing a metabolic switch. Long‐term treatment with rapamycin induced glucose intolerance in the 10‐week Acsl1 inactivation model,15 but this effect was not observed after 2 weeks of treatment with the mTOR inhibitor, indicating that inhibition of mTORC2‐mediated protein kinase B phosphorylation is circumvented in the partial Acsl1 inactivation model. We did not observe consistent activation of the mTORC1 downstream targets S6K or 4E‐BP1, suggesting that variability in attaining the threshold level required to activate mTORC1 at this early time point in the cardiac metabolic switch correlates with the variability observed in substrate uptake and metabolism in Acsl1 knockdown hearts. However, suppression of the mTORC1 inhibitors Deptor and Rptor and alterations in mTOR‐regulated autophagy genes (Figure S2), as well as the response to rapamycin treatment of various genes analyzed, confirmed that mTOR is indeed activated by partial Acsl1 ablation.
Despite the observed activation of mTORC1, metabolomic analyses at 2 weeks confirmed a downregulation of glycolytic intermediates during the early stages of the forced substrate switch, and an absence of the strong transcriptional glycolytic signal exhibited by Acsl1 knockout hearts. Instead, we observed an increase in alternative downstream uses of glucose; in addition to energy production, both our global gene expression and metabolomics analyses suggest that the excess imported glucose may be partially directed toward hexosamine and ECM biosynthesis, protein O‐/N‐GlcNAcylation, and the pentose phosphate shunt (Figure 8). Given that other studies have also observed changes in hexosamine biosynthesis and protein O‐GlcNAcylation in response to alterations in cardiac metabolism36 and hypertrophy,37 further study of these changes is warranted. In addition to the observed diversion of glucose toward alternative downstream fates, glucose metabolism correlated negatively with heart size. Thus, contrary to reports in other models that associate glucose metabolism with the development of mTOR‐mediated hypertrophy,13 our results suggest that increased use of glucose was more prominent in hearts that were not hypertrophied. Finally, rapamycin treatment blocked the development of hypertrophy, but had no effect on glucose uptake rates. Taken as a whole, these data indicate that in this model, at least initially, the use of glucose was not directly responsible for the development of hypertrophy.
As expected, the partial block in FA activation resulting from Acsl1 knockdown led to the accumulation of FA and other lipid intermediates, with a predominance of lysophospholipids. A second dominant class of metabolites included amino acids and intermediates of amino acid metabolism. Amino acids can regulate the activity and downstream effects of mTORC1 on protein translation and cell growth. Increased amino acid metabolism, particularly that of cysteine and glutathione, was also observed in the 10‐week Acsl1 ablation model and was presumed to be triggered to control excessive reactive oxygen species production from oxidative phosphorylation.18 Oxidative stress markers were not as consistently upregulated in the partial Acsl1 inactivation model, suggesting that the increase in amino acid metabolism was independent of oxidative stress. Therefore, the increase in cysteine and glutathione metabolism observed in the 10‐week ACSL1 knockout model may have been elicited in response to the initial switch toward the use of amino acids during the early stages of the metabolic switch.
Rapamycin treatment resulted in a marked attenuation of the Acsl1‐dependent alterations in the expression of various genes involved in metabolic processes. In addition, rapamycin had opposing effects on the expression of several genes involved in glucose and pentose phosphate metabolism, ketogenesis, and tricarboxylic acid cycle flux in Acsl1 knockdown hearts, suggesting that the ACSL1‐dependent generation of a specific metabolite may contribute to cardiac‐specific mTORC1 signaling in the early stages of the metabolic substrate switch. Confirming this notion, the opposing effects of rapamycin treatment were also observed in lysophospholipid, pentose phosphate shunt, and amino acid metabolites.
The premise of this study was to compare changes in gene expression and metabolites in Acsl1 knockdown versus Acsl1 knockout hearts in an effort to elucidate the precise mechanism underlying mTOR activation and the development of hypertrophy. Acsl1 knockdown hearts became hypertrophied during the first 2 weeks of the metabolic switch, and rapamycin treatment prevented this change. The striking induction of fibrosis‐related genes we observed exclusively in Acsl1 knockdown hearts seemed to be a likely candidate for such a regulator of hypertrophy; however, both the transcriptional induction of fibrosis and increased collagen staining are absent in ACSL1 knockout hearts.14 Therefore, we hypothesize that the transient effect on ECM remodeling observed in the early stages of the metabolic switch might occur to increase cell‐cell interactions and thereby enhance transport of nutrients from the surrounding endothelial cells. The concomitant upregulation of ECM synthesis and degradation genes in the partial Acsl1 inactivation model suggests that this is a highly dynamic process; rapid reversal might explain the lack of fibrosis once the switch to glucose metabolism is complete. Rapamycin treatment did not revert the altered expression of fibrosis genes to control levels, implying that this process is not mediated by mTOR. In contrast, mTOR inhibition prevented the development of hypertrophy in Acsl1 knockdown hearts, further suggesting that increased fibrosis does not underlie the development of hypertrophy observed.
Expression of pathological hypertrophy markers in Acsl1 knockdown hearts was not consistent with the results observed in the 10‐week total knockout model. Moreover, whereas pathologically stressed hearts may initially compensate, markers of physiological hypertrophy should not be in evidence. However, we observed a significant increase in myosin heavy chain‐β expression that did not alter the myosin heavy chain‐β/α ratio, an adaptive response thought to preserve function by increasing myocardial efficiency.38 Physiological and pathophysiological hypertrophies are generally described as separate conditions rather than as different phases. However, this dichotomous classification may be an oversimplification, particularly in the context of a model in which progression of the stimulus (long‐term ACSL1 inhibition) results in mTOR‐mediated pathological hypertrophy and diastolic dysfunction. We therefore posit that the hypertrophy observed in the early stages of a metabolic switch imposed by partial Acsl1 knockdown is more reminiscent of a physiological process and likely subject to distinct regulation by a novel mTOR‐regulated calcineurin B homologous protein 3/tescalcin (CHP3)‐glycogen synthase kinase‐3α (GSK3α)‐ CCAAT enhancer‐binding protein (C/EBP) axis. Knockdown of Chp3 expression results in cardiomyocyte hypertrophy, increased expression of the pathological hypertrophy marker atrial natriuretic factor (the single pathological marker observed to be significantly increased in the short Acsl1 ablation model), and elevated GSK3α phosphorylation, suggesting that CHP3 functions as a negative regulator of hypertrophy via inhibition of GSK3α/β phosphorylation and subsequent activation.39 In addition, the cAMP response element–binding protein (CREB) transcriptional activator, C/EBP, is thought to play a role in physiological cardiac growth and protect against pathological hypertrophy.40 Decreased CREB function is associated with pathological hypertrophy, whereas reactivation by C/EBP stabilizes adaptive hypertrophy.41 C/EBP is also known to be regulated by mTOR in adipocytes42 and has been proposed as a GSK3α regulator.43 We observed decreased CHP3, GSK3α, and C/EBP expression in our model of partial Acsl1 deficiency, whereas rapamycin treatment reverted the altered expression levels of GSK3α and C/EBP. Based on our findings, we hypothesize that mTOR‐mediated inhibition of GSK3α induces a partial hypertrophy phenotype. Understanding this molecular switch will be important for the development of future therapies, as it can help identify new molecular players in the progression of adaptive to pathological hypertrophy that may be suitable for therapeutic intervention.
Sources of Funding
This work was supported by NIH grants DK59935 and DK56598 (Coleman) and R37HL065619 (Schisler), the UNC Nutrition Obesity Research Center DK056350, and 12GRNT12030144 (Coleman) from the American Heart Association Mid‐Atlantic Division.
Data S1. Supplemental methods.
Figure S1. Normalization of metabolite concentrations results in Gaussian distributions. Prior to statistical analysis, raw metabolite concentrations were analyzed using Metaboanalyst (v3.0) run in the statistical package R (v3.3.2) (1–3). Concentrations were log transformed and unit scaled. The data distributions before and after normalization are provided both per metabolite (A) and per sample (B). These normalized data were used for PCA and 2‐way ANOVA analyses to identify metabolites that associated with genotype or rapamycin treatment (Figure 7, Table S3).
Figure S2. Expression of mTORC1 components and autophagy genes confirms mTOR activation in Acsl1H−/− knockdown hearts. A, Induction of mTORC1 inhibitor components Deptor and Rptor (4). B, Repression of various genes involved in autophagy in response to mTOR activation resulting from Acsl1H−/− knockdown. n=5 per group. The values reported are means±SEM (error bars). *P≤0.05; **P<0.01; †P<0.001; ‡P<0.0001 compared with littermate controls; ▲ denotes genes whose altered expression in Acsl1H−/− knockdown hearts was reverted by rapamycin treatment.
Figure S3. The effect of RNA technology platform on relative expression levels. Seven of the 13 genes from Figure 6A were identified on both the array and RNAseq platforms. The relative expression of these genes from each platform were mean centered on each control condition (Array‐ Acsl1flox/flox or RNAseq‐ Acsl1flox/flox) and are represented on the bar graph as the mean±SEM (N=3–5 biological replicates per condition). Two‐way ANOVA using genotype (Acsl1H−/− vs Acsl1flox/flox) and platform (Array vs RNAseq) as the factors identified that the genotype accounted for the variance seen across all 7 genes (estimate range 0.4–0.9, †, ††, and ††† correspond to P<0.05, 0.01, and 0.001, respectively). The expression difference in Ecm2 was also influenced by the platform (estimate 0.5, **P=0.0053) and the interaction between platform and genotype (estimate 0.4, ‡P=0.0060), which were about half of the estimate of the genotype effect (0.8, P<0.0001).
Table S1. Microarray Matrix of Gene Expression in Acsl1flox/flox Control and Acsl1H−/− Knockdown (2‐Week) and Knockout (10‐Week) Hearts
Table S2. Complete Table of Differential Gene Expression in Acsl1H−/− Knockdown Hearts
Table S3. Complete Table of Metabolites in Acsl1H−/− Knockdown Hearts
- ↵Sloan C, Tuinei J, Nemetz K, Frandsen J, Soto J, Wride N, Sempokuya T, Alegria L, Bugger H, Abel ED. Central leptin signaling is required to normalize myocardial fatty acid oxidation rates in caloric‐restricted ob/ob mice. Diabetes. 2011;60:1424–1434.
- ↵Dong F, Zhang X, Yang X, Esberg LB, Yang H, Zhang Z, Culver B, Ren J. Impaired cardiac contractile function in ventricular myocytes from leptin‐deficient ob/ob obese mice. J Endocrinol. 2006;188:25–36.
- ↵Trent CM, Yu S, Hu Y, Skoller N, Huggins LA, Homma S, Goldberg IJ. Lipoprotein lipase activity is required for cardiac lipid droplet production. J Lipid Res. 2014;55:645–658.
- ↵Howell JJ, Ricoult SJ, Ben‐Sahra I, Manning BD. A growing role for mTOR in promoting anabolic metabolism. Biochem Soc Trans. 2013;41:906–912.
- ↵Sen S, Kundu BK, Wu HC, Hashmi SS, Guthrie P, Locke LW, Roy RJ, Matherne GP, Berr SS, Terwelp M, Scott B, Carranza S, Frazier OH, Glover DK, Dillmann WH, Gambello MJ, Entman ML, Taegtmeyer H. Glucose regulation of load‐induced mTOR signaling and ER stress in mammalian heart. J Am Heart Assoc. 2013;2:e004796. doi: 10.1161/JAHA.113.004796.
- ↵Ellis JM, Mentock SM, Depetrillo MA, Koves TR, Sen S, Watkins SM, Muoio DM, Cline GW, Taegtmeyer H, Shulman GI, Willis MS, Coleman RA. Mouse cardiac acyl coenzyme A synthetase 1 deficiency impairs fatty acid oxidation and induces cardiac hypertrophy. Mol Cell Biol. 2011;31:1252–1262.
- ↵Li LO, Grevengoed TJ, Paul DS, Ilkayeva O, Koves TR, Pascual F, Newgard CB, Muoio DM, Coleman RA. Compartmentalized acyl‐CoA metabolism in skeletal muscle regulates systemic glucose homeostasis. Diabetes. 2015;64:23–35.
- ↵Schisler JC, Grevengoed TJ, Pascual F, Cooper DE, Ellis JM, Paul DS, Willis MS, Patterson C, Jia W, Coleman RA. Cardiac energy dependence on glucose increases metabolites related to glutathione and activates metabolic genes controlled by mechanistic target of rapamycin. J Am Heart Assoc. 2015;4:e001136. doi: 10.1161/JAHA.114.001136.
- ↵Xu L, Yates CC, Lockyer P, Xie L, Bevilacqua A, He J, Lander C, Patterson C, Willis MS. MMI‐0100 inhibits cardiac fibrosis in myocardial infarction by direct actions on cardiomyocytes and fibroblasts via MK2 inhibition. J Mol Cell Cardiol. 2014;77:86–101.
- ↵Zhan P, Zhao S, Yan H, Yin C, Xiao Y, Wang Y, Ni R, Chen W, Wei G, Zhang P. Alpha‐enolase promotes tumorigenesis and metastasis via regulating AMPK/mTOR pathway in colorectal cancer. Mol Carcinog. 2017;56:1427–1437.
- ↵Gelinas R, Mailleux F, Dontaine J, Bultot L, Demeulder B, Ginion A, Daskalopoulos EP, Esfahani H, Dubois‐Deruy E, Lauzier B, Gauthier C, Olson AK, Bouchard B, Des Rosiers C, Viollet B, Sakamoto K, Balligand JL, Vanoverschelde JL, Beauloye C, Horman S, Bertrand L. AMPK activation counteracts cardiac hypertrophy by reducing O‐GlcNAcylation. Nat Commun. 2018;9:374–390.
- ↵Watson PA, Reusch JE, McCune SA, Leinwand LA, Luckey SW, Konhilas JP, Brown DA, Chicco AJ, Sparagna GC, Long CS, Moore RL. Restoration of CREB function is linked to completion and stabilization of adaptive cardiac hypertrophy in response to exercise. Am J Physiol Heart Circ Physiol. 2007;293:H246–H259.
- ↵Park SA, Lee JW, Herbst RS, Koo JS. GSK‐3alpha is a novel target of CREB and CREB‐GSK‐3alpha signaling participates in cell viability in lung cancer. PLoS One. 2016;11:e0153075.