Reducing Fatty Acid Oxidation Improves Cancer-free Survival in a Mouse Model of Li-Fraumeni Syndrome

There is accumulating evidence that both the oxidation and synthesis of fatty acids contribute to tumorigenesis (1–4). Increased fatty acid β-oxidation (FAO) has been observed in diffuse large B-cell lymphoma as well as chronic lymphocytic leukemia (5, 6). Glioblastoma also rely on FAO for growth and reducing fatty acid availability leads to tumor senescence (7). Increased glycolysis is a well-known metabolic feature of cancer cells and dietary intake of its substrates, glucose and fructose, can enhance intestinal tumor growth in mouse models (8). Even in this model, the activation of glycolysis was associated with increased expression of enzymes involved in fatty acid synthesis that may support tumor growth. High-fat diet and fatty acid receptor, CD36, in cancer cells appear to play a role in metastasis (9). On the other hand, the consensus that fatty acid metabolism fuels cancer growth contrasts with proposals to consider high fat ketogenic diet as supplemental therapy against cancer based, in part, on its glucose and insulin lowering effects (10). Such conflicting information suggests the need to test whether modulating fat metabolism will affect tumorigenesis in vivo, for example, in mouse models of patients at high risk of developing cancer.

We previously reported that patients as well as a mouse model of the early-onset cancer disorder Li-Fraumeni syndrome (LFS) caused by germline mutations of TP53 display evidence of increased oxidative metabolism (11). Mice carrying the p53 R172H knock-in mutation (human TP53 R175H mutation homolog) showed increased aerobic exercise endurance with a lower metabolic respiratory exchange ratio (RER) compared with wild-type, indicating increased utilization of fatty acids (11). Other studies have suggested p53 regulation of fat metabolism, and a more recent work demonstrated that another LFS mutation can activate lipolysis, an important biochemical process for fatty acid utilization (12–14). Although the general inhibition of mitochondrial respiration was observed to delay tumorigenesis in the p53 R172H mouse model (15), delineating the contribution of the specific metabolic activities affected by the p53 mutation may permit the development of more targeted strategies for cancer prevention. Furthermore, given the many differences among the various studies examining fatty acid metabolism in cancer, such as cancer cell type, stage of development, and genetic mutations, the effect of inhibiting fatty acid metabolism on cancer development in a LFS mouse model remains unknown. As there is no chemoprevention treatment in LFS, it is imperative to identify potential therapeutic targets, such as the FAO pathway, to improve the cancer-free survival of patients affected by LFS. We, therefore, investigated the feasibility of inhibiting the upregulated FAO in the LFS mouse model. The results of this study, using two different genetic approaches, suggest that the inhibition of fatty acid oxidation may represent a potential strategy for delaying cancer onset in LFS.

All mice were used in accordance with, and with the approval of, an Institutional Animal Care and Use Committee (National Heart, Lung, and Blood Institute, Laboratory of Cardiac Energetics, NHLBI, NIH, Bethesda, Maryland, USA, NHLBI). Myoglobin-knockout (MB^−/−) mice (provided by R.S. Balaban and created by J. Schrader, Institute for Cardiovascular Physiology, Heinrich-Heine-University, Dusseldorf, Germany; ref. 16) were crossed with p53 R172H knock-in–mutant (p53^172H/H) mice (NCI Frederick Mouse Repository, Frederick, MD; ref. 17) to generate the MB^−/− p53^172H/H double-mutant mice. CD4-Cre transgenic mice (The Jackson Laboratory) were crossed with the p53 R172H mice to generate CD4-Cre p53^172H/H mice, which were, in turn, crossed with mice that had CPT2 alleles flanked by loxP (CPT2^fl/fl; ref. 18) to generate CD4⁺ T-cell–specific CPT2-knockout mice. All mice were in C57BL6 background or backcrossed into C57BL6 at least five generations. PCR primer sequences used for genotyping the mice are provided in Supplementary Table S1.

Patient samples were obtained after informed written consent as approved by the NHLBI-NIH Internal Review Board (Bethesda, MD, ClinicalTrials.gov identifier NCT00406445) and primary human myoblasts were prepared as described previously (11).

Measurements of lean tissue, fat, and fluid in live mice were performed on a Minispec NMR LF50 Body Composition Analyzer (Bruker).

Gastrocnemius skeletal muscle and thymus tissues were dissected and flash-frozen in liquid nitrogen. Small metabolite/limited lipid profiling and semiquantification in tissues were performed using gas chromatography–mass spectroscopy and LC/MS-MS platform by Metabolon, Inc. Lipidomic profiling and semiquantification in thymus tissue were performed by Human Metabolome Technologies America, Inc.

Lactate levels in tail blood of mice were measured using Nova Lactate Plus Meter and test strip according to the manufacture's protocol.

RNA sequencing (RNA-seq) was performed on thymus tissue from 10-week-old mice with no overt evidence of tumor. The samples were sequenced on the Illumina HiSeq 2500, and the reads were aligned to mouse reference genome mm9 using Bowtie 2. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis was performed with DAVID Bioinformatics Resources 6.8 (https://david.ncifcrf.gov/).

Antibodies used in this study were as follows: actin mouse mAb (A3853), tubulin mouse mAb (T5168), and actinin-α mouse mAb (A7811) (Sigma-Aldrich); cyclin D1 (DCS6) rabbit mAb (2978), acetyl (ac)-lysine rabbit pAb (9441), H3K9ac rabbit mAb (9649), H3K27ac rabbit mAb (8173), H3K4me3 rabbit mAb, phospho-AMPKα (Thr172) rabbit mAb (2535), AMPKα rabbit pAb (2532), phosphor-mTOR (Ser2481) rabbit pAb (2974), phosphor-mTOR (Ser2448) rabbit pAb (2971), mTOR rabbit pAb (2983), S6 ribosomal protein rabbit mAb (2217), and phosphor-S6 ribosomal protein (Ser235/236) rabbit pAb (4856) (Cell Signaling Technology); p53 (DO-1) mouse mAb (sc-126), myoglobin rabbit pAb (sc-25607), and hexokinase 2 goat pAb (sc-6521) (Santa Cruz Biotechnology); histone H3 rabbit pAb (ab1791), CPT1A mouse mAb (ab128568), CPT1B rabbit pAb (ab134988), and CPT2 rabbit pAb (ab181114) (Abcam); MAGL rabbit pAb (ABN1000); GMPR rabbit pAb (15683-1-AP) (Proteintech); GLUT4 rabbit pAb (3945-200) (BioVision); Cre rabbit pAb (NB100-56134) (Novus); VDAC rabbit pAb (600-401-882) (Rockland); and p21 mouse mAb (OP64) (Oncogene Science).

For immunoblotting, tissue samples were homogenized in ice-cold RIPA buffer with protease and phosphatase inhibitors (Roche Diagnostic Corp), resolved by Tris-glycine SDS-PAGE, and transferred to Immobilon-P Membrane (Millipore) for standard enhanced chemiluminescence development.

Thymi harvested from 10-week-old mice were cleaned off adhering tissues, rinsed in PBS buffer, and placed on a nylon mesh/filter (105 μm) in a 60-mm tissue culture dish with 3–4 mL PBS at 4°C. The nylon mesh was then folded in half over the thymus and gently disrupted using flat edged forceps until only residual connective tissue remained. The homogenized tissue was then triturated with a pipette, filtered through a cell strainer (BD Falcon, 40 μm), spun down, and resuspended for subsequent use.

Fatty acid oxidation was measured using ³H-palmitate as described previously (19). Briefly, 2 × 10⁶ freshly isolated thymic lymphocytes were resuspended in 1 mL of modified KHB medium (111 mmol/L NaCl, 4.7 mmol/L KCl, 1.25 mmol/L CaCl₂, 2 mmol/L MgSO₄, 1.2 mmol/L NaH₂PO₄, 2.5 mmol/L glucose, 0.5 mmol/L carnitine, and 5 mmol/L HEPES, pH 7.4) with or without 200 μmol/L etomoxir. A mixture of [9,10-³H] palmitate (53.7 Ci/mmol, PerkinElmer) and 10% BSA in KHB buffer was added to the cell suspension and incubated at 37°C for 2.5 hours. After incubation, the cells were spun down at 300 × g and assayed for protein content, while the supernatant was subjected to chloroform–methanol extraction to measure the ³H₂O product in the aqueous phase. The etomoxir-sensitive ³H₂O generated was used to calculate the β-oxidation rate and expressed as CPM/μg protein.

Primary human myoblasts transduced with control empty vector or human p53 R175H-mutant cDNA lentivirus, as described previously (20), were seeded in XF24 Cell Culture Microplates (Seahorse Bioscience) at 5 × 10⁴ cells/well. After incubating the cells in DMEM (GlutaMAX-I, Invitrogen) supplemented with 20% FBS and 2 μmol/L uridine at 37°C with 5% CO₂ overnight, it was replaced with KHB assay medium supplemented with 200 μmol/L palmitic acid/BSA with or without 200 μmol/L etomoxir. Whole-cell oxygen consumption was measured on an Agilent Seahorse XF24 Analyzer and normalized to protein content.

Whole-soleus skeletal muscle was embedded into islet capture microplate and assayed for oxygen consumption in DMEM (Agilent, 102353-100) supplemented with 10 mmol/L glucose and 1 mmol/L pyruvate using the Agilent Seahorse XF24 Analyzer according to the manufacture's protocol.

RNA was isolated from cultured cells using the poly(dT) Magnetic Bead System (Invitrogen), reverse transcribed using Superscript II (Invitrogen), and quantified using real-time RT-PCR with SYBR green fluorescence on a LightCycler Sequence Detection System (Roche). RT-PCR primer sequences are provided in Supplementary Table S1.

During the diurnal periods of increased activity, p53^172H/H mice were observed to have a lower RER compared with wild-type mice by metabolic chamber measurements, indicating increased oxidation of fatty acids in the mitochondria (11). Given the connections between fatty acid metabolism and cancer, we wondered whether inhibiting FAO activity could prevent tumorigenesis in this LFS mouse model. As a simple, but global, approach for reversing this fatty acid metabolic phenotype, we crossed p53^172H/H mice with Myoglobin-knockout (MB^−/−) mice, which are known to have a higher RER indicative of reduced FAO (21, 22). Because fatty acid oxidation has a greater stoichiometric requirement for oxygen compared with carbohydrates, the absence of oxygen reservoir, myoglobin, would be expected to result in substrate switching from fatty acids to carbohydrates. Myoglobin has been reported to be expressed at elevated levels in various cancers at their earliest stages and may promote neoplastic growth (23), so the deletion of myoglobin afforded an opportunity to test whether it affects the tumor susceptibility of p53^172H/H mice.

The absence of myoglobin was first confirmed in the skeletal muscle of MB^−/− p53^172H/H double-mutant mice by immunoblotting which also revealed increased levels of glucose transporter 4 (GLUT4) and hexokinase 2 (Fig. 1A). This suggested a compensatory increase in glycolysis because of reduced oxygen availability. Furthermore, the double-mutant MB^−/− p53^172H/H mice compared with p53^172H/H mice had a higher body fat composition, consistent with decreased fatty acid metabolism (Fig. 1B). Metabolomic profiling of skeletal muscle also suggested lower FAO and increased glucose utilization as reflected by lower ketone body metabolite, 1,2-propanediol, and higher lactate and citrate steady-state levels, respectively (Fig. 1C). In addition, multiple metabolites involved in one-carbon metabolism, known to be affected by mitochondrial function and to play a role in tumorigenesis (24, 25), were altered in MB^−/− p53^172H/H skeletal muscle (Fig. 1C). As in skeletal muscle, lactate levels were elevated in MB^−/− p53^172H/H mouse blood (Fig. 1D). Supporting diminished intracellular oxygen availability in the absence of myoglobin, the pale colored MB^−/− p53^172H/H soleus muscle, a slow fiber known to be enriched in mitochondria with a preference for fatty acid substrates (26), showed decreased ex vivo oxygen consumption rate (OCR) suggesting altered FAO (Fig. 1E; Supplementary Fig. S1). These observations were consistent with the metabolomic data and prompted us to investigate further whether modulating the aberrant fatty acid oxidative metabolism in MB^−/− p53^172H/H mice can affect cancer development.

Remarkably, the disruption of MB increased the median survival time of p53^172H/H mice from 131 to 188 days (∼40%; Fig. 2A). The tumor spectrum of the double-mutant MB^−/− p53^172H/H mice was not significantly changed, and most animals still died with thymic lymphoma, suggesting mainly a delay in tumorigenesis caused by the metabolic alteration (Fig. 2B). We performed gene expression analysis of p53^+/+, p53^172H/H, and double-mutant MB^−/− p53^172H/H thymi to examine globally how mutant p53 affects the tissue where the tumor develops and whether disrupting MB can rescue any of the changes. RNA-seq profiling and KEGG pathway analysis prior to the development of overt tumor revealed increased mitochondrial activity and ribosome biogenesis in p53^172H/H versus p53^+/+samples, while these increases were downregulated by MB deletion (Table 1). Notably, the upregulated one-carbon metabolism observed in p53^172H/H thymus was also decreased by MB disruption, consistent with the metabolomic results (Fig. 1C; Table 1). Furthermore, the thymic lymphocytes dissociated from p53^172H/H thymus showed increased mitochondrial FAO activity relative to wild-type, which could be reversed in the MB-deficient state (Fig. 2C). Deleting MB in the p53 wild-type state also reduced FAO, confirming the specific role of MB in facilitating FAO independently of p53 mutation (27).

To obtain further insights into how MB deficiency affects lipid metabolism and prevents tumorigenesis, we performed metabolomic profiling of thymic tissues from p53^+/+, p53^172H/H, and MB^−/− p53^172H/H mice. Of the 377 identified metabolites, only 14 were significantly changed in MB^−/− p53^172H/H mice compared with p53^172H/H mice when ranked by statistical significance (P < 0.05; Fig. 3A). The majority of these metabolites (11/14) were increased by MB deletion in p53^172H/H thymus and some stood out as being involved in lipid metabolism. The second most significantly ranked compound was 2-oleoylglycerol, a known substrate of monoacylglycerol lipase (MAGL) that has been implicated in promoting tumorigenesis (Fig. 3B; ref. 28). p53^172H/H thymus also showed significant reductions in 5′-GMP, which is converted to IMP via GMP reductase (GMPR), potentially activating lipolysis and mitochondrial activity (Fig. 3B; ref. 29). Consistent with the changes in 2-oleoylglycerol and 5′-GMP levels, the increased expression of MAGL and GMPR in p53^172H/H thymus was partially or completely reversed in the MB^−/− p53^172H/H genotype state (Fig. 3C). The FAO rate-limiting enzyme, carnitine palmitoyltransferase 1A (CPT1A), showed an expression pattern by p53 and MB genotypes, which was similar to that of MAGL and GMPR, consistent with its known induction by acetyl-CoA, the product of FAO (Fig. 3C; ref. 30). Furthermore, the levels of myoglobin were increased in p53^172H/H thymus, perhaps reflecting the increased oxygen requirement for FAO. These data suggested that the p53 R172H mutation may increase lipolysis and fatty acid metabolism, in part, through the concerted activities of multiple mediators such as MAGL, GMPR, and CPT1, and that these changes can be dampened by MB deletion.

The delayed tumorigenesis caused by MB deficiency paralleled previous observations whereby mild disruption of mitochondrial respiration was sufficient to improve cancer-free survival in p53^172H/H mice, potentially by inducing antiproliferative cell signaling caused by bioenergetic stress (15). Mutants of p53 can lose their capacity to repress cyclin D1 expression while acquiring gain-of-function to prevent the activating phosphorylation of AMPKα at Thr-172 by tumor suppressor LKB1 (31–33). On the other hand, inhibiting FAO can activate AMPKα by increasing Thr-172 phosphorylation (34). Indeed, the increased cyclin D1 and decreased phospho-AMPKα levels observed in p53^172H/H thymus were reversed by MB deletion in association with decreased FAO (Figs. 2C and 3C). Mutant p53 inhibition of AMPKα with subsequent activation of mTOR (Ser-2481 autophosphorylation) and its effector ribosomal protein S6 (RPS6 Ser-235/236 phosphorylation) were also reversed by MB deletion in p53^172H/H thymus (Fig 3C; ref. 35). Importantly, RPS6 activation plays a critical role in cancer initiation, as well as tumor growth, and is known to promote the upregulation of the ribosome biogenesis transcriptional program, which was identified independently by RNA-seq analysis (Table 1; refs. 36–38). Together, these screening studies suggest that the aberrant fatty acid metabolism caused by mutated p53 can be moderated indirectly by deleting MB, and that this metabolic inhibition activates AMPKα which, in turn, inhibits the mTOR pathway activated by mutant p53 with resultant tumor suppression.

To examine whether the in vivo observation of increased FAO in the LFS mouse model could apply to humans and to gain additional insights into mechanism, human myoblasts were transduced with lentivirus expressing human p53 R175H protein, homolog of the mouse p53 R172H mutation. Compared with control, expression of p53 R175H markedly promoted mitochondrial oxygen consumption using palmitate as FAO substrate in human myoblasts (Fig. 4A). Consistent with the observation in mice (Fig. 3C), mutant p53 increased the protein levels of some important enzymes involved in lipid metabolism while decreasing the levels of p21, the prototypical wild-type p53 target gene (Fig. 4B). Furthermore, the mRNA levels of FAO-related enzymes, CPT1B (muscle specific) and GMPR, were induced by mutant p53, while that of p21 was suppressed as expected (Fig. 4C). Thus, a human LFS-mutant p53 can cell autonomously promote the expression of more than one gene involved in FAO and fatty acid metabolism as they are complex processes requiring the concerted actions of multiple enzymes.

We next set out to examine whether FAO plays a cell autonomous role in the thymic T-cell lymphoma development of p53^172H/H mice. We targeted CPT2, which converts acyl-carnitine, produced by CPT1A on the outer mitochondrial membrane, back to acyl-CoA on the inner membrane for entry into the β-oxidation cycle in the matrix. To specifically knockout CPT2 in T cells, we took advantage of a previously described strategy for conditionally disrupting CPT2 using a cell type–specific promoter-driven Cre-loxP strategy (18). CPT2 was deleted in T cells by crossing CD4 promoter-driven Cre transgenic mice with p53^172H/H mice carrying floxed CPT2 alleles, resulting in heterozygous (CPT^+/fl) or homozygous (CPT2^fl/fl) CPT2-knockout states. Although the incidence and types of cancer were not significantly different at this sample size as with MB deletion, the LFS mouse model with heterozygous knockout of CPT2 (CD4-Cre CPT2^+/fl p53^172H/H) showed an approximately 30% improvement in survival time but, unexpectedly, not in the CPT2 homozygous–knockout state (Fig. 5A and B). Immunoblotting for CPT2 protein and measurement of FAO activity in the thymus showed that they were reduced in a CPT2 allele dose-dependent manner; ³H-palmitate oxidation was decreased by approximately 20% and approximately 50% in the CPT2^+/fl and CPT2^fl/fl genotype states, respectively (Fig 5C and D). Thus, the delay in tumorigenesis was not proportional to the degree of FAO inhibition, suggesting that the beneficial effect derived by mildly inhibiting FAO through CPT2 haploinsufficiency may be masked by the more severe consequences of complete CPT2 absence (Fig. 5A and D).

Immunoblotting revealed further differences between the CPT2^+/fl and CPT2^fl/fl genotypes. LFS mouse thymus with homozygous, but not heterozygous, knockout of CPT2 showed reductions in lysine acetylation, H3 histone acetylation, and methylation, complementing the previous observation of decreased histone acetylation upon loss of CPT1A with resultant depletion of acetyl-CoA levels (Fig. 5C; ref. 39). These results also suggested that severe inhibition of FAO in homozygous CPT2^fl/fl thymus can interfere with normal cellular process causing us to focus on the heterozygous CPT2^+/fl state to understand its tumor-suppressive effect under more physiologic conditions. The decrease in cyclin D1 caused by CPT2 deficiency was marginal, but the activation of AMPKα by Thr-172 phosphorylation, a sensor of bioenergetic stress, correlated well with CPT2 allele copy-number loss (Fig. 5C). These observations were consistent with the general consensus that fatty acids cannot be replaced entirely by glucose in vivo, both as a mitochondrial oxidative substrate and for nonbioenergetic activities (2). As observed in the MB^−/− p53^172/H/H double-mutant mice, CPT2 deficiency also suppressed the mTOR pathway in p53^172/H/H mice, presumably due to AMPKα activation (Fig. 5C; ref. 40). Thus, a mild reduction in FAO can activate AMPKα and delay tumorigenesis in LFS mice without causing significant changes in the common epigenetic modifications, such as histone acetylation and methylation, as observed in the CPT2^fl/fl homozygous–knockout state (Fig. 5C).

To confirm whether the mild inhibition of FAO caused by CPT2 haploinsufficiency measured in T cells in vitro is sufficient to alter lipid metabolism in vivo, we performed lipidomic profiling of CPT2 wild-type and heterozygous-knockout (CD4-Cre CPT2^+/fl p53^172H/H) thymus tissue. As expected with reduction of long-chain FAO in the mitochondria, there was accumulation of long-chain fatty acids in thymus of heterozygous CPT2^+/fl LFS mice (Fig. 5E). Furthermore, the steady-state levels of three acyl-carnitines (14:2, 14:3, and 16:2) were significantly decreased, while there was a pattern of increased longer chain acyl-carnitines in the CD4-Cre CPT2^+/fl p53^172H/H thymus, indicating decreased conversion of long-chain acyl-carnitines to acyl-CoA due to CPT2 deficiency (Fig. 5E; Supplementary Fig. S2). In summary, these results suggest that the partial inhibition of FAO activates AMPKα with subsequent downregulation of the mutant p53-activated mTOR pathway, which may contribute to delaying tumorigenesis (Fig. 6).

In this study, we have used two different genetic modalities to reduce fatty acid oxidation in a mouse model of LFS for which no cancer preventive treatments are currently available. We demonstrate that partial inhibition of FAO globally or specifically in CD4⁺ T cells, which are cancer prone, can significantly increase the cancer-free survival time of LFS mice. It should be noted that the mild inhibition of FAO appears to be beneficial in delaying tumorigenesis, while more severe inhibition results in the loss of this survival benefit. Although the health consequences of inhibiting FAO over a lifetime are unclear, the benefits of partially inhibiting FAO in this LFS mouse model appear to prevail and suggest its consideration as a potential strategy for cancer chemoprevention.

Because FAO impacts many different cellular processes including the generation of acetyl-CoA necessary for histone acetylation (39), the loss of such an essential metabolic function in the homozygous CPT2-knockout mice may compromise multiple cellular activities directly or indirectly related to cancer such as tumor suppressive immune function, confounding the interpretation of their survival time. Complete loss of CPT2 could also result in abnormal T-cell development as suggested by the significantly decreased thymic lymphocyte cell count in homozygous CPT2^fl/fl, but not heterozygous CPT2^+/fl or MB^−/− genotype states (ref. 41; Supplementary Fig. S3A). In addition to the synthesis of fatty acids, its catabolism by lipolysis and FAO is also appreciated to promote cancer cell survival as the FAO inhibitor, etomoxir, can promote apoptosis in leukemia cells (1, 26, 42). In vitro treatment of p53^172H/H thymic lymphocytes with etomoxir did not promote cell death, making it more likely that the cancer prevention mechanism by FAO reduction in the LFS mouse model is through alterations in cell signaling such as AMPK-mediated inhibition of the mTOR pathway (ref. 35; Supplementary Fig. S3B). In this light, the pharmacologic inhibition of FAO using clinically approved medications such as trimetazidine or ranolazine could be considered for chemoprevention in LFS.

A preponderance of clinical studies suggests that low-fat diet decreases the risk of cancer as well as cardiovascular diseases. A large randomized clinical trial recently showed that low-fat diet significantly decreases the risk of mortality from breast cancer, commonly observed in patients with LFS (43). Given the extremely high incidence of breast cancer in women with LFS, being on a low-fat diet may confer relatively greater benefit for cancer prevention in this specific group compared with the general population. On the basis of the observation that patients with LFS and a mouse model of this condition have increased oxidative phosphorylation and fatty acid utilization, we have now targeted both mitochondrial respiration and FAO in a LFS mouse model and observed decreased tumorigenesis and improved cancer-free survival (15). It remains to be seen whether inhibiting mitochondrial respiration, such as by metformin, and FAO together can have a synergistic effect on cancer prevention in LFS. This study reveals an alternative strategy for cancer prevention by targeting fatty acid metabolism in patients with LFS who, for example, cannot tolerate metformin for mitochondrial inhibition or wish to consider dietary methods for decreasing the lifetime risk of cancer.

M.J. Wolfgang reports grants from NIH during the conduct of the study. No disclosures were reported by the other authors.

P.-Y. Wang: Conceptualization, formal analysis, supervision, investigation, methodology, writing-original draft, writing-review and editing. J. Ma: Formal analysis, investigation, methodology, writing-review and editing. J. Li: Investigation, methodology. M.F. Starost: Formal analysis, investigation, methodology. M.J. Wolfgang: Resources, writing-review and editing. K. Singh: Data curation, software, formal analysis, investigation, writing-review and editing. M. Pirooznia: Data curation, software. J.-G. Kang: Formal analysis, investigation, methodology, project administration. P.M. Hwang: Conceptualization, formal analysis, supervision, funding acquisition, writing-original draft, project administration, writing-review and editing.

We wish to thank Robert S. Balaban/Jurgen Schrader and Toren Finkel for providing us with the MB^−/− and CPT2^fl/fl mice, respectively. We also thank Yuesheng Li (NHLBI DNA Sequencing and Genomics Core) and Martin P. Playford for their assistance and advice. This work was supported by the NHLBI-NIH Division of Intramural Research (HL005101-15 to P.M. Hwang).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.