Insulin Resistance in Skeletal Muscle Selectively Protects the Heart in Response to Metabolic Stress

Obesity causes insulin resistance which is implicated in the pathogenesis of metabolic diseases, including type 2 diabetes mellitus (T2DM) and cardiovascular disease (CVD) (1,2). However, despite the success of many commonly used antihyperglycemic therapies to control hyperglycemia in T2DM, the prevalence of heart failure remains very high in patients with diabetes, raising the possibility that additional factors beyond blood glucose might contribute to the increased heart failure risk (3). Moreover, a paradox exists with regard to the contradicting role of insulin resistance in the development of CVD. As the largest insulin-sensitive organ in the body, skeletal muscle accounts for up to 85% of glucose disposal following a glucose infusion, yet its role in diabetes and CVD is far from clear (4). Anecdotal evidence suggests that insulin resistance in skeletal muscle might protect the heart in response to metabolic stress. In support of this notion, overweight and obese patients with CVD, including coronary heart disease and heart failure, have better short- and medium-term prognoses compared with leaner patients with the same cardiovascular diagnoses, as demonstrated by numerous population studies and meta-analyses from across the world (5–7). Since most obese patients have metabolic syndromes including insulin resistance and hyperinsulinemia, the findings suggest that insulin resistance may benefit patients with CVD. Additionally, treatment of T2DM with thiazolidinedione, one of the most popular insulin-sensitizer drugs widely used for the treatment of T2DM, has led to increased risk of heart failure, as evidenced by multiple meta-analyses of clinical trials (8–11). This notion is further supported by experimental evidence from genetic studies of mouse models of insulin resistance and CVD. For example, mice with liver or islet β-cell–specific knockout of the insulin receptor developed insulin resistance and glucose intolerance (12,13). In contrast, MIRKO mice exhibited a relatively benign phenotype in glucose homeostasis despite rampant hyperinsulinemia (4). In further support of this notion, myocardial-specific deletion of Yme1l, a mitochondrial peptidase involved in the processing of OPA1, led to dilated cardiomyopathy and heart failure (14). This lethal phenotype was mitigated by either targeted deletion of Yme1l in skeletal muscle or by diet-induced obesity (DIO) (14). Together, these findings suggest that insulin resistance in skeletal muscle may benefit the heart under metabolic stress (14).

In this study, we tested the hypothesis that skeletal muscle insulin resistance protects the heart under metabolic stress. Using MIRKO mice as a mouse model of severe insulin resistance, we identified an unexpected role of insulin resistance as the key mediator between skeletal muscle and the heart in response to metabolic stress. We showed that insulin resistance in skeletal muscle protected the heart from the development of cardiac hypertrophy and left ventricle (LV) dysfunction in response to DIO. We further showed that skeletal muscle insulin resistance also prevented mitochondrial dysfunction in response to DIO, leading to a significant improvement in insulin sensitivity and fatty acid oxidation (FAO) in the heart but not in other metabolic tissues including the liver and adipose tissues.

The skeletal muscle–specific insulin receptor knockout mice (MIRKO) were generated by crossing insulin receptor loxP mice (IRlox) (stock no. 006955; The Jackson Laboratory) with skeletal muscle–specific Myl1 Cre mice (stock no. 024713; The Jackson Laboratory). Six-week-old male MIRKO and IRlox control mice were fed with a normal chow diet (ND) or a high-fat diet (HFD) (60% kcal from fat; Research Diets) for 24 weeks. All experiments involving animals were performed in compliance with approved institutional animal care and use protocols according to National Institutes of Health (NIH) guidelines (NIH publication no. 86-23 [1985]).

Data were analyzed with GraphPad Prism software (version 6.0). Statistical comparisons between two groups were done with a two-tailed nonpaired t test. Two-way ANOVA followed by Tukey multiple comparisons test was used for comparisons among multiple groups. The values were considered statistically significant at P values of *P < 0.05 and **P < 0.01 (see figure legends). Data are expressed as means ± SEM.

The data sets generated analyzed in the current study are available from the corresponding author upon reasonable request. No applicable resources were generated or analyzed during the current study.

Using the Cre-loxp system, we generated MIRKO mice with skeletal muscle–specific deletion of the insulin receptor (Insr) gene as detailed by a previous report (4). The approach led to a near total depletion of IR expression in skeletal muscle but not in any other tissues including the heart, as demonstrated by immunoblot analysis of insulin receptor α (IR-α) and β (IR-β) (Fig. 1A). The MIRKO mice and the IRlox control mice were fed with either ND or HFD for 24 weeks to induce obesity, followed by analysis of DIO-induced metabolic disorders. Although IR deficiency in skeletal muscle did not affect weight gain, the MIRKO mice exhibited increased content in fat mass relative to the IRlox control mice without significant changes in lean body mass or water content (Fig. 1B and C). As expected, the MIRKO mice also developed insulin resistance and glucose intolerance, which is supported by results from oral glucose tolerance and insulin tolerance tests (Fig. 1D and F, quantified in Fig. 1E–G). Moreover, the MIRKO mice also exhibited hyperinsulinemia, hypertriglyceridemia, and elevated levels of serum IGF-1 and total ketone bodies both under ND and under HFD (Fig. 1H–K).

We next determined the effect of skeletal muscle insulin resistance on DIO-induced LV dysfunction and adverse remodeling. As demonstrated by a representative echocardiograph (Fig. 2A), DIO caused LV contractile dysfunction, as shown by decreased LV fractional shortening (LVFS), LV ejection fraction (LVEF), LV posterior wall end-systole (LVPWs), and interventricular septal end-systole (IVSs) without significant effects on other parameters, including interventricular septal end-diastole, LV internal diameter end-diastole and end-systole, and LV posterior wall end-diastole (Fig. 2B–E and Supplementary Fig. 1A–D). Remarkably, the MIRKO mice were protected from the development of the LV contractile dysfunction, as supported by changes in LVFS, LVEF, LVPWs, and IVSs relative to the IRlox control mice (Fig. 2A–E). Consistent with the findings, DIO also caused cardiac hypertrophy in IRLox control mice, which is evidenced by results from wheat germ agglutinin (WGA) staining of cardiac sections, a measurement of cardiomyocyte size (Fig. 2F, quantified in Fig. 2G), and RT-PCR analysis of mRNA expression levels of hypertrophic biomarkers, including atrial natriuretic factor (Anf), brain natriuretic peptide (Bnp), and β-cardiac myosin heavy chain (β-Mhc) (Fig. 2H). Again, IR deficiency in skeletal muscle not only mitigated these defects but also prevented the onset of myocardial fibrosis in MIRKO mice, as evidenced by results from Masson’s trichrome staining of heart samples, a measurement of fibrosis of cardiomyocytes (Fig. 2I, highlighted by arrows, quantified in Fig. 2J), and by results from RT-PCR analysis of mRNA expression levels of both collagen I (Col1a1) and collagen III (Col3a1) (Fig. 2K).

The mammalian target of rapamycin (mTOR) signaling pathway controls cell growth and size. Hyperactivation of mTORC1 is often implicated in the pathogenesis of cardiac hypertrophy (15). We next determined whether IR deficiency in skeletal muscle would attenuate hyperactivation of mTORC1 signaling in the heart of MIRKO mice in response to DIO. Consistent with cardiac hypertrophy in DIO mice, consumption of HFD led to hyperactivation of the mTORC1 signaling pathway in the heart of IRlox control mice, which is supported by increased phosphorylation levels of the downstream effectors of mTORC1 signaling, including S6K and 4E-BP1 (Fig. 2L, quantified in Fig. 2M). DIO also moderately decreased the phosphorylation level of tuberous sclerosis complex (TSC2), a negative regulator of mTORC1 signaling, likely as a negative feedback response to mTORC1 hyperactivation. Again, these changes were normalized in the hearts of MIRKO mice.

To gain insight into the molecular mechanisms by which skeletal muscle insulin resistance prevents cardiac dysfunction in DIO mice, we next investigated the effect of MIRKO on insulin signaling in major metabolic tissues, including heart, skeletal muscle, inguinal white adipose tissue (iWAT), and liver. Consistent with insulin resistance, in MIRKO mice insulin-stimulated phosphorylation of AKT and GSK3β was totally blunted in the skeletal muscle when mice were fed with either ND or HFD, as evidenced by decreased levels of phosphorylated (p)AKT (Thr308), pAKT (Ser473), and pGSK3β (Ser9) (Fig. 3A, quantified in Fig. 3B). Strikingly, in spite of hyperinsulinemia and elevated levels of serum IGF-1, insulin sensitivity was significantly improved in the heart of MIRKO mice in response to DIO, as evidenced by increased phosphorylation levels of AKT (Thr308), a key phosphorylation site that mediates insulin sensitivity in metabolic tissues (Fig. 3A, quantified in Fig. 3C) (16). In contrast, hyperinsulinemia in MIRKO mice significantly impaired insulin sensitivity in other metabolic tissues, as suggested by decreased phosphorylation levels of AKT (pAKT, Thr308 and Ser473) and GSK3β (Ser9) in iWAT and liver (Supplementary Fig. 2A, quantified in Supplementary Fig. 2B and C). Together, these findings suggest that insulin resistance in skeletal muscle mediates a specific cross talk with the heart, but not other metabolic tissues, in response to metabolic stress associated with DIO.

Obesity causes chronic inflammation, which is also implicated in adverse cardiac remodeling (17–19). To gain further insight into the molecular mechanisms underlying the protective effects, we next investigated whether MIRKO would prevent myocardial inflammation and apoptosis, common defects associated with cardiac dysfunction. Consistent with cardiac dysfunction, DIO caused activation of multiple inflammatory pathways, including NLRP3, TXNIP, and cGAS-cGAMP-STING signaling pathways, as evidenced by results from Western blot analysis of key inflammatory biomarkers in the hearts of the IRlox control mice (Fig. 4A, quantified in Fig. 4B). NLRP3 is an intracellular sensor for various “danger” signals, including metabolic stress from DIO and adverse cardiac remodeling (17,18,20). TXNIP is an activator of NLRP3 and a key sensor of oxidative stress. Targeted deletion of TXNIP protects the myocardium from ischemia-reperfusion injury (21). The cGAS-cGAMP-STING pathway mediates innate immune responses to viral infection (22). The cyclic GMP-AMP synthase (cGAS) enzyme catalyzes the synthesis of cyclic GMP-AMP (cGAMP), a second messenger required for the activation of stimulator of interferon genes (STING) (22). Although the cGAS-cGAMP-STING pathway is primarily involved in antiviral responses, recent studies show that it can also be activated by cytosolic release of mitochondrial DNA (mtDNA) from metabolic stress, and therefore ablation of STING significantly attenuated chronic inflammation in DIO mice (18). The findings were further supported by increased mRNA expression levels of multiple proinflammatory cytokines, including Tnf, Nfkb, Il-1b, and Il-6 in the heart of IRlox mice (Fig. 4D). In further support of its potential benefit to the heart, MIRKO mice were protected not only from DIO-induced inflammation but also from apoptosis of cardiomyocytes, which is supported by results from Western blot analysis of major apoptotic and antiapoptotic biomarkers, including Bax, cleaved caspase 3 (C-cas3), cytochrome c (Cyt c), and Bcl-2 (Fig. 4A–D), and by results from TUNEL staining of apoptotic cells in the heart samples (Fig. 4E, highlighted by arrows, quantified in Fig. 4F).

Insulin resistance causes mitochondrial dysfunction in the heart, which is implicated in the pathogenesis of myocardial hypertrophy (23,24). As one of the highest-energy-demanding organs per gram of tissue weight in the body, the heart heavily depends on mitochondrial FAO to provide high-density fuel. During adverse cardiac remodeling, cardiac metabolism is reprogrammed toward an increased reliance on glycolysis, whereas FAO is downregulated (25). To further identify mechanisms underlying the beneficial effects of skeletal muscle insulin resistance to the heart, we next investigated whether MIRKO would cause metabolic reprogramming in the heart of DIO mice. As shown by results from electron microscopic analysis, DIO caused disarray of mitochondrial morphology, leading to significant accumulation of lipid droplets (LDs) in the hearts of IRlox control mice (Fig. 5A, quantified in Fig. 5B). Accumulation of mitochondria-bound LDs is associated with impaired FAO (26). Consistent with the notion, DIO also significantly increased intramyocardial triglyceride content in the heart, suggesting an impairment in FAO (Fig. 5C). The notion is corroborated by decreasing protein expression level of 3-hydroxyacyl-CoA dehydrogenase trifunctional multienzyme complex subunit-β (HADHB), which catalyzes the last three steps of mitochondrial β-oxidation of long-chain fatty acids, in both heart and skeletal muscle tissues of IRlox control mice (Fig. 5D, quantified in Fig. 5E and F). In contrast, MIRKO not only prevented accumulation of mitochondria-bound LDs (Fig. 5A, quantified in Fig. 5B) but also restored HADHB protein expression levels in response to DIO, leading to decreased triglyceride content in the heart (Fig. 5C and D, quantified in Fig. 5E). Consistent with metabolic reprograming, MIRKO also prevented DIO-induced depletion of HADHB in skeletal muscle and iWAT but not in the liver (Fig. 5D and F and Supplementary Fig. 3A–C).

In further support of metabolic reprogramming, MIRKO also significantly increased mRNA expression levels of major genes encoding FAO enzymes in the heart, including medium-chain acyl-CoA dehydrogenase medium chain (Acadm), fatty acid translocase (Cd36), carnitine palmitoyl transferase 1b and 2 (Cpt1b and Cpt2), fatty acid binding protein 3 (Fabp3), and peroxisome proliferatoractivated receptor γ coactivator 1α (Pgc-1α) (Fig. 5G and Supplementary Fig. 4A). MCAD is required for the initial step of β-oxidation. CD36 and FABP regulate membrane uptake of fatty acids, whereas CPT1 and CPT2 catalyze mitochondrial transport of fatty acids, a key step involved in FAO (27). PGC-1α is a coactivator of the PPARγ, a key transcriptional activator of lipid metabolic enzymes in the heart (28). Moreover, DIO also caused mitochondrial dysfunction in the heart of IRlox control mice, as evidenced by decreased mRNA expression levels of key regulators of mitochondrial biogenesis and oxidative phosphorylation, including transcription factor A (Tfam), pyruvate dehydrogenase kinase-isoenzyme 4 (Pdk4), and NADH dehydrogenase (ubiquinone) 1 α subcomplex 9 (Ndufa9), without significant effect on the expression of estrogen-related receptor α (Esrrα) and ubiquinol–cytochrome c reductase core protein 1 (Uqcrc1) (Supplementary Fig. 3F). In contrast, MIRKO not only restored mRNA expression levels of these mitochondrial regulators but also increased mtDNA copy number and citrate synthase activity, a key biomarker for intact mitochondria in the heart in response to DIO (Supplementary Fig. 3D and E). Consequently, MIRKO also significantly attenuated DIO-induced oxidative stress and lipid peroxidation, as suggested by decreased levels of malondialdehyde, a by-product of lipid peroxidation, and H₂O₂ (Fig. 5H and 5I).

Then comes the question of how insulin resistance in the MIRKO mice mediated specific cross talk with the heart. As the largest organ of the body in nonobese individuals, skeletal muscle is now considered to be an active endocrine organ, releasing a host of myokines that are part of a complex network that mediates communication between skeletal muscle and other organs, including the liver, iWAT, and heart tissues. To determine a role of myokines in mediating the protective effect to the heart in MIRKO mice, we next analyzed the effects of DIO and MIRKO on mRNA expression levels of several major myokines previously implicated in regulating heart function by RT-PCR analysis. The results show that DIO caused depletion of both myonectin (Erfe) and myostatin (Mstn), both of which are implicated to play a protective role in cardiac function (29,30). In contrast, DIO did not have any major effect on mRNA expression of other myokines in the skeletal muscle of IRlox mice, including osteocrin (Ostn), growth differentiation factor 11 (Gdf11), interleukin-6 (Il-6), insulin-like growth factor 2 (Igf2), and follistatin-like 1 (Fstl1) (Fig. 6C–G). Remarkably, MIRKO mimicked the effect of DIO by causing total depletion of both Erfe and Mstn in skeletal muscle when MIRKO mice were fed a chow diet (Fig. 6A and B), implicating a key role of insulin resistance in regulating the expression of both myokines.

To gain further insight into the role of MIRKO in regulating mRNA expression of other myokines, we next performed genome-wide RNA sequencing analysis to identify differentially expressed genes in the skeletal muscles between IRlox and MIRKO mice on HFD. The results show that MIRKO caused global changes in the mRNA expression profile (Supplementary Fig. 4A), including several myokines that are implicated in regulating cardiac function. Among them are chemokine receptor chemokine receptor 5 (Ccl5), bone morphogenetic protein −7 (Bmp7), and fibroblast growth factor 21 (Fgf21) (Supplementary Fig. 4B). Ccl5 is implicated in the pathogenesis of heart failure in a mouse model of chronic cardiac ischemia (31), whereas both Bmp7 and Fgf21 protect against fibrosis in hypertensive heart disease and LV remodeling under pressure overload (31,32). As expected, MIRKO also significantly changed the landscape of mRNA expression levels of other genes, including those involved in signal transduction, oxidoreductase activity, and lipid metabolism (Supplementary Fig. 4C–E). As a consequence of insulin resistance and metabolic reprogramming, MIRKO caused depletion of mRNA expression levels of several insulin effector proteins in skeletal muscles, such as Shc2 and Grb14, and significantly increased mRNA expression level of several lipid metabolic enzymes, including perilin1 (Plin1), angiopoietin-like protein 4 (Pnpla3), and fatty acid desaturase 3 (Fads3).

Insulin resistance is implicated as the common cause of various aging-related chronic diseases, such as T2DM, diabetes complications, sterile inflammation, and CVD. However, cumulative efforts in targeting insulin resistance for the treatment of these chronic conditions have not been met with great success to date. Consequently, the prevalence of heart failure remains very high in T2DM patients, despite the success of many commonly used antihyperglycemic therapies to control hyperglycemia over the last couple of decades. Therefore, key questions remain on how T2DM promotes the development of heart diseases. In this study, we identified a highly unexpected effect of insulin resistance in skeletal muscle in preventing cardiac dysfunction in DIO mice, which is supported by multiple lines of evidence. Accordingly, we show that the MIRKO mice were protected from development of LV hypertrophy, fibrosis, and apoptosis of cardiomyocytes in response to DIO. The findings are further underscored by underlying changes in signal transduction pathways that mediate these pathogeneses, including the hyperactivation of mTORC1, NLRP3, STING, and apoptotic signal transduction pathways in the heart.

Although the precise molecular mechanisms by which MIRKO benefits the heart remains elusive, our data suggest that MIRKO protects the heart in part by preventing cardiac insulin resistance in response to DIO. Myocardial insulin resistance is implicated in the pathogenesis of heart failure, which is supported by the phenotypes of mice with targeted deletion of insulin receptor gene (CIRKO) or Irs1 and Irs2 (H-DKO) in the heart as well as targeted deletion of Irs1 and Irs2 both in the heart and skeletal muscle (MDKO) (2,33–35). In support of this notion, we show in this study that MIRKO specifically improved insulin signaling in the heart, but not in any other tissues, despite rampant hyperinsulinemia in response to DIO. The importance of cardiac insulin signaling is further underscored by the phenotypes of the H-DKO and MDKO mice (2,34,35). Accordingly, the H-DKO mice developed heart failure and premature death at 6–9 weeks of age due to cardiac-specific deletion of both Irs1 and Irs2. This lethal phenotype is further exacerbated by targeted deletion of Irs1 and Irs2 both in heart and skeletal muscle, leading to accelerated death at 3 weeks of age of the MDKO mice. Together, these findings lend further support to our hypothesis that skeletal muscle insulin resistance prevents DIO-induced cardiac dysfunction by improving insulin signaling in the heart.

Insulin signaling also plays a critical role in maintaining mitochondrial function, which is again supported by the phenotypes of the H-DKO and MDKO mice. Accordingly, Irs1 and Irs2 depletion in skeletal muscle and the heart caused metabolic reprograming of MDKO mice that is also highly reminiscent of the metabolic defects associated with adverse myocardial remodeling and heart failure (25,35). The MDKO mice exhibited a dramatic shift in glucose utilization from oxidative phosphorylation to glycolysis, implicating a defective mitochondrial oxidative phosphorylation. Consistent with these findings, we show in this study that insulin resistance in the heart also caused mitochondrial dysfunction and metabolic reprograming in DIO mice. Accordingly, DIO significantly increased serum level of ketone bodies and depleted the protein expression of the mitochondrial trifunctional protein, which is required for FAO, leading to intramyocardial lipid accumulation. In contrast, MIRKO not only prevented mitochondrial fragmentation in response to DIO but also significantly increased mtDNA copy number and mRNA expression levels of proteins involved in mitochondrial respiration, leading to significant attenuation of oxidative stress and lipid peroxidation. Moreover, MIRKO also prevented accumulation of mitochondria-bound LDs in the heart of DIO mice by restoring the protein expression of mitochondrial trifunctional protein. Our findings are further corroborated by the phenotypes of CIRKO mice that exhibit myocardial insulin resistance, impaired FAO, and upregulated glycolysis in the heart (36).

Despite intensive research in the field, the potential benefit of insulin resistance, if any, in the survival of mammals remains elusive. Our findings reveal for the first time that insulin resistance in skeletal muscle mediates a specific cross talk with the heart to prevent cardiac dysfunction under metabolic stress. This notion is corroborated by anecdotal evidence from both T2DM patients with heart diseases and various mouse models, including those with targeted deletion of Glut-4 in skeletal muscle (muscle-G4KO). As Glut-4 is the key mediator of insulin-stimulated glucose uptake in skeletal muscle, one would expect that the muscle-G4KO mice would develop metabolic defects similar to those of MIRKO mice. Intriguingly, this has not been the case. In contrast to the benign phenotype of MIRKO mice, a subset of muscle-G4KO mice develop frank diabetes with severe insulin resistance. This phenotype was further exacerbated by targeted deletion of Glut-4 both in skeletal muscle and the heart, which resulted in glucose intolerance and insulin resistance as early as 8 weeks of age (37). Additionally, in contrast to hyperinsulinemia in MIRKO mice, neither the cardiac-G4KO nor the muscle-G4KO mice developed hyperinsulinemia or dyslipidemia, again suggesting that hyperinsulinemia may exert a protective effect to the heart in MIRKO mice (4,36). In support of this notion, myocardial-specific deletion of Yme1l, a mitochondrial peptidase involved in the processing of OPA1, led to dilated cardiomyopathy and heart failure (14). This lethal phenotype was mitigated by either targeted deletion of Yme1l in skeletal muscle or subjecting the mice to DIO (14). This notion is further corroborated by clinical observations that overweight and obese patients with CVD have better prognoses than leaner patients with the same cardiovascular diagnoses (5–7). Since most obese patients suffer from hyperinsulinemia and insulin resistance, the findings further contradict the commonly held belief that insulin resistance exacerbates heart failure. Finally, the potential benefit of hyperinsulinemia to the heart of MIRKO mice is further corroborated by findings from several recent prospective population-based studies, which suggest that hyperinsulinemia and risk of cardiovascular mortality may be independent (38).

Together, our findings revealed a highly unexpected role of insulin resistance in skeletal muscle as a double-edged sword in metabolic diseases. On one hand, it protected the heart from myocardial dysfunction in response to metabolic stress associated with DIO. On the other, it exerted a deleterious effect on glucose homeostasis by exacerbating DIO-induced insulin resistance in other metabolic tissues, such as liver and iWAT. Although the precise molecular mechanisms underlying the cross talk remain to be fully elucidated, our studies have provided the first glimpse into this complicated issue, calling for further investigation into this important but controversial phenomenon. Although our findings remain preliminary, if the mechanism(s) responsible for cardiac improvement can be separated from the negative impact on glucose tolerance, this would have enormous potential implications for the treatment of heart failure and other CVD. Moreover, how the development of skeletal muscle insulin resistance transmits its message to the heart remains unclear, since our effort on myokine expression generated mixed results. Likewise, why hyperinsulinemia failed to cause insulin resistance in the heart of MIRKO mice is a fascinating question that also deserves further investigation, since hyperinsulinemia is commonly believed to be the primary cause of insulin resistance.

Acknowledgments. The authors thank Drs. Ralph DeFranzo and Nicolas Musi for critically reading this manuscript and for providing valuable feedback and suggestions on the experimental design.

Funding. This work was supported in part by funding from the NIH (R01AG055747 [to Y.S.]), American Diabetes Association (1-18 IBS-329 [Y.S.]), Barth Syndrome Foundation (Y.S.), an endowment from Joe R. and Teresa Lozano Long Distinguished Chair in Metabolic Biology (Y.S.), National Institute on Aging T32 Training Grant (T32GM108563 [J.P.A.]), and National Natural Science Foundation of China (31671240 [Z.T.]). The Genome Sequencing Facility at Greehey Children’s Cancer Research Institute at UT Health San Antonio is supported by National Cancer Institute, NIH, grant P30 CA054174 (Cancer Center at UT Health San Antonio), NIH Shared Instrument grant 1S10OD021805-01 (S10 grant), and CPRIT Core Facility Award RP160732.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. D.J., J.Z., and X.L. performed the experiments and analyzed the data. Y.S. designed the experiments and wrote the manuscript. J.-P.A. and J.N. provided feedback and guidance on experimental designs. Z.T. provided mentorship to D.J. All authors reviewed results and commented on the manuscript. Y.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.