Compartmentalized Acyl-CoA Metabolism in Skeletal Muscle Regulates Systemic Glucose Homeostasis

Skeletal muscle switches between the use of glucose and lipid, depending on fuel availability and energy requirements. Metabolic inflexibility, a state of impaired capacity for fuel switching between fatty acids (FAs) and glucose (1), is associated with metabolic disorders such as insulin resistance, diabetes, and the cardiometabolic syndrome (2–5), whereas enhanced oxidation of FA spares glucose (6,7). Metabolic inflexibility occurs in human disorders in which β-oxidation is genetically impaired; these disorders are associated with hepatic steatosis, myopathy, cardiomyopathy, and episodes of fasting-induced nonketotic hypoglycemia (8). However, it remains uncertain how fuel selection is regulated in skeletal muscle and how fuel inflexibility specifically impairs exercise tolerance and leads to metabolic dysfunction (7,9).

Before FAs are used for oxidation or for the synthesis of cholesterol esters, phospholipids, or triacylglycerol (TAG), they must be activated to form acyl-CoAs by one of a family of five long-chain acyl-CoA synthetases (ACSLs), each encoded by a separate gene (10). We have proposed that each ACSL isoform determines whether FAs are channeled toward pathways of oxidation or complex lipid biosynthesis (11). Data from ACSL1 deficiencies in adipose, liver, heart, or macrophages suggest that channeling is tissue-specific, but that in highly oxidative tissues, the ACSL1 isoform directs FAs toward β-oxidation (11–14). Thus, mice deficient in brown adipose ACSL1 cannot metabolize FAs for heat production and fail to maintain a normal body temperature when placed in a cold environment (13). Likewise, hearts that lack ACSL1 cannot oxidize FAs for contractile energy, have an eightfold increased use of glucose, and develop mammalian target of rapamycin complex 1–mediated hypertrophy (14).

ACSL1 is located on the mitochondrial outer membrane, but it is unclear why its absence would block FA oxidation; acyl-CoAs are amphipathic molecules that are both soluble in water and able to integrate into membrane monolayers via their hydrophobic acyl chains. In addition, acyl-CoA binding protein and some isoforms of FA binding protein can bind and carry acyl-CoAs. Thus, an acyl-CoA should be able to enter any downstream pathway, unless the entry into one specific pathway versus another depend on the relative substrate affinities of the initiating acyltransferases.

The role of any ACSL isoform in skeletal muscle remains unknown (6,15). Given the importance of lipid metabolism in skeletal muscle and the key position of ACSL in initiating most lipid metabolic pathways, we hypothesized that a deficiency of ACSL1 would block the use of FAs for energy, thereby impairing fuel selection during exercise. To learn whether ACSL1 deficiency in skeletal muscle impairs exercise capacity or alters whole-body fuel metabolism, we studied the metabolic adaptation of skeletal muscle–specific Acsl1 knockout mice (Acsl1^M−/−) to physiological challenges. Unexpectedly, acyl-CoA content rose markedly during exercise, revealing that acyl-CoAs synthesized by non-ACSL1 isoforms were retained in a pool that was inaccessible for mitochondrial oxidation.

Homozygous floxed mice (Acsl1^loxP/loxP) (12) were mated with mice expressing Cre recombinase controlled by the muscle-specific human skeletal actin (HSA) promoter (catalog #006149; The Jackson Laboratory) to achieve muscle-specific deletion of Acsl1. HSA-Cre is expressed in the early embryo (16), and mice were born with deficient ACSL1. Acsl1^M−/− (Acsl1^loxP/loxP-HSACre) were mated with Acsl1^loxP/loxP mice to obtain Acsl1^M−/− and littermate controls (Acsl1^loxP/loxP). Because most measurements were qualitatively similar in male and female Acsl1^M−/− mice compared with controls, we report measurements from only one sex. Procedures were approved by the University of North Carolina Institutional Animal Care and Use Committee. Mice were kept in a pathogen-free barrier facility (12-h dark/light cycle) and fed standard chow (Prolab RMH 3000; LabDiet, Richmond, IN). Between 8 and 25 weeks of age, some mice were fed a high-fat diet (45%; D12451) or a matched low-fat (10%; D12450B) standard diet (Research Diets). Oral glucose (2.5 g d-glucose/kg body wt) and intraperitoneal insulin tolerance tests (control diet 0.35 units/kg body wt; high-fat diet 0.50 units/kg body wt) (12) and glycerol tolerance tests (2 mg/kg body wt) (17) were performed.

Metabolic analysis was performed in PhenoMaster/LabMaster chambers (TSE Systems, Chesterfield, MO). After 24 h of acclimatization, measurements were acquired every 27 min for VO₂, VCO₂, respiratory exchange ratio (RER), food intake, and physical activity.

Mice (16–26 weeks of age) were acclimatized to a motorized, speed–controlled treadmill with an electric shock stimulus (Exer-3/6; Columbus Instruments, Columbus, OH) for 3 consecutive days, starting from 0 to 16 m/min and from 0 to 10% grade. On the study day, mice were fasted for 2 h after 10:00 a.m. (“rested” state) to prevent potential carbohydrate-induced increases in insulin with a subsequent decrease in adipose lipolysis (18). Speed was increased from 8 m/min by 2 m/min every 15 min until reaching 18 m/min. Mice were maintained at this speed until they refused to run. Tail blood was taken for measurement of glucose and lactate levels. For measurement of plasma metabolites and tissue glycogen, trunk blood and tissues were collected at exhaustion. Blood and tissues were taken from additional control mice after they had run for the same length of time as the exhausted Acsl1^M−/− mice.

Mice were fasted for 4 h (basal state) or 17 h. Plasma metabolites were measured colorimetrically (glucose, nonesterified FA; Wako, Richmond, VA; triglyceride; Stanbio Laboratory, Elkhart, IN) or enzymatically (β-hydroxybutyrate, creatine kinase; Stanbio Laboratory), with a mouse insulin ELISA kit (Millipore, Billerica, MA) or with a lactometer (Lactate Plus; Nova Biomedical, Waltham, MA). Glycogen was measured in snap-frozen tissues (19). TAG was measured in tissue lipid extracts (Stanbio Laboratory) (20,21). Gastrocnemius muscle was fixed and stained (22).

Tissue proteins were blotted with primary antibodies to detect ACSL1, Akt, and pAkt (Cell Signaling Technology, Danvers, MA); ACSL4 (Santa Cruz Biotechnology, Santa Cruz, CA); and GLUT1 and GAPDH (Abcam, Cambridge, MA).

Mice were injected with PBS or insulin (2 units/kg body wt i.p.). Glucose was measured by glucometer to ensure adequate lowering. Ten minutes later, tissue supernatants were prepared, electrophoresed, and blotted for antiphospho-Akt (Thr-308) and total Akt (Cell Signaling Technology) (20).

Genomic/mitochondrial DNA and total RNA were extracted from tissues (TRIzol; Invitrogen, Carlsbad, CA; AllPrep DNA/RNA kit; Qiagen, Valencia, CA). cDNA was synthesized (Applied Biosystems). Gene expression and mitochondrial DNA content were determined with Absolute QPCR Sybr Green Fluorescein Mix (Thermo Scientific, West Palm Beach, FL). Expression of target genes (Supplementary Table 1) was normalized to endogenous β-actin, tubulin, or H19 and calculated using the 2^−ΔΔCt method (23).

ACSL-specific activity was measured in tissue homogenates from 13- to 16-week-old mice (24). CO₂ and acid soluble metabolites (ASM) were measured from gastrocnemius homogenates incubated for 30 min in 100 mmol/L KCl, 40 mmol/L Tris-HCl, 10 mmol/L Tris base, 5 mmol/L MgCl₂⋅6H₂O, 1 mmol/L EDTA, and 1 mmol/L ATP, pH 7.4, and [1-¹⁴C]palmitate or [1-¹⁴C]palmitoyl-CoA (25). Amino acids, acyl-carnitines, acyl-CoAs, and ceramides were analyzed by flow injection tandem mass spectrometry (26,27). Organic acids were analyzed by gas chromatography–mass spectrometry (27).

For whole-muscle incubations, soleus and extensor digitorum longus muscles were removed (fed state) and incubated (28) with gassing in Krebs-Henseleit buffer, 5 mmol/L d-glucose, 1 mmol/L l-carnitine, 12.5 mmol/L HEPES, and 1% BSA. Media were changed after 30 min. Contralateral muscles were incubated with 1 mmol/L [1-¹⁴C]oleate (2 μCi/ml) or with 5 mmol/L [U-¹⁴C]glucose (2 μCi/ml), 100 nmol/L insulin, and 0.75 mmol/L palmitate for 90 min. Media were transferred to culture tubes containing 1 N NaOH CO₂ traps. ¹⁴CO₂ and [¹⁴C]ASM were assessed (28).

Data were analyzed using a two-tailed Student t test or two-way ANOVA where appropriate (SAS, Inc.) and were expressed as means ± SE. Statistical significance was defined as P < 0.05.

In Acsl1^M−/− mice, weight and activity did not differ from those in littermate controls, and female reproduction was normal. ACSL1 protein was absent in Acsl1^M−/− gastrocnemius muscle but was present in the heart (Fig. 1A) and liver (Supplementary Fig. 1A). Acsl1 mRNA expression in gastrocnemius muscle was 86% lower than that in controls, with no compensatory increases in other ACSL isoforms (Fig. 1B). Compared with controls, the mRNA content of Acsl4 was 30% lower in Acsl1^M−/− gastrocnemius muscle and 35% higher in soleus muscle (Supplementary Fig. 1B), but ACSL4 gastrocnemius muscle protein content was unchanged (Supplementary Fig. 1C). ACSL-specific activity was unchanged in the heart but was diminished by >90% in Acsl1^M−/− gastrocnemius muscle (Fig. 1C). Residual ACSL activity represents the combined activities of all remaining ACSL isoforms. These data show that ACSL1 contributes the majority of total ACSL activity in skeletal muscle.

Deficiency of ACSL1 caused a mild myopathy with occasional central nuclei and cells surrounded by macrophages (Fig. 1D and E), but plasma creatine kinase activity was not statistically elevated (Table 1). Compared with controls, the sum of complete oxidation of [1-¹⁴C]palmitate to CO₂ plus incomplete oxidation to ASM was 83% and 63% lower, respectively, in Acsl1^M−/− gastrocnemius and soleus muscles (Fig. 1F). Unimpaired oxidation of [¹⁴C]palmitoyl-CoA to CO₂ (Fig. 1G) and the presence of normal mitochondrial DNA content (Supplementary Fig. 1D) indicated that ACSL1 deficiency did not block the downstream FA oxidation pathway or diminish mitochondrial respiratory function. The significant decrease in FA oxidation in Acsl1^M−/− muscle suggests that the ACSL1 isoform is required to activate FAs before they can be converted to acyl-carnitines, enter the mitochondrial matrix, and undergo β-oxidation; in contrast, acyl-CoAs were oxidized.

Because the preparation of homogenates or isolated mitochondria alters normal organelle relationships, we performed additional studies in permeabilized muscle fibers. FA oxidation in permeabilized fibers was only 1% of the oxidation in homogenates (compare Fig. 1F and H), and neither glucose nor FA oxidation or FA incorporation into complex lipids was impaired (Fig. 1I and J), suggesting that maximal FA transport, similar to that of glucose, may require specific transporters or ACSL isoforms that are activated by muscle contraction. Experiments using homogenates or isolated mitochondria would eliminate the transport step and permit ACSL1 to directly activate FAs and allow them to be converted to acyl-carnitines for mitochondrial entry. Alternatively, in static muscle, perhaps only very small amounts of FA enter the mitochondria for oxidation, and these small amounts of acyl-CoAs are supplied by non–ACSL1 isoforms such as ACSL5, which is also present on the outer mitochondrial membrane (10).

To examine how the limitation in skeletal muscle FA oxidation affects whole-body energy metabolism, we challenged Acsl1^M−/− mice with an overnight fast, a state in which FAs released from adipose become the major fuel source for most organs. Acsl1^M−/− mice became mildly hypoglycemic (plasma glucose level: 82 ± 6 mg/dL vs. 137 ± 24 mg/dL in controls) (Fig. 2A). Compared with controls, basal plasma FA concentrations were 64% higher in Acsl1^M−/− mice and rose further after an overnight fast (0.58 ± 0.04 vs. 0.25 ± 0.05 mmol/L, respectively) (Fig. 2B). After fasting, the plasma TAG level was 46% higher in Acsl1^M−/− mice (Fig. 2C), and the plasma β-hydroxybutyrate level was 71% higher (Fig. 2D), consistent with increased hepatic β-oxidation of FA. Compared with fasting controls, fasting Acsl1^M−/− mice stored 53% more TAG in the liver (Fig. 2E and F), suggesting that the inability of Acsl1^M−/− muscle to metabolize FA caused adipose-derived FAs to relocate to the liver for re-esterification and storage (29,30).

Overnight fasting depleted liver glycogen stores in both control and knockout mice (Fig. 2G). Although mRNA abundance of glucose-6-phophatase and phosphoenolpyruvate carboxykinase increased similarly in both genotypes (Fig. 2H), which is consistent with appropriate gluconeogenic upregulation, this upregulation was insufficient to counteract hypoglycemia in Acsl1^M−/− mice, probably because of diminished substrate availability since a glycerol tolerance test showed enhanced gluconeogenesis (Fig. 2I). The gastrocnemius muscle of Acsl1^M−/− mice contained more GLUT1 protein (Fig. 2J). Weight gain, body composition (Supplementary Fig. 2), and glucose tolerance test results showed no genotype differences in mice fed control or high-fat diets (Fig. 2K), but less insulin was secreted by Acsl1^M−/− mice (Fig. 2L). Enhanced insulin sensitivity was confirmed by insulin tolerance tests (Fig. 2M), high basal pAkt (T308) levels in the liver, and increased pAkt levels in the liver and gastrocnemius muscle after insulin injection (Fig. 2N). Taken together, these data show that, during fasting, the inability of Acsl1^M−/− skeletal muscle to switch from glucose to FA use resulted in an overflow of FAs to plasma and liver, thereby increasing liver TAG and plasma VLDL levels, diminishing plasma glucose levels, and increasing insulin sensitivity.

The RER (VCO₂/VO₂) was consistent with impaired fuel switching from glucose to FA oxidation in Acsl1^M−/− muscle (Fig. 3A). The RER did not differ between genotypes when mice were fed chow (60% carbohydrate), suggesting that muscle FA oxidation may not be critical without exercise, but when food was removed, the responses of the two genotypes diverged. Whether or not food is available at the start of the dark cycle, the RER normally rises; this rise in RER occurred in both control and Acsl1^M−/− mice (Fig. 3B and Supplementary Fig. 3). However, although fasting decreased the total RER during both the day and night cycles, at every point during the dark cycle the RER of the Acsl1^M−/− mice remained higher than that of the controls, indicating increased systemic use of glucose by Acsl1^M−/− mice, despite the fast (Fig. 3A and B). With refeeding, the RER rose similarly in both genotypes, and the amount of food eaten and heat produced were similar (Fig. 3C and D). These data indicate that during fasting when FA is normally the major fuel source, Acsl1^M−/− mice relied predominantly on nonlipid fuels; this requirement was less prominent in the nonfasted state, primarily because exercise was minimal in both genotypes.

During endurance exercise, Acsl1^M−/− mice ran only 46–48% of the distance run by controls (1,522 ± 122 vs. 3,175 ± 327 m) (Fig. 4A and Supplementary Fig. 4A). Blood lactate levels increased minimally (Fig. 4B), indicating that the aerobic threshold had not been exceeded, but both genotypes ceased to run when their blood glucose concentrations dropped to ∼40 mg/dL (Fig. 4C). Although glucose concentrations declined more slowly in control mice, at exhaustion, liver and muscle glycogen levels were equally depleted in both genotypes, and both were similarly hypoglycemic (Table 1). At exhaustion, plasma creatine kinase levels were similar in both genotypes.

At the time point when Acsl1^M−/− mice refused to run further, running control mice had blood glucose values of 87 ± 7 mg/dL, despite depleted liver glycogen levels and undetectable muscle glycogen levels (Table 1). Because the basal content of liver and muscle glycogen was similar in both genotypes (Fig. 2G) and since liver glycogen stores were almost totally absent in the running controls, it appeared that gluconeogenesis in control mice was sufficient to avoid severe hypoglycemia, but Acsl1^M−/− mice, despite similar increases in the expression of hepatic gluconeogenic genes (Fig. 4D), were unable to produce enough glucose to compensate for the increased muscle demand.

During prolonged exercise, hydrolyzed adipose FA becomes a major fuel source for most tissues (31). Endurance exercise increased plasma nonesterified FA levels in both genotypes (Table 1). Although an increase in plasma FA level normally increases liver TAG content during exercise, Acsl1^M−/− mice accumulated less hepatic TAG than did controls, probably because more FA was used for energy in the liver (Table 1). Similarly, although plasma TAG levels increased in both genotypes, there was less of an increase in the Acsl1^M−/− mice (Table 1). Intramuscular TAG is believed to be a major fuel source in muscle during endurance exercise (32), and, at exhaustion, control muscle had lost twice as much TAG (Table 1), which is consistent with their ability to use lipid for energy.

To understand how FA use differs in exercising muscle, we profiled plasma and muscle metabolites. Consistent with absent ACSL1 activity, resting Acsl1^M−/− muscle contained lower levels of long-chain acyl-CoAs (Fig. 5A and Supplementary Fig. 4A). Compared with controls, acyl-CoA species of 14–18 carbons were 31–42% lower, but little difference was observed for short-chain acyl-CoAs, which are primarily products of amino acid degradation. Basal levels of long-chain acyl-carnitine species were similar in both genotypes (Fig. 5B). Acetyl-CoA is the final product of pathways of FA oxidation, glycolysis, and amino acid degradation. Basal acetyl-CoA levels were higher in Acsl1^M−/− muscle than in that of controls (Fig. 5A). Because FA oxidation was severely impaired in Acsl1^M−/− muscle and basal tricarboxylic acid (TCA) cycle intermediates were not limiting (Fig. 5C), this elevation in acetyl-CoA content, together with the lower basal content of total amino acids in Acsl1^M−/− muscle, suggests an increase in the use of amino acids and glucose for energy, even without exercise (Fig. 5D). Levels of amino acids and acyl-carnitines in plasma did not differ between genotypes (Supplementary Fig. 5B and C).

With endurance exercise, many acyl-CoA and acyl-carnitine metabolites increased markedly. In muscle from exhausted control mice, the levels of total acyl-CoA and long-chain acyl-CoA species each rose threefold (Fig. 5A), which is consistent with the normally enhanced influx of adipocyte-derived FA and its metabolism to acyl-CoA. Surprisingly, the increases of total acyl-CoA and long-chain acyl-CoA were even greater in Acsl1^M−/− muscle than in that of controls, with 4- and 10-fold increases, respectively. Total long-chain acyl-CoA levels were 89% higher in exhausted knockout muscle than in exhausted control muscle, despite the absence of ACSL1, the 79% lower ACSL-specific activity, and the absence of increases in mRNA expression of other Acsl or Fatp isoforms (Fig. 6A) (33).

Acyl-carnitines, which normally correspond to acyl-CoA levels, differed greatly by genotype. Endurance exercise increased total and long-chain acyl-carnitines 5.5- and 23-fold, respectively, in control muscle, but only 3- and 12-fold, respectively, in Acsl1^M−/− muscle. Thus, in exhausted Acsl1^M−/− muscle, the levels of total and individual long-chain acyl-carnitine species were 33–77% lower than in exhausted controls, consistent with impaired FA oxidation (Fig. 5B), and levels of medium-chain acyl-carnitines were 30–77% lower than those in controls (Supplementary Fig. 5). These data indicate that, despite the marked decrease in skeletal muscle ACSL-specific activity in Acsl1^M−/− mice, during endurance exercise plasma FAs taken up into muscle were converted to acyl-CoAs by other ACSL isoforms; these acyl-CoAs, however, remained in cytosolic pools that were inaccessible to carnitine palmitoyltransferase (CPT) 1 and β-oxidation enzymes in mitochondria and peroxisomes (34).

Exercise increases the demand for carnitine as a carrier for acyl groups (18). Although exercise diminished muscle carnitine content in both genotypes (Fig. 5B), free carnitine levels remained higher in Acsl1^M−/− than in control muscle, both in the resting state and at exhaustion, indicating that the availability of carnitine had not limited the conversion of acyl-CoAs to acyl-carnitines. Because acyl-carnitines often reflect the flux of acyl-CoAs that enter the β-oxidation pathway, the observation that muscle from exhausted Acsl1^M−/− mice contained higher amounts of long-chain acyl-CoAs but lower amounts of long-chain acyl-carnitines (Fig. 5A and B) strongly suggests that muscle contains separate pools of acyl-CoA. Acsl3 mRNA was 40 and 30% higher, respectively, in resting and exhausted Acsl1^M−/− muscle, but none of the other Acsl isoforms was specifically upregulated by exercise (Fig. 6A). In control muscle, the total specific activity derived from all ACSL isoforms was unchanged by exercise; however, in muscle from exhausted Acsl1^M−/− mice, ACSL-specific activity more than doubled (0.06 ± 0.01 to 0.17 ± 0.04 nmol/min/mg protein), suggesting that one or more ACSL isoforms had been post-translationally activated by exercise (Fig. 6B). Lack of change in intramuscular TAG content in Acsl1^M−/− mice (Table 1 and Fig. 6C) suggests that the accumulating long-chain-CoAs were not used for net TAG synthesis.

Intense exercise induces autophagy with the use of proteolyzed amino acids for both fuel within muscle and as a substrate for hepatic gluconeogenesis (35–37). With exercise, control muscle content of proline, valine, leucine/isoleucine, aspartate/asparagines, and tyrosine rose 57–115% (Fig. 5D); plasma branch-chain and aromatic amino acids increased two- to threefold (Supplementary Fig. 6B); and amino acid degradation products (short-chain C3–C6 acyl-carnitines) rose 3.5-fold (Supplementary Fig. 5B). In exhausted Acsl1^M−/− muscle, however, the levels of proline, alanine, aspartate/asparagines, and the branched-chain amino acids were low (Fig. 5D), probably because these amino acids were diverted toward energy production within muscle, as indicated by the marked rise in short-chain acyl-CoA amino acid degradation products (Supplementary Fig. 5A). TCA intermediates were not depleted (Fig. 5C).

Our data show that acyl-CoAs are poorly available for β-oxidation in muscle if they have not been synthesized by ACSL1 (Fig. 7). Endurance exercise in Acsl1^M−/− mice caused a marked increase in muscle content of long-chain acyl-CoAs. These acyl-CoAs, generated by other ACSL isoforms, were not in a pool that was accessible to the multiprotein CPT1-voltage-dependent anion channel-ACSL1 complex (38,39), and could not be converted to acyl-carnitines or enter the mitochondrial matrix. Thus, acyl-CoAs are located in separate and distinct pools within the cell. This is a surprising result, because amphipathic long-chain acyl-CoAs are water soluble and can interact with membranes and with lipid binding proteins that should allow them to move freely within cells.

Because exercise-activated AMPK downregulates lipid synthesis, lipid synthetic enzymes were apparently unable to handle the excess FA that enters muscle during endurance exercise. Exercise also enhances lipolysis (18,40). Consistent with its critical importance in facilitating acyl-CoA entry into mitochondria and in concert with other peroxisome proliferator–activated receptor α targets, Acsl1 was the only Acsl mRNA upregulated by endurance exercise in control mice (41,42). No change in expression of other Acsl isoforms was observed in Acsl1^M−/− muscle, but exercise increased the low basal specific activity of ACSL in gastrocnemius muscle by 166%, suggesting possible post-transcriptional activation.

A second important finding was that the inability of skeletal muscle to oxidize FA made Acsl1^M−/− mice vulnerable to systemic hypoglycemia. The dependence of Acsl1^M−/− muscle on glucose limited glucose availability for other tissues, and gluconeogenesis was unable to supply sufficient glucose during fasting or exercise. The metabolic profile observed is consistent with muscle that is unable to oxidize FA and, instead, uses hydrolyzed amino acids for energy rather than releasing them as substrates for hepatic gluconeogenesis. Thus, Acsl1^M−/− muscle burned more glucose during exercise, and its demand for nonlipid fuel impaired systemic energy homeostasis. The increased demand for glucose during an overnight fast was accompanied by increased muscle GLUT1 protein and enhanced insulin sensitivity in both the liver and muscle.

The marked decrease in FA β-oxidation in Acsl1^M−/− muscle resulted in a fasting phenotype of hypoglycemia, elevated plasma FA levels, and fatty liver similar to that of humans and animals with genetic defects in FA oxidation (43,44). However, Acsl1^M−/− mice differed from other models of defective FA oxidation in that muscle TAG did not accumulate. Further, because the liver retained normal ACSL activity, fasting-induced hypoketonemic hypoglycemia did not occur (43).

Metabolic limitations on endurance exercise have been variously attributed to low intramuscular fuel stores (45), diminished liver glycogen stores (41), lack of glucose or ketone production by the liver (43), or an impaired ability of muscle to take up and oxidize available fuels, including glucose (46) and FA (43). Diminished endurance in rats is associated with hypoglycemia (45) and with low glycogen content in discrete cerebral regions (47), which may contribute to exhaustion (48).

Failure of muscle to switch to FA use rendered Acsl1^M−/− mice unable to run optimally, despite normal hepatic FA oxidation and comparable upregulation of mRNA for hepatic gluconeogenic genes. Although a mild myopathy was present, the most striking difference between the running genotypes was the concentration of plasma glucose. Acsl1^M−/− mice refused to run when their glucose concentrations dropped to ∼40 mg/dL, while the controls, with glucose concentrations of ∼90 mg/dL, continued to run, despite levels of liver and muscle glycogen, muscle TAG, and plasma ketones, FA, and amino acids equivalent to those of the knockouts. Thus, hypoglycemia, causing central fatigue and an inadequate peripheral energy supply, was the deciding feature in the refusal of both genotypes to run.

In summary, ACSL1 is essential for FA oxidation in skeletal muscle, and acyl-CoA synthesis by other ACSL isoforms cannot compensate (Fig. 7). ACSL1 is located on both the endoplasmic reticulum and the mitochondrial outer membrane in rat liver (38,49) and 3T3-L1 adipocytes (50), but the subcellular location of ACSL1 in muscle has not been investigated. The absence of ACSL1 resulted in an inappropriate dependence on glucose for muscle energy, such that systemic glucose homeostasis became severely compromised. In human and animal models with defective FA oxidation, the hypoglycemia that occurs with fasting or exercise has been attributed to a lack of hepatic FA metabolism and diminished ATP production to power gluconeogenesis. In Acsl1^M−/− mice, however, hypoglycemia occurred despite the fact that hepatic FA oxidation was not impaired. Instead, the muscle-specific defect in FA oxidation drained the glucose supply, and the likely increased use of amino acids within muscle precluded amino acid availability as a substrate for hepatic gluconeogenesis. Thus, during both fasting and endurance exercise, the liver was unable to synthesize sufficient glucose to prevent hypoglycemia, a dramatic example of the interdependence of muscle and liver fuel metabolism.

Funding. This work was supported by National Institutes of Health grants T32HL069768 (T.J.G.), DK058398 (C.B.N.), AG028930 (D.M.M.), and DK59935 and DK59935(ARRA) (R.A.C.); the UNC Nutrition Obesity Research Center grant P30DK56350; and Ellison Medical Foundation grant AG-NS-0548-09 (T.R.K.).

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

Author Contributions. L.O.L. designed and performed the research, wrote the manuscript, and contributed to the writing of the final submitted version of the manuscript. T.J.G., D.S.P., T.R.K., and F.P. performed the research and contributed to the writing of the final submitted version of the manuscript. O.I. performed the research. C.B.N. and D.M.M. contributed to the writing of the final submitted version of the manuscript. R.A.C. designed the research, wrote the manuscript, and contributed to the writing of the final submitted version of the manuscript. R.A.C. 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.