Partial Disruption of Lipolysis Increases Postexercise Insulin Sensitivity in Skeletal Muscle Despite Accumulation of DAG

Skeletal muscle accounts for a large part of insulin-stimulated glucose disposal and is an important tissue in regard to peripheral glucose uptake (1,2). The general perception has been that insulin resistance can be attributed to the accumulation of diacylglycerol (DAG) in liver and skeletal muscle (3–5). The mechanism by which DAG is believed to cause insulin resistance is by the activation of protein kinase C (PKC) ε and PKC θ leading to phosphorylation and inhibition of insulin receptor substrate 1 and the insulin signaling cascade in both liver (6,7) and skeletal muscle (8–10). However, not all studies have been able to show these associations among the accumulation of DAG, impaired insulin signaling, and insulin resistance in skeletal muscle (11–13).

In adipose tissue and skeletal muscle, adipose triglyceride lipase (ATGL) has been found to be the major triacylglycerol (TAG)-lipase and hormone-sensitive lipase (HSL), the major DAG lipase (14–16). Consequently, a partial disruption of lipolysis by knocking out the HSL gene leads to the accumulation of DAG (14). DAGs are, in skeletal muscle, generated from the following three different pathways: from hydrolysis of TAG by ATGL, from esterification of fatty acids (FAs) to the phosphoglycerol backbone, and from hydrolysis of phospholipids (17). In addition, DAG can occur in three different stereoisomers, each with unique biological properties, and their structure and localization differ depending on their origin. In COS7 cells, it was shown that ATGL has a preference for hydrolyzing TAG in the sn-2 position, producing primarily sn-1,3 DAGs located in the lipid droplet membrane, whereas re-esterification of plasma FA and the hydrolysis of phospholipids primarily generates sn-1,2 DAG located in the endoplasmic reticulum membrane and at the plasma membrane, respectively (18). Furthermore, in rats fed a high-fat diet it was shown that elevated DAG production from hydrolysis of TAG by lysosomal lipase (LIPA) in the liver was located in two different localizations: in the dense and the light lysosomes and, to a larger extent, in the light lysosomes (19); however, whether lipase has the same preference as ATGL for hydrolyzing TAG in the sn-2 position and producing sn-1,3 DAGs is not known. It could therefore be speculated that the role of DAG in lipid-induced insulin resistance may be dependent on cellular localization, stereo chemical structure, and cell type. No studies have evaluated the effect of the different cellular localization and stereochemical structure of DAG on insulin resistance and insulin-stimulated glucose uptake in skeletal muscle. Therefore, the aim of the current study was to test the hypothesis that the described function of DAGs as mediators of lipid-induced insulin resistance is dependent on the origin and presence of the different DAG isomers. To test this hypothesis, we studied the HSL knockout (KO) mice and their wild-type (WT) littermates. There have been conflicting findings in the literature as to whether mice lacking HSL are insulin resistant. Earlier findings (20,21) suggested that lack of HSL led to impaired insulin sensitivity in skeletal muscle. In contrast, when a hyperinsulinemic-euglycemic clamp combined with D-3-H-glucose were applied, no significant differences were observed in insulin-mediated whole-body glucose uptake between HSL KO and WT mice (22). In addition, HSL KO mice actually are protected from high-fat diet–induced insulin resistance in skeletal muscle (23). Because of these conflicting findings we investigated insulin-stimulated glucose uptake in skeletal muscle in the basal resting state and during the recovery period after an in vivo exercise bout and took advantage of the fact that skeletal muscle subjected to exercise displays increased insulin sensitivity up to 48 h postexercise (POST EX) (2,24,25) and that HSL KO mice accumulate DAG POST EX (26). We combined this and measured insulin-stimulated glucose uptake ex vivo using radiolabeled glucose 90 min after treadmill running in HSL KO mice and their WT littermates.

HSL KO mice were generated by targeted disruption of the HSL gene as previously described (20,27). Female HSL KO mice and WT littermates on a mixed genetic background (C57BL/6 and SV129) were generated by heterozygous breeding and used for experiment at 16–25 weeks of age. The animals were kept in standard laboratory cages on a 12-h light/dark cycle at room temperature, and received standard chow diet (Brogården Aps, Lynge, Denmark) and water ad libitum. All experiments were approved by the Danish Animal Experimental Inspectorate and complied with the “European Convention for the Protection of Vertebrate Animals Used for Experiments and Other Scientific Purposes” (Council of Europe no. 123, Strasbourg, France, 1985).

All animals were acclimatized to treadmill running (TSE Treadmill; TSE Systems, Bad Homburg, Germany) before all experimental testing. During acclimatization, the mice were allowed to rest at a still treadmill for 5 min, and then were set to run for 5 min at 7 m/min followed by 5 min at 14 m/min, and the speed was increased over a 2-min period. All running was performed at 0% incline. The intensities used for acclimatization were low to avoid training adaptation. A shocker at the bottom of the treadmill was used to encourage running if necessary. On the third day, the mice were subjected to a maximal running capacity test in which the mice ran at 10 m/min for 2 min; thereafter, the speed was increased by 2 m/min every second minute until fatigue. Fatigue was defined as spending >10 s at the lower part of the treadmill despite manual encouragement. Three days later, the mice were subjected to an endurance running test at 70% of the individual maximal running speed until fatigue. The main experiment was performed 3 days after the endurance test. On the basis of the results from the maximal running capacity and endurance tests, the mice were set to run for 40 min at 65% of their individual maximal running speed during the experiment. During the main experiment, mice in the basal group were transferred to a still treadmill for 40 min. All the mice were fasted 4 h prior to the experiment.

Basal and insulin-stimulated glucose uptake at rest and 90-min POST EX were measured ex vivo. After the completion of treadmill running, mice were anesthetized by an intraperitoneal injection of sodium pentobarbital (6 mg/100 g body wt). The soleus (SOL) muscle and extensor digitorum longus (EDL) muscle were dissected free, and silk sutures were tied around the intact tendons and transferred to incubation chambers (Multi Myograph System 610M; Danish Myo Technology A/S, Aarhus, Denmark) and suspended at resting tension. SOL and EDL were preincubated in Krebs–Henseleit–Ringer buffer (pH 7.4) containing 7 mmol/L mannitol, 2 mmol/L pyruvate, and 0.1% BSA at 30°C and continuously gassed with a mixture of 5% CO₂ and 95% O₂ for 75 min. SOL and EDL muscles were insulin stimulated 75 min after completion of exercise for 15 min in preincubation buffer containing 100, 200, or 10,000 µU/mL insulin (Actrapid; Novo Nordisk, Bagsværd, Denmark). The SOL and EDL muscles from the contralateral leg were used to evaluate basal glucose uptake without insulin.

Basal and insulin-stimulated glucose uptake were measured ex vivo 90 min after treadmill running by changing the incubation media to the preincubation buffer containing the desired insulin concentration, 1 mmol/L 2-deoxyglucose, 0.056 MBq/mL 2-[³H]deoxyglucose, and 0.0167 MBq/mL [¹⁴C] mannitol (PerkinElmer, Buckinghamshire, U.K.) for 10 min. After tracer incubation, the muscles were washed in Krebs–Henseleit–Ringer buffer, frozen in liquid nitrogen, and stored at −80°C until further analysis.

Glucose uptake was measured on muscle lysate as the accumulation of 2-[³H]deoxyglucose, and [¹⁴C] mannitol was used as a marker of the extracellular space, as described previously (28).

FA metabolism experiments were conducted in incubation reservoirs (Radnoti, Monrovia, CA) using procedures that have been described previously (29,30). Isolated SOL muscles were incubated at 30°C in Krebs–Henseleit–Ringer buffer, pH 7.4, containing 2 mmol/L pyruvate, 2% FA-free BSA, and 0.5 mmol/L palmitic acid. After incubation at resting tension (4–5 mN) for 20 min, the incubation buffer was refreshed and supplemented with 0.5 µCi/mL [1-¹⁴C] palmitate (Amersham BioScience, Buckinghamshire, U.K.). For resting experiments, FA metabolism was measured for 25 min. Contracting experiments of SOL muscle (50 Hz, 350 ms pulse duration, 6 tetani/min) was performed for 25 min. Oxidation of [1-¹⁴C] palmitate was measured in resting and contracting SOL muscles from gaseous ¹⁴CO₂ produced from the exogenous oxidation of [1-¹⁴C] palmitate, as previously described (31). The incorporation of exogenous FAs into DAG and intracellular TAG was measured by thin-layer chromatography (TLC), as described below.

Body composition was measured 3 days prior to the experiment on unanesthetized mice using quantitative MRI by EchoMRI (Echo Medical Systems, Houston, TX).

The quadriceps muscle was freeze dried and dissected free of visible adipocytes under a microscope, as described previously (32), and muscle DAG was measured by TLC. Initially, the lipids were extracted in chloroform-methanol (2:1) using the method of Folch et al. (33) and were dissolved in chloroform. The lipids were separated by TLC using two different separate mobile phases consisting of chloroform-methanol-acetic acid-water (50:50:5:5) followed by petroleum ether-diethyl ether-acetic acid (120:25:1.5). Butylated hydroxytoluene (50 mg/L) was added to both of the mobile phases. The lipids were developed by a 10% copper sulfate pentahydrate and 8% phosphoric acid solution at 120°C for 15 min, and subsequently were visualized and analyzed using a Kodak image station (2000MM; Kodak, Glostrup, Denmark). Specific standards were used to identify sn-1,2 DAG (catalog #D9135; Sigma-Aldrich) and sn-1,3 DAG (catalog #D1639; Sigma-Aldrich).

Intramyocellular TAG (IMTG) was measured by a biochemical method on 1 mg of freeze-dried and dissected quadriceps muscle, as described previously (34,35).

Muscle glycogen was measured as glycosyl units after acid hydrolysis on 30 mg w/w quadriceps muscle, as described previously (36).

Entire SOL and EDL muscles were homogenized and subjected to SDS-PAGE and Western blotting analysis, as described previously (37,38).

A detailed list of the primary antibodies used to determine phosphorylation status and total protein expression is presented in Supplementary Table 1.

All data are expressed as the mean ± SEM. Statistical analysis was performed using IBM SPSS Statistics 20.0 (SPSS, Chicago, IL) and Sigma Plot 12.3 (Systat Software Inc., Erkrath, Germany). A three-way ANOVA with repeated measures was used for comparison of insulin, 90-min POST EX, and genotype. Two-way ANOVA with repeated measures was used for comparison of 90-min POST EX and genotype. Two-way ANOVA was used for comparison of exercise/contraction and genotype in Fig. 3A–F and Fig. 7A–F. Finally, a t test was used for comparison of body composition, and running differences between groups were considered statistically significant at P < 0.05.

There were no differences in body weight (WT 24.7 ± 0.3 g; HSL KO mice 25.5 ± 0.4 g), fat mass (WT 2.6 ± 0.2 g; HSL KO mice 2.9 ± 0.1 g), or lean body mass (WT 19.8 ± 0.2 g; HSL KO mice 20.6 ± 0.2 g) between WT and HSL KO mice. In addition, maximal running capacity (WT 33.4 ± 0.6 m/min; HSL KO mice 33.6 ± 0.7 m/min), running endurance at 70% of maximal running speed (WT 3,018 ± 106 s; HSL KO mice 3,079 ± 97 s), and experimental running speed (WT 21.05 ± 0.53 m/min; HSL KO mice 20.26 ± 0.62 m/min) measured on a treadmill at 0% incline were not different between the genotypes.

Insulin was found to dose-dependently increase glucose uptake in both SOL and EDL muscles, and in both genotypes (Fig. 1A and B). There were no significant differences in glucose uptake between genotypes at basal conditions with no insulin in the media or at 100 or 10,000 µU/mL insulin.

Glucose uptake without insulin exposure, measured 90-min POST EX, was similar to that of resting muscles in both genotypes and in both SOL and EDL muscles (Fig. 2A and C). When stimulated with insulin (100 µU/mL), muscle glucose uptake was significantly higher compared with no insulinat rest, and POST EX in both SOL and EDL muscles in both WT and HSL KO mice (P < 0.01). Furthermore, insulin-stimulated glucose uptake was higher in pre-exercised compared with nonexercised SOL and EDL muscles in both WT and HSL KO mice (P < 0.05) (Fig. 2A and C). In addition, in pre-exercised muscle, insulin-stimulated glucose uptake as well as delta glucose uptake was significantly higher in HSL KO mice compared with WT mice both in SOL and EDL muscles (P < 0.01) (Fig. 2A–D).

Total muscle DAG content at rest was not different between genotypes. After exercise, DAG content remained unchanged compared with at rest in WT mice, whereas an increased DAG content was observed in HSL KO mice (P < 0.05) (Fig. 3A). The majority of the DAG content in the muscle samples was of the sn-1,3 DAG isomer origin both in resting and pre-exercised muscle, and sn-1,3 DAG content was more than 17-fold higher than sn-1,2 DAG content in both WT and HSL KO mice (Fig. 3B). To further visualize the changes within sn-1,3 DAG and sn-1,2 DAG, the data are displayed separately in Fig. 3C and D, respectively. In the resting, nonexercised muscle, sn-1,3 DAG content was similar between genotypes. In the exercised muscle, sn-1,3 DAG content was similar to the resting muscle in WT, whereas an increase in sn-1,3 DAG was observed in exercised muscle from HSL KO mice (P < 0.05) (Fig. 3C). In addition, muscle sn-1,3 DAG content was 43% higher in HSL KO mice compared with WT after exercise (P < 0.01). Muscle sn-1,2 DAG content was significantly higher in HSL KO mice at rest and after exercise (P < 0.05) and was reduced by exercise in both genotypes (P < 0.05) (Fig. 3D).

IMTG content was similar in resting muscle between genotypes. A similar reduction in IMTG content (P < 0.01) was observed after exercise in WT and HSL KO mice (Fig. 3E).

Muscle glycogen levels at rest were similar between genotypes, and were 29% and 46% reduced after exercise in WT and HSL KO mice, respectively, compared with at rest (P < 0.01). There was no difference in glycogen use between genotypes during in vivo treadmill running (Fig. 3F).

Phosphorylation of AKT thr³⁰⁸ and AKT ser⁴⁷³ was increased by insulin stimulation to the same extent in the nonexercised and pre-exercised muscles in both genotypes (Fig. 4A, B, D, and E). Total AKT2 protein expression was not affected by either of the interventions (90-min POST EX and insulin) or by genotype (data not shown).

Insulin stimulation increased AS-160 thr⁶⁴² phosphorylation by ∼139% and 67% in SOL and EDL muscle, respectively, in both WT and HSL KO mice (Fig. 4C and F) (P < 0.05) with no effect of the intervention in either SOL or EDL muscle, and with no differences between the genotypes.

Phosphorylation of AMPK on thr¹⁷² was not affected by the interventions either in SOL and EDL muscles or in genotypes (Fig. 5A and C). Acetyl-CoA carboxylase (ACC) ser²¹² phosphorylation, a downstream target of AMPK, confirmed that AMPK was not activated in either of the interventions in both SOL and EDL muscles in both genotypes (Fig. 5B and D).

GLUT4 protein expression was ∼24% (P < 0.05) and ∼32% (P < 0.01) higher in HSL KO mice compared with WT mice in SOL and EDL muscles, respectively (Fig. 6A and D). Both hexokinase (Fig. 6B and E) and ATGL (Fig. 6C and F) protein expression were similar in both genotypes, and were unaffected by insulin and exercise.

Exogenous palmitate uptake and oxidation in SOL muscle at rest were similar in WT and HSL KO mice (Fig. 7A and B). During muscle contraction, exogenous palmitate uptake and oxidation increased in both genotypes. Exogenous palmitate incorporation into muscle DAG and TAG was similar between genotypes, at rest and with contractions (Fig. 7C and D). The ratio between palmitate esterification and oxidation was similar between genotypes at rest and with muscle contractions, but the ratio was significantly lower with contractions than at rest, indicating that the exogenous FAs are directed toward oxidation rather than esterification during muscle contractions (Fig. 7E).

It is well accepted that in skeletal muscle ATGL is the major TAG lipase (15,16), and HSL is mainly responsible and rate limiting for the hydrolysis of DAG to monoacylglycerol (14). Accordingly, activating lipolysis in the absence of HSL activity is expected to lead to an accumulation of DAGs in skeletal muscle, which was previously found in overnight fasted and exercised HSL KO mice (14,26). In the current study, we demonstrated that insulin-stimulated glucose uptake in skeletal muscle was increased after exercise in both mice lacking HSL mice and their WT littermates. However, this increase was significantly higher in muscles from HSL KO mice than WT mice despite an increased content of DAG in HSL KO muscles. These findings may seem counterintuitive given the proposed inhibitory effect of intracellular DAG on insulin sensitivity and signaling (3). However, a further analysis of the DAG isomer content in the muscles revealed that only the sn-1,3 DAG isomer was increased by exercise and only in the HSL KO muscles. Together, these findings suggest that the increased DAG content originating from ATGL-mediated lipolysis does not impair insulin signaling or effect.

In the current study, total DAG content, measured in the quadriceps muscle was ∼40% higher in HSL KO mice after exercise compared with WT mice. In the gastrocnemius muscle, an even higher total DAG content (∼2.5-fold) was obtained after exercise by Fernandez et al. (26), compared with the present finding, which probably is due to a more energy-demanding exercise protocol in the study by Fernandez et al. (26) or to muscle-specific differences.

Given that DAGs exist in different stereoisomers, depending on how they are generated (18), we examined the two most abundant DAG isomers, sn-1,2 DAG and sn-1,3 DAG, in skeletal muscle to determine their relative contribution to the increased DAG content after exercise in skeletal muscle. We found that the majority of DAG present in skeletal muscle samples were of the sn-1,3 DAG stereoisomers and that the overall level of sn-1,2 DAG represents only a minor fraction of the total DAG pool. Even though not quantified in the study by Fernandez et al. (26), their data suggest a higher sn-1,2 DAG content in the gastrocnemius muscle of HSL KO mice compared with the present findings, resulting in a more equitable distribution between sn-1,2 DAG to sn-1,3 DAG content than in the current study. The reason for this could be due to differences in the metabolic state of the animals and the muscle type being analyzed, as mentioned previously. Only the sn-1,3 DAG levels in skeletal muscle were increased after exercise in HSL KO mice in the current study, whereas the sn-1,3 DAG level was unaltered by exercise in WT mice. This finding is in line with a recent study by Eichmann et al. (18), who showed that ATGL in COS7 cells had a strong preference for the hydrolysis of FA esters at the sn-2 position of the glycerol backbone, resulting in the formation of sn-1,3 DAG.

The increased content of the sn-1,3 DAG isomers apparently had no inhibitory effect on insulin signaling and glucose uptake in skeletal muscle. Support for this notion are the previous in vitro findings showing that only sn-1,2 DAG is able to activate conventional and novel PKCs, whereas the others, sn-1,3 DAG and sn-2,3 DAG, lack this bioactivity (39,40). We observed that the muscle content of sn-1,2 DAG represented only a very small part of the total DAG content in skeletal muscle in the present situation. Nonetheless, sn-1,2 DAG content in skeletal muscle was reduced in both genotypes after exercise and was significantly higher in HSL KO mice both in the basal state and POST EX. Nevertheless, insulin-stimulated glucose uptake was similar in the resting state and higher POST EX in HSL KO mice compared with WT mice, respectively. This indicates that the minor sn-1,2 DAG content in skeletal muscle did not play a role in determining insulin sensitivity in the basal state or after exercise.

The exercise effect on insulin-stimulated glucose uptake in skeletal muscle could be attributed to AMPK activation, as it was shown that activation of AMPK by AICAR or hypoxia increased insulin-stimulated glucose uptake ex vivo in the rat skeletal muscle 3.5 h after removal of the stimulus (41). In addition, recent findings by Kjøbsted et al. (38), demonstrating that insulin-stimulated glucose uptake in skeletal muscle was increased 4 h after AICAR stimulation in WT mice, but not in mice where AMPK activity was blunted, further supporting the notion that AMPK activation is sufficient to increase skeletal muscle insulin sensitivity. Since we observed that AMPK and ACC phosphorylation were similar between genotypes, both at basal and POST EX and with insulin stimulation in both SOL and EDL muscles, AMPK activation cannot account for the higher insulin-stimulated glucose transport in the skeletal muscle of HSL KO mice. Part of insulin sensitization after exercise stems from an effect on GLUT4-mediated glucose transport (42,43). Interestingly, we observed a significantly higher GLUT4 protein expression in HSL KO mice compared with WT mice in both SOL and EDL muscles, which may explain the higher glucose uptake POST EX in HSL KO mice than in WT mice. On the other hand, higher GLUT4 levels have previously been connected to higher insulin responsiveness (44), which did not differ between genotypes in the current study. The mechanisms for the higher GLUT4 protein content in these mice are unknown, as this, to our knowledge, is the first study to measure GLUT4 protein expression in HSL KO mice in skeletal muscle. An upregulation of genes involved in carbohydrate metabolism was previously found in SOL muscles of HSL KO mice, compared with WT mice (26). It could be speculated that the higher GLUT4 expression is due to a compensatory upregulation of GLUT4 protein due to the partial disruption of lipolysis in HSL KO mice in adipose tissue (45) and skeletal muscle (15,26), thereby decreasing the availability of circulating FA and FA generated locally from lipolysis in the muscle, for oxidation, and hence increasing the reliance on carbohydrate as a substrate.

Randle et al. (46) were the first to propose that reduced insulin sensitivity was caused by elevated fat oxidation; however, studies from the last 2 decades reveal that reduced insulin sensitivity is associated with reduced fat oxidative capacity (47). It was previously demonstrated that HSL KO mice during exercise have reduced plasma FA levels and decreased lipid metabolism (27,44). Together with this, the results of the current study suggest that reduced FA availability for mitochondrial oxidation plays a role in the observed increase in POST EX insulin sensitivity in HSL KO mice. One could argue that the reduced lipolysis, and thus the decreased FA oxidation, may be due to a reduced capacity of mitochondrial FA metabolism in HSL KO mice. We measured exogenous FA uptake and oxidation in skeletal muscle at rest and with muscle contractions in HSL KO mice and their WT littermates. The data clearly demonstrated a similar handling of exogenous FAs in both genotypes, which is in line with previous findings (44).

It could be hypothesized that decreased lipid availability should be preferential in increasing insulin sensitivity rather than increasing lipid use. This hypothesis is supported by studies in rodents (48), patients with type 2 diabetes (49,50), and healthy male subjects (51). Our findings suggest that partial disruption of lipolysis in white adipose tissue and skeletal muscle could be beneficial in increasing insulin sensitivity and glucose uptake by increasing the reliance on carbohydrate to maintain the metabolic demand of the cell.

In conclusion, we show that HSL KO mice accumulate DAG in skeletal muscle after exercise and that insulin-stimulated glucose uptake in muscle POST EX was higher in HSL KO mice compared with WT. The sn-1,3 DAG isomer was the DAG isoform with the highest concentration in skeletal muscle and was higher in HSL KO mice after exercise than in WT mice. The sn-1,2 DAG isomer was only found at low levels in both genotypes. Accordingly, these results highlight the importance of taking the chemical structure and cellular localization of DAG into account when evaluating the role of DAG in lipid-induced insulin resistance in skeletal muscle and that the accumulation of sn-1,3 DAG originating from lipolysis does not inhibit insulin-stimulated glucose uptake. This increase in insulin-stimulated glucose uptake could be associated with an upregulation of GLUT 4 protein in HSL KO mice, increasing the capacity to take up glucose upon insulin stimulation. The data suggest that a reduction in FA availability and oxidation should be favored in order to increase insulin sensitivity in skeletal muscle.

Acknowledgments. The authors thank Irene B. Nielsen and Betina Bolmgren (Department of Nutrition, Exercise and Sports, University of Copenhagen) for their skilled technical assistance. The authors also thank Jørgen F.P. Wojtaszewski (Department of Nutrition, Exercise and Sports, University of Copenhagen) and Jonas Thue Treebak (The Novo Nordisk Foundation Center for Basic Metabolic Research, Section of Integrative Physiology, University of Copenhagen) for helpful insight into the experimental protocol.

Funding. This study was financially supported by the Novo Nordisk Foundation (grant 10807) and the UNIK Project Food, Fitness & Pharma for Health and Disease (see www.foodfitnesspharma.ku.dk), which was supported by the Danish Ministry of Science, Technology and Innovation; the Swedish Research Council (project 11284 to C.H.); and the Novo Nordisk Foundation Center for Basic Metabolic Research. The Novo Nordisk Foundation Center for Basic Metabolic Research is an independent Research Center at the University of Copenhagen that is partially funded by an unrestricted donation from the Novo Nordisk Foundation (www.metabol.ku.dk).

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

Author Contributions. A.K.S. and T.J.A. conducted the laboratory experiments, contributed to the analysis of the data, and wrote the first version of the manuscript. T.J.A. and B.K. designed the study. A.B.J. conducted the laboratory experiments and contributed to the analysis of the data. P.S. contributed to the analysis of the data. C.H. donated the hormone-sensitive lipase knockout mice. B.K. wrote the manuscript. All authors contributed to the interpretation of the results, revised the manuscript, and approved the final version of the manuscript. B.K. 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.