Changes in muscle cell proteins linked to insulin sensitivity after high-fat diet and exercise

Regular exercise has beneficial effects on most organs of the body¹ and current public health organizations promote exercise as a cornerstone in prevention, management, and treatment of numerous chronic conditions, including hypertension, obesity, coronary heart disease, type 2 diabetes mellitus, and age-related muscle wasting². The mechanisms underpinning these benefits are complex. Notably, adaptations in exercised skeletal muscle play a key role in locomotion, exercise performance, metabolic responses and regulation of whole body metabolic homeostasis³. Skeletal muscle comprises ∼40% of total body mass and accounts for ∼30% of the resting metabolic rate in adult humans⁴. Importantly, it is also the predominant (∼80%) site of glucose disposal under insulin-stimulated conditions⁵. Accordingly, it is not surprising that insulin-resistant muscle represents a major challenge to maintaining healthy glycaemia and thus is a risk factor for developing diabetes. Exercise improves insulin sensitivity of skeletal muscle, which translates to whole-body glycemic benefits^6,7. This beneficial effect of exercise is recognized in the treatment and prevention of type 2 diabetes, since exercise is often more successful than pharmacological interventions⁸.

Some key exercise signaling proteins have been identified. For example, the 5′ AMP-activated protein kinase (AMPK) has been shown to be an essential signaling nexus for exercise-induced increase in insulin sensitivity^9,10, while the peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1α) is important for mitochondrial biogenesis^11,12. Notwithstanding these important discoveries, recent studies have demonstrated that the signaling response to an acute exercise bout is far more complex than previously assumed^13–16. Whether changes in the protein landscape following chronic exercise are equally dramatic is relatively unknown. Identifying the molecular mechanisms underlying improved insulin sensitivity following exercise training could provide novel therapeutic targets to treat insulin-resistant and type-2 diabetics.

In rodents, a high-fat diet can lead to obesity and insulin resistance and thus represents a model for the human metabolic syndrome¹⁷, an established risk factor for development of type 2 diabetes. Therefore, amelioration of insulin resistance by physical activity might be particularly relevant in conditions of obesity and insulin resistance. Furthermore, the current obesity crisis can in part be attributed to a pervasive availability of palatable, calorie-dense foods and the composition of the diet consumed during training influences both enzymatic adaptations and improvement in exercise performance^18,19. This makes it relevant to understand how ET interacts with diet on the molecular level. Accordingly, we profiled adaptations in the skeletal muscle proteome of mice after 20 weeks of ET and exposure to either chow or HFD. Both ET and HFD increase the lipid oxidative capacity in skeletal muscle^20–22. However, adaptations in lipid oxidative capacity are unlikely candidates for the specific beneficial effects of exercise, since only ET-, not HFD-induced lipid adaptations promote health benefits. A side-by-side comparison of the ET- and the HFD-controlled proteome can identify proteins that are differentially regulated by HFD and ET, thereby generating a unique list of candidate proteins that could have particular health-related bearing on skeletal muscle metabolism.

Sedentary (SED) and ET mice were fed either chow or HFD for 20 weeks. SED mice fed a HFD were ∼25% heavier compared to SED chow-fed mice (Fig. 1A). Both ET groups ran the same distance per day (∼6 km/day) (data not shown). ET lowered body weight in both diet groups (−7%) (Fig. 1A). Fat mass accounted for ∼40% of body mass in HFD-fed mice and for 12% in chow-fed mice (Fig. 1B). ET reduced fat mass and increased lean body mass in both diet groups (Fig. 1B,C). Maximal running speed increased with ET, independent of diet (+17%) (Fig. 1D). HFD decreased insulin action in muscle and importantly, ET increased in vivo insulin-stimulated glucose uptake into skeletal muscle in both diet groups (Fig. 1E), demonstrating a clear benefit of ET on muscle insulin action irrespective of diet.

Proteomic analysis of gastrocnemius muscles identified 2,884 proteins at a 1% peptide and protein false discovery rate (FDR) with 2,391 proteins quantified in at least two biological replicates (Supplemental Table 1). ET in chow-fed mice altered 170 skeletal muscle proteins compared to SED chow-fed animals (q < 0.05 and fold-change +/−50%). Of those, 146 were up- and 24 down-regulated (Fig. 2A). Among these regulated proteins were known ET-responsive proteins, like GLUT4 (+90%) and hexokinase II (HKII; +50%) (Fig. 3). Significantly enriched gene ontology (GO) biological terms in ET chow-fed mice were dominated by biological functions related to fatty acid metabolism, oxidative phosphorylation, and mitochondrial adaptations (Fig. 2D). HFD altered the abundance of 56 proteins (39 up- and 17 down-regulated) compared to SED chow-fed mice (q < 0.05 and fold-change +/−50%) (Fig. 2B). Significantly enriched GO biological terms for the HFD-regulated proteins implicated changes in lipid metabolism pathways (Fig. 2D). ET during HFD feeding (ET on HFD) altered only 29 proteins (20 up- and 9 down-regulated) compared to SED HFD-fed controls (Fig. 2C) and the only significantly enriched GO biological term for ET on HFD was fatty acid metabolism (Fig. 2D).

There was a considerable overlap between the proteins regulated by HFD and by ET, as ET altered 26 of the 56 proteins altered by HFD in the same direction (Fig. 2E, Fig. 4A). For example, PDK4, which promotes FA oxidation by inhibiting the PDH-mediated conversion of glycolysis-derived pyruvate into acetyl-CoA²³, was increased with both HFD and ET (Fig. 3). PDK4 is often considered a key regulator of metabolic flexibility allowing the muscle to quickly adjust to different fuel sources, like fatty acids (FAs) and glucose (reviewed in²⁴). FABP3, which is involved in intracellular fatty acid transport²⁵, was also increased with both HFD and ET (Fig. 3). These changes in PDK4 and FABP3 protein abundance are consistent with several studies showing that both exercise and HFD feeding increase fatty acid metabolic capacity^20,21. This similarity is also evident in the GO terms, as the three biological pathways enriched in the HFD-regulated proteome (lipid metabolism, FA metabolism, and FA beta-oxidation) were enriched to a similar extent with ET (Fig. 2D).

These commonalities help explain why ET resulted in fewer regulated proteins during a HFD (compared to SED HFD mice). This makes the 29 proteins regulated by ET on HFD particularly interesting, because these are associated with improvements in insulin sensitivity based on enhanced insulin action in muscle (Fig. 1E) and their regulation by ET is likely independent of HFD-related processes. In addition, given the considerable overlap of HFD- and ET-regulated proteins, we surmised that the proteins underpinning the detrimental effects of HFD or, conversely, the beneficial effects of ET are the ones differentially regulated by HFD and ET. Such divergently regulated proteins would correlate with impaired and improved insulin sensitivity, making a causal relationship more likely, albeit not definite.

To this end, we sought to identify candidate proteins underpinning the exercise benefit (e.g. improved insulin sensitivity) by filtering for proteins that were either regulated by ET irrespective of diet or that were differentially regulated by HFD and ET (i.e. down with HFD but up with ET, or vice versa). This yielded 15 proteins (Fig. 4B). We tested proteins for which already validated antibodies were available and found that protein abundance of Myoglobin (MB), SLIRP and ACOT13 all increased with ET irrespective of diet (Fig. 4D–G) confirming the proteome data. The proteome data also indicated that the major urinary protein 1 (MUP1) was the most decreased protein on HFD, while ET blunted this HFD effect (Fig. 4B, Fig. S1A). This was confirmed by western blotting, which showed a marked decrease in MUP1 on HFD, while ET increased MUP1 protein abundance (Fig. 4C). The detection of MUP1 protein in skeletal muscle was intriguing, because MUP1 is reported to be a liver-specific protein that may play a role in regulating insulin sensitivity^26,27. A possible explanation is that we had detected liver-secreted MUP1 taken up from the circulation by skeletal muscle. However, Mup1 transcripts were detected in cultured mouse C2C12 muscle cells and in mouse quadriceps muscle (Fig. S1B), suggesting that Mup1 is endogenously expressed in mouse skeletal muscle. Compared to the liver, skeletal muscle Mup1 gene expression is substantially lower, but still robust and several magnitudes higher compared to levels of Mup1 in the hypothalamus (Fig. S1B,C). In line with the alterations in protein abundance, skeletal muscle Mup1 gene expression was 10-fold lower after 20 weeks of HFD, while ET had the opposite effect, increasing Mup1 gene expression 3-fold and 8-fold in chow-fed and HFD-fed mice respectively (Fig. 4H).

Having found that MUP1 is an ET- and HFD-regulated protein in skeletal muscle, we next investigated upstream regulation. A key regulator of energy homeostasis that is activated by exercise and required for many of the exercise-related health benefits is AMPK²⁸. We therefore investigated whether the increased Mup1 expression with ET requires functional AMPK by measuring Mup1 gene expression in skeletal muscle from SED and ET wildtype (WT) or transgenic mice overexpressing an AMPKα2 kinase dead (AMPK KD) construct in striated muscle²⁹. AMPK KD mice exhibit a reduced running capacity, but with ET both WT and AMPK KD mice can adapt to ET and improve their running capacity to similar extent³⁰. In SED WT and SED AMPK KD mice, Mup1 expression was similar (Fig. 4E). With ET, Mup1 expression increased in WT (+250%), but not in AMPK KD mice (Fig. 4E), suggesting that AMPK activation is required for the ET-mediated increase in skeletal muscle Mup1 gene expression.

The induction by ET, possibly via AMPK, makes MUP1 an intriguing candidate that may play a role in the insulin sensitizing effects of exercise. In support of the notion, a previous study reported an improved glucose tolerance in obese diabetic mice following pharmacological treatment with recombinant MUP1²⁷, but it is unknown whether this reflected improved insulin sensitivity at the level of the muscle. To understand whether MUP1 has a direct effect on muscle insulin sensitivity, we employed L6 muscle cells that stably overexpress a myc-tagged version of the insulin-responsive glucose transporter 4 (GLUT4). This allows for quantification of GLUT4 translocation from the inside of the cell to the membrane and this model has been widely used to investigate alterations in insulin sensitivity. When we pretreated these cells with recombinant MUP1 protein, basal GLUT4 translocation was increased (Fig. 5A). Furthermore, MUP1 potentiated submaximal insulin-mediated GLUT4 translocation in a dose-dependent manner, suggesting that MUP1 has direct insulin sensitizing properties in muscle cells (Fig. 5A). Akt signaling, AMPK activity (judged by p-ACC Ser212), and p38 phosphorylation were not altered by MUP1 (Fig. 5B), suggesting that these pathways do not mediate the effect of MUP1 on GLUT4 translocation.

In the present study, we provide a unique resource of proteins regulated by exercise training (ET), high-fat diet (HFD), and ET during HFD in mouse skeletal muscle. This side-by-side comparison allowed us to create a list of candidate proteins that may be particularly relevant for transducing the beneficial metabolic adaptions of exercise. The usefulness of this resource is demonstrated by the identification of MUP1 as a HFD- and ET-regulated skeletal muscle protein that improves insulin sensitivity in muscle cells.

In healthy insulin-sensitive individuals, 50–60% of carbohydrates in a meal are taken up by skeletal muscle³¹. With this large contribution to postprandial glucose disposal, insulin resistance of skeletal muscle⁵ represents a serious challenge to whole body glucose homeostasis. Consequently, maintaining muscle mass and metabolic function is essential for health. Exercise improves skeletal muscle insulin sensitivity^6,7,32, making exercise an efficacious treatment for type 2 diabetes mellitus and numerous other chronic conditions². However, at the population level, the adherence to exercise training is poor. As we are facing an unprecedented obesity crisis with serious comorbidities, like diabetes, deciphering the molecular blueprint that underpins the metabolic benefits of exercise possesses therapeutic potential.

Therefore, we characterized global protein changes in response to ET. The inclusion of HFD and ET on HFD groups is a major strength of this study. Both ET and HFD increase lipid oxidative capacity in skeletal muscle^20,21, but only ET leads to health benefits, suggesting that proteins similarly regulated by ET and HFD are unlikely candidates to transduce the beneficial effects of exercise training. Thus, we were interested in identifying proteins that are differentially regulated by HFD and ET, such as MUP1, which was markedly decreased by HFD but increased with ET.

MUP1 belongs to the lipocalin family, which comprises a large group of small extracellular proteins with diverse functions. Lipocalins have been shown to bind and transport small hydrophobic molecules as well as soluble macromolecules and some lipocalins bind to specific cell-surface receptors³³. For example, lipocalin 2, secreted from bone, is thought to act on the melanocortin 4 receptor (MC4R) in the hypothalamus to reduce appetite³⁴. MUP1 is thought to be predominantly expressed in and secreted from the liver and circulating MUP1 levels are decreased in mouse models of obesity^26,27. Pharmacologically restoring MUP1 levels in male obese db/db mice improves whole body insulin sensitivity²⁷. How this is mediated by MUP1 and on what organs MUP1 is acting is unknown. Our data indicate that MUP1 is expressed in skeletal muscle, and, akin to the circulating MUP1 levels, protein abundance is down-regulated in obesity. Interestingly, we show that MUP1 improves insulin sensitivity of GLUT4 translocation in cultured muscle cells, but it is unclear if this is mediated via specific receptors or if MUP1 is taken up by the muscle to trigger signaling events inside the cell. Our signaling data appear to rule out MUP-1 effects on Akt, AMPK and p38, which all have been implicated in the regulation of GLUT4 translocation in muscle³⁵. Alternatively, MUP1 could also act as a binding partner that is required for other proteins to function. We also found that the ET-dependent induction of Mup1 gene expression required functional AMPK. Since AMPK is not necessary for Mup1 expression in the sedentary state, it is most likely that AMPK-specific pathways activated with exercise regulate Mup1 expression, but the details of this regulation remain to be unraveled.

In addition to MUP1, two other hits stand out to us. One is the SRA Stem-Loop Interacting RNA Binding Protein (SLIRP), which has recently been identified as a novel regulator of mitochondrial protein translation³⁶. In our study, SLIRP was increased with both ET and ET on HFD, while HFD alone had no effect on SLIRP protein abundance. This leads to the intriguing hypothesis that mitochondrial adaptations and possibly improvements in exercise performance with ET are in part regulated by SLIRP. Another intriguing hit is the guanylate binding protein 4 (GBP4), which was increased with HFD whereas both ET and ET during HFD decreased GBP4 protein. GBPs are thought to play a role in the innate immune response and their expression is generally increased by pro-inflammatory stimuli³⁷. Thus the expression pattern may reflect an HFD-mediated inflammatory state that can be ameliorated by ET. Whether genetic downregulation of the protein would confer metabolic protection during HFD feeding in skeletal muscle would be an interesting hypothesis to test.

A limitation of our findings is the use of an animal model. Although there is considerable overlap between humans and mice in metabolic processes and regulation thereof, there are also considerable distinctions, such as the vast difference in metabolic rate between mice and humans¹⁷. Therefore, it remains to be seen whether exercise and diet regulate the same proteins and whether these proteins in turn regulate distinct metabolic processes in human skeletal muscle. This, however, can now be addressed following our contribution in targeted hypothesis driven follow-up studies.

In conclusion, since obesogenic HFD and exercise training have opposite impact on health, we believe that proteins that are differentially regulated by exercise training and HFD are particularly promising candidate proteins to transduce exercise-induced health benefits. This rationale is supported by the identification of MUP1 as a novel diet-, ET- and AMPK-regulated skeletal muscle protein that improves insulin sensitivity of GLUT4 translocation in muscle cell culture. This warrants future investigations into MUP1 and also into the other unique molecular targets we have provided.

We acknowledge the skilled technical assistance of Betina Bolmgren, Nicoline Resen Andersen (Molecular Physiology Group, Department of Nutrition, Exercise and Sports, University of Copenhagen, Denmark) and Emilija Malogajski (Institute for Diabetes and Obesity, Helmholtz Diabetes Center at Helmholtz Zentrum München). We also acknowledge Sara Brandt for her help with editing the manuscript (Institute for Diabetes and Obesity, Helmholtz Diabetes Center at Helmholtz Zentrum München). L6-GLUT4myc muscle cells were a kind gift from Amira Klip (Cell Biology Program, Research Institute, The Hospital for Sick Children, Toronto, Canada). The study was supported by several grants. Danish Research Council (DFF – 4004-00233 to L.S.; DFF – 5053-00155 to M.K.; and DFF-6108-00203A to E.A.R.). The Novo Nordisk Foundation (grant number: 27274 to E.A.R.). Novo Nordisk Foundation Excellence project grant to T.E.J. (#15182) and L.S. (#32082). B.L.P. is an NHMRC Early Career Fellow. D.E.J. is an NHMRC Senior Principal Research Fellow and this work was supported by an NHMRC Project Grant (APP1122376). The contents of the published material are solely the responsibility of the individual authors and the Administering Institution and do not reflect the views of the NHMRC.

M.K., E.A.R., and L.S. designed the study. M.K., B.L.P., P.P., S.H.R. and X.H. conducted the experiments and performed the laboratory analysis. E.A.R., T.E.J., D.E.J., and L.S. took part in conducting the experiments and/or developing the methods. All authors commented on and approved the final version of the manuscript. L.S. is the guarantor of this work and, as such, has full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analyses.

Erik A. Richter, Email: erichter@nexs.ku.dk

Lykke Sylow, Email: lshansen@nexs.ku.dk.