The gut microbiota influences skeletal muscle mass and function in mice

Altered metabolism in skeletal muscle of GF mice

Given that skeletal muscle of GF mice showed signs of atrophy, we next investigated the oxidative metabolic capacity of skeletal muscle. Histological staining revealed reduced activity of the mitochondrial enzyme succinate dehydrogenase (SDH; Fig. 3A) and reduced expression of the Sdh gene (P < 0.05; Fig. 3) in GF mouse muscle compared to muscle of PF mice. Mitochondrial DNA content was also reduced in GF mouse muscle but was restored when GF mice were transplanted with the gut microbiota ofpathogen-free mice (C-GF) (P < 0.05; Fig. 3C). In addition, we observed reduced gene expression of mitochondrial biogenesis markers, such as peroxisome proliferator-activated receptor γ coactivator 1α (Pgc1α) and mitochondrial transcription factor A (Tfam) in skeletal muscle of GF mice compared to PF mice (P < 0.05; Fig. 3D). Moreover, we noticed reduced expression of genes encoding different mitochondrial oxidative phosphorylation complexes in skeletal muscle of GF mice compared to PF mice, including genes encoding cytochrome oxidase subunits of complex IV (CoxVa, CoxVIIb, and CytC) known to be involved in the electron transport chain (P < 0.05; Fig. 3E). Similar observations supporting dysfunctional mitochondrial biogenesis and oxidative capacity were also observed in two other skeletal muscle subtypes, the soleus (oxidative) and extensor digitorum longus (glycolytic) muscle of GF mice compared to PF mice (P < 0.01; fig. S2, A to D). The protein expression profile of the different oxidative phosphorylation complexes of the electron transport chain (fig. S6A) was also altered in muscle from GF mice compared to PF mice. Despite a possible reduction in oxidative metabolic capacity, GF mice performed as well as PF mice when challenged to exercise until exhaustion (fig. S2E).

Open in a separate window

Fig. 3.

Oxidative capacity of the skeletal muscle of GF mice.

(A) Representative images of TA muscle sections from PF, GF, and C-GF mice stained for the enzyme SDH. (B) Shown is expression of the Sdh gene in TA muscle from PF (n = 7), GF (n = 7), and C-GF (n = 9) mice. (C) Quantitative analysis of the ratio of mitochondrial DNA (mtDNA) to nuclear DNA in gastrocnemius muscles from PF (n = 4), GF (n = 5), and C-GF (n = 5) mice. (D and E) Shown are changes in expression of the Pgc1α and Tfam genes (D), and the CoxVa, CoxVIlb, and CytC genes (E) in TA muscles of PF (n = 6), GF (n = 7), and C-GF (n = 9) mice. (F) Shown are changes in expression of genes involved in glucose metabolism (Pfk, Pk, Ldh, and Pdh) in TA muscles of PF (n = 6), GF (n = 7), and C-GF (n = 9) mice. (G) Shown are changes in expression of genes involved in the fatty acid oxidation pathway (Lead, Mcad, and Cpt1b) in TA muscles of PF (n = 7), GF (n = 7), and C-GF (n = 9) mice. (H) Shown is the amount of glycogen in quadriceps muscles of PF (n = 4), GF (n = 3), and C-GF (n = 4) mice. Data are expressed as means ± SEM. Data were analyzed using ANOVA followed by Tukey’s post hoc test and were considered statistically significant at *P < 0.05, **P < 0.01, and ***P < 0.001 between indicated groups.

These results prompted us to undertake further metabolic analyses in GF mice. Analysis of serum metabolic markers revealed reduced quantities of glucose and insulin in GF mice compared to PF mice (P < 0.05; fig. S2F). Further investigation of metabolic pathways for fatty acid and glucose metabolism in GF mouse muscle revealed reduced expression of glycolytic genes encoding the enzymes phosphofructokinase (Pfk), pyruvate kinase (Pk), lactate dehydrogenase (Ldh), and pyruvate dehydrogenase (Pdh) in tibialis anterior muscle from GF mice compared to PF mice (P < 0.05; Fig. 3F). Expression of the genes encoding these enzymes was restored when GF mice were transplanted with the gut microbiota of pathogen-free mice (C-GF) (P < 0.01; Fig. 3F). A similar trend of reduced expression of glycolytic genes was also observed in the soleus and extensor digitorum longus muscles of GF mice compared to PF mice (P < 0.01; fig. S2, G and H). No differences in expression of genes encoding medium-chain acyl– coenzyme A dehydrogenase (Mcad) and muscle-specific carnitine palmitoyltransferase 1b (mCpt1b) involved in fatty acid oxidation (Fig. 3G) or phosphorylation of acetyl–coenzyme A carboxylase (fig. S2, I and J) were observed in tibialis anterior muscle of GF mice compared to PF mice. Cholesterol and free fatty acids in serum also remained unaltered in GF mice compared to PF mice (fig. S2F). Further analysis of glucose uptake in skeletal muscles, using FDG-PET/MRI (fluorodeoxyglucose–positron emission tomography/magnetic resonance imaging) imaging, revealed no obvious differences in glucose uptake for the back and hindleg muscles of GF mice compared to PF mice (fig. S2, K and L). However, we observed an increased accumulation of glycogen in the quadriceps (fast glycolytic) muscles of GF mice compared to PF mice (P < 0.001; Fig. 3H), implying possible impaired utilization of glucose by the quadriceps muscle of GF mice.

To further corroborate our findings, we assessed whether disruption of the gut microbiota–host environment using antibiotics had effects on skeletal muscle. We monitored muscle samples from PF mice exposed to low-dose penicillin to examine consequences of disruption of host metabolic functions mediated by antibiotics through alterations in microbial community composition (16, 17). Whereas low-dose penicillin increased FoxO3 gene expression and reduced MyoD gene expression in skeletal muscle of PF mice (P < 0.05; fig. S3A), the expression of Atrogin-1, Murf-1, Pgc1α, and Tfam genes remained largely unaffected under these experimental conditions (fig. S3, A and B). However, electron transport chain genes (CoxVa, CoxVIIb, and CytC) were down-regulated in PF mice treated with low-dose penicillin compared to untreated animals (P < 0.05; fig. S3C), indicating reduced oxidative metabolic capacity in skeletal muscle.

Altered metabolites in skeletal muscle, liver, and serum of GF mice

To characterize the metabolic phenotype of GF mice in detail, we measured proton nuclear magnetic resonance (NMR) spectra at 600 MHz for the skeletal muscle, liver, and serum samples from GF, PF, and C-GF mice (Fig. 4). Cross-validated principal components analysis and orthogonal partial least squares analysis models revealed differences in ¹H NMR metabolic profiles for the skeletal muscle, liver, and serum from GF mice compared to PF mice and GF mice compared to C-GF mice (fig. S4, A to C). Transplanting GF mice with gut microbiota of PF mice, as observed in C-GF mice normalized the metabolite profiles to those observed in PF mice.

Open in a separate window

Fig. 4.

Metabolite analysis of the muscle, liver, and serum from GF mice.

(A to C) Shown is the average ¹H NMR spectrum of hydrophilic phase after Folch extraction for 25 metabolites. The ¹H NMR spectrum is shown for (A) gastrocnemius muscle from PF (n = 8), GF (n = 8), and C-GF (n = 10) mice; (B) liver tissue from PF (n = 7), GF (n = 8), and C-GF (n = 10) mice; (C) serum from PF (n = 8), GF (n = 8), and C-GF (n = 9) mice. 1, taurocholic acid; 2, bile acids; 3, low-density lipoprotein (LDL); 4, very-low-density lipoprotein (VLDL); 5, leucine; 6, 3-hydroxybutyrate; 7, alanine; 8, acetate; 9, glutamine; 10, glutamate; 11, pyruvate; 12, glutathione; 13, hypotaurine; 14, dimethylamine; 15, sarcosine; 16, trimethylamine; 17, dimethylglycine; 18, unknown (δ 3.11) (s); 19, choline; 20, glycerophosphorylcholine; 21, taurine; 22, betaine; 23, glycine; 24, unknown (δ 3.59) (d); and 25, unknown (δ 3.71) (s). Metabolites in red were found in higher concentrations in GF mice compared to C-GF or PF mice; metabolites in blue were found in lower concentrations in GF mice compared to C-GF or PF mice. s, singlet; d, doublet; a.u., arbitrary units.

Our metabolite profiling identified large quantities of amino acids such as alanine and glycine in skeletal muscle of GF mice (Fig. 4A and Table 1). The expression of the gene encoding alanine transaminase (Alt) that results in transamination of alanine was also increased in muscle of GF mice (P < 0.01; fig. S5A). Typically, the amino acids alanine and glycine, and their carbon skeletons, enter different metabolic pathways to generate energy. However, we did not observe any change in adenosine 5′-triphosphate (ATP) concentrations in the skeletal muscle of GF mice compared to PF mice or C-GF mice (fig. S5B). We next tested whether higher alanine concentrations in the muscle of GF mice served as a source for hepatic gluconeogenesis through the glucose-alanine shuttle. Reduced alanine (Fig. 4B and Table 1) along with low expression of the gene encoding glucose-6-phosphatase (G6Pase), a gluconeogenic marker, was observed in the liver of GF mice compared to PF mice (P < 0.05; fig. S5C).

A large amount of glycine was observed in skeletal muscle of GF mice compared to PF mice (Fig. 4A and Table 1). Apart from a possible increase in protein degradation of muscle tissue, glycine can be derived from intermediates of glycolysis (18). However, no changes in expression of genes encoding key enzymes of this intermediate pathway, such as phosphoglycerate dehydrogenase (Phgdh) and serine hydroxymethyltransferases (Shmt1 and Shmt2) were observed in the muscle of GF mice (fig. S5D). Glycine can also be generated from choline via betaine, dimethylglycine, and sarcosine. Reduced quantities of choline and dimethylglycine along with higher amounts of glycine were found in the serum of GF mice compared to PF mice (Fig. 4C and Table 1). Intermediates in the choline metabolic pathway, glycerophosphocholine and betaine, were also reduced in the liver of GF mice (Fig. 4B and Table 1). However, no changes in expression of the genes encoding the enzymes involved in these intermediate steps: Choline dehydrogenase (Chdh), betaine homocysteine S-methyltransferase (Bhmt), Bhmt2, and sarcosine dehydrogenase (Sardh) were observed in the skeletal muscle of GF mice when compared to PF mice (fig. S5E).

The metabolite profile of the liver from GF mice was characterized by high taurine, hypotaurine, and tauro-conjugated bile acids (taurocholic acid), showing that bile acid metabolism was affected in GF mice compared to PF mice. We also observed increased dimethylamine, a product of choline degradation, in the liver of GF mice compared to PF mice (Fig. 4B and Table 1). Furthermore, glycerophosphocholine, betaine (trimethylglycine), and oxidized glutathione were also present in lower amounts in the liver of GF mice, indicating potential perturbation of the cysteine metabolic pathway although the amount of sarcosine remained high. In addition, glutamine, alanine, leucine, and valine along with glutamate were reduced in the liver of GF mice compared to PF mice. Pyruvate was lower in the liver of GF mice compared to PF mice. Expression profiling of genes encoding enzymes involved in mitochondrial oxidative and glucose metabolism revealed reduced expression of several genes involved in these pathways, indicating metabolic dysfunction in the liver of GF mice compared to PF mice (P < 0.01; fig. S5, F and G).

The amount of glycine was high in serum and muscle of GF mice compared to PF or C-GF mice (Fig. 4C and Table 1). However, the amounts of choline, trimethylamine, and dimethylglycine were lower in serum of GF mice compared to PF or C-GF mice. In addition, acetate, pyruvate, and 3-hydroxybutyrate were lower in serum of GF mice. We also observed differences in serum lipid profiles, with high amounts of very-low-density lipoproteins and low amounts of low-density lipoproteins in GF mice compared to PF mice (Fig. 4), in agreement with an earlier report (19). Overall, the metabolite data indicated disrupted energy homeostasis in GF mice compared to PF mice, with a marked perturbation of the amino acid metabolic pathway (Fig. 5).

Open in a separate window

Fig. 5.

Multicompartment metabolic reaction network.

Metabolites are connected on the basis of the shortest paths of reactions that are mediated by enzymes encoded in the Mus musculus genome or on the basis of nonenzymatic reactions from the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. Metabolites in orange were found in higher concentrations in GF mice compared to PF or C-GF mice. Conversely, metabolites in blue were found in lower concentrations in GF mice compared to PF and C-GF mice. Alanine was higher in the skeletal muscle (gastrocnemius) and lower in the liver. The background shading indicates the three different subnetworks for gastrocnemius muscle (purple), liver (green), and serum (pink). Overlap exemplifies similarity between affected metabolic compartments.

Altered expression of genes encoding NMJ proteins in GF mice

Given that we observed low quantities of choline in the serum of GF mice (Fig. 4C and Table 1) and that choline is a substrate for the derivation of the neurotransmitter acetylcholine and is essential for membrane integrity (20), we screened an array of molecules that affect NMJ development and function including acetylcholine receptors (21). We observed reduced expression of genes encoding different acetylcholine receptor subunits (Chrna1, Chrnb, Chrne, and Chrnd) in tibialis anterior muscle of GF mice compared to PF mice (P < 0.05; Fig. 6A), suggesting a potential impairment of acetylcholine receptor assembly. We therefore monitored the expression of genes associated with formation, maturation, and maintenance of NMJs, including Rapsyn, a 43-kDa receptor-associated synaptic protein (22) and the low-density lipoprotein receptor–related protein 4 (Lrp4), both reported to be important for the development and assembly of acetylcholine receptors in NMJs (23). Expression of both Rapsyn and Lrp4 genes was reduced in GF mice, and this change was reversed when GF mice were transplanted with the gut microbiota of PF mice (C-GF) (P < 0.05; Fig. 6B). Lrp4 associates with muscle-specific kinase (MuSK) to form a receptor complex necessary for agrin to bind (24). Whereas we noted elevated expression of the MuSK gene, no change in expression of the Agrn gene encoding agrin was observed in muscle of GF mice compared to PF mice (P < 0.05; Fig. 6B). The reduced expression of the gene encoding troponin in skeletal muscle of GF mice compared to PF mice and C-GF mice suggested possible impairment of myofiber contractility (P < 0.01; Fig. 6C). We therefore evaluated muscle strength in GF mice. GF mice displayed reduced muscle strength compared to PF mice when examined in a weights test (P < 0.05; Fig. 6D). GF mice also exhibited reduced locomotor and rearing activity compared to PF mice (P < 0.01 and P < 0.001; Fig. 6, ​,EE and ​andF,F, respectively) and C-GF mice (P < 0.05 and P < 0.05; Fig. 6, ​,EE and ​andF,F, respectively).

Open in a separate window

Fig. 6.

Differential expression of mouse neuromuscular junction proteins between GF and PF mice.

(A) Shown are changes in expression of genes encoding acetylcholine receptor subunits (Chrn) in the TA muscle of GF, PF, and C-GF mice. Genes include Chrna1 (PF, n = 7; GF, n = 7; C-GF, n = 9), Chrnb (PF, n = 7; GF, n = 7; C-GF, n = 8), Chrnd (PF, n = 7; GF, n = 7; C-GF, n = 7), and Chrne (PF, n = 7; GF, n = 7; C-GF, n = 7). (B) Shown are changes in expression in TA muscle of PF, GF, and C-GF mice of genes encoding the receptor-associated protein of the synapse (Rapsyn; PF, n = 7; GF, n = 7; C-GF, n = 9), low-density lipoprotein receptor-related protein 4 (Lrp4), and Agrin (Agrn) (PF, n = 7; GF, n = 7; C-GF, n = 9). (C) Shown are changes in expression of the gene encoding fast-twitch troponin (Tnn) in TA muscle of PF (n = 6), GF (n = 7), and C-GF (n = 9) mice. (D) Analysis of hindlimb grip strength using the weights test in PF, GF, and C-GF mice (n = 6 per group). (E to F) Shown is the spontaneous activity of GF, PF, and C-GF mice in the open-field test measured by cumulative distance traveled (E) and cumulative vertical activities (F). PF (n = 8), GF (n = 9), and C-GF (n = 8) mice were monitored over a 2-hour period. All data are expressed as means ± SEM. Data were analyzed using ANOVA followed by Tukey’s post hoc test and were considered statistically significant at *P < 0.05, **P < 0.01, and ***P < 0.001 between indicated groups.

Animals

GF and PF C57BL/6J male mice (6 to 8 weeks old) were raised in the Core Facility for GF Research, Karolinska Institutet, Sweden and Lee Kong Chian (LKC) School of Medicine GF Research facility, Singapore, in accordance with the regulatory standards of each institution. All experiments were approved by respective local ethical committees. GF mice were raised in special sterile isolators. Isolator sterility was analyzed weekly by plating fecal homogenates onto different types of agar plates to detect aerobic and anaerobic bacteria and fungi. To transplant GF with gut microbiota from PF mice, GF mice were conventionalized (C-GF) after weaning by gavage of 100 μl of homogenate generated by dissolving two fecal pellets from PF mice in 1 ml of phosphate-buffered saline (PBS). After gavage, C-GF mice were cohoused with PF mice for 21 days before they were euthanized. The control group was gavaged with sterile PBS. Mice were euthanized under isoflurane anesthesia. All animals were maintained on autoclaved R36 Lactamin chow (Lactamin), provided with sterile drinking water ad libitum, and kept under 12-hour light/dark cycles. Three independent cohorts of PF, GF, and GF mice transplanted with C-GF mice were used with a minimum of five mice in each group. The behavioral assays were conducted in accordance with national guidelines for the care and use of laboratory animals for scientific purposes, with approved protocols from the Institutional Animal Care and Use Committees of Duke–National University of Singapore (NUS) Graduate Medical School, Singapore.

SCFA treatment of GF mice

GF mice (6 to 8 weeks of age) were treated with a cocktail of SCFAs as per a previously published protocol (38). Briefly, mice (n = 6 per group) were fed a mix of SCFAs (67.5 mM sodium acetate, 40 mM sodium butyrate, and 25.9 mM sodium propionate) in their drinking water for a period of 4 weeks. No decrease in water intake was noted, and mice were monitored for signs of dehydration. Sterility of the GF and SCFA-treated GF cohort was assessed twice a week and maintained throughout this study. After 4 weeks, mice were euthanized by inhalation of isoflurane, and muscles were harvested. Tibialis anterior (TA) muscles were used for quantitative polymerase chain reaction (qPCR) analyses.

Antibiotic treatment

C57BL/6J mice were given low-dose penicillin (1 μg/g body weight) via drinking water or no antibiotics (control) for 4 weeks after weaning as previously reported (16).

Weights test

Weights tests were performed to measure muscular strength as described previously (39). Briefly, each mouse was held by the middle of the tail and slowly lowered to grasp the first weight (26 g) in a well-lit fume hood. Upon grasping the wire scale, mice were raised until the weight was fully raised above the bench. The criterion was met if the mouse could hold the weight for 3 s. If the weight was dropped in before 3 s, then the time was noted, and the mouse rested for 10 s before performing a repeat trial. Each mouse was allowed a maximum of five trials before being assigned the maximum time achieved; if it successfully held the weight for 3 s, then it was allowed to progress to the next heaviest weight. The apparatus comprised six weights, weighing 26, 33, 44, 63, 82, and 100 g. All cage mates of each group were tested using a given weight before progressing to the next heaviest weight. The total score for each mouse was then calculated as the product of the number of links in the heaviest chain held for the full 3 s, multiplied by the time (seconds) it was held.

Activity in the open-field task

Locomotor and rearing activities were monitored for individual mice using an automated Omnitech Digiscan apparatus (AccuScan Instruments, Columbus, OH, USA), as described previously (40). Mice were monitored in the open field for 60 min. Locomotion was measured as total distance traveled and rearing as vertical activity.

Treadmill running task

Eight-week-old mice were subjected to exercise on a treadmill (Columbus, OH, USA) using a previously reported exercise regimen (41). Mice in all groups were acclimated to moderate treadmill running (10 m/min for 15 min; inclined at 5 degrees) for 2 days. After acclimation, for the exercise test, mice ran on a treadmill set at a gradually increasing speed from 0 to 15 m/min and then maintained at a constant speed until exhaustion. Mice were removed from the treadmill upon exhaustion and had free access to food and water after the running protocol was completed. Mice were euthanized 3 hours after the exercise test.

Positron emission tomography/magnetic resonance imaging

Mice were fasted for 4 hours before the imaging session. They were imaged under anesthesia using a nanoScan PET/MRI scanner (Mediso, Hungary) under anesthesia (42). The whole-body MRI images were acquired with the 1-T MRI component of the nanoScan PET/MRI scanner. All images and data analyses for the FDG PET images were performed using PMOD (version 3.5; PMOD Technologies). PET acquisition: A 40-min-long three-dimensional (3D) static PET scan was performed after intravenous injection of FDG; the animals were moving freely during the waiting period. The injected radioactivity was 10.0 to 15 megabecquerels, and the maximum injected volume was 0.1 ml.

Cell culture and treatments

Murine C2C12 myoblasts (American Type Culture Collection, USA) were cultured and were considered myotubes after 96 hours of differentiation. Differentiated cells were then treated with 1 mM dexamethasone (Sigma-Aldrich, USA) for 24 hours, along with a cocktail of SCFAs, such as acetate, propionate, and butyrate, based on the molar ratio of their availability in the plasma (60:25:15 for acetate:propionate:butyrate) (43).

In vitro cellular respiration assay

Oxygen consumption in differentiated C2C12 cells in response to bacterial metabolites (i.e., SCFAs) was measured using XF96 equipment (Seahorse Bioscience), as described previously (44).

Metabolite analyses of skeletal muscle, liver, and serum

¹H NMR spectroscopy was performed on the aqueous-phase extracts from the skeletal muscle (gastrocnemius), liver, and serum at 300 K on a Bruker 600-MHz spectrometer (Bruker BioSpin, Karlsruhe, Germany) as described in (45). Metabolic network was generated using the freely available MetaboNetworks software (46).

Sample treatment

Hydrophilic metabolites were extracted from the liver and skeletal muscle (gastrocnemius). Samples were homogenized (64,000 rpm for 2 min) in 7.5 ml of chloroform:methanol (2:1). The homogenate was combined with 1 ml of water, mixed by vortexing, and centrifuged at 13,000g for 5 min. The upper aqueous phase was separated from the lower organic phase and dried using nitrogen gas. Frozen serum samples (−80°C) were thawed and then centrifuged at 2700g for 10 min to remove particulates and precipitated proteins. A total of 300 μl of individual serum samples was prepared with pH 7.4 phosphate buffer, as described previously for high-resolution ¹H NMR spectroscopy (45).

¹H NMR metabolic profiling analysis

The hydrophilic phase of the liver and skeletal muscle was reconstituted in 540 μl of D₂O and 60 μl of phosphate buffer (pH 7.4, 80% D₂O) containing 1 mM of the internal standard, 3-(trimethylsilyl)-[2,2,3,3,-2H4]-propionic acid (TSP). ¹H NMR spectroscopy was performed on the aqueous phase extracts at 300 K on a Bruker 600-MHz spectrometer (Bruker BioSpin, Karlsruhe, Germany) using the following standard 1D pulse sequence with saturation of the water resonance: relaxation delay (RD) – gz,1 − 90° – t − 90° – tm – gz,2 − 90° – acquisition time (ACQ), 90° represents the applied 90° radio frequency pulse, t1 is an interpulse delay set to a fixed interval of 4 μs; RD was 2 s, and tm (mixing time) was 100 ms (45). Water suppression was achieved through irradiation of the water signal during RD and tm. For the liver and muscle samples, each spectrum was acquired using four dummy scans followed by 32 scans and collected into 64K data points. A spectral width of 20,000 Hz was used for all samples. Before Fourier transformation, the free induction decays (FIDs) were multiplied by an exponential function corresponding to a line broadening of 0.3 Hz. ¹H NMR spectra were manually corrected for phase and baseline distortions and referenced to the TSP singlet at δ 0.0. Spectra were digitized using an in-house MATLAB (version R2014a, MathWorks Inc.; Natwick, MA, USA) script. Spectra were subsequently referenced to the internal chemical shift reference (TSP) at δ 0.0. Spectral regions corresponding to the internal standard (δ −0.5 to 0.5) and water (δ 4.5 to 5.5) were excluded.

The serum samples were analyzed using ¹H NMR spectroscopy with two different pulse sequences (45). The first pulse sequence, the so-called NOESY preset sequence, provides an overview of all proton-containing species and yields sharp peaks for small molecule species, broad bands from the lipoproteins (used later for lipoprofile analysis, see below), and a broad largely featureless background from proteins, the most abundant being albumin. The second pulse sequence, the so-called Carr-Purcell-Meiboom-Gill (CPMG) sequence, takes advantage of the different nuclear spin relaxation times between large and small molecules to attenuate the peaks from large molecules (those with shorter spin-spin relaxation times) to leave mainly small-molecule metabolite peaks. The latter have been used for the main analysis. Experiments were performed using a standard 1D ¹H NMR and the CPMG pulse sequence [with the form RD-90°-(t-180°-t) n-ACQ] to attenuate peaks from larger and more slowly moving molecules (such as lipoproteins and lipids) with a view to identifying low-molecular weight metabolites. An RD of 4 s, a tm of 0.01 s, a spin-echo delay of 0.3 ms, 128 loops, and an FID ACQ of 3.067 s were applied. A total of 32 scans were recorded into 73K data points with a spectral width of 20 parts per million (ppm). Each spectrum was normalized to probabilistic quotient normalization. Multivariate statistical analysis based on pattern recognition techniques such as principal components analysis and orthogonal partial least squares discriminant analysis (OPLS-DA) was performed with Pareto scaling in SIMCA-P+14 applied to all data variables. OPLS-DA was used to compare independently each dataset of ¹H NMR metabolic profiles. The robustness of the models was evaluated on the basis of R² (explained variance) and Q² (capability of prediction) values as well as sevenfold cross-validation and class permutation validation.

Metabolite identification

Confirmation of NMR metabolite identities was obtained using 1D and 2D NMR experiments [spike-in of chemical standards, J-resolved spectroscopy (known as JRES), TOtal Correlation SpectroscopY (known as TOCSY), and heteronuclear single quantum coherence (known as HSQC) spectroscopy].

Histological assessment of skeletal muscle fiber morphology and oxidative capacity

For histological staining, serial cross sections (10 μm) were cut from the midpart of the tibialis anterior muscle, fixed in optimal cutting temperature (OCT) compound and frozen in isopentane cooled by liquid nitrogen. Hematoxylin and eosin staining was used for muscle fiber cross-sectional area measurements. The histochemical assay for SDH activity on the tibialis anterior muscle sections was used to distinguish between oxidative and nonoxidative fibers, as described previously (47).

Glycogen measurements in muscle

Glycogen content was estimated in quadriceps muscle tissues using a glycogen assay kit (BioVision, USA) and following the manufacturer’s instructions.

Serum analyses

Serum was prepared from blood collected and stored at −70°C until assayed using the Roche/Hitachi 902 robot system (Roche Diagnostics, Rotkreutz, Switzerland) following the manufacturer’s instructions.

Corticosterone estimation

Corticosterone was estimated in serum using the corticosterone enzyme-linked immunosorbent assay kit (Abcam) following the manufacturer’s instructions.

Gene expression analysis by real-time quantitative RT-PCR

Total RNA was extracted from cells and muscle tissues using TRIzol reagent following the manufacturer’s instructions. For SYBR Green–based quantitative reverse transcription PCR (qRT-PCR), isolated RNA was reverse-transcribed using Superscript II and random primers (Invitrogen). qRT-PCR assays were performed in triplicate, and data were normalized to Rpl27 and Rpl13 and analyzed using qBase software. Details regarding gene-specific primers are provided in table S1.

Mitochondrial DNA copy number quantification

The relative copy number of mitochondrial DNA per nuclear DNA ratio was measured by qPCR.

Western blot analysis

Tissue extracts (50 μg per lane) were prepared in lysis buffer [10 mM tris-HCl (pH 8), 150 mM NaCl, 1% Triton X-100, and protease inhibitors and PhosSTOP (Roche)] and subjected to Western blotting. The Western blot was probed with anti-p-AMPK and anti-AMPK, p-ACC and ACC, p-mTOR and mTOR, p-S6 and S6 (Cell Signaling), and tubulin antibodies. Immunodetection with an appropriate secondary peroxidase-conjugated antibody (DAKO) was followed by chemiluminescence (Amersham). Protein band quantification was performed using LabImage software.

We thank V. Arulampalam and T. Mak for helpful suggestions. We also thank T. W. Ling, A. A. Amoyo-Brion, S. S. Wee, P. Wong, A. Samuelsson, and J. Aspsater for technical assistance. We thank the Center for Germ-Free Research, Karolinska Institutet, Stockholm, Sweden; Mouse Metabolic Evaluation Facility, Genomics Technology Facility, Center for Integrative Genomics, Lausanne, Switzerland; and Duke-NUS Behavioral Phenotyping Core Facility, Duke-NUS vivarium, and SingHealth Experimental Medical Centre for technical help and support.

Funding: This project was supported by grants from the Seventh European Union (EU) program TORNADO (S.P. and W.W.); the Swiss National Science Foundation (SNSF) (W.W. and J.A.); Fondation Suisse de Recherche sur les Maladies Musculaires and Fondation Marcel Levaillant (J.A.); Vetenskapsradet, EU project “Molecular Targets Open for Regulation by the Gut Flora—New Avenues for Improved Diet to Optimize European Health,” Hjärnfonden, Merieux Institute, and Singapore Millennium Foundation (S.P.); and the startup grant from the Lee Kong Chian School of Medicine, Nanyang Technological University (S.P. and W.W.). We also thank the SNSF (P300P3_151157) and European Society for Clinical Nutrition and Metabolism for providing grant support to S.L. I.G.-P. is supported by an NIHR career development research fellowship (NIHR-CDF-2017-10-032). M.J.B. was supported in part by NIH grant number R01-DK090989. J.M.P. is supported by a Rutherford Fund Fellowship at Health Data Research, UK (MR/S004033/1). E.H. is supported by the U.K. Dementia Research Institute at Imperial College, London. The views expressed are those of the authors and not necessarily those of the funders, UK National Health Service, the NIHR, or the U.K. Department of Health.