Adiponectin Corrects High-Fat Diet–Induced Disturbances in Muscle Metabolomic Profile and Whole-Body Glucose Homeostasis

Male AdKO mice (12) on a C57bl6 background and littermates were routinely bred and genotyped in-house prior to being allocated to experimental groups. Animal facilities met the guidelines of the Canadian Council on Animal Care, and all protocols used were approved by the Animal Care Committee, York University. At 6 weeks of age, animals were fed either a regular chow diet or 60% HFD diet for a period of 6 weeks. At the end of the 6 weeks, chow- and HFD-fed AdKO animals received either saline or full-length adiponectin (3 µg/g body wt i.p.) twice daily (at 9:00 a.m. and 6:00 p.m.) for an additional 2 weeks. Adiponectin was prepared in saline as previously described (22). At the time of experimental analysis, animals were simultaneously injected with adiponectin or saline, deprived of food for 5–6 h, and then subjected to hyperinsulinemic-euglycemic clamp. Acute insulin stimulation was via a bolus insulin (4 units/kg body wt) injection via tail vein, and tissues were harvested 5 min later. Skeletal muscle tissue samples collected were snap-frozen in liquid nitrogen (N₂) and together with serum samples from the same animals kept at −80°C until analysis.

Four days prior to the clamp procedure, catheters were inserted into the right jugular vein and left carotid artery, and then the hyperinsulinemic-euglycemic clamp was performed essentially as previously described (23). Serum samples collected at experimental time = −90, −30, −20, −10, 0, 75, 80, 85, 90, 95, 100, 105, 110, 115, and 120 min were used to determine the R_d, R_a, and glycolytic rate. For the glucose tolerance test, the animals were starved 5–6 h before receiving a bolus injection of glucose (2 g/kg body wt i.p.). Tail-vein blood samples were collected after 15, 30, 60, and 90 min, and blood glucose level (millimoles per liter) was determined with a glucometer (Conture; Bayer). Free fatty acid level was assessed using the NEFA-HR(2) kit (Wako, Richmond, VA).

Skeletal muscle strips were snap-frozen in liquid nitrogen and prepared for analysis essentially as described by us previously (24). Tissue-specific triglyceride and ATP content was analyzed by a colorimetric Triglyceride Quantification kit (cat. no. K622) and ATP Assay kit (cat. no. K355) purchased from Biovision (Mountain View, CA). For transmission, electron microscopy soleus muscles were carefully isolated and cut into small pieces. The tissues were fixed in 2.5% glutaraldehyde in Sorenson phosphate buffer for 1 h at room temperature and postfixed with 1% osmium tetroxide prior to dehydration and embedding in Spurr resin in a longitudinal orientation. Ultrathin sections (60–80 nm) were cut and mounted on copper grids, contrasted with 1% uranyl acetate and lead citrate, and observed with a Philips EM210 transmission electron microscope. For ultrastructural analysis, 10–15 random fields of view were obtained from each of three to four mice from each group.

Global biochemical profiles were determined in gastrocnemius muscle as fee for service by Metabolon with additional lipidomic analysis by us as previously described (25–28). Briefly, for global metabolomic analyses the extracted samples were split into equal parts for analysis on the gas chromatography–mass spectrometry and liquid chromatography–tandem mass spectrometry platforms. Also included were several technical replicate samples created from a homogeneous pool containing a small amount of all study samples (Client Matrix). Instrument variability was determined by calculating the median relative SD for the internal standards that were added to each sample prior to injection into the mass spectrometers. Overall process variability was determined by calculating the median relative SD for all endogenous metabolites (i.e., noninstrument standards) present in 100% of the Client Matrix samples, which are technical replicates of pooled client samples. Compounds were identified by comparison with library entries of purified standards or recurrent unknown entities. Identification of known chemical entities was based on comparison with metabolomic library entries of purified standards. The data extraction of the raw mass spectrometry data files yielded information that was examined using bioinformatic approaches, and appropriate quality-control limits were imposed. The contents of DAGs in extracted lipids were quantified by liquid chromatography–tandem mass spectrometry. Five species of DAGs (di-16:0, di-18:1, 16:0–18:1, 18:0–18:2, and 18:0–20:4) were separated by high-performance liquid chromatography with C6-phenyl column, ionized in positive atmospheric pressure chemical ionization mode, and quantified by a multiple reaction–monitoring method. Sphingolipids—ceramides (C14:0, C16:0, C18:0, C18:1, C20:0, C24:0, and C24:1), sphinganine, sphingosine, sphingosine-1-phosphate, and sphingomyelins (C16:0, C18:0, and C18:1)—were separated by high-performance liquid chromatography with C18 columns (3.5 μm, 2.1 × 50 mm, XTerra C18) and ionized in positive electrospray ionization mode. Metabolites of sphingolipids were monitored for multiple reaction–monitoring quantification.

All data were calculated as means ± SEM and further analyzed using nonparametric one-way ANOVA with Tukey posttest or Welch’s two-sample t tests as appropriate without Bonferroni correction. Differences were considered statistically significant at P < 0.05 and further stratified to P < 0.01 and P < 0.001, respectively.

wt or AdKO mice were fed either commercial chow or the 60% HFD at the age of 6 weeks for a further 6 weeks, resulting in four subgroups: wt-chow, wt–60% HFD, AdKO-chow, and AdKO–60% HFD. There was an exaggerated development of whole-body insulin resistance in AdKO mice in response to an HFD, indicated by glucose tolerance tests (Fig. 1A). We then sought to determine whether restoration of adiponectin in AdKO mice would significantly revert HFD-induced metabolic deterioration. HFD-fed AdKO mice were given recombinant full-length adiponectin (fAd) (containing high–, medium–, and low–molecular weight forms) at 3 µg/g body wt i.p. twice a day for an additional 2 weeks after the initial 6-week HFD feeding period and compared with untreated HFD-fed AdKO mice, with chow-fed AdKO mice as the genotype-diet control. Importantly, the fAd treatment protocol used here could successfully bring the circulating level of adiponectin back to a normal range seen in wt mice, and this level was maintained for at least 6 h postinjection (data not shown). During the experimental period, mice continued to gain weight normally with slightly, but not significantly, less weight gain in adiponectin-treated animals (data not shown). We also performed more detailed analyses of whole-body insulin sensitivity and glucose homeostasis using the hyperinsulinemic-euglycemic clamp to determine whether adiponectin could correct HFD-induced insulin resistance. Figure 1B shows that HFD caused a significant decrease in glucose infusion rate and that fAd treatment partially and significantly reversed this defect, indicating that adiponectin treatment can indeed correct HFD-induced whole-body insulin resistance. Upon insulin infusion during the clamp procedure, R_a from liver is suppressed while R_d into skeletal muscle and other peripheral tissues is enhanced. Figure 1C shows that the defect in both R_a and R_d (post–insulin infusion over basal) that develops in response to HFD in AdKO mice was significantly reversed by adiponectin administration. Although adiponectin treatment significantly improved HFD-induced changes in whole-body glucose homeostasis, the changes were not complete reversal, as the negative impact of HFD is so severe as to be dominant (Fig. 1B, C, E, and F). Glycolytic rate represents the ability of the body to use glucose as an oxidative substrate, and the decrease in basal glycolytic rate in mice treated with adiponectin (Fig. 1D) suggests that adiponectin can switch energy substrate preference to using more fatty acids rather than glucose. This decreased basal level of glycolytic rate upon adiponectin treatment corresponded with decreased basal R_a and R_d level (Fig. 1E and F). Interestingly, HFD induced no difference in glycolytic rate in response to insulin stimulation (Fig. 1D). When insulin-stimulated phosphorylation of insulin receptor substrate (IRS) (Y612), Akt (T308), and Akt (S473) was examined in skeletal muscle, we found decreased phosphorylation after HFD, and all of these were significantly recovered in mice treated with adiponectin (Fig. 1G, H, and L). The enhanced levels of circulating nonesterified fatty acid (NEFA) induced by HFD were corrected by adiponectin (Fig. 1J). This correlated with a similar profile of changes in total levels of intramuscular triglyceride (Fig. 1K) and DAG (Fig. 1L), although HFD-induced changes in total ceramide were not corrected by adiponectin (Fig. 1M).

To gain further insight into changes in individual molecular species, we conducted unbiased metabolomic and lipidomic profiling of skeletal muscle samples from the above three groups of mice. Metabolomic analysis included all detectable metabolites (biochemicals) of known identity, and the dataset in this study comprised a total of 267 named biochemicals. In summary, when we compared HFD with chow animals we found statistically significant (P < 0.05) changes for 87 metabolites (70 increased and 17 decreased). Treatment of HFD mice with adiponectin was found to significantly change levels of 61 metabolites (4 increased and 57 decreased). Importantly, a key feature of this dataset was a signature of relatively normalized metabolism across multiple metabolic pathways with adiponectin supplementation in the HFD condition. When we focused on individual DAG species, we observed the most prominent increases in 16:0–18:1, 18:1, and 18:0–18:2 DAGs (Fig. 2). There was a clear trend for adiponectin treatment to reduce intramuscular DAG, in particular 16:0–18:1 DAG (Fig. 2). Significant increases were observed in 18:0, 16:0, and 18:1 ceramides after HFD; however, no significant correction or trend was observed in adiponectin-treated mice (Fig. 2). Other changes of note in response to HFD were elevated 18:1 sphingomyelin, 1-oleoylglycerol, glycerophosphorylcholine, glycerol-3-phosphate, and numerous lysolipids, all of which were attenuated after adiponectin treatment (Fig. 2). We also examined metabolic profiles of biochemicals related to fatty acid metabolism and found clear increases in the majority of long-chain fatty acids and essential fatty acids—trends that were clearly opposed by adiponectin supplementation in these HFD-fed mice (Fig. 3A). We also observed a striking increase in muscle acetyl coA levels, which was significantly reduced by adiponectin (Fig. 3B). In terms of carbohydrate metabolism, increased levels of glucose-6-phosphate and glycerate after HFD were not influenced by adiponectin (Fig. 3B). We observed an unexpected increase in glucose metabolism via the pentose phosphate pathway after HFD, and this was generally opposed by adiponectin, with significant correction of changes in ribulose and xylulose observed (Fig. 3B).

We next used transmission electron miscrocopy (TEM) and showed that in the chow-fed group intermyofibrillar mitochondria showed characteristic distribution and numbers (Fig. 4A). The HFD with or without adiponectin treatment did not alter the intermyofibrillar mitochondrial alignment or number. In contrast, the HFD led to strong pathological changes in subsarcolemmal mitochondrial morphology characterized by mitochondrial swelling, vacuolization, and profound matrix loss. However, no alterations in subsarcolemmal mitochondrial density were observed in response to the HFD challenge. Interestingly, adiponectin treatment alleviated the pathological changes in the mitochondrial morphology caused by the HFD. With HFD feeding, we did not observe strong changes in markers of muscle mitochondrial biogenesis, such as expression of peroxisome proliferator–activated receptor γ coactivator (PGC)1α and various complexes of the mitochondrial electron transport chain, as shown in Fig. 4B–G. Although HFD did not change the ATP production, a dramatic increase of ATP production was observed upon adiponectin treatment (Fig. 4J). To examine whether adiponectin promoted lipid storage in nonmuscle tissues, we examined adipocyte size from these animals. Epididymal fat pad analysis showed that HFD increased adipocyte size, together with profound inflammatory cell infiltration. Adiponectin treatment resulted in significantly smaller adipocyte size and less apparent inflammatory cell infiltration compared with the HFD group (Fig. 4H and I).

Study of amino acid metabolism by metabolomic profiling revealed a trend of significantly increased levels of branched-chain amino acids (BCAAs) after HFD, which were reduced by adiponectin treatment (Fig. 5A). In particular, this was apparent for the valine, leucine, and isoleucine group as well as for serine and threonine. HFD-induced decrease in betaine levels was not corrected by adiponectin. Finally, metabolites of purine and pyrimidine were measured and a decrease in allantoin and increase in cytidine were detected in muscle of mice fed an HFD, and these were corrected by adiponectin (Supplementary Fig. 1B). Other highlights from metabolomic analysis indicating metabolites increased by HFD and corrected by adiponectin include the eicosanoid 15-hydroxy-5Z,8Z,11Z,13E-eicosatetraenoic acid (15-HETE) and the acylcarnitines stearoylcarnitine, oleoylcarnitine, and palmitoylcarnitine (Supplementary Fig. 1A). Bioinformatic analysis of metabolite profiles allowed us to extrapolate changes in biochemical enzymatic activities that were altered by HFD with or without adiponectin. From this, it was apparent that HFD-induced changes in major groups of enzymes such as oxidoreductases, transferases, hydrolases, lyases, and ligases were reversed after adiponectin treatment (data not shown).

This study used metabolomic profiling to identify precise changes in metabolic parameters in skeletal muscle of AdKO mice fed an HFD. AdKO mice have been very informative, shedding light on the beneficial metabolic effects of adiponectin, and here we first validated the changes in metabolism and insulin sensitivity in this model by performing hyperinsulinemic-euglycemic clamp studies to assess whole-body insulin sensitivity and glucose metabolism. This showed that HFD-induced changes in R_a and R_d were significantly, but not fully, improved by adiponectin. This is in agreement with several previous studies showing exaggerated insulin resistance in these mice after HFD feeding (12,29,30). The influence of adiponectin on skeletal muscle metabolism correlated with improved insulin-stimulated tyrosine phosphorylation of IRS and also activation of Akt. Adiponectin also enhanced removal of NEFA from the circulation (12,31). In addition, our powerful unbiased approach uncovered many novel insights into the principal in vivo targets underlying adiponectin’s antidiabetes effect, as discussed in more detail below.

Unbiased metabolic profiling has now been used in several studies related to diabetes (32–39). For example, in a study of mice after 12 weeks on an HFD, changes in numerous metabolic pathways were reported to contribute to insulin resistance (38). A comprehensive metabolomic analysis in liver of mice fed an HFD for 10 weeks showed that lipid metabolites including fatty acids, phosphatidylcholines, and lysophosphatidylcholines; lipid metabolism intermediates (betaine, carnitine, and acylcarnitines); amino acids; acidic compounds; monosaccharides; and serotonin underwent major changes (35). To gain insight into precisely where metabolic changes occurred in the mice used in our study, we used an untargeted metabolomic profiling platform to analyze intermediate metabolites involved in glucose, fatty acid, and amino acid metabolism (27). From all available metabolomic data, our conclusions were based primarily on focusing on the most statistically significant changes, together with additional factors such as biochemical functional similarity, inclusion in a common pathway, and significance with respect to other aspects of the study.

We focused on changes in lipid species that are well established to cause insulin resistance after an HFD, including ceramides and DAGs. Until recently, the functional differences of lipid metabolites with different chain lengths were not distinguished; however, there is now a clear appreciation of the distinct roles of various forms of DAGs and ceramides (40), and these can be studied by metabolomic analysis. In our study, an HFD was associated with multiple impacts on lipid metabolism in muscle tissue in the AdKO mouse. As expected, an HFD caused an increase in circulating NEFA and total lipid accumulation in muscle, both of which were significantly reduced in mice treated with adiponectin. A similar trend was observed when DAG levels were determined. In particular, the major HFD-induced increases were in 18:0–18:2, 16:0–18:1, and 18:1 DAG, while the most pronounced effects of adiponectin occurred in 16:0–18:1 and 16:0 DAG. Our analyses also showed that HFD induced increases in 18:1, 18:0, and 16:0 ceramides, although adiponectin did not correct these changes. We did not find major changes in very long–chain ceramides. Adiponectin has been shown to activate ceramidase and reduce hepatic ceramide levels with formation of sphingosine-1-phosphate (31). However, we found that under the conditions used in our study, adiponectin did not significantly alter ceramide levels in skeletal muscle. Rather, other effects such as decreased plasma NEFA and improved fatty acid oxidation by adiponectin contribute to sensitization of insulin action in muscle.

Data from our unbiased profiling approach is in support of the incomplete fatty acid oxidation concept (6). Complete oxidation of fatty acids to CO₂ involves β-oxidation, tricarboxylic acid cycle (TCA), and oxidative phosphorylation, and defects in any one of these steps will lead to an incomplete oxidation of fatty acids. Our study indicated that HFD disrupted mitochondrial function, including the coupling between β-oxidation and TCA, rather than altering mitochondrial number. Adiponectin supplementation reduced intramuscular content of various free fatty acids, as well as several TCA intermediate metabolites. Several studies have indicated that adiponectin can improve mitochondrial function and oxidative capacity via enhanced biogenesis indicated by elevated expression level of PGC1α (13,14), although this was not observed in our study. Previous studies using transmission electronic microscopy showed that reduced number and damage of subsarcolemmal mitochondria were closely associated with insulin resistance and type 2 diabetes (41). In our chow-fed mice, transmission electron microscopy demonstrated intact sarcomere assembly and mitochondrial structure with clear mitochondrial membrane and an electron-dense matrix. In contrast, a 6-week HFD feeding resulted in mitochondrial swelling, decreased electron density of matrix, and cristae loss, which were particularly evident in the subsarcolemmal region. However, mitochondrial number appeared unchanged upon HFD feeding. Interestingly, adiponectin treatment alleviated the pathological changes in subsarcolemmal mitochondria structure associated with HFD feeding. We detected a significantly elevated level of acetyl-CoA in muscle with HFD feeding, supporting the notion that elevated free fatty acids were processed by β-oxidation in this tissue but not efficiently coupled to the TCA. In agreement with findings of recent studies, we detected increased acylcarnitines in skeletal muscle of mice fed an HFD, and these were prevented by adiponectin. Importantly, we also observed that a robust increase in muscle ATP levels was brought about by adiponectin after HFD. The decreased acetyl-CoA accumulation in presence of adiponectin may indicate more efficient coupling between β-oxidation and TCA in mitochondria, and this correlated with improved ATP production in response to adiponectin. In agreement with recent studies from Kahn and colleagues, who used an HFD for 8 weeks (39), we observed increases in muscle glucose and glucose-6-phosphate content after an HFD at the basal level. Interestingly, we observed activation of the pentose phosphate pathway by HFD, although this pathway is typically not active in muscle cells and only observed in cases such as tissue damage or disease (42). Together, these data suggest that improved mitochondrial function may be another underlying mechanism via which adiponectin exerts its insulin-sensitizing effect. In the absence of altered PGC1α levels in response to adiponectin, the improved mitochondrial function may be due to direct effects on major metabolic enzymes regulating substrate oxidation.

Our metabolomic profiling highlighted an interesting change in amino acid metabolism. The essential BCAAs leucine, isoleucine, and valine are of particular interest based upon our results. BCAA, when combined with overnutrition, have been shown to induce insulin resistance (43). Indeed, a metabolomic analysis performed by Newgard et al. (19) indicated that changes in BCAA profiles in obese subjects contribute to development of obesity-associated insulin resistance. Elevated BCAA metabolism has also been suggested to increase transamination of pyruvate to alanine, with increases in this highly gluconeogenic amino acid subsequently contributing to glucose intolerance. Circulating pyruvate and alanine levels are altered in obese subjects (44,45), and here we observed significantly increased muscle alanine content, which was prevented by adiponectin. BCAA have been show to attenuate insulin signaling via stimulation of mammalian target of rapamycin and S6K1 with phosphorylation of IRS on serine residues (46), which is in keeping with our observation of reduced IRS tyrosine phosphorylation and downstream signaling in HFD-fed animals. It has also been suggested that in HFD-fed mice, BCAA may saturate capacity for mitochondrial fuel oxidation in muscle leading to incomplete oxidation due to lipid-induced mitochondrial stress. Our work thus identifies for the first time reversal of changes in BCAA metabolism as a direct consequence of adiponectin treatment in mice.

Our metabolomic studies also highlight a thus far unappreciated connection between adiponectin and glycerolipid metabolism. The latter is enhanced by nutrient excess, and in recent years the roles of glycerolipids in health and disease have become established (47). Glucose and lipid metabolism converge via the glycerolipid free fatty acid cycle, which involves the substrates glycerol-3-phosphate and fatty acetyl-CoA from glucose and fatty acids, respectively. Numerous studies have demonstrated that increased flux through glycerolipid synthesis impairs insulin action (48,49). Mechanistically, it has recenty been shown that phosphatidic acid synthesized via the glycerol-3-phosphate pathway inhibits mTORC2 activity by decreasing the association of rictor and mTOR, thus impairing insulin signaling (48). We observed a metabolomic profile indicating increases in numerous glycerolipid metabolites in mice fed the HFD, in particular glycerol-3-phosphate. There was a general trend toward correction of changes in glycerolipids mediated by adiponectin. This is in keeping with the fact that enzymes such as glycerol-3-phosphate acyltransferase-1 are downregulated acutely by AMP kinase (50). Thus, our studies identified HFD-induced changes in glycerolipid profiles in skeletal muscle of AdKO mice. They also suggest that adiponectin mediates important corrective effects on these changes in glycerolipid metabolism.

We also observed changes in lysolipid profiles (51). Many lysolipids showed elevated levels with HFD alone and amelioration with adiponectin supplementation. Head group biochemical precursors to major classes of membrane phospholipids, including choline, ethanolamine, and myo-inositol, were observed at higher levels in muscle with HFD than with chow, and adiponectin supplementation was associated with levels for these biochemicals that were similar to those observed in chow-fed animals. A similar pattern was observed for potential markers of phospholipid breakdown, glycerol-3-phosphate (which is also a precursor in the pathways leading to DAG) and, to a lesser degree, glycerophosphorylcholine. Together, these changes in biochemicals associated with phospholipid metabolism suggest increased membrane breakdown in muscle under the HFD condition, with significant amelioration by adiponectin. Additional changes that we observed that are worthy of note include an HFD-induced increase in pyroglutamic acid (5-oxoproline)—a cause of metabolic acidosis via excess consumption of bicarbonate (52). Another unexpected observation was a large increase in kynurenine, a product of tryptophan-degrading enzymes that can be induced by proinflammatory stimuli (53,54). Prior metabolomic analysis of early insulin resistance in mice found lower levels of serum kynurenine (55), and it has been suggested that this defect is a common indication of fatty liver disease and insulin resistance (54). We show here for the first time that adiponectin regulates kynurenine production, although the significance of altered muscle kynurenine levels is unclear.

In summary, in the AdKO mouse model, we found that HFD and adiponectin supplementation were associated with broad-ranging impacts on lipid, carbohydrate, and amino acid metabolism in skeletal muscle. Many of the metabolic changes were anticipated, but we also discovered new insights into the principal in vivo targets underlying the antidiabetes effect of adiponectin. A key signature of adiponectin supplementation in the context of HFD was to normalize metabolism, generally ameliorating to some degree the HFD-induced alterations compared with global metabolism under the chow dietary condition with genetic adiponectin deficiency. Overall, the results support the idea that detrimental health effects associated with metabolic syndrome and type 2 diabetes may be improved by therapeutic strategies targeting adiponectin action. Understandably, there is great interest in therapeutic exploitation of adiponectin’s metabolic effects, and determining changes in metabolomic profiles induced by adiponectin, or lack of adiponectin, will likely have multiple impacts across diagnostics, prognostics, and target identification.

This work was supported by an operating grant from the Canadian Institutes of Health Research to G.S. The authors also acknowledge the Collaborative Research Fund of Hong Kong Research Grant Council (HKU4/CRF/10) to A.X.

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

Y.L. performed all experimental work (except as indicated below), analyzed data, and wrote the first draft of the manuscript. S.T. contributed to data of Fig. 4. T.P. performed lipidomic analysis, analyzed data, and reviewed the manuscript. N.J.M. provided bioinformatic analysis of the metabolomic dataset and reviewed the manuscript. Y.D. reviewed and edited the manuscript. A.X. discussed and advised regarding the study design and reviewed and edited the manuscript. G.S. initiated and designed the study, secured funding, and edited and finalized the writing of the manuscript. G.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank Dr. Saber Saleuddin (Department of Biology, York University, Toronto, Ontario) for advice on transmission electronic microscopy studies.