Intramyocellular Lipid Is Associated With Resistance to In Vivo Insulin Actions on Glucose Uptake, Antilipolysis, and Early Insulin Signaling Pathways in Human Skeletal Muscle

A total of 20 healthy men (42 ± 2 years), with a BMI ranging from 21 to 32 kg/m² (25.7 ± 0.7 kg/m², means ± SE) and no family history of type 2 diabetes, were studied on three occasions. On the first visit, IMCL content was measured using proton spectroscopy of the vastus lateralis muscle. On this visit, visceral and subcutaneous fat were also quantified using MRI. The subjects were divided into two groups: low IMCL (LoIMCL; 3.0 ± 0.5 resonance of intramyocellular CH₂ protons in lipids/resonance of CH₃ protons of total creatine [IMCL/Cr_T]) and high IMCL (HiIMCL; 9.5 ± 0.9 IMCL/Cr_T) based on their median IMCL content (6.1 IMCL/Cr_T). The LoIMCL and HiIMCL groups were comparable with respect to several parameters thought to cause insulin resistance (see results). On the second study occasion, maximal oxygen consumption (Vo_2max) was measured directly using an incremental bicycle ergometer test. On the third occasion, insulin sensitivity was measured using the euglycemic-hyperinsulinemic clamp technique. Before and after 30 min of insulin infusion, biopsies of the left and right vastus lateralis muscles were taken, respectively, for determination of early insulin signaling steps. All subjects were right-handed.

Magnetic resonance (MR) images for localization and ¹H-MR spectroscopy were acquired using a 1.5 T MR system (Magnetom Vision; Siemens, Erlangen, Germany). A loop surface coil was used for detection. The subjects lay supine, with the left leg immobilized by firm padding. One-third of the distance from the trochanter major of the femur to the middle of the patella was measured, and the center of the coil was placed in contact with that spot. The ¹H-spectra were obtained from quadriceps femoris (vastus lateralis) muscle. The volume of interest, voxel (13 × 13 × 20 mm³), was placed inside the lateral part of the vastus lateralis muscle to ensure parallel fiber orientation. The position of the voxel was chosen on the T₁-weighted MR images so that it did not contain any visible fat or fasciae known to contain significant amounts of adipocytes (23), which would affect the resonance-frequency shift data. Spatial localization was achieved by using a stimulated echo acquisition mode applied with a repetition time of 3,000 ms, with an echo time of 20 ms, and a mixing time of 30 ms. We used 128 excitations with water presaturation. The resonance at 1.5 ppm originates from the extramyocellular CH₂-protons of lipids (triglycerides and fatty acids) and the resonance at 1.3 ppm from intramyocellular CH₂-protons of lipids (23,24,25) (Fig. 1). Resonance intensities were expressed relative to the resonance at 3.0 ppm, which is derived from the CH₃-protons of total creatine (Cr_T; creatine and phosphocreatine). The spectral data were fitted in time domain with the program MRUI, based on the solving of nonlinear least squares problems (26). The IMCL data were quantified by using Cr_T signal intensity as a reference. The total water-to-Cr_T ratios were constant between subjects (signal intensity ratio 829 ± 46, correlation coefficient for signal intensity of water vs. Cr_T 0.96) and the Cr_T-to-water ratio (×10⁴) averaged 12.9 ± 0.9 and 12.3 ± 1.2 in the LoIMCL and HiIMCL groups, respectively (NS). We chose IMCL/Cr_T for grouping criterion because both signals are of intramyocellular origin and of comparable magnitude of intensity (Fig. 1).

A series of T1-weighted transaxial scans for the determination of visceral and subcutaneous fat were acquired from a region extending from 4 cm above to 4 cm below the 4th and 5th lumbar interspace (16 slices, field of view 375 × 500 mm², slice thickness 10 mm, and breath-hold repetition time divided by the echo time 138.9 ms/4.1 ms), as previously described (27). Visceral and subcutaneous fat areas were calculated using an image analysis program (available at www.perceptive.com/alice.htm). A histogram of pixel intensity in the intra-abdominal region was displayed, and the intensity corresponding to the nadir between the lean and fat peaks was used as a cut point. Visceral adipose tissue was defined as the area of pixels in the intra-abdominal region above this cut point. For calculation of subcutaneous adipose tissue area, a region of interest was first manually drawn at the demarcation of subcutaneous adipose tissue and visceral tissue.

Maximal aerobic power was measured directly, using an incremental work-conducted upright exercise test with an electrically braked cycle ergometer (Ergometer Ergoline 900ERG; Ergoline GmbH, Bitz, Germany) combined with continuous analysis of expiratory gases and minute ventilation (Vmax 229 series; SensorMedics). Exercise was started at a workload of 50 W. The workload was then increased by 50 W every 3 min, until perceived exhaustion or a respiratory quotient of 1.10 was reached. Maximal aerobic power was defined as the Vo₂ during the last 30 s of exercise.

After an overnight fast, whole-body insulin sensitivity was measured using a euglycemic-hyperinsulinemic clamp technique as previously described (28). Briefly, two indwelling catheters were inserted, one in an antecubital vein for infusion of insulin and glucose and another in a heated hand vein for sampling of arterialized venous blood. The rate of the continuous-rate insulin infusion was 40 mU/m² · min. The rate of the glucose infusion was adjusted to maintain euglycemia based on plasma glucose measurements that were performed every 5 min from arterialized venous blood.

Muscle biopsies of vastus lateralis were obtained under local lidocaine anesthesia using a Bergström needle as previously described (29). Biopsies were frozen and stored in liquid nitrogen until analyzed. To minimize variability due to non–muscle tissue contaminants, such as blood, blood vessels, connective tissue, and fasciae and fat deposits, the muscle biopsy (∼50 mg) was lyophilized (2–4 mbar, Edward freeze-dryer) for 14 h and macroscopically dissected free of nonmuscle contaminants using forceps and scalpel at room temperature conditions before homogenization.

Freeze-dried and purified muscle fibers (∼10 mg) were homogenized (HETO, Birkerod, Denmark) for 2 min using a Teflon pestle in 1 ml of buffer A containing (in mmol/l) 25 Tris-HCl (pH 7.4), 10 sodium vanadate, 100 sodium fluoride, 10 sodium pyrophosphate, 10 EGTA, 10 ethylenediaminetetraacetic acid, one phenylmethylsulfonyl fluoride, and 5 μg/ml leupeptin, 5 μg/ml aprotinin, and 1% Nonidet P40. After centrifugation (275,000g, 1 h, plus 4°C) the supernatant was transferred to another tube, and its protein content was measured (BCA assay; Pierce, Rockford, IL).

To determine insulin receptor (IR) expression and IRS-1 associated PI 3-kinase enzymatic activity, the lysates were immunoprecipitated with an appropriate antibody (anti-IR or anti–IRS-1). Rotation with 40 μl of 50% protein A-Sepharose slurry (6MB) for 2 h at 4°C was used to trap the polyclonal antibodies. The beads were washed three times with buffer A and dried quickly. For detection of IR tyrosine phosphorylation and expression, 40 μl Laemmli’s sample buffer was added, and the samples were boiled for 3 min, after which they were applied to SDS-PAGE (30). For IRS-1–associated PI 3-kinase detection, the dried samples were frozen and stored at −80°C before proceeding.

Immunoprecipitates stored at −80°C in Laemmli sample buffer were boiled for 3 min before protein separation by SDS-PAGE. After transferring the proteins to polyvinylidene difluoride membranes (Immobilon-P; Sigma, St. Louis, MO), the membranes were blocked using a 5% milk solution. Western blotting was performed using appropriate antibodies (against phosphorylated tyrosine [pY20] or IR) and enhanced chemiluminescence (ECL) detection. The films were scanned with a tabletop scanner and analyzed by using Image Gauge software (Version 3.12; Fuji Photo Film, Tokyo).

The dried protein-A Sepharose beads stored at −80°C were washed twice with PI 3-kinase reaction buffer containing 20 mmol/l Tris-HCl, pH 7.4, 100 mmol/l NaCl, and 0.5 mmol/l EGTA and then dried quickly and resuspended in the same buffer supplemented with the PI 3-kinase substrate PI (Avanti Biolipids, Alabaster, AL), which was freshly suspended in 50 μl of PI 3-kinase reaction buffer by sonication for 10 min at 4°C. The PI 3-kinase reaction was initiated by the addition of 5 μl of MgCl₂-ATP mixture (200 mmol/l MgCl₂ and 200 μmol/l ATP) containing 5 μCi of [γ-³²P]ATP to the immunoprecipitates and incubated for 25 min at room temperature. The reaction was terminated by adding 150 μl of chloroform–methanol–11.6N HCl (100:200:2 dilution). After adding 120 μl of chloroform and 120 μl of methanol–1N HCl (1:1 dilution) twice, the organic phase was separated by centrifugation at 14,000 rpm for 1 min. The resulting organic phase was dried in SpeedVac (UVS400A; Savant Instruments, Holbrook, NY), dissolved in 15 μl of chloroform, spotted on a silica gel thin layer chromatography plate (Kieselgel 60; Merck, Darmstadt, Germany), and developed in chloroform–methanol–28% ammonium hydroxide–water (43:38:5:7 dilution) for 1 h. The phosphorylated products were visualized and quantified by a Bio-Imaging analyzer (BAS-5000; Fuji, Tokyo).

Polyclonal antibodies against IR (JD 15, JD 433) and IRS-1 (JD 288) were a gift from Professor C. Ronald Kahn (Joslin Diabetes Center, Harvard Medical School, Boston, MA). Horseradish-peroxidase rabbit anti–mouse antibodies, ECL kit, and [¹²⁵I]protein A and were from Amersham Pharmacia Biotech (Buckinghamshire, U.K.). The [γ-³²P]ATP was from NEN Research Products (Boston, MA). PI was from Avanti Polar Lipids (Alabaster, AL). Silica gel thin-layer chromatography plates were from Merck (Darmstadt, Germany). Immobilon-P transfer membranes were purchased from Millipore (Bedford, MA). Protein A-Sepharose 6MB was from Amersham Pharmacia Biotech (Uppsala, Sweden). Reagents for SDS-PAGE were from Bio-Rad (Richmond, CA). All other reagents were purchased from Sigma.

Plasma glucose concentrations were measured in duplicate using the glucose oxidation method (Glucose Analyzer II; Beckman Instruments, Fullerton, CA) (31). Serum free insulin concentrations were determined by double-antibody radioimmunoassay (Insulin RIA Kit; Pharmacia, Uppsala, Sweden) after precipitation with polyethylene glycol. Serum FFAs were determined by the fluorometric method of Miles et al. (32), and serum triglycerides were determined by an enzymatic colorimetric assay (33) with an automated analyzer (Cobas Mira; F Hoffmann la Roche Diagnostica, Basel, Switzerland).

Whole-body insulin sensitivity was calculated from the glucose infusion rate needed to maintain euglycemia divided by body weight. Comparison of HiIMCL and LoIMCL groups was performed using Student’s t test. For comparison of FFA concentrations between the groups, analysis of variance for repeated measures was used. Correlation coefficients were calculated using Pearson’s correlation coefficient. All data are given as the means ± SE.

The LoIMCL and HiIMCL groups were comparable with respect to age, BMI, and physical fitness (Table 1, Fig. 2). The groups also had similar volumes of visceral (3.4 ± 0.5 vs. 2.9 ± 0.3 l, LoIMCL versus HiIMCL, NS) and subcutaneous (2.9 ± 0.3 vs. 2.5 ± 0.5 l, NS) fat. The ratios of IMCL versus Cr_T were comparable between the HiIMCL (15 ± 4) and LoIMCL (17 ± 8, NS) groups. Plasma triglyceride concentrations were slightly higher in the HiIMCL group than in the LoIMCL group, but other lipid concentrations were comparable (Table 1.)

The fasting plasma glucose concentrations (Table 1) as well as the steady-state mean plasma glucose concentrations during the euglycemic-hyperinsulinemic clamp study (5.2 ± 0.1 vs. 5.2 ± 0.1 mmol/l, HiIMCL versus LoIMCL, NS) were comparable between the groups. The HiIMCL group had a 46% higher fasting insulin concentration (Table 1), and the insulin concentration remained slightly but significantly higher during the euglycemic-hyperinsulinemic clamp in the HiIMCL group (450 ± 24 vs. 382 ± 18 pmol/l, HiIMCL versus LoIMCL, P < 0.05), despite a similar insulin priming and infusion protocol. The serum insulin concentrations at the time of the muscle biopsy (30 min) were comparable between the groups (382 ± 26 vs. 340 ± 20 pmol/l, HiIMCL versus LoIMCL, respectively, NS).

Basally, before the insulin infusion, serum FFA concentrations were comparable between the groups (Table 1). During hyperinsulinemia, the FFA concentrations decreased slower (0.47 ± 0.04 vs. 0.31 ± 0.21 mmol/l at 20 min, HiIMCL versus LoIMCL, P < 0.01) and remained higher (0.23 ± 0.02 vs. 0.17 ± 0.02 mmol/l at 120 min) in the HiIMCL compared with the LoIMCL group (Fig. 3). IMCL was significantly correlated with serum FFA at all time points during the insulin infusion (data not shown) and with the mean FFA concentration during hyperinsulinemia (r = 0.53, P < 0.02) but not with the fasting FFA concentration (r = 0.22, NS).

The rate of glucose uptake expressed per body weight, fat-free mass, or body surface area was significantly lower in the HiIMCL compared with the LoIMCL group (3.0 ± 0.4 vs. 5.1 ± 0.4 mg/kg [total weight] · min, 3.8 ± 0.5 vs. 6.1 ± 0.5 mg/kg [fat-free mass] · min, P < 0.005; 124 ± 17 vs. 202 ± 16 mg/m² · min, P < 0.005; HiIMCL versus LoIMCL, P < 0.005) (Fig. 2). The correlation coefficient between IMCL and glucose uptake per body weight was −0.63 (P < 0.01).

Insulin-induced IR tyrosine phosphorylation was determined using Western blotting from anti-IR immunoprecipitates after SDS-PAGE by determining the increase above basal phosphorylation in each subject. There were no detectable differences in the IR expression level by direct Western blotting from the tissue homogenates (data not shown). Normalized for respective protein expression level in each subject, the HiIMCL group had a blunted insulin–stimulated IR tyrosine phosphorylation (from 0.28 ± 0.07 to 0.44 ± 0.10, from basal to insulin-stimulated, respectively; NS) compared with the LoIMCL group (from 0.22 ± 0.08 to 0.55 ± 0.07, P < 0.05). The insulin-stimulated increase in tyrosine phosphorylation of the IR above basal was significantly blunted in the HiIMCL group (57 ± 7%) compared with the LoIMCL group (142 ± 8%, P < 0.001) (Fig. 4). The insulin-stimulated increase in the IRS-1–associated PI 3-kinase activity above basal activity was markedly decreased in the HiIMCL group (42 ± 13% above basal) compared with the LoIMCL group (88 ± 13% above basal, P < 0.05) (Fig. 4).

In the present study, we found two in vivo actions of insulin to be impaired in healthy men with a high content of IMCL (HiIMCL) compared with equally fit men with a similar BMI but a low content of IMCL (LoIMCL). Insulin-stimulated whole-body glucose uptake, which occurs primarily in skeletal muscle (34), was significantly blunted, as was the ability of insulin to decrease circulating FFA concentrations. Both insulin action on glucose and FFA metabolism were significantly correlated with IMCL independent of obesity, age, and physical fitness. We also measured early insulin signaling from vastus lateralis muscle biopsies taken before and after 30 min of in vivo hyperinsulinemia. Insulin stimulation of tyrosine phosphorylation of the IR and IRS-1–associated PI 3-kinase activity were significantly blunted in the HiIMCL group compared with the LoIMCL group.

The HiIMCL and LoIMCL groups had virtually identical BMIs and amounts of visceral and subcutaneous fat, as determined by MRI. This can be considered the first demonstration of an independent association between insulin sensitivity and IMCL in healthy subjects. This may be because previous studies either failed to separately quantitate intra- and extramyocellular fat (4,9,10,35), studied relatives (7) or offspring (11) of patients with type 2 diabetes, studied mixed sexes (5), or failed to find an independent association when BMI was included in the analysis (6). In the latter study, the number of subjects was similar, but the range of BMIs was greater than in the present study (6). The previous studies did not distinguish between intra- and extramyocellular fat, and obese subjects had higher muscle lipid contents than nonobese subjects in some (9,10) but not all (8) of those studies. Taken together, these data support the idea that IMCL can be related to insulin sensitivity independent of obesity, but also that obesity itself is associated with an increase in IMCL. Because the percentage of muscle that is lipid (extramycellular lipid plus IMCL) is ∼1% in lean individuals and 2–3% in obese subjects (10), and because the IMCL component is usually smaller than the extramyocellular peak (25), variation in IMCL does not per se change total fat mass.

In addition to overall adiposity, fat distribution in visceral versus subcutaneous regions and physical fitness are established independent causes of variation of insulin sensitivity (36,37). Visceral and subcutaneous fat have previously been quantitated in one study in which a correlation was found between IMCL and measures of obesity, including BMI, percent fat, and visceral fat (6). In the present study, where an identical number of subjects were studied, such an association was not found, possibly because of a narrower range of obesity. During endurance exercise, muscle triglycerides represent an important source of energy. During low-intensity exercise, most of the oxidized fatty acids are derived from circulating fatty acids (38). During moderate-intensity exercise, the relative contribution of intramyocellular triglycerides increases and can represent nearly half of all fat that is oxidized (38). Physical training increases the capacity of muscle to oxidize fat (39) via mechanisms such as an increase in 1) the density of mitochondria, 2) the activities of respiratory enzymes, and 3) the amount of carnitine palmityltransferase (CPT), which facilitates fatty acid transport across the mitochondrial membrane (40). These adaptations are opposite those found in obese human skeletal muscle, in which lipid oxidation (41), activities of muscle mitochondrial enzymes (42,43), and CPT-1 activity (41) are reduced. Data on IMCL levels are sparse but suggest, perhaps unexpectedly, that muscle triglyceride levels are similar or even higher in trained athletes than in sedentary control subjects (39,44,45,46). Electron microscopy studies have, however, revealed that the anatomical localization of lipid droplets within the myocyte differs between trained and untrained human subjects (47,48). In trained dogs, lipid droplets have been reported to surround and be in direct contact with mitochondria, whereas such an association is not observed in untrained animals (49). In sedentary subjects, lipid droplets may be located in the neutral lipid domains of the plasma membrane itself (47,48), which could interfere with membrane-associated signaling pathways, such as IR tyrosine phosphorylation. These data suggest that in addition to total IMCL content, study of the intramyocellular localization of fat might be of interest. Thus, although there was no correlation between Vo_2max and IMCL in our untrained subjects, a correlation might have been found between physical fitness and the amount of lipid localized in the mitochondrial compared with the plasma membrane compartment.

In the present study, both IR tyrosine phosphorylation and IRS-1–associated PI 3-kinase activation by insulin were significantly lower in the HiIMCL group than in the LoIMCL group. The defect in IRS-1–associated PI 3-kinase activity is similar to that recently found in fatless mice with a twofold increase in muscle triglycerides (12). In humans, an elevation of circulating FFA acutely, about fourfold above basal, decreases whole-body glucose uptake by 50% and abolishes insulin stimulation of IRS-1–associated PI 3-kinase activity (18). IR tyrosine phosphorylation was not measured in these two studies. The present data imply that regardless of whether FFAs originate from intramyocellular triglycerides or from the circulation, they may similarly influence early insulin signaling events. In the present study, the concentration of serum FFAs during hyperinsulinemia, but not under fasting conditions, was significantly correlated with IMCL. This finding is consistent with acute FFA infusion studies in which IMCL increased during hyperinsulinemia but not under basal conditions, perhaps because the combined effects of increased FFA availability and increased inhibition of intramyocellular lipolysis by insulin favor IMCL accumulation (15). It is thus possible that insulin resistance of lipolysis in adipose tissue or other sites, which contribute to circulating FFA, may be responsible for IMCL accumulation. On the other hand, because FFA fluxes were not quantitated in the present study, it is also possible that impaired insulin stimulation of FFA uptake in skeletal muscle or adipose tissue (50) was responsible for the higher FFA levels during hyperinsulinemia in the HiIMCL group compared with the LoIMCL group. Whether it is IMCL or circulating FFA that causes the signaling defects, or whether the signaling defects precede lipid accumulation, cannot be determined in a cross-sectional study.

Several possible mechanisms could link IMCL or circulating FFA to impaired insulin signaling. These include activation of signaling cascades known to downregulate either IR tyrosine phosphorylation or IRS-1 association with PI 3-kinase, such as activation of the hexosamine pathway (51), various protein tyrosine phosphatases (52), or protein kinase C (PKC) (53). The hexosamine pathway can be activated by infusion of either lipid or glucosamine (19). Both agents also decrease IRS-1–associated PI 3-kinase activity (18,51). Activation of the hexosamine pathway also increases O-glycosylation of IRS-1 (51), which may reduce IRS-1 phosphorylation and association with PI 3-kinase, but whether the effect of FFAs on the hexosamine pathway is important in mediating FFA-induced alterations in early insulin signaling and glucose uptake is not known and requires more definite experiments. Finally, it is also possible that increases in metabolites, such as malonyl-CoA (54) and long-chain fatty acid–CoA (LCACoA), may mediate effects of increased lipid availability on glucose metabolism and early insulin signaling. Infusion of FFA in rats increases muscle LCACoA (55). This increase has been suggested to increase PKC θ, a serine/threonine kinase, which is associated with decreased tyrosine and increased serine phosphorylation of IRS-1 (56). However, in humans, PKC θ is lower in obese subjects than in lean subjects (57). LCACoA may also interact directly with glucose metabolism through the inhibition of enzymes such as glycogen synthase (21) and hexokinase (22).

In summary, IMCL appears to be a feature of insulin resistance in skeletal muscle, which is associated with insulin resistance of antilipolysis and with signaling defects similar to those found to be acutely induced by infusion of FFA. It is unclear at present to what extent FFA derived from intra- versus extramyocellular sources regulate insulin signaling. If IMCL is important, it is an attractive therapeutic target because total myocellular lipid can be altered by weight loss (58), a low fat diet (46), and exercise (59,60). Although these manipulations may also change insulin sensitivity by mechanisms other than IMCL, it is of interest that drugs such as troglitazone in Zucker diabetic fatty rats prevent overexpression of genes such as the sterol regulatory element–binding protein-1c (SREBP-1c; a lipogenic transcription factor) and ectopic fat accumulation, insulin resistance, and diabetes, despite increasing body weight and fat mass (61).

This study was supported by grants from the Sigrid Juselius Foundation (to A.V. and H.Y.-J.), Liv och Hälsa (to A.V.), the Academy of Finland (to E.K., S.V., and H.Y.-J.), and the Finnish Diabetes Research Foundation (to A.V.)