Increased Reactive Oxygen Species Production and Lower Abundance of Complex I Subunits and Carnitine Palmitoyltransferase 1B Protein Despite Normal Mitochondrial Respiration in Insulin-Resistant Human Skeletal Muscle

The purpose, nature and potential risks of the study were explained to the participants, and written consent was obtained. The protocol was approved by the institutional review board of Arizona State University. Additional information is provided in an online appendix (http://diabetes.diabetesjournals.org/cgi/content/full/db10-0174/DC1).

Participants underwent a euglycemic-hyperinsulinemic clamp with basal muscle biopsies, regarding which the online appendix provides additional information.

Myosin heavy-chain (MHC) isoform composition was determined as previously described (17). MHC I, IIx, and IIa protein bands were visualized using Coomasie stain and quantified using densitometry.

The online appendix provides information regarding the small-scale isolation of mitochondria.

Oxygen uptake by isolated mitochondria was measured using a DW1 Electrode Chamber (Hansatech Instruments, Norfolk, U.K.) in 250 μl respiration medium containing (in millimolar): 100 KCl, 50 3-[N-morpholino] propanesulfonic acid, 10 K^.PO₄, 20 glucose, 10 MgCl₂, 1 EGTA, and 0.2% BSA, pH 7.0, at 37°C. Indexes of mitochondrial function were assessed using the following:: 1 mmol/l pyruvate, 10 mmol/l glutamate, 10 μmol/l palmitoyl-dl-carnitine (PC), and 5 mmol/l succinate. All substrates included 1 mmol/l malate to prime the citrate cycle. Mitochondria (10–50 μg) and substrates were added to respiration medium, and maximal (state 3) O₂ consumption rate (Jo₂) was stimulated with 125 nmol ADP (0.5 mmol/l). Complete phosphorylation of this ADP bolus elicited transition to resting Jo₂ (state 4), which allowed calculation of the ADP-to-O ratio and the state 3–to–state 4 ratio (respiratory control ratio).

Citrate synthase activity was measured using the method of Srere (18). The online appendix provides further information.

The sensitivity of mitochondrial respiration to kinetic and thermodynamic control signals (free ADP and ATP free energy, respectively) was assessed using the “creatine kinase energy clamp” of Messer et al. (19). Mitochondria (40–60 μg protein) were added to respiration medium containing 1.22 mmol/l phosphocreatine (PCr), 5.0 mmol/l creatine (Cr), and 5.0 mmol/l ATP. Creatine kinase was added in excess (30 units creatine kinase; Sigma Chemical, St. Louis, MO) to maintain thermodynamic equilibrium. Free [ADP] ([ADP]_f) was calculated using the equilibrium constant of the creatine kinase reaction (K_CK), assumed to be 150 (20):

The [ADP]_f with the known phosphate ([P_i]) of the medium provided the extramitochondrial free energy of ATP (ΔG_ATP): ΔG_ATP = RTLn(Γ/K_ATP), where R is the gas constant, T is degrees Kelvin, Γ is the mass action ratio of the ATP hydrolysis reaction ([ADP]_f[P_i]/[ATP]), and K_ATP is the equilibrium constant of ATP hydrolysis (225,000, based on the study of Teague and Dobson [20]).

Stepwise additions of PCr established progressively lower [ADP]_f and higher (more negative) ΔG_ATP, suppressing mitochondrial respiration (supplemental Fig. 1). Eadie-Hofstee analysis of the [ADP]_f:Jo₂ relation yielded the K_MADP for respiration. Plots of ΔG_ATP:Jo₂ were near linear (thermodynamic force-flow relation) and were extrapolated through the abscissa to calculate the theoretical ΔG_ATP at zero respiration (static head energy) (Fig. 3 A).

Substrates (see mitochondrial energetics for concentrations of substrates [10 μmol/l rotenone and 1 μg/μl oligomycin]) were added to respiration medium. Horseradish peroxidase (20 mU) and 10-Acetyl-3,7-dihydroxyphenoxazine (17 ng/μl) were added to a 96-well plate. ROS production was initiated by the addition of 10 μg mitochondrial protein. Optical density was measured every 5 min for 30 min (λex: 550 and λem: 590 nm). Due to limited material, 10 μg rat mitochondria were added to the H₂O₂ standard curve to correct for signal quenching by mitochondria.

Forty micrograms of mitochondrial protein were separated by one-dimensional SDS polyacrylamide gel (12%). Proteins were visualized with Coomassie blue (Sigma Chemical, St. Louis, MO). Gel lanes (supplemental Fig. 2) were cut into 20 slices. Each slice was destained, dehydrated, and digested with trypsin as previously described in detail (21). The volume of the extracted peptide solution was reduced to 5 μl by vacuum centrifugation, and 20 μl of 0.1% trifluoroacetic acid/1% formic acid/2% acetonitrile was added.

High-performance liquid chromatography–electrospray ionization–tandem mass spectrometry was performed on a hybrid linear ion trap–Fourier Transform ion cyclotron resonance mass spectrometer (LTQ FT; Thermo Fisher; San Jose, CA) fitted with a PicoView nanospray source (New Objective, Woburn, MA) (refer to Supplemental Information).

Assignment of individual proteins was completed using Mascot (version 2.2; Matrix Science, London, U.K.) and Scaffold (version 02_00_06; Proteome Software, Portland, OR) as previously described (21) (refer to Supplemental Information).

Normalized spectral abundance factors (NSAFs) were calculated for every protein detected in each isolated mitochondria sample as previously described by Hwang et al. (22), who confirmed that this methodology can be applied to complex protein mixtures in a reproducible fashion. In addition, loading 30–120 μg protein lysate generates similar NSAF values (22). Briefly, NSAF values were log₂ transformed to yield a Gaussian distribution. For every protein, the average lean log₂ NSAF values were subtracted from all log₂ NSAF values (both lean and obese groups), resulting in a log₂-fold change for the obese group relative to the lean group. The online appendix provides information regarding immunoblotting.

Lean and obese participants were matched for age, sex, and aerobic capacity (Vo_2max) (Table 1). Obese individuals had higher BMI (P < 0.01), fasting plasma insulin (P < 0.01), A1C (P < 0.01), and fasting plasma glucose (P < 0.05). Insulin-stimulated glucose disposal was lower in the obese group (7.9 ± 0.6 vs. 3.7 ± 0.3 mg · kg⁻¹ · min⁻¹; P < 0.01); suppression of endogeneous glucose output was complete in both groups (0.1 ± 0.1 vs. 0.0 ± 0.1 mg · kg⁻¹ · min⁻¹). Obese participants had a higher proportion of myosin heavy-chain IIx (MYHI) (P < 0.05). No differences in MHC IIa (MYH2) or I (MYH7) were observed. Total mitochondrial protein and citrate synthase activity isolated per muscle wet weight were lower in obese individuals (P < 0.01), although the proportion of mitochondria recovered in both lean and obese groups was similar (∼65%).

ADP-to-O and respiratory control ratios were within previously published ranges for isolated mitochondria from human muscle and did not differ between groups (Table 2). Maximal and resting respiration in response to NADH-linked (pyruvate, malate, and glutamate) and FADH₂-linked (succinate) substrates did not differ between the groups (Fig. 1 A and B). PC oxidation also was similar between both groups.

Rates of ROS production were measured in the presence of oligomycin in a situation where the ETC complexes were highly reduced by the influx of electrons provided by NADH (pyruvate and PC) and FADH₂ (PC). Theoretical sites of ROS production are shown in Fig. 2,A. With pyruvate and PC (both including malate), insulin-resistant muscle mitochondria produced ROS at higher rates than did insulin-sensitive muscle (Fig. 2 B). Rotenone addition to inhibit passage of electrons from complex I to ubiquinone typically results in increased ROS production (23). However, rotenone raised ROS production only in the mitochondria from insulin-sensitive individuals.

Steady-state respiration rates spanning the aerobic range were established by experimentally manipulating extramitochondrial energy phosphate levels (ADP) with the creatine kinase energy clamp. Supplemental Fig. 1 shows a typical experiment using stepwise addition of phosphocreatine. Eadie-Hofstee plots of Jo₂ vs. Jo₂/[ADP] yielded similar values for the K_MADP for respiration (35 ± 2 μmol/l for lean vs. 31 ± 4 for obese; P = 0.33). Figure 3,A shows a typical relationship between ΔG_ATP and oxygen consumption. The ΔG_ATP:Jo₂ (force flow) relation extrapolated through the abscissa indicated a higher (more negative) static head ΔG_ATP in insulin-resistant muscle mitochondria (P = 0.003) (Fig. 3 B). The slope of the force-flow relationship in the lean group showed a trend to be higher than that in the obese group (95.4 ± 7.1 vs. 75.4 ± 8.8 nmol O₂ · min⁻¹ · kcal⁻¹ · mol⁻¹ (P ≤ 0.10) (supplemental Fig. 3).

The number of unique proteins assigned per mitochondrial isolate was similar between the groups (lean 425 ± 19 vs. obese 422 ± 29 proteins; P = n.s.), as were the number of total known mitochondrial proteins (lean 248 ± 8 vs. obese 236 ± 13 mitochondrial proteins; P = n.s.). A composite list assigned 691 unique proteins. In addition, the total spectra generated solely by known mitochondrial proteins did not differ between groups (lean 19,683 ± 1,421 vs. obese 18889 ± 1,238; P = n.s.). Approximately 60% of total spectra were assigned to known mitochondrial proteins.

Table 3 depicts the number of assigned subunits in at least half of all participants. Complex II (succinate dehydrogenase) and III (ubiquinone: cytochrome c reductase) subunits were present in equal numbers per mitochondrial mass in insulin-sensitive and insulin-resistant muscle. A number of complex I subunits (P ≤ 0.05: NDUFC2, NDUFA8, NDUFA11, NDUFS2, and NDUFB8; P ≤ 0.10: NDUFS5, NDUFB3, and MTND1) were observed to be lower in abundance in mitochondria from the obese group. The iron-sulfur cluster scaffold protein (NFU1) was also lower in the obese group (P < 0.05). Only one subunit from complex IV was significantly decreased in obesity: COX5B (Table 4).

Carnitine palmitoyltransferase 1B (CPT1B) was significantly lower in the obese group. Methyl-malonate-semialdehyde dehydrogenase (ALDH6A1) and propionyl-CoA carboxylase β were also lower in mitochondria from the obese group. These proteins are involved in valine and valine/isoleucine oxidation.

MYH2, coding for myosin type IIa, was higher in the obese versus lean participants; MYH7 (myosin type I) was unchanged. MYH1, coding for myosin type IIx, was higher in the obese participants. In agreement with the myosin isoforms, the Ca²⁺ ATPase fast twitch isoform (ATP2A1) was significantly higher and the slow twitch isoform (ATP2A2) tended to be lower (P = 0.07) in mitochondria from obese individuals. CASQ2, the major calsequestrin isoform in cardiac muscle, comprises 50% of the calsequestrin population in slow oxidative muscle fibers. CASQ2 relative abundance was lower in mitochondria from obese individuals.

Mitochondrial lysates were probed for COXIV abundance, and no difference was observed between lean and obese groups (P = 0.4), as predicted by the mass spectrometry data (Fig. 4). ALDH6A1 abundance normalized to COXIV (P = 0.06) (ALDH6A1 abundance without normalization to the COXIV loading control resulted in P = 0.04) showed a 59% lower abundance in obesity by immunoblot—comparable with the 52% lower abundance determined by mass spectrometry.

It has been proposed that excess lipid availability combined with a lower muscular capacity to oxidize fat (24) may lead to accumulation of intramyocellular lipids (25). This in turn is thought to lead to insulin-signaling abnormalities and insulin resistance (1). Lower oxidative capacity could be due to lower mitochondrial content, a defect intrinsic to the individual mitochondrion, or both. Mitochondrial content was significantly lower in the obese individuals, consistent with previous reports (2,,–5). To determine the presence of intrinsic defects, mitochondria were isolated from lean, insulin-sensitive and obese, insulin-resistant human vastus lateralis muscle. The lower mitochondrial content but similar Vo_2max values in the obese versus lean volunteers may appear contradictory. However, Vo_2max primarily reflects cardiovascular, rather than mitochondrial, parameters (26,27).

With regard to intrinsic characteristics of the mitochondria, we used the substrate combinations of pyruvate, glutamate, and pyruvate plus glutamate (all including malate), which rely on complex I activity, whereas the addition of succinate recruits complex II activity. Fatty acid oxidation requires complexes I and II and the electron transfer flavoprotein shuttling electrons from dehydrogenases within the β-oxidation pathway to the electron transport chain (ETC). Basal and maximal respiration rates, regardless of substrates, did not differ in the obese participants—a result similar to that in a recent study (11). These findings are in accordance with some (9,28) but not all (8,10) studies assessing mitochondrial function in insulin resistance and type 2 diabetes. Differences between the present results and findings of decreased mitochondrial function in diabetes could be due to the possibility that insulin resistance itself is not associated with gross changes in mitochondrial function or mitochondrial protein abundance; rather, mitochondrial dysfunction may be more associated with hyperglycemia. Results of a whole muscle proteomics comparison of lean and obese normoglycemic and type 2 diabetic individuals showed that decreases in ETC components and other mitochondrial proteins are most pronounced in the diabetics, with more moderate changes in the obese group (22). In addition, most of the earlier studies did not report aerobic fitness measurements (8,9,28), whereas one study did match the control and diabetic subjects for whole-body aerobic power (Vo_2max) (10). The lean and obese individuals in the present study had comparable Vo_2max and glycemia, and their muscle mitochondria displayed similar maximal substrate oxidation rates, coupling, and respiratory control—in contrast to the study of Phielix et al. (10). These differences suggest that the progression from insulin resistance to type 2 diabetes may result in functional defects in SM mitochondria. On the other hand, our data suggest that reports of decreased in vivo ATP turnover in resting insulin-resistant muscle (6,7,29) most likely reflect decreased cellular ATP demand, given that 1) in vivo mitochondria rest at <2% of their oxidative potential (30), 2) the differences in observed mitochondrial content were rather small, and 3) insulin resistance appears to improve, not degrade, the cellular energy state defended by muscle mitochondria. If higher ΔG_ATP had any influence on resting ATP turnover, it would be predicted to be in the opposite direction (31). Our results show that there are no intrinsic defects in mitochondrial oxidative capacity in insulin resistance without hyperglycemia. Taken together, these data indicate that with regard to substrate oxidation, the findings of decreased ATP turnover in vivo (6,7,29) likely are explained on a gross level by a decrease in mitochondrial content in insulin-resistant muscle.

Nevertheless, there were more subtle intrinsic differences in the function of mitochondria from insulin-resistant muscle. In resting muscle, mitochondrial oxidative flux is extremely low (32), with very low [ADP]_f and robust (highly negative) ΔG_ATP (31). Contractile activity can elicit extremely high oxidative flux, revealing a cellular aerobic scope spanning two orders of magnitude (30). The metabolic signals that control respiration over this range include both kinetic (e.g., [ADP] or K_MADP) and thermodynamic (ΔG_ATP) components. The creatine kinase energy clamp provides an in vitro evaluation of the mitochondrial sensitivity to these signals. In the present study, the K_MADP determined using the creatine kinase clamp for respiration was not different between groups. However, extrapolation of the thermodynamic force-flow relation to the hypothetical zero Jo₂ resulted in significantly more negative (higher) static head ΔG_ATP in mitochondria isolated from insulin-resistant muscle. This finding is consistent with the observation of Petersen et al. (6), who used ³¹P–nuclear magnetic resonance to show that the inorganic phosphate–to–phosphocreatine ratio is lower in insulin-resistant than in insulin-sensitive muscle. Lower inorganic phosphate and more robust ΔG_ATP are characteristics that differentiate type II from type I mammalian muscle cells (33). It is therefore noteworthy that insulin-resistant muscle showed trends toward greater expression of type II myocyte proteins, e.g., MYH2 (type IIa), MYH1 (type IIx), ATP2A1 (Ca²⁺ ATPase fast twitch isoform), and higher PYGM (glycogen phosphorylase).

Taken together, these data suggest that, at the low metabolic turnover rates associated with rest, mitochondria from insulin-resistant muscle maintain higher thermodynamic driving forces. Because the protonmotive force and the oxidation-reduction energy (redox pressure) down the electron transport chain are the energetic precursors to ΔG_ATP, higher ROS production rates by mitochondria from the obese group would be the predicted outcome. This result was experimentally observed. In the absence of rotenone (Fig. 2 A), ROS production was higher in the obese individuals, suggesting that muscle mitochondria from obese individuals produce more ROS at complex I and/or complex III. Complex III has been labeled as a higher contributor of ROS in SM mitochondria (14). Secondly, the level of long-chain fatty acid oxidation intermediates within the matrix, which have been postulated to be elevated in insulin resistance due to incomplete β-oxidation (34), dictates the amount of ROS produced (35) specifically at the level of complex III and the electron transfer flavoprotein (35). Higher rates of ROS production in insulin resistance have been reported previously (16) and precede the onset of the diabetic state in mice on a high-fat diet (36). Lower rates of ROS production in obese individuals, but not type 2 diabetic individuals, have been reported (37). However, this study did not report maximal ROS production rates because mitochondria were not incubated with oligomycin during the measurements. On the other hand, the insulin-sensitizing effect of exercise training is abrogated upon consumption of antioxidants, supporting a positive role for ROS in the modulation of insulin sensitivity (38). Additionally, mice lacking glutathione peroxidase are protected from insulin resistance induced by high-fat diets (39). Therefore, it is possible that higher ROS production rates observed in isolated mitochondria from obese individuals may be a compensatory mechanism to increase insulin action; this possibility warrants further investigation because ROS have also been identified as a causal factor of insulin resistance (12).

In order to guide future research efforts studying the role of the mitochondrion in insulin resistance, we performed an analysis of the abundance of individual proteins in mitochondria (22,40). We used equal amounts of isolated mitochondrial protein for proteomic analysis, resulting in a comparison of relative protein abundance per mitochondrial mass. Several individual subunits of ETC complex I were significantly reduced in obese, insulin-resistant individuals. Complex I is a 980 kDa conglomerate comprised of 46 subunits in mammals (41). Fourteen of these 46 subunits are the core subunits, responsible for the location of the redox centers (flavin mononucleotide and eight Fe-S cluters) (41). The remaining 32 subunits are less well characterized: they may be involved in maintaining the structure of the complex, in linkage of the complex to the inner mitochondrial membrane, or even in altering ROS generation (41). Five complex I subunits were significantly lower in the obese individuals. These subunits were distributed among the four subcomplexes of complex I (NDUFB3/B8/C2 in subcomplex Iβ, ND1 in Iγ, NDUFA8/A11/S5 in Iα, and NDUFS2 in Iλ) (41). A negative correlation has previously been found between the degree of insulin resistance and complex I activity (4). NDUFS2 and ND1 are two subunits found within the core of complex I. A lower abundance of these two subunits did not translate to a detectable decrease in NADH-linked respiration.

The present study also shows a lower abundance of the muscle outer-mitochondrial membrane carnitine palmitoyltransferase 1B (CPT1B). Cpt1B transesterifies long-chain fatty acyl-CoAs to fatty acylcarnitines, which are transported into the mitochondrial matrix for β-oxidation (42). Its expression is dependent on peroxisome proliferator–activated receptor coactivator-1α (PGC-1α) (43) transcription factor activity. Reduced Cpt1B protein level could be a direct consequence of lower PGC-1α gene expression reported in insulin-resistant muscle (44,45). Furthermore, increasing Cpt1B protein levels in SM increased fatty acid oxidation and lowered the severity of insulin resistance in rats on a high-fat diet (46). In the present study, the utilization of PC in the assessment of fatty acid–mediated mitochondrial respiration bypassed the highly regulated CPT system. Our results therefore show that β-oxidation activity per se is not impaired in insulin-resistant muscle but, rather, that the protein determining the rate of fatty acid entry into mitrochondria, Cpt1B, was decreased.

The last set of proteins altered in the proteomics analysis involved enzymes involved in BCAA catabolism. MMS dehydrogenase and propionyl-CoA carboxylase are required for valine and isoleucine metabolism before entry into the tricarboxylic acid (TCA) cycle as succinyl-CoA. Decreased BCAA oxidation, together with increased proteolysis in insulin resistance (47), could lead to increased intracellular BCAA concentrations. Such an increase could activate mammalian target of rapamycin signaling, which can inhibit insulin signaling and produce insulin resistance (48). Moreover, recent metabolomic studies in humans and BCAA/high-fat feeding studies in rodents indicate that alterations in BCAA metabolism are associated with insulin resistance (49,50). Alternatively, or in addition, it is possible that decreased entry of BCAA into the TCA cycle could reduce TCA cycle flux and lead to accumulation of incompletely oxidized fatty acids, as suggested by Koves and Muoio (34). A decreased CPT1B abundance then could be a compensatory effort to protect the cell from accumulation of incompletely oxidized fatty acids and other fatty acid metabolites. A caveat is that this scenario would imply a rate-determining role of BCAA-derived TCA intermediates.

In conclusion, mitochondria isolated from skeletal muscle from obese and overweight insulin-resistant volunteers do not have an intrinsic abnormality in fuel oxidation. However, measurements of citrate synthase activity and mitochondrial protein indicated a decrease in mitochondrial mass. Therefore, in order to maintain similar respiration rates, we propose that mitochondria in insulin-resistant muscle undergo adaptations that allow them to generate a more negative ΔG_ATP. A consequence of this higher ΔG_ATP would be increased ROS production rates at rest (51) as found here. We propose a model for these data shown in Fig. 5.

We also defined new abnormalities in mitochondria from human insulin-resistant muscle using mass spectrometry. A number of complex I subunits were found to be lower in mitochondria from insulin-resistant muscle, but the lack of characterization of function of most complex I subunits makes it difficult to interpret the consequences of this finding. However, these differences may be related to differences in ROS production. Mitochondria from insulin-resistant muscle also was found to contain less carnitine palmitoyltransferase 1β, offering an explanation for the intramuscular lipid accumulation observed in insulin-resistant muscle (25). In addition, enzymes involved in BCAA metabolism were lower in abundance in insulin-resistant muscle. In addition to the lower mitochondrial content in insulin-resistant muscle, perturbations intrinsic to mitochondria could alter ROS production, limit the entry of fatty acids into the matrix and affect the overall oxidation rate of BCAA.

This study was supported in part by National Institutes of Health grants R01DK47936 (to L.J.M.), R01DK66483 (to L.J.M.), R01DK081750 (to Z.Y.), and the American Diabetes Association.

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

N.L. researched data, contributed to discussion, wrote the manuscript, and reviewed and edited the manuscript. B.G. researched data. B.B. researched data, contributed to discussion, wrote the manuscript, and reviewed and edited the manuscript. W.T.W. researched data, wrote the manuscript, and reviewed and edited the manuscript. Z.B. researched data. E.A.D.F. researched data. C.B. researched data. C.M. researched data. K.H. researched data. Z.Y. researched data, wrote the manuscript, and reviewed and edited the manuscript. L.J.M. researched data, contributed to discussion, wrote the manuscript, and reviewed and edited the manuscript.

The authors gratefully acknowledge the technical assistance of Kenneth Kirschner, Arizona State University, and the excellent nursing skills of Alexandra Meyer, Julie Robbins, and Dianne Denardo, Arizona State University. The authors are particularly indebted to their enthusiastic participants for willingness to participate in this study.