Rapid Repression of ADP Transport by Palmitoyl-CoA Is Attenuated by Exercise Training in Humans: A Potential Mechanism to Decrease Oxidative Stress and Improve Skeletal Muscle Insulin Signaling

Skeletal muscle, by virtue of its mass and high rate of insulin-stimulated glucose transport, represents an important tissue in the development of insulin resistance (1). Understanding the mechanisms that result in peripheral insulin resistance is critical to developing novel therapies. Within skeletal muscle, a large emphasis has been placed on establishing a mechanistic link between reactive lipid accumulation and the induction of insulin resistance. In particular diacylglycerol (DAG) (2) and ceramides (3) have been proposed to antagonize insulin signaling by mediating protein kinase C serine phosphorylation of the insulin receptor substrate and Akt phosphorylation, thereby attenuating insulin-stimulated glucose uptake (4–9). Similar to DAG and ceramides, long-chain fatty acyl-CoA (LCFA-CoA) accumulation is associated with insulin resistance in humans, and consumption of a high-fat diet rapidly accumulates LCFA-CoA moieties in association with insulin resistance (10,11). Collectively, these studies suggest a causal link between intramuscular LCFA-CoA accumulation and insulin resistance; however, a potential mechanism remains to be elucidated.

In addition to regulating reactive lipid accumulation, increased mitochondrial reactive oxygen species (ROS) emission has been implicated in the development of insulin resistance. Although the exact manner by which mitochondrial ROS causes insulin resistance has not been delineated, it has been suggested that activation of the NF-κB/IκB/IKKβ pathway results in serine phosphorylation of the insulin receptor substrate 1 (IRS1), attenuating insulin signaling at a proximal step (12,13). Two lines of evidence suggest that mitochondrial ROS has a primary role in the etiology of insulin resistance. First, mitochondrial ROS generation increases in response to fatty acid exposure (14,15) and high-fat feeding (16) in association with the development of insulin resistance. Second, attenuating mitochondrial ROS emission using either a mitochondrial-targeted antioxidant (SS31) or overexpression of mitochondrial antioxidant enzymes (catalase [CAT] and SOD2) prevents diet-induced insulin resistance (16–18). Therefore treatment modalities that target both reactive lipids and mitochondrial ROS may be particularly beneficial at recovering insulin sensitivity. ADP binding to F₁F₀ ATP synthase decreases membrane potential and the overall rate of superoxide production (15,19,20) while simultaneously increasing rates of substrate oxidation. Therefore, attenuations in mitochondrial ADP sensitivity can influence both reactive lipid accumulation and mitochondrial ROS emissions. Supporting this notion, we have recently shown in Zucker diabetic fatty rats that submaximal ADP-stimulated respiration is impaired in association with increased mitochondrial ROS emission in the presence of ADP (21). More recently, it was predicted in humans that adenine nucleotide translocase 1 (ANT1), which is required for ADP/ATP exchange across the inner mitochondrial membrane, is inhibited through acetylation of lysine 23 (22). In addition, ANT1 lysine 23 acetylation was reduced after acute exercise (22), suggesting that chronic exercise interventions may increase mitochondrial ADP sensitivity in association with improving insulin sensitivity. However, this remains to be determined, and the effect of exercise training on submaximal ADP transport kinetics is currently ambiguous. For instance, a cross-sectional analysis in humans with varying training statuses (23), as well as a chronic training program in rats (24), suggests that training decreases the sensitivity to ADP in muscle, as the apparent K_m for ADP increased approximately threefold (23). In contrast to these in vitro assessments of mitochondrial ADP sensitivity, a classic response of exercise training is a decrease in free ADP concentrations during exercise (25,26), indicating an improvement in ADP sensitivity. These conflicting data suggest additional regulation exists on ADP kinetics that is not reflected in the in vitro environment.

One potential regulatory mechanism is palmitoyl-CoA (P-CoA), which has previously been suggested to competitively bind with ADP on ANT (27). Theoretically, the increase in P-CoA content observed in the skeletal muscle of insulin-resistant individuals could impair ADP kinetics, accounting for the strong correlation between P-CoA content and insulin resistance (11). Decreased mitochondrial ADP transport could increase mitochondrial ROS emission and reactive lipid accumulation, and therefore represents a mechanism converging the two prominent working models for the development of insulin resistance. However, although acyl-CoA concentrations are increased with insulin resistance (10,11), exercise training does not reduce acyl-CoA content in association with an improvement in insulin sensitivity (28), challenging the direct relationship between P-CoA content and insulin resistance. Therefore, the purpose of the current study was to determine if exercise training altered the sensitivity of mitochondria to the inhibitory effects of P-CoA on mitochondrial ADP transport, and the potential mechanistic link with mitochondrial ROS emission and redox balance. These data highlight a novel mechanism by which elevated P-CoA can influence skeletal muscle insulin sensitivity in humans.

Middle-aged males were screened using a medical questionnaire and were excluded if they were diagnosed as diabetic, were taking medications to control blood glucose, or had a fasting blood glucose of >7 mmol/L. All participants indicated that they were sedentary prior to commencing training by self-report. No participant changed medication during, or for the 4 months prior to commencing, the experiment. However, two participants were taking antidepressant medication, and two other participants were taking medication for hypertension. Specifically, one was taking an ACE inhibitor, and the other was taking a calcium channel blocker with a diuretic. Participants (n = 14) provided written informed consent prior to experiments, all of which conformed to the Declaration of Helsinki and were approved by the University of Guelph and Hamilton Integrated Research Ethics Boards.

Participants reported to the University of Guelph after a 12-h overnight fast. Weight and height measures were taken and a catheter was inserted into the antecubital vein by trained phlebotomists. Blood was collected before (t = 0) and 15, 30, 60, 90, and 120 min after the ingestion of a 75-g dextrose drink (Trutol; Thermo Scientific). Posttraining oral glucose tolerance tests (OGTTs) were performed ∼72 h after the last exercise session. HDL cholesterol, total cholesterol, serum triglycerides, and hs-CRP were measured from serum samples processed by a certified clinical laboratory (LifeLabs, Guelph, Ontario, Canada). Fasting plasma samples were analyzed in-house for nonesterified fatty acids and insulin using commercially available kits (Wako Diagnostics and Millipore, respectively). Plasma glucose was measured at all time points during the OGTT using a standard plate assay. VO₂peak was measured using a MOXUS metabolic cart (AEI Technologies) and an electronically braked cycle ergometer (Lode) using standard protocols before and after 6 weeks of supervised exercise training.

Muscle samples were obtained from the vastus lateralis using a Bergstrom needle at 8:00 a.m. after an overnight fast (29). Two biopsies were taken under basal conditions: a first sample was used for the preparation of permeabilized muscle fibers and a second sample was immediately frozen in liquid nitrogen. A third sample was taken 30 min after participants consumed the same 75-g dextrose drink and was immediately frozen for subsequent Western blotting.

Subjects began training 1 week after muscle biopsy procedures. Training sessions were supervised, 5 days/week for 6 weeks. Participants completed endurance sessions on Monday, Wednesday, and Friday and high-intensity interval training sessions on Tuesday and Thursday. Sessions were completed on Monark bicycles and included a 5-min warm up and cool down with very low resistance. Only one session was missed by one participant due to a work-related emergency. Full details on the training progression are outlined in Supplementary Fig. 1.

Western blotting was performed on whole-muscle homogenate as previously described (21) using the following commercially available antibodies: α-tubulin (Ab7291; Abcam), ANT1 (MSA02; MitoSciences), ANT2 (AP1057; Millipore), 4HNE (HNE11-S; Alpha Diagnostic International), OXPHOS (MS604; MitoSciences), COXIV (Invitrogen), CAT (AB1877; Abcam), and SOD2 (AB11889; Abcam), and for protein carbonylation, the OxyBlot Protein Oxidation Detection Kit (S7150; Millipore) was used. Ponceau staining was used to confirm equal loading for antibodies that required the entire membrane (e.g., 4HNE and protein carbonylation). In addition, Western blotting was performed on recovered permeabilized fibers following respiration protocols as previously described (30). All samples for a given protein were detected on the same membrane using chemiluminescence and the FluorChem HD imaging system (Alpha Innotech, Santa Clara, CA).

Permeabilized fiber preparation and high-resolution respirometry (Oroboros, Innsbruck, Austria) were performed as described previously (29,30). The following substrate concentrations were used: pyruvate (10 mmol/L), malate (5 mmol/L), ADP (5 mmol/L), glutamate (10 mmol/L), succinate (10 mmol/L), and creatine (20 mmol/L). Titration experiments were stopped at these concentrations. Additionally, ADP respiratory kinetics were measured in the presence of 60 μmol/L P-CoA or palmitate complexed to BSA (31). All titrations were performed with saturating pyruvate and malate concentrations. Cytochrome C was added to all experiments to ensure respiration was stimulated <10% to confirm mitochondrial membrane integrity.

Mitochondrial H₂O₂ emission was measured fluorometrically (Lumina; Thermo Scientific) in a constantly stirring cuvette at 37°C (Peltier controlled) containing a standard reaction buffer (Buffer Z) supplemented with 25 μmol/L blebbistatin, 40 units/mL CuZnSOD, 10 μmol/L Amplex Red (Invitrogen), and 0.5 units/mL horseradish peroxidase in the presence of various substrate combinations (60 μmol/L P-CoA, 100 μmol/L ADP, and 10 mmol/L succinate).

The red gastrocnemius muscle from male C57 mice (8–10 weeks old, 22.3 ± 0.7 g) from our mouse-breeding colony were used, and all protocols were identical to those outlined above, unless specified. ADP-stimulated respiration was induced in the presence of 5 mmol/L malate + 10 mmol/L pyruvate, and the response to lipids (250 μmol/L palmitate, 250 μmol/L palmitate + 1 mmol/L CoA, and 60 μmol/L P-CoA) was determined. In separate experiments, fibers were incubated in Oxygraph chambers with MiR05 and various lipids in the presence or absence of 5 mmol/L ADP + 2 mmol/L l-carnitine for 15 min. Fibers were then washed in a chamber with standard MiR05 for 15 min before determining state III respiration. These experiments were approved by the University of Guelph Animal Care Committee and conform to the Guide for the Care and Use of Laboratory Animals published by the U.S. National Institutes of Health.

All values are presented as means ± SEM. Apparent K_m values were determined using Prism (GraphPad Software, Inc., La Jolla, CA) using Michaelis-Menten kinetics for standard ADP titrations. However, ADP titrations in the presence of P-CoA did not display Michaelis-Menten kinetics, and therefore a one-phase association was used to estimate the concentration of ADP required to reach half-maximal respiration. V_max is presented as the highest respiration rate directly measured in titrations. Pre- versus posttraining measures were compared using a paired Student t test, with P < 0.05 considered statistically significant. A two-way ANOVA with a least significant difference post hoc analysis was also used to compare the effects of P-CoA on ADP sensitivity before and after training.

We first aimed to characterize the training adaptations in our participants to ensure that improvements in classical responses occurred. Specifically, training reduced BMI and increased VO₂peak and peak power during a VO₂peak test (Table 1). Exercise training also reduced fasting blood glucose, LDL cholesterol, LDL/HDL cholesterol, and hs-CRP (Table 1). There was a strong trend (P = 0.05) for an improvement in whole-body insulin sensitivity (HOMA) (Table 1), a finding supported by a reduction in the area under the curve during a 2-h OGTT (Fig. 1A and B). Exercise training also increased skeletal muscle Akt serine 473 and threonine 308 phosphorylation by ∼30% after glucose ingestion (Fig. 1C), suggesting an improvement in skeletal muscle insulin sensitivity. Altogether, the current training program improved several indices of cardiorespiratory fitness and glucose homeostasis.

Given the improvement in skeletal muscle insulin signaling after exercise training, we next aimed to determine the influence of exercise on several aspects of mitochondrial bioenergetics. We first used a mixed substrate protocol to examine the capacity of the electron transport chain in permeabilized muscle fibers. Mitochondrial oxygen consumption was increased ∼60% in all conditions (Fig. 2A), suggesting training-induced increases in mitochondrial content. This interpretation is supported by the finding that the content of several subunits of the electron transport chain were increased ∼50% in muscle homogenates (Fig. 2B), as well as in muscle fiber bundles used for respirometry protocols (Supplementary Fig. 2). The protein content of ANT1 increased similarly to markers of mitochondrial content, whereas in contrast, ANT2 was not altered (Fig. 2B). Once again, these patterns were verified in fiber bundles recovered from the respirometer (Supplementary Fig. 2). Altogether, ANT1 increased proportionally to mitochondrial content, whereas ANT2 was unaltered by exercise training.

Genetic ablation of ANT1 decreases basal uncoupled/leak respiration (32), raising the possibility in the current study that the observed lack of change in ANT2 protein may alter basal respiration or ROS emission. Whereas exercise training improved the redox environment of skeletal muscle, as assessed by a reduction in 4HNE (Fig. 2C) and protein carbonylation contents (Fig. 2D), mitochondrial H₂O₂ emission was not altered (Fig. 2E). Whereas CAT did not change, the mitochondrial antioxidant enzyme SOD2 increased after training (Fig. 2F), potentially explaining this discrepancy. However, given the ability of lipids to interact with ANT and induce uncoupling (33,34), we also examined the potential for training to increase the ability of P-CoA to induce mitochondrial uncoupling and decrease ROS emission rates. However, regardless of the training status, P-CoA did not alter state IV respiration (Fig. 2G) or mitochondrial H₂O₂ emission rates (Fig. 2H). Combined, these data suggest that in the absence of ADP, P-CoA does not affect mitochondrial bioenergetics.

We next aimed to determine the respiratory sensitivity to ADP in permeabilized fibers as a potential alternative functional consequence to the observed alterations in ANT1/ANT2 ratios. Regardless of the presence of creatine, titrating ADP in the presence of saturating pyruvate and malate resulted in a typical Michaelis-Menton kinetic curve (Fig. 3A and D), which could be used to estimate maximal respiration and ADP sensitivity. Similar to the mixed substrate protocol, exercise training increased maximal respiration (V_max) in a creatine-independent manner (Fig. 3B and E). In addition, exercise training increased the apparent ADP K_m ∼40% in the presence of creatine (Fig. 3C), and there was a strong trend for an increase in the apparent ADP K_m in the absence of creatine (Fig. 3F). These data suggest that exercise training decreased mitochondrial ADP sensitivity in association with improvements in insulin sensitivity, contrary to our working model. We therefore next aimed to determine external regulation on ANT that could be influenced by exercise training.

Intramuscular P-CoA concentrations are elevated in insulin-resistant skeletal muscle and are known to inhibit ANT function (11,27). We therefore next aimed to directly determine if training altered the ability of P-CoA to attenuate ADP sensitivity. Before training, the presence of 60 μmol/L P-CoA almost completely prevented the ability of ADP to stimulate respiration (Fig. 4A). After training, the V_max was increased ∼75% (Fig. 4B) and the ability of ADP to stimulate respiration was approximately fourfold greater (Fig. 4A), suggesting decreased sensitivity to P-CoA inhibition. In support of this, the ADP concentration required to reach half-maximal respiration was reduced by ∼50% after training when P-CoA was present (Fig. 4C). In addition, respiration in the presence of 100 μmol/L ADP, which represents a submaximal free ADP concentration within muscle, was increased after training in the presence of P-CoA (depicted in Fig. 4A). Altogether, whereas P-CoA impaired ADP sensitivity, training improved the sensitivity to ADP when P-CoA was present, a response that is in contrast to typical assessments of ADP sensitivity in permeabilized fibers (Fig. 4D).

To further examine a potential mechanism of action for how P-CoA alterations in mitochondrial bioenergetics could influence insulin sensitivity, we conducted several experiments in wild-type mice. During state III respiration, the real-time addition of 60 μmol/L P-CoA rapidly depressed state III respiration by ∼50% within 15 min (Fig. 5A and B). The subsequent addition of dinitrophenol recovered respiration rates up to ∼90% of maximal ADP-stimulated respiration (Fig. 5A and B), indicating that P-CoA was not affecting electron flux through the electron transport chain, implicating a primary inhibition of ADP transport. The presence of palmitate similarly inhibited ADP-supported respiration (Fig. 5C). Furthermore, these effects are not complex I specific, as other substrate protocols (provision of succinate and G3P) yield the same inhibition pattern (Supplementary Fig. 3). Together, these data suggest that lipids in general attenuate ADP transport, supporting previous findings that lipids competitively bind to the ADP binding site on ANT (27).

Given that we observed an exercise-training interaction with P-CoA–mediated attenuations in ADP sensitivity, we reasoned that this process was not simply regulated by competitive inhibition as classically suggested (27). Therefore, we next performed experiments whereby mitochondria were exposed to various lipids for 15 min, washed for 15 min in a lipid-free environment, and then stimulated to respire in the presence of saturating ADP and pyruvate + malate. Using this approach, ADP-stimulated respiration was not affected by a 15-min preincubation with palmitate (Fig. 5D). However, when CoA was included with palmitate in the preincubation media, subsequent state III respiration was reduced similarly to the effects of P-CoA (Fig. 5D). These data suggest that, unlike palmitate, P-CoA inhibition of ADP transport is not simply through competitive inhibition, and suggests a more rigid interaction with ANT. Intriguingly, when l-carnitine was included in the preincubation media, the negative influence of P-CoA was prevented (Fig. 5C), suggesting that the availability of P-CoA in the intermembrane space, and not the matrix, is required for the observed inhibition of ADP-stimulated respiration.

To investigate the potential link between reactive lipid inhibition of ADP transport and ROS production, we examined the impact of P-CoA on the propensity to generate mitochondrial ROS in the presence of ADP. Exposure of mitochondria to P-CoA did not alter maximal succinate-supported H₂O₂ emission (Fig. 2G [humans] and Fig. 6 [mice]). Whereas 100 μmol/L ADP consistently decreased mitochondrial H₂O₂ emission rates in rodents, exposure to P-CoA attenuated this response (Fig. 6). These data suggest that P-CoA inhibition of ADP transport can influence mitochondrial ROS emissions. Combined, these data provide a plausible mechanism for the observed reduction in oxidative stress, improved insulin sensitivity, and the association between skeletal muscle Akt phosphorylation and ADP-stimulated respiration in the presence of P-CoA that was apparent after aerobic training in the current study.

In the current study, we show that the ability of P-CoA to inhibit mitochondrial ADP-stimulated respiration is dramatically attenuated by exercise training. This response is strongly associated with skeletal muscle Akt phosphorylation, whole-body insulin sensitivity, and an improvement in the oxidative state of the muscle. Altogether, the present data provide a plausible mechanism for how P-CoA can influence insulin sensitivity and highlight mitochondrial ADP transport as a potential nexus point between reactive lipids and mitochondrial ROS during the development of insulin resistance.

The role of mitochondrial ADP transport as a regulator of mitochondrial bioenergetics has remained largely unexplored, although recent literature has elucidated biologically relevant ADP K_m values (35), solidified the regulation of mitochondrial ADP transport by mitochondrial creatine kinase (36,37), and uncovered the functional consequence of acute exercise (29). In the current study, we advance our understanding of the regulation of ADP transport by providing evidence that chronic exercise training in obese individuals attenuates the apparent sensitivity of mitochondria to ADP, supporting a previous cross-sectional analysis in healthy humans (23) and treadmill training in rodents (24). These data are in contrast to the well-established decrease in free ADP that occurs after training (25). However, although the reduction in ADP respiratory sensitivity after training appears perplexing, in the current study, respiration at a given submaximal ADP concentration was higher after training, an index that may have greater relevance to whole-body physiology. Nevertheless, the increase in the apparent ADP K_m that occurred with training suggests that an alteration occurred within this system that attenuated the sensitivity to ADP. This likely occurred on ANT, and not mitochondrial creatine kinase, given the apparent creatine-independent response. Therefore, potential mechanisms for the observed impairment in ADP sensitivity include an increase in the acetylation of lysine 23 (22), a decrease in ANT1 tyrosine 194 phosphorylation (38), and a decrease in glutathionylation/carbonylation of ANT (39,40). Alternatively, an alteration in the expression of ANT isoforms may mediate this response, as in the current study, exercise training did not alter the content of ANT2. The functional effects of ANT isoforms are not currently known; however, in rodent skeletal muscle, a reduction in ANT2 protein is associated with impaired submaximal, but not maximal, ADP-stimulated respiration (21). It is therefore tempting to suggest that ANT2 displays a greater ADP sensitivity, and the relative abundance of ANT2/ANT1 observed in the current study influenced the overall ADP respiratory sensitivity. Regardless of the potential mechanism, exercise training consistently displays a reduction in mitochondrial ADP sensitivity when determined in the absence of lipid moieties (current study and Zoll et al. [23], Zoll et al. [24], and Guerrero et al. [41]).

The shift in the ADP sensitivity after training may suggest external regulation on ANT. To examine one potential mechanism, we determined ADP respiratory sensitivity in the presence of P-CoA. We used this substrate as palmitate represents one of the most abundant dietary lipids, changes in intramuscular P-CoA content are most closely associated with muscle insulin sensitivity (42), and ANT has previously been shown to have the highest sensitivity to inhibition by P-CoA (27). Similar to previous reports, P-CoA dramatically, and almost instantaneously, decreased ADP-stimulated respiration (43). However, whereas previous reports suggest this inhibition occurred as a result of competitive inhibition of ADP binding (27,43), the current study suggests another regulatory mechanism is at play, as the removal of P-CoA from the respiration media did not recover ADP-stimulated respiration. Given that training dramatically attenuated the ability of P-CoA to prevent ADP-stimulated respiration, this system appears to be sensitive to exercise, and corresponding posttranslational modifications to ADP binding proteins (such as ANT) represent one candidate for this regulation. Alternatively, since ANT2 protein did not increase with exercise training, it is tempting to speculate that ANT2 is less susceptible to P-CoA inhibition, and future work should examine this possibility. Regardless, the present data provide novel basic understanding on the regulation of ADP sensitivity and the response to exercise training. To investigate the potential link between P-CoA inhibition of ADP transport and oxidative stress, we examined the impact of P-CoA on the propensity to generate mitochondrial ROS in the presence of ADP. Using this approach, we provide evidence that the ability of ADP to lower mitochondrial ROS production was significantly diminished after mitochondria were exposed to P-CoA. These data are the first to link P-CoA inhibition of ADP transport to increased ROS production and establish a novel mechanism detailing the association between intramuscular P-CoA content, ROS production, and potentially insulin resistance. Although these responses likely contribute to an improvement in insulin sensitivity after training, the increase in antioxidants (SOD2 in current study and reviewed in Powers and Lennon [44]) and known reductions in DAGs and ceramides (45,46) all contribute to the beneficial effects of exercise.

Mitochondrial ROS production is intricately linked to an imbalance between energy supply and energy demand by the interrelated rise in mitochondrial membrane potential (Δψ_m). In this context, the current data strongly suggest that an accumulation of cytosolic P-CoA can influence redox balance by attenuating mitochondrial ADP transport. In addition, previous literature shows that elevations in cytosolic P-CoA can promote an increase in carnitine palmitoyl-transferase I (CPT-I) activity by reducing the inhibitory effectiveness of malonyl-CoA (31). Therefore, an increase in cytosolic P-CoA would be expected to inhibit mitochondrial ADP transport while simultaneously increasing matrix P-CoA concentrations and membrane potential, dramatically affecting redox balance. It is likely for these reasons in the current study that the attenuation in P-CoA inhibition of ADP-stimulated respiration was associated with decreased 4HNE and protein carbonylation content and insulin signaling. Altogether, the current data strongly implicate P-CoA as regulator of mitochondrial bioenergetics, positioning P-CoA inhibition of ADP transport as a convergence point between reactive lipid accumulation, mitochondrial ROS emission, and insulin resistance—interactions that are attenuated with exercise training.

Funding. This work was funded by the Natural Sciences and Engineering Research Council of Canada (G.P.H.) and infrastructure was purchased with the assistance of the Canadian Foundation for Innovation (G.P.H.) as well as the Ontario Research Fund (G.P.H.).

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

Author Contributions. A.L. and G.P.H. designed the study, performed experiments, and wrote the manuscript. S.P., B.K.S., E.A.F.H., M.K.A., and G.J.H. performed experiments, interpreted data, and contributed to writing the manuscript. P.D.N. designed experiments and edited the manuscript. G.P.H. 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.