AIFM2 Is Required for High-Intensity Aerobic Exercise in Promoting Glucose Utilization

Exercise requires a large amount of energy (ATP) for muscle contraction (1). However, the intramuscular stores of ATP are small (1), so ATP needs to be resynthesized to sustain contractile activity for extended periods. Skeletal muscle replenishes ATP by activating the three major energy systems: 1) phosphagen, 2) glycolysis, and 3) mitochondrial respiration (1). The phosphagen system breaks down phosphocreatine and transfers the phosphate group to ADP and is responsible for ATP regeneration during several seconds (1). As exercise continues for >10–15 s, the glycolysis rate increases through using muscle glycogen and glucose from the blood (2). Exercise that lasts more than several minutes activates the oxidative metabolism of carbohydrates and fats and produces acetyl-CoA, which enters the tricarboxylic acid cycle for further energy production via mitochondrial respiration (3). The relative contribution of different types of fuel is determined through exercise intensity and duration. Overall, carbohydrate oxidation dominates at higher exercise intensities, whereas fat oxidation is more important at lower intensities (3). Despite the importance of fuel utilization in exercise, the molecular mechanisms that determine fuel efficiency during exercise remain largely unknown.

NAD⁺ is an obligatory substrate in glycolysis, and thus regeneration of NAD⁺ is critical during exercise. Mitochondrial inner membrane is impermeable to NAD⁺. Recently, it was found that cytosolic NAD⁺ can be imported into the mammalian mitochondrion via MCART1/SLC25A51 transporter, in addition to yeast and plant mitochondria (4). However, no specific transporter system exporting NAD⁺ out of mitochondria has been identified (5,6). In skeletal muscle, the cytosolic NADH is oxidized back to NAD⁺ mainly through the glycerol-3-phosphate (G-3-P) shuttle system (6), which uses electrons bound for the mitochondrial electron transport chain to catalyze a series of reactions that result in NAD⁺ regeneration. In addition, lactate dehydrogenase (LDH) can replenish NAD⁺ in the cytosol (7,8) by converting pyruvate to lactate under anaerobic conditions. However, exercise changes LDH activity only to a very small degree (9). Because intense exercise increases skeletal muscle glucose uptake by up to 50-fold (10), we hypothesized that these known NAD⁺ regeneration systems are insufficient to meet NAD⁺ demand and that there must be alternative pathways to supply NAD⁺ for robust glycolysis.

Yeast, bacteria, and plant contain at least two NADH dehydrogenases that may couple the oxidation of NADH to the mitochondrial respiratory chain in yeast, bacteria, and plant (2,3,5,6). The first one is internal NADH dehydrogenase (NDI), whose catalytic site faces the mitochondrial matrix. Thus, it can oxidize the intramitochondrial NADH generated by the pyruvate-dehydrogenase complex and the tricarboxylic acid cycle (2). The other one is referred to as external NADH dehydrogenase (NDE), the catalytic site of which faces the intermembrane space (11), thus allowing continuing reoxidation of NAD⁺ in the cytosol. AIFM2 is a flavin adenine dinucleotide–dependent NADH/NAD⁺ oxidoreductase that belongs to the apoptosis-inducing factor (AIF) family of proteins (12). A recent study identified a surprising role for AIFM2 as the first brown adipose tissue (BAT)-specific mammalian NDE. AIFM2 is highly enriched in BAT and is required to support the high rate of glycolysis that fuels thermogenesis in BAT, especially under feeding conditions where circulating glucose is high (12). Given that skeletal muscle requires a large amount of glucose oxidation during endurance exercise, we undertook the current studies to test the hypothesis that AIFM2 in skeletal muscle supports robust glycolysis by providing NAD⁺ during high-intensity exercise.

Mice were maintained under a 12-h light/12-h dark cycle at constant temperature (23°C) with free access to food and water. All animal work was approved by the University of California, Berkeley, Animal Care and Use Committee.

Aifm2 floxed (Aifm2f/f) mice were generated as previously described (12). Muscle-specific Aifm2 haploinsufficient mice were generated by crossing of Aifm2f/f mice with ACTA1-Cre/Ers1 mice (025750; The Jackson Laboratory). Both Aifm2^f/+ and ACTA1^Aifm2+/− mice were injected with tamoxifen by (100 mg/kg i.p.) for 5 consecutive days. Animal studies were conducted after a 2-week washing period to avoid potential effects of tamoxifen. Male mice age 2–4 months were used. Tissue and blood samples were taken right after exercise except for the measurement of blood glucose at recovery.

Low- and high-intensity exercise exhaustion studies were performed as previously described (13). In brief, all mice were acclimated to the treadmill prior to the exercise test session. For each session, food was removed 2 h before exercise. Acclimation began at a low speed of 5–8 m/min for a total of 10 min on day 1 and was increased to 5–10 m/min for a total of 10 min on day 2. The experiments were performed on day 3. For the low-intensity treadmill test, the treadmill began at a rate of 12 m/min for 40 min. After 40 min, the treadmill speed was increased at a rate of 1 m/min every 10 min for a total of 30 min and then increased at the rate of 1 m/min every 5 min until the mice were exhausted. The high-intensity treadmill test was conducted on the same open-field six-lane treadmill, set at a 10% incline. Following a 5-min 0 m/min acclimation period, the speed was raised to 6 m/min and increased by 2 m/min every 5 min to a maximal pace of 20 m/min until exhaustion. Mice were considered exhausted when they were unable to respond to continued prodding with a soft brush. For experiments involving the same duration of exercise, the exercise regimen was a high-intensity exhaustion test up to 30 min.

Cell or tissue culture samples were lysed in radioimmunoprecipitation assay lysis buffer containing 50 mmol/L Tris-HCl (pH 7.5), 150 mmol/L NaCl, 1% Triton X-100, 0.1% SDS, and protease inhibitor cocktail (cOmplete Mini EDTA-free, 11836170001; Roche) for 30 min on ice with rotating samples. Protein concentration was quantified with Bradford or BSA analysis, and 40–60 μg proteins were used for the assays. Proteins were size fractionated with SDS-PAGE and then transferred to polyvinylidene difluoride or nitrocellulose membrane. After blocking with 5% nonfat dried milk in TBS-Tween (0.25%), the membranes were incubated overnight with the appropriate primary antibodies, followed by secondary antibody incubation for 1 h at room temperature. After washing with TBS-Tween for 30 min, signals were detected with use of an enhanced chemiluminescence substrate assay kit (cat. no. NEL104001EA; PerkinElmer) using an iBright CL1500 Imaging System (Invitrogen) and quantified with image J. Antibodies were purchased from Sigma-Aldrich (AIFM2, SAB4503389, 1:300), Santa Cruz Biotechnology (GAPDH, SC-25778, 1:2,000, and HSP90, sc-13119, 1:2,000), and Abcam (His-tag antibody, ab18184, 1:1,000, and GFP, Ab290, 1:1,000).

Total RNA was extracted from tissues using TRIzol reagent according to the manufacturer’s instructions. cDNA was reverse transcribed from 1 μg RNA using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Quantitative PCR was performed with AccuPower 2X GreenStar qPCR Master Mix (Bioneer) with QuantStudio 5 (Applied Biosystems). The relative amount of mRNA normalized to cyclophilin B was calculated with the 2^−ΔΔCt method. Primer sequences are listed in Supplementary Table 1.

NAD⁺/NADH from muscle tissues (no. ab65348; Abcam) and C2C12 samples (no. K337; BioVision) was measured with commercial kits according to the manufacturer’s instructions. Briefly, differentiated C2C12 cells (2 × 10⁵) or muscle tissues were washed and homogenized with 400 µL NADH/NAD Extraction Buffer. The samples were then centrifuged at 14,000g for 5 min at 4°C and supernatants were separated into two tubes. One of the tubes were heated at 60°C for 30 min to decompose NAD⁺ and used to determine NADH level. The supernatants in another tube were used to determine total NAD level. These kits provide a convenient tool for sensitive detection of the intracellular nucleotides: NADH, NAD, and their ratio. The NAD cycling enzyme mix in the kit specifically recognizes NADH/NAD in an enzyme cycling reaction.

C2C12 cells were plated and differentiated in XF24 Cell Culture Microplates. On the day of experiments, the cells were treated with either glucose or palmitate. Cells were then incubated for 1 h at 37°C without CO₂ prior to analysis on the XF24 analyzer. Extracellular acidification rate (ECAR) was determined with an XF24 Extracellular Flux Analyzer (Seahorse Bioscience). Mitochondrial respiration was determined with oligomycin (4 μmol/L) and FCCP (4 μmol/L). Antimycin A and rotenone (4 μmol/L each) were used to inhibit complex III– and complex I–dependent respiration. Basal ECAR refers to the starting rates prior to the addition of an agent. Maximal ECAR was calculated as the difference between ECAR following the injection of oligomycin and the basal ECAR reading.

C2C12 myoblasts were maintained in DMEM (containing 4 mmol/L l-glutamine, 4.5 g/L d-glucose, 110 mg/L sodium pyruvate) supplemented with 10% FBS, 100 units/mL penicillin, and 100 μg/mL streptomycin. Culture conditions were maintained in a humidified incubator under an atmosphere of 5% CO₂ at 37°C. For differentiation into myotubes, myoblasts at subconfluency were supplemented with DMEM with 2% horse serum every other day for 4 days. For overexpression (OE) studies, day 4 mature myotubes were treated with genome copies per mL (GC/mL) (GFP) and AIFM2-His adenoviruses (000541A and 115990540300, respectively; Applied Biological Materials) overnight. Two days later, the infected cells were subjected to individual experiments.

Blood was collected from the tail tip from mice after exhaustion running. Blood plasma was separated by centrifugation for 15 min at 4,000 rpm. Lactate levels in the plasma were measured with a Lactate Colorimetric Assay Kit (Abcam).

Blood was collected from the tail tip from mice after exhaustion running. Blood plasma was separated by centrifugation for 15 min at 4,000 rpm. Free fatty acid (FFA) levels in the plasma were measured with a NEFA-HR Assay Kit (National Mouse Metabolic Phenotyping Centers).

Eight-week-old C57BL/6 mice or ACTA1^Aifm2+/− mice were anesthetized with an intraperitoneal injection of 91 mg/kg ketamine and 9.1 mg/kg xylazine, after which hindlimbs were shaved, and the extensor digitorum longus (EDL) muscles were injected with 30 μL hyaluronidase solution (which was prepared through resuspending bovine placental hyaluronidase [Sigma-Aldrich] in sterile injectable 0.9% NaCl at a concentration of 0.4 units/μL). Mice were anesthetized 2 h later, and the EDLs were injected with 180 μg plasmid DNA in sterile saline. After injection of plasmid DNA, the hindlimbs were placed between two-paddle electrodes and subjected to 10 pulses (20 ms) of 175 V/cm (with 480-ms intervals between pulses) with use of an ECM 830 electroporator (BTX Harvard Apparatus, Holliston, MA). The mice were returned back to home cages, and experiments were performed 7 days after electroporation.

Oxymax/Comprehensive Lab Animal Monitoring System (CLAMS) was used to measure individual mice: their oxygen consumption (VO₂), carbon dioxide production (VCO₂), activity, and feed intake. We calculated the respiratory exchange ratio (RER) by dividing the volume of CO₂ produced by the volume of O₂ consumed. Mice were housed individually at 23°C under a 12 h light/12 h dark cycle. Food and water were available ad libitum.

Mice was placed on the treadmill belt immediately after the intraperitoneal injection of saline containing 10 μCi ³H-2-deoxyglucose (2-DG). After exercise, mice were sacrificed and tissues were excised and snap-frozen in liquid nitrogen. 2-DG uptake was measured as previously described (14). Briefly, frozen tissues were homogenized in 0.5% perchloric acid and centrifuged for 20 min at 2,000g. After, supernatants were neutralized with KOH, and 2-DG radioactivity was then measured with liquid scintillation counting using Beckman LS 5000TD.

Where appropriate, data are expressed as mean ± SEM. The data were statistically analyzed with unpaired Student t test with use of GraphPad Prism, version 8. The P values <0.05 were considered statistically significant.

Data sets generated from the current study are available from the corresponding author.

To study the role of skeletal muscle AIFM2 during exercise, we first compared Aifm2 expression in soleus, gastrocnemius (GA), and EDL muscles at rest and after a single bout of treadmill running under varying exercise regimens (15,16). We found that low-intensity exercise, which is powered by the oxidation of fatty acids (17), induced Aifm2 mRNA and protein expression by ∼2.5-fold and ∼1.8-fold in EDL muscles, respectively, but not in soleus or GA muscles (Fig. 1A–C). Concurrently, EDL had increased expression of genes involved in glucose transport and utilization, such as Gpdh, Pfk, and Glut4 (Fig. 1A). During high-intensity exercise, which more relies on carbohydrates, Aifm2 mRNA and protein expression increased drastically by more than 10-fold and 2.5-fold, respectively, in EDL muscles (Fig. 1D–F). This again coincided with increased expression of genes involved in glucose transport and utilization, such as Gpdh, Pfk, and Glut4 (Fig. 1D). Although Aifm2 mRNA expression in GA significantly increased by twofold after a high-intensity exercise, the differences in its protein levels were not detected (Fig. 1D–F). Together, these data show that AIFM2 expression is highly induced in skeletal muscles, especially in the glycolytic EDL muscle, during high-intensity exercise.

Next, we sought to determine whether AIFM2 regulates the NAD⁺-to-NADH ratio and, subsequently, glycolysis in vitro. We introduced AIFM2 expression vector versus GFP in mature C2C12 myotubes via adenovirus (Fig. 2A–C) and measured the NAD⁺-to-NADH ratio using a biochemical assay kit that is based on lactate dehydrogenase cycling (12). Remarkably, AIFM2-overexpressing (AIFM2-OE) C2C12 myotubes had a ∼2.8-fold increase in NAD⁺-to-NADH ratio compared with control GFP cells (Fig. 2D). We next investigated whether the altered NAD⁺-to-NADH ratio impacts the glycolytic rate by performing an extracellular acidification rate assay (ECAR), which mainly depicts glycolytic flux to lactate. AIFM2-OE C2C12 myotubes displayed a significantly increased basal and maximal ECAR compared with control cells (Fig. 2E–G). Congruently, their intracellular lactate levels were increased by 31% (Fig. 2H) and their intracellular FFA levels were also increased by 11% (Fig. 2I), implying that fatty acid utilization was diminished in favor of glycolysis.

In addition, we probed the effect of OE of AIFM2 on the mitochondrial respiration in these cells in the presence of either glucose or palmitate (a saturated fatty acid) as a fuel source. We found that OE of AIFM2increased the overall mitochondrial respiration rate compared with control cells when fueled by glucose (Fig. 2J and K), but not by palmitate (Fig. 2L and M), which is consistent with the idea that AIFM2 enhances glucose-driven mitochondrial oxidative metabolism. Together, these data show a cell-autonomous role for AIFM2 in altering the NAD⁺-to-NADH ratio and promoting glucose oxidation in myotubes.

Glycolysis is the major source of energy for high-intensity aerobic exercise (18). Given our in vitro results showing AIFM2 increased the glycolytic rate, we hypothesized that AIFM2 increases in vivo exercise performance. To test whether AIFM2 is required for exercise capacity through promoting glucose utilization, we delivered shRNA against Aifm2 or control shRNA to the EDL muscle via in vivo electroporation in 8-week-old C57BL/6 mice and achieved a highly selective knockdown (KD) of in EDL by ∼50% reduction at the mRNA and protein levels (Fig. 3A–C). First, we performed an exhaustion test by putting the transfected mice on the treadmill under high-intensity exercise. Remarkably, mice with Aifm2 KD in the EDL muscle ran for ∼50% less time under the high-intensity regimen (Fig. 3D). In association with reduced exercise competency in Aifm2 KD mice, we noted that after exhaustion from exercise, they had lower NAD⁺ levels and a lower NAD⁺-to-NADH ratio (Fig. 3E and F) and lower intramuscular lactate levels in the EDL (Fig. 3G) but higher plasma glucose levels (Fig. 3H) compared with control mice. The difference in blood glucose levels was sustained after 30 min recovery (Fig. 3H), implying that the difference in glucose utilization may slowly normalize to basal levels. At the same time, we found the Aifm2 KD mice had lower serum FFA levels (Fig. 3I), implying increased utilization of FFA, likely due to the insufficient increase in the glycolytic rate during exercise.

Since Aifm2 KD mice run less in the exhaustion test, we cannot tell whether such differences in metabolic profile between genotypes are mainly due to the consequence of differences in exercise capacity or due to the metabolic fuel efficiency of glucose. To investigate this question, we put the animals on the treadmill for equal amounts of time, 30 min, under the same high-intensity regimen and measured their metabolic profile after exercise. Remarkably, Aifm2 KD mice in the EDL still displayed lower NAD⁺ levels and a lower NAD⁺-to-NADH ratio (Supplementary Fig. 1A and B) and lower intramuscular lactate levels in the EDL (Supplementary Fig. 1C), while they had higher plasma glucose levels (Supplementary Fig. 1D) than control mice. On the contrary, serum FFA levels did not show a significant change between genotypes when animals ran equally (Supplementary Fig. 1E). This might be because the metabolic shift from carbohydrates to lipids was not yet completed over the relatively shorter running time (30 min), in comparison with the exhaustion test, where control mice were allowed to run much longer. Next, we assessed exercise-stimulated glucose uptake by administrating 2-DG to a cohort of control and Aifm2 KD mice prior to exercise and measured 2-DG levels in various muscles on equal duration of exercise for 30 min. As a result, we found that EDL muscles in the KD mice, but not soleus or GA muscles, showed a significant reduction of glucose uptake (Supplementary Fig. 1F). Together, these data indicate that Aifm2 KD in the EDL leads to inefficient glucose utilization.

Next, we sought to investigate whether EDL-targeted expression of AIFM2 is sufficient to increase exercise performance and the glycolytic rate. To do that, we electroporated the AIFM2 expression plasmid or GFP control plasmid into C57BL/6 mice, achieving physiological levels of AIFM2 OE mainly in EDL muscles (Fig. 3J–L). Strikingly, AIFM2-OE mice ran for a longer period of time compared with control mice under the high-intensity regimen by approximately twofold (Fig. 3M). In association with increased exercise capacity, the AIFM2-OE mice displayed indications of an increased glucose utilization, having higher NAD⁺ levels and NAD⁺-to-NADH ratio (Fig. 3N and O), elevated levels of lactate during exercise, and lower plasma glucose levels during exercise and recovery (Fig. 3P and Q). These mice had higher serum FFA levels (Fig. 3R), indicating decreased fatty acid oxidation. To assess metabolic efficiency in the AIFM2-OE mice, we measured their metabolic profile on equal duration of exercise for 30 min and found that the AIFM2-OE mice still showed indications of increased glucose utilization, as shown by higher NAD⁺ levels and a higher NAD⁺-to-NADH ratio (Supplementary Fig. 1A and B), elevated levels of lactate (Supplementary Fig. 1C), and lower plasma glucose levels (Supplementary Fig. 1D) compared with control mice but without significant difference in serum FFA levels (Supplementary Fig. 1E). Also, importantly, we noted an increase in exercise-stimulated glucose uptake selectively in EDL muscles in the AIFM2-OE mice (Supplementary Fig. 1F). Overall, these results show that AIFM2 is both necessary and sufficient to regulate the glucose utilization and exercise capacity by regulating NAD⁺ supply in mice.

To firmly establish the in vivo role of AIFM2, we next performed loss-of-function studies using a genetic mutant mouse model. We crossed Aifm2f/f mice with human α-skeletal actin (ACTA1)-Cre-Esr1 mice (Fig. 4A), in which Cre expression is induced in a tamoxifen-dependent manner in muscle tissues (19). We injected tamoxifen (100 mg/kg i.p. for 5 days) into 8-week-old Aifm2^f/+ mice with and without the ACTA1 Cre and detected a high efficiency of muscle-specific Aifm2 deficiency in various muscles of the tamoxifen-treated Aifm2^f/+ mice with Cre (ACTA1^Aifm2+/−) compared with control mice (Fig. 4B–D). Notably, despite the fact that Aifm2 was equally knocked down in all three muscles, NAD⁺ levels and the NAD⁺-to-NADH ratio were only altered in EDL, not in soleus or GA muscles (Fig. 4E and F). There were no discernible changes in overall muscle histology between genotypes (not shown). Remarkably, we found that ACTA1^Aifm2+/− mice had reduced exercise capacity, running for ∼40% less time than control mice (Fig. 4G). Consistently, after exhaustion from exercise, ACTA1^Aifm2+/− mice had inefficient glucose utilization but increased fatty acid utilization, as evidenced by higher levels of plasma glucose and intramuscular glycogen (Fig. 4H and I) and decreased levels of lactate (Fig. 4J) and FFA (Fig. 4K). These altered metabolite levels in ACTA1^Aifm2+/− mice during exercise raise the possibility that these mutant mice had a shift in fuel preference. To test this, we subjected a cohort of animals to CLAMS to measure their RER. Indeed, we found that the ACTA1^Aifm2+/− mice had significantly reduced RER compared with control mice (Supplementary Fig. 2A), indicating a preference toward fatty acids as the main source of fuel. This was without discernable differences between the genotypes in food intake, ambulatory movement, and energy expenditure (Supplementary Fig. 2B–E).

Given that AIFM2 functions as a NADH oxidoreductase, we asked whether reverting the NAD⁺-to-NADH ratio rescues the exercise defect in ACTA1^Aifm2+/− mice. We delivered an expression plasmid for control GFP and for NDE1, a mitochondrial external NADH oxidoreductase from Saccharomyces cerevisiae (20), via electroporation into ACTA1^Aifm2+/− mice (Fig. 4L and M). We first confirmed that OE of NDE1 in ACTA1^Aifm2+/− mice completely rescued NAD⁺ levels, nearly to those of wild-type mice (Fig. 4N and O). Remarkably, the ACTA1^Aifm2+/− mice that received NDE1 had 28.6% restored exercise capacity (Fig. 4P) and largely normalized tissue and blood parameters of glucose and fatty acid utilization (Fig. 4Q–S). These results demonstrate that the NADH oxidoreductase activity of AIFM2 is critical in fully facilitating its role in glucose utilization and exercise capacity.

During high-intensity exercise, there is a high demand for glucose utilization (21). AIFM2 is a flavin adenine dinucleotide–dependent NADH/NAD⁺ oxidoreductase that belongs to the AIF family (12). Here, we demonstrate that AIFM2 is induced, especially in glycolytic EDL muscles, during high-intensity exercise, thereby rapidly providing NAD⁺ to support a high rate of glycolysis.

Glucose metabolism is designed for fast ATP production. Carbohydrates are the only fuel that can be used for both aerobic and anaerobic ATP production, both of which are activated very quickly during transitions from rest to high-power exercise (2). Carbohydrates can also provide all the fuel during exercise at a power output that elicits ∼100% VO_2max, and it is a more efficient fuel than fat (22). Muscle glycogen is the initial predominant source for glycolysis through glycogenolysis in moderate- to high-intensity exercise; however, as exercise proceeds, glycogenolysis declines in parallel with glycogen depletion (23). Thus, muscle glucose uptake increases during extended exercise and can reach values that are 30–50 times higher than those at rest (24). The increase in muscle glucose uptake is related to the workload and duration of exercise and corresponds to the output of glucose from the liver (24). Various studies have shown the importance of glucose utilization in exercise tolerance through genetic manipulation of the genes involved in glucose utilization. For instance, Glut4-deficient mice have impaired exercise tolerance due to insufficient supply of glucose (25), and OE of hexokinase (HK)II, which phosphorylates glucose in the first committed step in glucose metabolism, increases exercise capacity in both Glut4-deficient mice and wild-type mice (25, 26). Mutant mice bearing a deficiency of mitochondrial peptidyl prolyl isomerase cyclophilin-D have increased exercise capacity due to increased glucose uptake and utilization with increased AMPK activity (27). Our data are in line with these studies, as we found that increased glycolysis by muscle-targeted OE of AIFM2 enhances exercise capacity, while reduced glycolytic rate due to muscle-targeted depletion of Aifm2 via in vivo transfection of shAifm2 and muscle-specific Aifm2 haploinsufficiency impairs exercise capacity. Although we had relatively efficient KD using the Aifm2 muscle-targeted haploinsufficient mice, future studies are necessary to assess the effects of a full, muscle-specific deficiency of Aifm2.

We have noted that, although NDE1 OE fully normalized NAD⁺ levels and other metabolic parameters to the levels of wild-type control mice, the rescuing effect on exercise capacity was not full. This indicates that AIFM2 may have additional mechanism(s) to affect exercise capacity in addition to acting as a NADH oxidase. It was previously thought that AIFM2 contributes to apoptosis as a target gene of P53 in the presence of bacterial and viral DNA (28). Yet, Aifm2 KO mice did not develop tumors. A previous study indicated that OE of AIFM2 in brown adipocytes did not cause apoptosis (12). Thus, we speculate that the potential role of AIFM2 in apoptosis is not engaged with its regulatory role in exercise. Future studies are warranted to explore the additional mechanistic basis of how AIFM2 supports high-intensity exercise.

In previous studies investigators found that AIFM2 was detected primarily in BAT tissues and not in other tissues including quadricep muscle (12). In the current study, we found that the basal level of AIFM2 was barely detectable in various muscles, except for soleus. However, AIFM2 was highly induced, especially in EDL muscles, during a bout of endurance exercise, concurrent with increased demand for glucose oxidation. Induction of AIFM2 was more robust as the exercise intensified, requiring a high rate of glucose utilization. Even though a moderate level of AIFM2 was detected in soleus at rest, it was little changed after either low- or high-intensity exercise. Our result is conceptually in line with the report of a previous study that AIFM2 expression is induced in BAT by feeding and cold exposure, both of which induce glucose-fueled thermogenesis (12). The same study also identified that cAMP-inducible transcription factor CREB directly regulates Aifm2 transcription (12). However, we found that this regulation does not appear to be conserved in myotubes. Moreover, Aifm2 expression in C2C12 myotubes was not regulated by caffeine or the AMPK activator, AICAR, which are known activators of signaling pathways that facilitate exercise (data not shown). This suggests that AIFM2 may be regulated by other signaling pathways and that it may also require in vivo context to modulate muscle contraction and exercise, selectively operating in EDL. Along this line, we found that Aifm2 KD altered NAD⁺ levels and the NAD⁺-to-NADH ratio selectively in EDL but not in soleus or GA. Further studies are necessary to identify upstream regulators of AIFM2 expression and mechanisms that confer EDL-selective regulation of glycolytic activity.

EDL is composed of mostly fast-twitch glycolytic muscle fibers and expresses a higher level of glycolytic enzymes, such as phosphofructokinase-1 and lactate dehydrogenase, compared with soleus (29); thus, it is an ideal location for glycolysis. Even though EDL has a small total mass, the absolute amount of glucose utilization is significantly greater in EDL compared with soleus (30). Also, notably, EDL-specific reconstitution of Glut4 expression in Glut4-null mice completely rescues their defect in insulin-stimulated glucose uptake (31). These findings strongly support the idea that altered metabolism in EDL alone can shift whole-body glucose homeostasis. In that sense, it is not entirely surprising to find that EDL-targeted OE and KD altered blood parameters of glucose during high-intensity exercise. However, we note that the nature of our gene delivery method (via electroporation) does not allow for achievement of complete muscle-depot specificity; thus, we cannot completely exclude a potential contribution from other muscle types.

Exercise both prevents and mitigates metabolic syndrome; thus, AIFM2’s role in promoting exercise capacity itself could be beneficial to patients. Other than its role in mobility and exercise, skeletal muscle is an important metabolic tissue playing a key role in glycemic control. Patients with obesity and diabetes have defects in shifting from oxidative to glycolytic metabolism in response to insulin (32,33). It has been suggested that reinforcing glucose utilization can serve as an adaptive response to improving metabolic dysregulation in diabetes conditions. For example, inhibiting myostatin signaling leads to glycolytic muscle growth and improved glucose homeostasis (34). Human studies showed that resistance training improves the metabolic profile of patients with diabetes by promoting the glycolytic program in fast-twitch muscles (35,36). Similarly, a mouse genetic study showed that switching from oxidative to glycolytic metabolism due to transgenic expression of Baf60c in muscle confers protection from diet-induced insulin resistance (37). Thus, it will be intriguing and important to test whether increased AIFM2 function at rest, even when dissociated with exercise, plays a beneficial role in the context of insulin resistance and glucose intolerance.

Overall, we found that skeletal muscle AIFM2 functions as an NADH oxidoreductase and plays a critical role in sustaining high-intensity exercise by supporting a high rate of glycolysis in skeletal muscles during exercise. A better understanding of the regulatory mechanisms of fuel utilization during exercise and diet-induced adaptation may shed light on novel therapeutic targets and strategies to improve exercise inefficiency.

Acknowledgments. The authors thank Dr. David Wasserman (Vanderbilt University) for sharing protocols for in vivo glucose uptake assay.

Funding. Work was funded by National Institute of Diabetes and Digestive and Kidney Diseases (R01 DK116008 and Pilot and Feasibility P30DK116074 [Stanford Diabetes Research Center] to S.K. and R01 DK123843 to H.S.S.).

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

Author Contributions. H.S.S. and S.K. supervised experiments, and S.D.V. and S.K. co-wrote the manuscript. C.J. consulted on data analysis and interpretation. Experiments were carried out by H.P.N., S.D.V., B.C.J., D.Y., F.L., D.Y., A.P., K.M., S.J., and S.-H.P. S.K. 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.