Contraction-Mediated Glucose Transport in Skeletal Muscle Is Regulated by a Framework of AMPK, TBC1D1/4, and Rac1

In the presence of insulin, skeletal muscle accounts for up to ∼85% of peripheral glucose disposal from the blood (1). In addition to insulin stimulation, glucose uptake into skeletal muscle is increased by contraction during exercise. Both stimuli lead to the redistribution of the facilitative GLUT4 from intracellular storage vesicles to the cell surface, resulting in increased glucose clearance from the blood stream (2,3). The process of contraction-mediated GLUT4 translocation is tightly regulated and, up to now, not fully understood (4). A major mechanism linking muscle contraction to GLUT4 translocation involves the activation of the serine/threonine protein kinase AMPK by AMP (5,6). In a contracting muscle, cellular AMP levels increase as a result of the amplified energy demand. AICAR is a known chemical activator of AMPK and is thus frequently used as an in vitro contraction mimetic (7). However, contraction signaling is clearly more complex and involves more players in addition to AMPK. For instance, AMPK-related kinases (ARKs) have been discussed to facilitate AMPK-independent contraction- or exercise-induced glucose uptake. Kinases belonging to this group share the commonality of being activated by their upstream kinase liver kinase B1 (LKB1) (8,9).

As another very relevant component of contraction-mediated glucose transport into skeletal muscle, the Rho GTPase Ras-related C3 botulinum toxin substrate 1 (Rac1) has emerged recently. Rac1 is activated by mechanical stress or stretching of the muscle during contraction, a process that is independent from the mainly metabolic signaling pathways involving AMPK (10–12). Both AMPK and Rac1 display an extraordinary degree of evolutionary conservation, emphasizing the physiological relevance of at least two redundant mechanisms to secure glucose clearance from the blood stream upon physical activity (13,14). When combined, Rac1 and AMPK inhibition constitute nearly the complete contraction-dependent glucose disposal into skeletal muscle (15). Activation, and thus phosphorylation, of AMPK eventually triggers a complex signaling cascade, resulting in an insulin-independent elevation of glucose transport. Two main downstream effectors of AMPK are the RabGTPase-activating proteins (RabGAPs) TBC1D1 and TBC1D4. Previously, we demonstrated fiber type–dependent regulation of the two RabGAPs through AICAR stimulation of muscles (16). Whereas Tbc1d1-deficient (D1KO) animals display substantially impaired insulin- and AICAR-stimulated glucose transport into glycolytic extensor digitorum longus (EDL) but not oxidative soleus muscle, adverse observations have been described for Tbc1d4 knockout (D4KO) mice. These results are in line with the RabGAPs’ expression pattern, where TBC1D1 is the predominant isoform in glycolytic skeletal muscle and TBC1D4 shows the highest abundance in oxidative skeletal muscle and adipocytes (16–18). Likewise, GLUT4 protein abundance is reduced in oxidative skeletal muscle and white adipose tissue from D4KO and in glycolytic skeletal muscle from D1KO mice (19–22). Of note, fiber type specificity of Tbc1d1 and Tbc1d4 expression patterns appear to differ between mouse and human skeletal muscle (23,24). Double-deficient Tbc1d1/4 (D1/4KO) mice present a combined muscular phenotype with impaired insulin- as well as AICAR-mediated glucose transport and reduced GLUT4 content in all muscle types. In addition, these mice display deteriorated whole-body glycemia, unlike the single-knockout mice, and presumably because of a compensatory action of the remaining RabGAP (16,25).

Interestingly, several seemingly redundant cellular pathways have been identified in addition to the activation of AMPK mediating contraction-stimulated glucose uptake, including, but not limited to, Rac1/actin and CAMK signaling (26–28). For instance, a transgenic mouse strain overexpressing a dominant inhibitory mutant of the AMPKα2 subunit (AMPK-DN) in muscular tissues shows completely ablated AICAR-stimulated glucose uptake into skeletal muscle, whereas a contraction stimulus still results in a reduced, but significantly elevated transport of glucose into the muscle (29,30). The complexity of this regulation supposably represents the evolutionarily conserved need to maintain muscle energy supply in situations of high energy demand, such as physical activity. The aim of the current study was to segregate the individual contribution of TBC1D1, TBC1D4, and AMPK to skeletal muscle glucose metabolism in vivo as well as in different skeletal muscle types.

Chemicals and buffer ingredients are listed in Supplementary Table 1.

All animal experiments were approved by the ethics committee of the State Ministry of Agriculture, Nutrition and Forestry (State of North Rhine-Westphalia, Germany). Three to six male (12–30 weeks old) mice per cage were housed at 22°C on a 12-h light-dark cycle (lights on at 6:00 a.m.) with ad libitum access to food and water. After weaning, animals received a standard chow diet containing 22.1% (wt/wt) protein (30% of calories), 4.5% fat (11% of calories), and 53.35% carbohydrates (53% of calories) containing 3.3 kcal/g energy (ssniff-Spezialdiäten, Soest, Germany). Unless otherwise indicated, all interventions were conducted during the morning and in random-fed animals. Generation of C57BL/6J D1KO, D4KO, and D1/4KO mice has previously been described (16). Transgenic mice overexpressing AMPK-DN, a kinase-dead α₂-subunit of the AMPK enzyme, under the control of the muscle creatine kinase promotor (29) were a gift from Morris J. Birnbaum (University of Pennsylvania, Philadelphia, PA) and crossbred with RabGAP-deficient animals to obtain the five experimental genotypes: wild type (WT), transgenic AMPKα2-DN (DN), AMPKα2-DN-D1KO (D1KO-DN), AMPKα2-DN-D4KO (D4KO-DN), and AMPKα2-DN-Tbc1d1/Tbc1d4 double-deficient (D1/4KO-DN) mice. Sequences of all genotyping primers are listed in Supplementary Table 2.

For glucose tolerance tests, sterile glucose (2 g/kg body weight, 20% solution) was injected intraperitoneally into 16-h fasted animals. Compared with humans, a 16-h period is a comparatively long time for mice to fast. However, to reduce variability in basal blood glucose levels, it is beneficial to expose mice to an overnight fast (31). For AICAR tolerance tests, nonfasted mice were injected intraperitoneally with AICAR (250 mg/kg body weight; Toronto Research Chemicals, Toronto, Ontario, Canada), and blood samples were taken from the tail tip at 0, 15, 30, 60, and 120 min. Blood glucose was determined with a glucometer (Contour; Bayer, Leverkusen, Germany). On the basis of fasting plasma, an insulin- and glucose-level HOMA of insulin resistance (HOMA-IR) index was calculated according to the following equation (32): HOMA-IR = fasting insulin (ng/mL) × fasting blood glucose (mg/dL)/405.

Plasma insulin was measured with ELISA (Insulin Mouse Ultrasensitive ELISA; DRG Instruments, Marburg, Germany). Nonesterified fatty acids (NEFAs) in plasma were determined by an NEFA-HR(2) enzymatic colorimetric assay (Wako Chemicals, Richmond, VA) according to the manufacturer’s instructions.

Mice were familiarized with a calorimetric treadmill system (TSE Systems, Bad Homburg, Germany) at low speed (15 cm ∗ s⁻¹, 5° incline) for 5 min the day before performing the exercise test. Acute exercise testing consisted of a run to exhaustion starting at 15 cm ∗ s⁻¹ (5° incline) for the first 2 min and then continuously increasing speed by 5 cm ∗ s⁻¹ every 2 min. Time to exhaustion was defined as the time point when the mouse was no longer able to maintain its normal running position and showed frequent contact with the grid at the rear of the treadmill. VO_2max (mL ∗ min⁻¹ ∗ kg^−0.75) and respiratory exchange ratios (RERs) (VCO₂ ∗ VO₂⁻¹; average RER during the exercise test or RER at VO_2max) were determined with PhenoMaster software (TSE Systems).

Recombinant full-length His₆-Tbc1d1 and His₆-Nuak1 (AMP-activated protein kinase–related kinase 5 [ARK5]) were cloned, expressed in Sf9 cells using the baculovirus system, and purified by IMAC using Ni-NTA resins (QIAGEN, Hilden, Germany) as previously described (33). Activation of purified His₆-ARK5 was confirmed by Western blotting of pThr211 (data not shown). Phosphorylation reactions were carried out at room temperature for 20 min in the presence of 3 pmol ARK5, 40 mmol/L Tris-HCl (pH 7.4), 8 mmol/L MgCl₂, 200 μmol/L AMP, 2 mmol/L ATP, and 0.4 mmol/L dithiothreitol in a volume of 100 μL as previously described (33,34).

Glucose uptake was assessed by the accumulation of [³H]2-deoxyglucose (Hartmann Analytic, Braunschweig, Germany) in muscle with the use of [¹⁴C]mannitol (PerkinElmer, Waltham, MA) as an extracellular marker. Mice were anesthetized (500 mg/kg Avertin [2,2,2-tribromoethanol] via intraperitoneal injection), and intact EDL and soleus muscles were removed. Isolated muscles were mounted in a Muscle Strip Myograph chamber (Danish Myo Technology, Aarhus, Denmark) and incubated for 15 min in preoxygenated (95% oxygen/5% carbon dioxide) Krebs-Henseleit buffer (Supplementary Table 3) containing 5 mmol/L HEPES and supplemented with 8 mmol/L pyruvate and 15 mmol/L mannitol. Rac1 inhibition was achieved after 50-min preincubation with NSC23766 (200 μmol/L; Sigma-Aldrich) or DMSO as vehicle control, both diluted in Krebs-Henseleit buffer containing 1% BSA, 15 mmol/L mannitol, and 8 mmol/L pyruvate. All incubation steps were conducted under continuous gassing (95% oxygen/5% carbon dioxide) at 30°C. After recovery, a basal mechanical tension was applied (2–3 mN to EDL and 4–5 mN to soleus muscle), and muscles were again incubated for 30 min. Then, muscles were electrically stimulated to contract with 300-ms trains of 0.1-ms pulses at 160 Hz every second in the presence of 1 mmol/L [³H]2-deoxyglucose and 19 mmol/L [¹⁴C]mannitol. After 10 min of stimulation and an additional 10 min of radioactive incubation, muscles were immediately frozen in liquid nitrogen and stored at −20°C. Cleared protein lysates were used to determine incorporated radioactivity by scintillation counting. [¹⁴C]Mannitol counts were used to correct for the extracellular space.

Protein lysates (10–30 μg) were separated by 8–12% SDS-PAGE and transferred by tank blotting onto nitrocellulose membranes (Amersham Protran 0.45 μm). Membranes were blocked for 1 h with 5% fat-free powdered milk in Tris-buffered saline with Tween (Supplementary Table 3), incubated with primary antibodies and secondary horseradish peroxidase–conjugated antibodies as described in Supplementary Table 4, and developed with ECL reagent (PerkinElmer). Protein abundances were normalized to the housekeeping protein GAPDH.

RNA was extracted using QIAzol Lysis Reagent (QIAGEN, Hilden, Germany) according to the manufacturer’s instructions, and cDNA was synthesized with a GoScript Reverse Transcriptase system (Promega, Madison, WI). Real-time quantitative PCR was performed using self-designed SYBR Green PCR primers (Supplementary Table 5), and data were normalized to Gapdh (gastrocnemius muscle) or Rps29 (liver) according to the 2^−ΔΔCt method (35).

Male mice were sacrificed in the morning after 4 h of fasting, and plasma samples and homogenized tissue (40 mg liver and gastrocnemius muscle) were analyzed using a triglycerides assay kit (Trigs; Randox, Crumlin, U.K.) according to the manufacturer’s guidelines. Glycogen content was determined using the amyloglucosidase method (36) using a glucose oxidase–based colorimetric detection kit (Glucose liquicolor; HUMAN Diagnostics, Taunusstein, Germany) according to the manufacturer’s instructions.

Unless otherwise stated, data are reported as mean ± SEM. Significant differences were determined by one-way or two-way ANOVA (post hoc test, Tukey correction, Šidák correction, or Fisher least significant difference) or paired two-tailed Student t test, as indicated in the figure legends. P < 0.05 was considered statistically significant.

Both TBC1D1 and TBC1D4 have been shown to play crucial roles in skeletal muscle glucose transport. To investigate the contribution of each of the two RabGAPs to glycemic control, we subjected D1KO, D4KO, or D1/4KO mice to an overnight fast followed by 1 h of ad libitum refeeding. While there were no alterations in blood glucose or plasma insulin levels among the genotypes in the fasted state, D4KO and D1/4KO mice showed markedly increased blood glucose and plasma insulin levels compared with WT littermates after 1 h of refeeding. In contrast, D1KO mice displayed neither increased blood glucose levels nor elevated plasma insulin levels in the refed state (Fig. 1A and B). We and others previously demonstrated disturbed insulin- and AICAR-stimulated glucose transport into oxidative soleus muscle from D4KO and D1/4KO mice and glycolytic EDL muscle from D1KO and D1/4KO mice, respectively (16,25). However, glucose transport in response to contraction has not been investigated in D4KO and D1/4KO animals. Thus, we sought to investigate the specific roles of TBC1D1 and TBC1D4 in the regulation of contraction-mediated glucose transport into different skeletal muscle fiber types, subjecting isolated EDL and soleus muscles from D1KO, D4KO, and D1/4KO mice to ex vivo analyses of contraction-stimulated glucose uptake in a myograph chamber. D1KO and D1/4KO mice showed ∼50% reduced contraction-stimulated glucose uptake compared with WT controls in glycolytic EDL muscle. Basal glucose transport was also impaired in EDL muscle from D1KO and D1/4KO but not D4KO mice (Fig. 1C). In contrast, basal glucose uptake was unaltered among the genotypes in oxidative soleus muscle, and contraction-stimulated glucose transport was reduced exclusively in muscles from D1/4KO mice (Fig. 1D). Strikingly, skeletal muscle lacking either TBC1D1, TBC1D4, or both RabGAPs displayed significant residual stimulation of glucose transport in response to contraction, indicating spare signaling capacity. Contraction force determined as time for half-capacity was unaltered among the genotypes (Supplementary Fig. 1).

We generated a new mouse line by crossbreeding muscle-specific AMPK-DN mice with RabGAP-deficient animals on a C57BL/6J background (16,29). Overexpression of an inactive AMPKα2 subunit did not alter the abundance of TBC1D1 in D1KO or D4KO mouse skeletal muscles, and all genotypes overexpressing the AMPK-DN allele displayed a strong signal for AMPK (Fig. 2A and B). Despite some tendencies, both 6-h fasted blood glucose and plasma insulin levels were not significantly altered between the genotypes (Fig. 2C and D). The estimated HOMA-IR index revealed an impaired insulin sensitivity as a result of a combined inactivation of AMPK and Tbc1d1-deficiency (D1KO-DN). Interestingly, D4KO-DN completely ablated the increased HOMA-IR in D1KO-DN mice (Fig. 2E). No major differences were determined in body weight and body composition between the genotypes. Moreover, glucose tolerance was not altered because of AMPK or RabGAP deficiency (Supplementary Fig. 2A–D). During an intraperitoneal AICAR tolerance test, DN mice presented slightly elevated blood glucose levels after AICAR injection (Fig. 2F). In accordance with our previous findings, GLUT4 abundance was reduced in gastrocnemius skeletal muscle from D1KO-DN and D1/4KO-DN but not DN and D4KO-DN animals, and GLUT1 abundance was unchanged among the groups (Fig. 2G and Supplementary Fig. 2E).

Next, we subjected all five genotypes of the AMPK-DN RabGAP mouse line to a fasting-refeeding experiment with 16-h overnight fasting and 1 h of ad libitum refeeding. Fasted plasma blood glucose as well as fasted and refed plasma insulin levels were equal in all genotypes. In contrast, following 1 h of ad libitum refeeding, postprandial blood glucose levels were increased in D4KO-DN and D1/4KO-DN animals compared with WT controls (Fig. 3A and B). Following 16 h of fasting, plasma triglycerides were significantly increased in DN mice compared with WT littermates, an effect that was almost completely abolished by additive deficiency of both RabGAPs (D1/4KO-DN), whereas there were no differences among the genotypes in the refed state (Fig. 3C). Interestingly, the increased plasma free fatty acids in DN and D4KO-DN animals following 16 h of fasting were restored to WT levels in D1KO-DN mice (Fig. 3D).

We analyzed liver triglyceride and glycogen content as well as the abundance and phosphorylation status of regulatory enzymes in all genotypes. Liver triglycerides were not different in DN mice compared with WT littermates but increased in D1KO-DN, D4KO-DN, D1/4KO-DN mice (Fig. 4A). In contrast, protein abundance of PEPCK1 (PCK1) was only slightly elevated in D1/4KO-DN mice but not in the other genotypes, and no changes were observed for PEPCK2 (PCK2) (Fig. 4B). Liver glycogen content was slightly reduced in D4KO-DN and D1/4KO-DN animals compared with the DN genotype but unaltered in DN and D1KO-DN mice compared with WT controls (Fig. 4C). D1/4KO-DN animals displayed increased glycogen synthase (GS) protein abundance, but the ratio of phosphorylated (Ser641) to total GS (pGS/GS) was slightly decreased (Fig. 4D).

We determined the exercise capacity of DN, D1KO-DN, D4KO-DN, and D1/4KO-DN mice on treadmills. Muscle-specific inactivation of AMPKα2 (DN mice) led to a significantly reduced time to exhaustion compared with WT animals. Interestingly, both D1KO-DN and D1/4KO-DN mice showed an additional impairment in running capacity compared with the DN animals. In contrast, additional ablation of Tbc1d4 in D4KO-DN mice did not affect exercise performance to a larger extent than AMPK inactivation alone (Fig. 5A). These results were also reflected by significantly lower VO_2max values of DN and D1KO-DN mice compared with WT animals (Fig. 5B). The RER and the RER at VO_2max, however, were not changed among the five different genotypes (Fig. 5C and D). No genotype-dependent differences were observed for skeletal muscle triglyceride or glycogen content (Fig. 5E and F).

Intact isolated EDL and soleus muscles from WT, DN, D1KO-DN, D4KO-DN, and D1/4KO-DN mice were ex vivo contracted in a myograph chamber, and [³H]2-deoxyglucose uptake was measured. Compared with WT animals, EDL muscles from AMPK-inactive DN mice displayed only a very minor reduction in the basal state but substantially decreased glucose transport in the contracted state. In accordance with the impaired exercise performance in vivo, D1KO-DN and D1/4KO-DN, but not D4KO-DN, animals demonstrated substantially impaired contraction-mediated glucose uptake into EDL muscle in addition to the reduction caused by AMPK inactivation (Fig. 6A). In isolated soleus muscle, AMPK inactivation led to a clearly impaired contraction-stimulated glucose transport rate. However, RabGAP deficiency did not result in any further decline in contraction-stimulated glucose uptake (Fig. 6B). Ablation of both RabGAPs in combination with muscle-specific AMPK inactivation (D1/4KO-DN) resulted in a blunted contraction response in the oxidative soleus muscle. In contrast, contraction-mediated glucose transport was still moderately increased in the glycolytic EDL muscle from D1/4KO-DN mice compared with the basal state (Fig. 6A and B). We found reduced GLUT4 abundance in EDL muscle from D1KO-DN and D1/4KO-DN mice, whereas GLUT4 abundance in DN and D4KO-DN mice was unaltered (Fig. 6C). In contrast, GLUT4 content was diminished in soleus muscle from D4KO-DN and D1/4KO-DN but not from D1KO-DN mice (Fig. 6D). Inactivation of AMPK alone, however, did not affect GLUT4 content in either EDL or soleus muscles. An overview on the different responses in contraction-mediated glucose uptake between the two muscle types can be found in Fig. 6E.

We investigated the residual contraction response in EDL muscle from D1/4KO-DN animals and determined ex vivo contraction-mediated glucose uptake after preincubation with either DMSO or the Rac1 inhibitor NSC23766. The small Rho GTPase Rac1 has been described as an important factor in AMPK-independent contraction signaling (13,15), and Rac1 inhibition led to a pronounced increase of basal glucose transport into EDL muscles from WT mice. Contraction-stimulated glucose transport, in contrast, was substantially impaired in WT muscles as a result of Rac1 inhibition. Similarly, both D1/4KO and D1/4KO-DN muscles had markedly reduced contraction-stimulated glucose uptake. Of note, WT muscles after Rac1 inhibition as well as D1/4KO and D1/4KO-DN muscles treated with DMSO still showed a significant response to the ex vivo contraction stimulus. Additional Rac1 inhibition in D1/4KO and D1/4KO-DN muscles, in contrast, completely ablated the residual contraction-stimulated glucose transport (Fig. 6F). Rac1 inhibition as well as the applied contraction tension did not affect Rac1 protein abundance in EDL muscle (Supplementary Fig. 3A). In addition, the GTPase activity of Rac1 was not directly altered by TBC1D1 or TBC1D4 in vitro (Supplementary Fig. 3B and C). Supplementary Fig. 4 summarizes the observed mouse phenotypes that depend on either the AMPK inactivity or the RabGAP deficiency.

We analyzed mRNA and protein expression of AMPK-related kinases (ARKs) downstream of LKB1, Nuak1, Nuak2, and Sik1, 2, and 3, respectively, but found no genotype-dependent alterations in D1/4KO-DN muscles (Fig. 7A–C). However, to study possible kinase/substrate interactions, we expressed and purified both ARK5 kinase and full-length TBC1D1 using the baculovirus system and conducted phosphorylation assays in vitro as described in research design and methods. As illustrated in Fig. 7D and E, incubation of TBC1D1 with recombinant ARK5 led to phosphorylation of AMPK target sites Ser660 and Ser700 but not Ser231.

In the current study, we investigated the contribution of AMPK, TBC1D1, and TBC1D4 to contraction-stimulated glucose transport by using novel mouse models that combine deficiency in both AMPK and RabGAP functions. Our results identify a nonredundant fiber type–specific framework of GTPase signaling that controls contraction-dependent glucose transport into skeletal muscle.

We recently analyzed insulin- and AICAR-stimulated glucose transport into intact isolated skeletal muscle from mice lacking either one or both RabGAPs (16,25). Because AICAR may only partially act as a contraction mimetic in skeletal muscle, we conducted in the current study ex vivo electrical stimulation and contraction of muscles from KO mice and measured glucose transport. In concordance with AICAR stimulation, contraction-induced glucose uptake in EDL muscle was strongly reduced in D1KO and D1/4KO but not in D4KO muscles, indicating that TBC1D1 has a critical function in glycolytic muscle and that AICAR indeed mimics contraction. However, contraction-induced glucose uptake was normal in the oxidative soleus muscle, whereas AICAR-stimulated glucose uptake was markedly reduced (16), hence dissociating the stimulatory effect of AICAR from muscle contraction. Only combined lack of TBC1D1 and TBC1D4 led to strong reduction in contraction-stimulated glucose transport in soleus muscle.

These findings indicate, at least in oxidative fibers, that 1) TBC1D4 is not essential for contraction-mediated glucose disposal ex vivo and 2) the single ablation of each RabGAP isoform leads to compensatory action of the remaining TBC1D protein and that only the complete lack of RabGAPs is sufficient to block the contraction response. Of note, contraction-stimulated glucose uptake was not completely ablated in either muscle type from D1/4KO animals, indicating the involvement of additional factors in this process, one of them most likely being AMPK. To delineate the AMPK-RabGAP signaling axis, we generated animals with muscle-specific AMPKα2 inactivity (DN) in combination with the deficiency in either one or both RabGAPs. Expectedly, inactivation of the AMPKα2 subunit led to a substantial decrease in contraction-mediated glucose transport into both soleus and EDL muscle (29). However, in EDL muscle, combined inactivation of AMPK and both RabGAPs was not sufficient to fully abolish contraction-stimulated glucose transport, in contrast to soleus muscle. This indicates that the relevance of the RabGAP-AMPK signaling axis may depend on divergent metabolic properties of muscle fibers, expression of AMPK subunit isoforms, and intensity/duration of the applied contraction stimulus (38–40). The residual contraction response in D1/4KO-DN EDL muscle, however, suggests the presence of an additional pathway circumventing canonical AMPK-RabGAP signaling, presumably to ensure the integrity of contraction-mediated glucose transport.

The small Rho GTPase Rac1 becomes activated independently of AMPK by mechanical stress or stretching of skeletal muscle during contraction (10–12). Inhibition of Rac1 or KO partially decreased contraction-stimulated glucose uptake in EDL muscle (14). Moreover, Sylow et al. (15) demonstrated that combined inhibition of Rac1 and AMPK only partially prevents glucose uptake in response to contraction, indicating the involvement of multiple signaling routes in this pathway. Here, we show that pharmacological Rac1 inhibition completely abolished contraction-stimulated glucose transport into D1/4KO and D1/4KO-DN EDL muscles, demonstrating that TBC1D RabGAPs and Rac1 operate in concert with AMPK to regulate the GLUT4 translocation machinery. While RabGAPs are believed to regulate critical Rabs localized to GLUT4-containing vesicles, Rac1 is involved in remodeling of the actin cytoskeleton in skeletal muscle (14). Another potentially relevant compound of contraction-mediated glucose transport we identified is the AMPK-related kinase ARK5 as a novel upstream regulator of TBC1D1. ARK5 phosphorylated TBC1D1 at two of the three known AMPK phosphorylation sites, Ser660 and Ser700, while phosphorylation of Ser231, another major AMPK site (41), was not significantly increased by ARK5, possibly because of prephosphorylation of this site.

Previous studies have shown that either muscle contraction, AICAR, or insulin leads to enhanced ARK5 phosphorylation (42). It has been implicated that a muscle-specific loss of ARK5 can prevent high-fat diet–induced glucose intolerance, presumably by suppressing glucose uptake through negative regulation of insulin signaling in skeletal muscle (43). Thus, the ARK5-TBC1D1 interaction may present a key pathway linking muscle contraction to insulin sensitivity. Further studies need to focus on the interface between the two pathways.

Of note, there are clear discrepancies in muscle physiology between ex vivo and in vitro contraction and in vivo exercise. Because of the lack of neuronal interconnectedness and substrate provisioning by the capillary system, both in vitro and ex vivo systems do not fully reflect the complexity of skeletal muscle metabolism. However, they present useful tools to study specific cellular pathways in a reliable way that may not be possible during in vivo studies where circulating and environmental factors add an extensive degree of complexity to the system, potentially hampering the clarity of data interpretation (44,45). Consistent with the impaired contraction-stimulated glucose uptake into isolated skeletal muscles, we found that DN mice exhibit reduced exercise capacity on treadmills (46). Notably, time to exhaustion was further reduced in D1KO-DN and D1/4KO-DN but not D4KO-DN animals, implying a higher relevance for TBC1D1 in the AMPK-mediated control of physical fitness. In addition, VO_2max and skeletal muscle glycogen content (by trend) were exclusively decreased in DN and D1KO-DN animals, again highlighting the tight relationship between this particular RabGAP isoform in controlling exercise performance. Our data concur with that of Stöckli et al. (47), who demonstrated impaired exercise endurance in Tbc1d1 KO mice, and AMPK-DN mice were shown to be unable to perform high-intensity, but not low-intensity, treadmill exercise (48). Moreover, it has been observed that the activation of the AMPK-TBC1D1 signaling nexus is increased in cyclophilin-D KO mice, thus enabling higher exercise endurance in these mice (49). Nevertheless, reduced mitochondrial oxidative capacity, associated with decreased mitochondrial gene expression of peroxisome proliferator–activated receptor γ coactivator 1α, cytochrome C, and citrate synthase, may also contribute to impaired exercise tolerance in muscle-specific AMPKα1,α2 double KO mice (50).

We found marked differences in the impact of Tbc1d1 or Tbc1d4 deficiency on postprandial glycemia and exercise physiology. Lack of TBC1D4, but not TBC1D1, leads to increased postprandial blood glucose and insulin levels, indicating a more prominent role of TBC1D4 in the regulation of postprandial blood glucose levels and insulin action. Accordingly, postprandial hyperglycemia is also strongly associated with a loss-of-function variant of TBC1D4 that is common in Arctic populations (51). Interestingly, this effect is more pronounced in mice deficient in both RabGAPs (D1/4KO mice), pointing toward potential RabGAP interactions.

After 6 h of fasting, the combined Tbc1d1 deficiency and AMPK inactivation (D1KO-DN) led to impaired insulin sensitivity, whereas the HOMA-IR index was normal in D4KO-DN and D1/4KO-DN mice. Thus, AMPK, but not TBC1D1, is required for maintaining skeletal muscle insulin sensitivity in the fasted state. In contrast, depletion of Tbc1d4 leads to a restoration of insulin sensitivity despite the muscular AMPK inactivation, emphasizing the importance of the AMPK-TBC1D4 axis in the regulation of glycemia during fasting. The disparate impact of Tbc1d1 and Tbc1d4 deficiency on fasting glycemia and AICAR tolerance, however, cannot be explained by opposite changes in GLUT4 abundance. Instead, GLUT4 abundance in the different muscle types follows the previously observed expression pattern dependent on the RabGAP isoform (16), without further impact because of AMPK inactivation. In accordance with previous results (16), an adverse distribution of energy substrates is associated with RabGAP deficiency but independent from AMPK inactivation. Presumably secondary to the skeletal muscle insulin resistance, liver glycogen stores were depleted, whereas triglyceride levels were increased in RabGAP-deficient animals. In addition, muscle-specific overexpression of the AMPK-DN variant alone affects hepatic lipid metabolism, as reflected by increased fasting plasma free fatty acid and triglyceride levels in our study. These findings are in accordance with data from other metabolic studies in muscle-specific transgenic mice, demonstrating a complex regulatory pattern of whole-body glycemia involving a diversity of target tissues and organ crosstalk (52–54).

Collectively, our data demonstrate a nonredundant function for TBC1D1 and TBC1D4 in exercise- and insulin-mediated regulation of whole-body glycemia and, more specifically, contraction-induced glucose transport into skeletal muscle. We identify a new signaling axis consisting of two independent GTPase regulators, Rac1 and TBC1D1/4, both direct/indirect targets of AMPK and presumably other kinases, that act in concert to achieve the full contraction-mediated glucose uptake response in skeletal muscle. Shedding light onto these regulatory circuits represents one more step toward the understanding of mechanisms involved in glucose homeostasis.

Acknowledgments. The authors thank Angelika Horrighs, Anette Kurowski, Antonia Osmers, Carina Heitmann, Jennifer Schwettmann, Peter Herdt, and Cornelia Köllmer, German Diabetes Center (DDZ), Leibniz Center for Diabetes Research, Heinrich Heine University Düsseldorf, Germany, for expert technical assistance, support, and consultation. They also thank Dr. Annette Schürmann, Department of Experimental Diabetology, German Institute of Human Nutrition Potsdam-Rehbruecke, Nuthetal, Germany, and German Center for Diabetes Research (DZD), Partner Potsdam-Rehbruecke, München-Neuherberg, Germany, for providing GLUT1 antibodies and Dr. Morris Birnbaum, Pfizer Inc., Cambridge, MA, for providing the transgenic AMPK-DN mouse model. The authors thank Dr. James Hastie, Division of Signal Transduction Therapy, University of Dundee, Dundee, Scotland, U.K., for providing phosphorylated TBC1D1 antibodies.

Funding. This work was supported in part by grants from the Deutsche Forschungsgemeinschaft (DFG-RTG 2576 vivid to H.A.-H. and A.C.; CH1659 to A.C.), Deutsche Diabetes Gesellschaft (DDG), EFSD/Novo Nordisk Programme for Diabetes Research, the Ministry of Science and Research of the State North Rhine-Westphalia, and the German Federal Ministry of Health.

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

Author Contributions. C.d.W., L.E., S.E., C.S., and L.T. performed the experiments and analyzed data. C.d.W., L.E., A.C., and H.A.-H. were involved in the study design, contributed to the data analysis and interpretation, and wrote the manuscript. A.S., A.D.B., T.B., S.C., and T.S. contributed to the data interpretation. A.C. 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.