How is this effect achieved in vivo

Using this value, the reciprocal of the integral time, , was increased until the step response error was determined to converge to a near zero value to ensure a proper transient response. Through appropriate tuning, the system may exhibit relatively small or no overshoot during the muscle contraction response. In this design, seen in Fig 4 , the control effort is a function of the adaptive gains, reference input, and current state.

The adaptive gains, likewise, depend on the reference input, current state, and error between the unknown plant and reference system. The control effort is computed using these gains, which then drives the mass-spring-muscle system to the reference system.

The reference input, r t , is then chosen so that the reference system x r t tracks the desired trajectory,. It is assumed that ideal gains, and , exist to drive the system to the reference model through an ideal control law,. By substituting u id t into Eq 6 , the following relationship is obtained, 8. The error is given by.

Choosing and , the closed-loop system expression simplifies to the reference model and the error dynamics are given by 9. However, since a and b are unknown, the ideal gains and cannot be computed and the ideal control law u id t cannot be implemented.

Theorem 1. Then, the closed loop system given by Eqs 6 , 7 , 10 and 11 with the adaptive control law 12 is Lyapunov stable and the tracking error e ad t converges to zero.

Define the difference between the adaptive and ideal gains as and. By substituting the adaptive control input Eq 12 into Eq 6 and introducing the definitions of , , , and , we obtain, Accordingly, the error dynamics are given by computing the difference between Eqs 13 and 7 as follows, Next, introduce the following Lyapunov function candidate [ 39 ], Using Eqs 14 , 10 and 11 , the Lyapunov derivative is determined by computing the time derivative of Eq 15 along the trajectories of Eq 14 , While larger tuning constants will speed the rate of adaptation, high adaptive rates will lead to aggressive oscillatory behavior in the response, which is undesirable in muscle stimulation.

On the other hand, limiting the rates of adaptation will also limit the rate of system convergence. This alteration allowed for the use of less aggressive adaptive rates by starting each experimental run with conditions closer to a converged state, but also required an additional tuning effort. In order to maintain many of the overall benefits observed with the implementation of the PI controller but facilitate the advantages of adaptation, the ADP-PI controller presented in [ 41 ] was chosen.

Alterations, however, were made to this strategy to support a more heuristic tuning methodology which was used with the classical PI controller. The design of this controller is given in Fig 5. The system assumes a single input-single output SISO form where x t is the state vector. At the core of the algorithm, a classical PI controller provides a linear contribution to the overall control strategy, augmented with an adaptive control effort which depends on adaptive gains, current system state, desired trajectory, and a set of structured nonlinear functions of the state.

The adaptive gains are continuously updated based on system state, trajectory, error between the reference system, and the same set of nonlinear functions. Since the PI control relies on the integration of the error between current and desired states, an additional integral state is introduced,.

This state is then augmented to the current model to provide an overall representation, 18 where. To implement the adaptive control algorithm, a reference system needs to be introduced.

In particular, we chose the known dynamics of Eq 18 which can be written as 19 where. A standard PI controller is chosen to guarantee that Eq 19 tracks the desired trajectory, that is 20 where. Substituting Eq 20 into Eq 19 provides the following closed-loop system, which will be used as the reference system for the design of the adaptive controller 21 where A r is defined as, By designing k P and k I through classical techniques, A r is guaranteed to be Hurwitz.

This, however, is under the assumption that both a and b are known. In reality, however, k P and k I are tuned in the same manner discussed with the PI controller. Once satisfactory values are determined, approximate values for a and b are found by examining the step response to achieve a similar trend in simulation. To design a controller which guarantees that the augmented system Eq 18 converges to the reference system Eq 21 , we introduce the following tracking error, 23 and analyze its dynamics, obtained as follows, Next, we define the control effort u t as the sum of a linear part and an adaptive one, that is , where u lin t is given by Eq 20 and u ad t needs to be designed.

Accordingly, Eq 24 can be simplified to the following expression, Since A r is designed to be Hurwitz, by choosing u ad t as follows 26 the error dynamics Eq 25 simplifies to which guarantees asymptotic stability and hence convergence of e a t to 0.

Hence, the ideal control law needs to be modified into an adaptive one which, instead of those ideal values, implements adaptive variables and. Theorem 2. The closed loop system given by Eqs 18 , 21 , 27 and 28 with the adaptive control law 29 is Lyapunov stable and the tracking error e ad t converges to zero. By defining the difference between the ideal gains and the adaptive gains as and , the error dynamics can be rewritten as follows, Next, the following Lyapunov function candidate is introduced, 31 where is a positive definite matrix.

Using Eqs 30 , 27 and 28 , the Lyapunov derivative is determined by computing the time derivative of Eq 31 along the trajectories of Eq 30 32 where the last inequality follows from the fact that is positive definite since it satisfies the Lyapunov algebraic equation with P positive definite and A r Hurwitz.

Hence, LaSalle-Yoshizawa Theorem guarantees that e a t will converge to zero as time goes to infinity and all signals stay bounded [ 40 ]. Since open-loop systems performances are not guaranteed to remain within acceptable limits, while closed-loop systems are known to be more accurate due to feedback, no open-loop systems were considered in experimentation. Thus, the main focus of the paper is the application of an optimal approach, i.

In our experiments, the fast contracting EDL mouse muscle was activated at Hz to evoke tetanic contractions. Cages were maintained by vivarium staff as needed. Mean body mass was Mice were euthanized with CO 2 to carry out the experiments. One muscle was used per animal. The approval was obtained prior to the start of the study. All efforts were made to minimize animal suffering.

Since isolated muscle viability ex vivo decreases with increased temperature, we selected room temperature to maximize muscle viability [ 42 ]. The muscle aligned equidistant between two platinum electrodes. The EDL muscle was set at 1g of resting tension which corresponds to , prior to the initial stimulation. Feedback was captured from the instantaneous position to minimize the error, e t , between the desired predefined trajectory and the actual muscle contraction. Each contraction trial was executed at a stimulation frequency of Hz with a duration ranging from 7s to 10s.

The constant voltage applied from the stimulator Aurora Scientific Inc. Measurements were recorded, which primarily included muscle force, muscle contraction, and twitch responses. The latter were recorded at the beginning and between trials to monitor muscle fatigue and assess muscle decay.

Furthermore, resting periods of one minute were provided between trials to ensure the quality of the muscle. After the conclusion of each trial, physical measurements of the muscle including muscle mass and muscle length at resting tension were recorded.

The average muscle mass and muscle length were 0. A timeline of events describing the step by step process for an experiment is detailed on Supporting Information S3 Table. The board recorded and sampled at Hz.

The FPGA outputted a digital pulse width modulation PWM signal at a frequency of Hz, which consists of a succession of pulses in which the pulse duration is determined by the feedback controllers to output the adequate stimulation. The FPGA sends the PWM stimulation signal via the digital output ports to the external trigger on the current stimulator, which are connected to the platinum electrodes.

The control effort, u t , determines the parameters of the duty cycle transmitted to the FPGA, which onsets the muscle contraction. The signal from the encoder, comprised of two square waves in quadrature, was detected by the FPGA through the digital input pins at a rate of 1MHz. The high sample rate ensures all encoder pulses are considered.

Simultaneously, the controller, running on the microprocessor, updated the output and pulse width parameters from the FPGA using the measured position.

Four controllers one open-loop and three closed-loop were evaluated in simulation and the closed-loop controllers were validated experimentally. Four trajectories were implemented per controller. In the case of the simulations, tuned trajectories were attained deterministic results.

For the experimental results, trials were carried out for all trajectories. STE is determined as,. In addition, mean settling time for the step trajectory was calculated across all experiments per controller. To ensure normality, a square root transformation for the ramp and square trajectories was required. Since the data did not meet the homogeneity of variances assumption, we implemented the Games Howell post hoc test.

The statistical analyses focused on identifying differences between all controllers. The number of samples for each trajectory are detailed on Table 1. Tuning of parameter gains were largely dependent on each muscle. To determine the appropriate gains for the control effort, we applied first a step input for tuning the muscle gains. All the trajectory responses exhibited friction, which was mainly attributed to the mass motion along the rail.

Results are displayed in Figs 7 — The PI controller gains, k P and k I , are user-defined parameters which drive the response of the system. The input control effort obtained through the PI controller generated the appropriate PWM signal for muscle contraction.

Examples of the pulse duration input signal and its corresponding output are shown in S2 — S3 Figs in the Supporting Information. The MRAC controller required six tuning parameters. A sample of these parameters are listed for simulation and experiments on Table 2. Specifying these tuning parameters allowed the MRAC controller to determine the required control input signal generating the pulse duration for muscle contraction.

The pulse duration and its corresponding trajectory response are shown in Supporting Information S4 and S5 Figs. This nonlinear controller, which defines the control input, u t , is composed of a linear and adaptive component. The known state constant a and the known input constant b were determined through computational simulations using previously calculated PI gains, k P and k I.

These were then used to determine the P matrix by solving , prior to the start of each trial. The ADP-PI controller converged at a rate proportional to muscle fiber contraction, thus minimizing the error between the reference trajectory and the actual trajectory.

The controller required tuning of ten parameters, which are listed in Table 3 , for the simulation and experimental validation. First, the linear gains, k P and k I were computed using a method similar to the Ziegler—Nichols Critical Gain method presented in [ 35 ]. The constants, a and b are parameters determined for the linear muscle model behavior as described by Eq In this way, the system can converge quickly through adaptation.

Moreover, S6 and S7 Figs Design in the Supporting Information displayed the pulse duration input signal and its corresponding output for all trajectories in experiment. The results obtained implementing the step input trajectory for the three controllers are displayed in Fig 7A—7C.

Overall, no overshoot was observed for all the controllers. The mean settling time using the PI controller was , while the mean settling time using the MRAC controller exhibited fluctuations and did not settle during the 8s of muscle stimulation. The sine trajectories are shown in Fig 9A—9C.

The experimental and simulation responses using the PI controller, shown in Fig 9A , exhibited some oscillatory behavior during relaxation at interval sets [3s, 5s] and [8s, 10s] as opposed to the open-loop simulation. The PI simulation was within bounds of the experimental trajectories while the open-loop case did not remain within the bounds during the initial contraction. The oscillations during muscle relaxation were attributed to the extension spring in the setup.

This effect likely indicated that the spring force after muscle contraction was pulling the muscle faster than the time required for the muscle to return to its original length after stimulation. The MRAC sine trajectories from simulation and experiment were also characterized by oscillatory behavior during adaptation, as shown in Fig 9B.

Both responses followed the set reference trajectory as opposed to the open-loop case, which was characterized by a higher amplitude and did not capture these fluctuations. These trajectories followed the reference system in both experiment and simulation, but lagged after 3s of stimulation. Oscillations prevailed for the ramp trajectories in experiment and simulation as shown in Fig 10A—10C as well as a delayed response of approximately 1s for all controllers.

For the PI controller Fig 10A , the mean ramp response and distribution attenuated after 6. The MRAC controller in experiment and simulation showed highly oscillatory behavior throughout the 10s of stimulation as observed in Fig 10B. On the contrary, the ramp trajectories using the ADP-PI controller in experiment and simulation refer to Fig 10C attenuated its oscillatory behavior after 5. The square trajectories are illustrated in Fig 11A—11C. The experimental and simulations using the MRAC controller sustained oscillatory behavior throughout the 9s of stimulation as observed in Fig 11B.

Moreover, we observed that no experimental samples restored to its original position after the initial contraction. This behavior could be an indication of muscle fatigue and friction between the mass and rail of the setup. In addition, the closed-loop simulations presented similar behavior as the experiments as opposed to the open-loop simulation which did not capture the nonlinearities affecting the system.

To evaluate the treatment of the controllers for muscle stimulation in experiment, the mean STE was used to test the following hypotheses: 1 the condition of the controller is significant and has an effect on muscle stimulation, and 2 the ADP-PI controller is the optimal controller to implement for muscle stimulation compared to PI and MRAC.

To assess controller tracking performance, we analyzed the mean STE values between the three controllers for all trajectories. Box plots of the step and sine trajectories are shown in Fig The sine trajectories satisfied the normality condition.

For the ramp and square trajectories, box plots are displayed in Fig For these trajectories, we performed a square root transformation to ensure normality, which was validated through Q-Q plots of the residuals. Lastly, the square trajectories satisfied the normality condition.

Here, we have presented numerical simulations of muscle stimulation via adaptive and linear closed-loop controllers as well as experimental validation of these systems ex vivo. By testing the efficacy of the controllers using the EDL muscle in isolation, we can measure contractions solely due to the stimulation delivered by the control effort. The three feedback controllers are effective in tracking a variety of reference trajectories. In addition, the model simulated with the closed-loop controllers generally matched well with the experimental results, both in terms of quantitative metrics and qualitative behavior.

Overall, experiments validated controller simulations since it captured nonlinearities such as delays and friction behaviors, which were not present in the open-loop responses.

All closed-loop controllers worked in simulation to track signals. To build the simulations, linear dynamics were assumed for muscle behavior while nonlinear controllers, with the exception of the PI controller and open-loop case, were applied. The successful outcomes of this comparative study highlights the advantage of feedback controllers in electrical stimulation without the requirements of higher order muscle models that carry great computational demands.

Nevertheless, challenges still persist when implementing control systems for FES treatments, such as the nonlinearities of the muscle system, appropriate coordination of muscle group activation, non-physiological recruitment eliciting fatigue associated with muscle contractions, time delays of the biochemical processes between stimulation and the start of muscle contraction, spasticity, just to mention a few [ 12 , 16 , 43 — 46 ]. All closed-loop controllers are consistent with experiments.

The closed-loop muscle system simulations qualitatively captured the behavior of the EDL mouse muscle as evidenced by the trajectory outputs in the Results section. Several assumptions were made when building, testing and coding the experimental system, which included Coulomb friction and no moment effects, among others. Moreover, spring reaction forces and friction contributed to the stick-slip behavior observed in all trajectories during muscle relaxation. These effects observed in experiments were captured by the closed-loop controllers in both simulation and experiments, as opposed to the open-loop cases.

The rate of muscle fiber contraction and adaptation worked in synergy to output adequate tracking of the reference system with minimal system error. Although ten parameters characterized the system, the linear and nonlinear components could be tuned separately for optimal output. Thus, the ADP-PI controller stimulated the muscle efficiently by providing real-time changes to the gains and by reducing the influence of perturbations in the system e.

Overall, adaptation enabled fastest tracking error convergence and remained stable while the parameters varied in time. The ADP-PI controller best tracked the set trajectories by first using the PI component to approach the area of convergence, thus minimizing the system error while the adaptive component fine-tuned the controller.

In this way, trajectory responses were characterized by minimal oscillatory behavior, as compared to the MRAC controller which lacks this synergy. Closed-loop systems are capable of using the feedback information to minimize the system error and tracking of muscle contraction.

However, comparing the closed-loop controllers in experiments through statistical analyses revealed a significant effect of controller type on muscle stimulation for the sine and ramp trajectory inputs. On the other hand, MRAC showed faster adaptation compared to the rate of muscle fiber activation as characterized by overshoot and oscillations captured for both simulation and experiment.

By incorporating the linear and adaptive components via the ADP-PI controller, we were able to apply a robust feedback system capable of handling system uncertainties and unmodeled dynamics. From a theoretical perspective, closed-loop controllers mimic the feedback inherent in the nervous system, since our nerves carry out both commands to muscles and feedback to the central nervous system; thus closed-loop control may be better integrated with, biological systems.

The open-loop approach is not a natural process and simulations for this case failed to capture the key characteristics observed in our FES experiments. Although the same muscle-mass-spring system was applied for all simulations, the open-loop case was unable to assess the muscle response through tracking and capture key behaviors, such as friction and response delays. Conversely, tuning capabilities through closed-loop systems have the ability to capture these behaviors and compensate for muscle fatigue by changing the control input and outputting the appropriate PWM signal.

Nevertheless, these controllers warrant further investigation, for example, to study the mismatch between the experimentally-observed control effort and that predicted by the model. We observed that the experimental control effort was generally smaller for experiments than predicted by simulations. During experiments, the muscle was suspended in a bath submerged in ionized solution which was not represented in simulation. The differences between these control input signals and muscle response will be explored in future studies.

In conclusion, the application of feedback controllers, very rarely used in FES applications, has the potential to become an effective tool for treatments regarding skeletal muscle stimulation. Adaptive controllers, especially the ADP-PI controller, provide a beneficial strategy for muscle treatment by providing appropriate stimulation to the affected muscle and considering the muscle response to set stimulation.

This proof of principle can be extrapolated to a clinical setting as a rehabilitation tool since it can provide patients with an efficient method of muscle activation.

From a translational research perspective, next steps involve two major components: 1 medical device design to incorporate these controllers in a user-friendly rehabilitation tool comprising of electrodes, stimulator, and embedded systems tracking joint movement, and 2 standard input trajectories for tracking joint movement to ensure best practices for patient rehabilitation. Both aspects introduce new challenges to guarantee reliable medical device design, ensure safety procedures, and integrate human factors for future clinical trials.

Different values for k P and k I are determined to identify the appropriate tuning parameters. Experimental sample applying the PI controller for muscle contraction compared to the muscle system simulation for the PI and open-loop case. Trajectories include A the step and B sine functions.

Trajectories include C ramp and D square functions. Special thanks to Dr. This material is based upon work supported by while serving at the National Science Foundation. Funding acquisition: AL. Project administration: AL RG. Resources: RG AL. Writing — original draft: PJC. Browse Subject Areas? Click through the PLOS taxonomy to find articles in your field. Abstract Functional Electrical Stimulation is a promising approach to treat patients by stimulating the peripheral nerves and their corresponding motor neurons using electrical current.

Kellermayer, Semmelweis Egyetem, HUNGARY Received: December 16, ; Accepted: February 9, ; Published: March 8, This is an open access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose.

Introduction Functional Electrical Stimulation FES is a rehabilitation technique used to treat patients with absence of a functioning nervous system by maintaining blood flow and directly stimulating peripheral nerves [ 1 , 2 ]. Materials and methods The Materials and Methods section is divided into three main components: 1 the Skeletal Muscle Simulation, detailing the computational muscle system, 2 Controller Design, outlining the three controllers implemented in experiment and simulation, as well as 3 Experimental Design, which focuses on the explanation of the mouse muscle trials performed and the statistical analyses of these results.

Download: PPT. Fig 1. Fig 2. Controller design For the muscle model simulation, an open-loop system and three closed-loop controllers were used to control muscle contraction. Proportional integral controller. Fig 3. Block diagram of the closed-loop system using a Proportional Integral controller. Fig 4. Block diagram of the closed loop system using a Model Reference Adaptive Controller.

Adaptive augmented PI control. Wortmannin was from Sigma-Aldrich St. Extensor digitorum longus and soleus muscles were homogenized in 0. Plasma insulin concentrations were measured by enzyme immunoassay kits Morinaga Institute of Biological Science, Yokohama, Japan. Aliquots of the supernatants were used for the immunoblot analysis. Louis, MO , respectively. When two mean values were compared, analysis was performed using an unpaired t test. Skeletal muscle fibers are classically divided into either type I slow-twitch, oxidative or type II fast-twitch, glycolytic.

Under resting conditions, the temperatures of the EDL rich in type II fibers and soleus rich in type I fibers muscles were Heat stimulation in vivo facilitates muscle glucose uptake in vivo and in vitro.

Immediately after the cessation of heat stimulation, the temperatures of extensor digitorum longus EDL and soleus muscles were measured in situ a , and then both muscles were snap-frozen.

Temp temperature. Elevation of muscle temperature per se directly facilitates muscle glucose uptake in vitro. During the heat stimulation in vivo, the glucose uptake measured in vivo was significantly increased in the EDL and soleus muscles Fig.

This thermal effect was observed again when glucose uptake was measured in vitro in both isolated muscles immediately after the heat stimulation in vivo Fig. When blood parameters were measured immediately after the heat stimulation in vivo, blood glucose levels were comparable between groups, while higher blood lactate levels were observed in heat-stimulated rats Table 1.

We also found that heat stimulation in vivo significantly increased plasma insulin concentrations Table 1. In order to exclude any systemic interference, we further examined the effect of heat stimulation in vitro on glucose uptake in vitro. When isolated EDL muscles were incubated in water set at various controlled temperatures, muscle temperature increased with water temperature Fig.

To examine whether the thermal effect on glucose uptake represented a controlled physiological response, rather than an uncontrolled inflow of glucose due to cellular damage e. Compared to glucose uptake measured during heat stimulation incubation 2 in Fig. Re-stimulation again increased glucose uptake to the level observed in the first heat stimulation incubation 5 in Fig. These tightly controlled responses of glucose uptake to heat stimulation suggest that the thermal effect on glucose uptake is a specific cellular response to temperature.

Time course for glucose uptake in vitro in response to heat stimulation, cessation of stimulation, and re-stimulation in vitro. Incubation schedules for the time-course experiment are shown in a. Table 2 summarizes the effect of heat stimulation on EDL muscle metabolites. This facilitated glycogenolysis was accompanied by an accumulation of G6P and lactate levels, showing promotion of glycolysis. Figure 4 shows the effect of heat stimulation on muscle signal transduction pathways associated with glucose uptake.

Elevation of muscle temperature has direct impacts on muscle signal transduction pathways associated with glucose uptake. A representative blot is shown above each bar.

In parallel experiments, the effects of pharmacological inhibitors of PI3K-independent pathways on heat stimulation-induced glucose uptake were examined in EDL muscles Fig. In agreement with the immunoblotting results, compound C, an AMPK inhibitor, partially but significantly inhibited heat stimulation-induced glucose uptake. AMPK inhibitor partially blocked the heat stimulation-induced muscle glucose uptake in vitro.

If heat stimulation-induced glucose uptake occurs through only PI3K-independent pathways, it would be expected that the thermal effect and the maximal effect of insulin together on glucose uptake would be additive [ 12 ]. However, when the two effects were measured in combination, we observed only a partially additive result Fig.

Given that heat stimulation increased Akt phosphorylation, we expected that wortmannin, a PI3K inhibitor might inhibit heat stimulation-induced glucose uptake. In accord with this hypothesis, the thermal effect on glucose uptake was partially inhibited in the presence of wortmannin Fig.

PI3K inhibitor partially blocked the heat stimulation-induced muscle glucose uptake in vitro. Under conditions where accumulations of glucose transport-inhibitory substrates, including intracellular G6P, are present, the phosphorylation step by HK as well as the glucose transport step across the cell membrane could be a rate-limiting step for 2DG6P accumulation. In the balance between the inhibitory effect of HK by G6P accumulation and the stimulatory effect of HK by the Q 10 effect, we found a significant increase in 2DG6P accumulation in the heat-stimulated muscles measured in vivo and in vitro.

The present results confirm that elevation of muscle temperature per se can directly stimulate glucose uptake in skeletal muscle. Also, this effect was observed regardless of the composition of muscle fiber types. In the present study, despite the thermal effect on muscle glucose uptake in vivo, we did not observe a hypoglycemic effect in vivo during heat stimulation. This may be due to the promotion of hepatic glucose output, as reported [ 17 ].

We demonstrated here that elevation of muscle temperature has direct and acute impacts on signal transduction pathways associated with glucose uptake in intact skeletal muscle. However, because of the complexity of skeletal muscle, i. These results suggest a role of the AMPK-dependent pathway for the thermal effect on muscle glucose uptake.

The precise mechanisms by which heat brought about the activation of AMPK are not clear, but allosteric regulation via a reduction in ATP, PCr, and glycogen levels may have contributed [ 1 , 3 ]. The compound C has been frequently used as an inhibitor of AMPK, however, it also has inhibitory effects on a number of protein kinases [ 18 ].

Although the physiological role of these kinases other than AMPK for the muscle glucose transport system is not defined, the result with compound C needs to be interpreted with caution. One of most accepted concepts regarding glucose uptake is that muscle contraction in vitro induces glucose uptake through a PI3K-independent pathway.

This concept is largely based on evidence that wortmannin does not inhibit glucose uptake induced by muscle contraction in vitro [ 19 — 22 ]. In contrast to these data obtained in vitro, Wojtaszewski et al. These results might be interpreted to mean that glucose uptake induced by contractile activity in vivo, but not in vitro, contains a component that is inhibitable by wortmannin.

Importantly, we noticed in the present study that heat stimulation-induced glucose uptake was partially inhibited by wortmannin. An inhibitory effect of LY, another PI3K inhibitor, on the thermal effect on glucose uptake was also observed. Another possible interpretation is that wortmannin and LY blunted the thermal effect by inhibiting mammalian target of rapamycin mTOR rather than PI3K inhibition.

This is based on previous studies showing that muscle-specific deletion of rictor rapamycin-insensitive companion of mTOR , a component of mTORcomlex 2, have impaired insulin-stimulated glucose uptake [ 24 ] and that wortmannin and LY can prevent mTOR activity as a direct action [ 25 ]. The use of more specific inhibitors or gene-modulating animals will be necessary to clarify the wortmannin- and LYsensitive mechanisms for the thermal effect on glucose uptake.

It is possible that increased oxidant activity generated in response to heat stimulation [ 26 ] could impact insulin signaling and glucose uptake. One study [ 19 ], but not another [ 28 ], showed that NO-induced glucose uptake was partially inhibited by wortmannin. However, our experiment using an NOS inhibitor demonstrated that there was not a link between NO and heat stimulation-induced glucose uptake. Another oxidant, hydrogen peroxide H 2 O 2 , a major ROS, can also induce muscle glucose uptake [ 30 — 32 ].

Nevertheless, ROS contribution to the thermal effects observed in the present study are still unclear because AMPK seems to be more sensitive than Akt in response to heat stimulation. We also measured the thermal effect on signal transduction pathways and the effects of inhibitors on glucose uptake only in EDL muscles rich in type II fibers.

Thus, we cannot rule out the possibility that we might have obtained different results using soleus muscles rich in type I fibers. However, that study measured glucose uptake in vitro 15 min after the cessation of exercise, presumably enough time for the thermal effect to be lost. Therefore, the molecular mechanisms underlying the thermal effect on glucose uptake in type I fiber-rich muscles remain to be examined. Moon et al. Although we cannot explain this discrepancy between results, our results are partially supported by a previous study by Holloszy et al.

However, the range of temperature they used is not in a physiological range observed in mammalian skeletal muscle. In addition, the glucose transport system in amphibian muscle seems markedly different from mammalian muscle, because Holloszy et al. We have demonstrated herein that elevation of muscle temperature in a physiological range stimulates glucose uptake in rat skeletal muscle.

We showed that: 1 elevation of muscle temperature per se is a stimulatory factor to increase muscle glucose uptake, 2 the thermal effect on glucose uptake is very rapid and sensitive. Furthermore, 3 elevation of muscle temperature per se stimulates AMPK and Akt, and 4 the thermal effect on muscle glucose uptake is inhibitable by compound C, wortmannin, and LY J ApplPhysiol — CAS Google Scholar.

Acta Physiol — Proc Nutr Soc — Merry TL, McConell GK Skeletal muscle glucose uptake during exercise: a focus on reactive oxygen species and nitric oxide signaling. Curr Biol — J Appl Physiol — Moon B, Duddy N, Ragolia L, Gegum N Stimulation of glycogen synthesis by heat shock in L6 skeletal-muscle cells: regulatory role of site-specific phosphorylation of glycogen-associated protein phosphatase 1.

Biochem J — Biochem Biophys Res Commun — PubMed Article Google Scholar. Rev Physiol Biochem Pharmacol — Ueyama A, Sato T, Yoshida H, Magata K, Koga N Nonradioisotope assay of glucose uptake activity in rat skeletal muscle using enzymatic measurement of 2-deoxyglucosephosphate in vitro and in vivo. Biol Signals Recept — Metabolism — J Physiol Sci —9. Anal Biochem — A practical guide.

Humana, Totowa. Book Google Scholar.