Insulin Sensitization Following a Single Exercise Bout Is Uncoupled to Glycogen in Human Skeletal Muscle: A Meta-analysis of 13 Single-Center Human Studies

A single bout of exercise leads to enhanced insulin action on glucose uptake and glycogen synthesis in prior contracted muscle (Fig. 1). By introducing the percutaneous needle muscle biopsy technique (1), the Swedish scientists Bergström and Hultman pioneered the search for exercise-induced factors facilitating muscle insulin action, as initially proposed in 1926 by Lawrence (2). They took advantage of the one-legged exercise model to demonstrate that a single bout of exercise subsequently led to enhanced glycogen synthesis in prior contracted muscle (3). The authors concluded that the “persistent enhancing effect” represents a phenomenon “localized to the exercised muscle by unknown factors.” Moreover, they speculated that “some of the beneficial effects of exercise in healthy as well as in diabetic subjects are mediated by these factors” (3). Subsequently, it was demonstrated that a single bout of exercise improves insulin action on glucose uptake in rat (4) and human (5) skeletal muscle by enhancing glucose membrane permeability (6) through recruitment of glucose transporters (GLUT4) to the surface membrane (7,8) (Fig. 1). A series of both rodent and human studies demonstrated an intracellular interplay between Akt (insulin) and AMPK (exercise) at the level of TBC1D4, and by use of transgenic models both AMPK and TBC1D4 were found to be necessary for exercise to improve insulin action on muscle glucose uptake (9–14) (Fig. 1). From studies in humans, it is also apparent that local redistribution of blood flow (perfusion) is necessary to accommodate the increased myocellular glucose disposal (12). The insulin-sensitizing effect on glucose uptake can persist for several days (15), and during the early period of recovery from exercise, the effect is reported to be at least 3 times higher than what is observed following 12 weeks of endurance exercise training (16). The phenomenon of insulin sensitization following exercise has been observed across mammalian species (4,5,9,17–19) and thus likely represents an evolutionarily conserved phenomenon ensuring adequate glucose supply for the highly prioritized glycogen resynthesis in recovery following exercise. The time window in which improved sensitivity for insulin is reported after exercise is very short in rodents (up to a few hours) (4) but rather long in humans (days) (15). If the enhanced sensitivity for glucose uptake is mainly related to glycogen metabolism, such difference in time window may relate to the difference in postexercise glycogen resynthesis and/or glycogen storage capacity, which is vastly different between muscle from rodents and human (20).

The profound influence of muscle insulin sensitivity on whole-body glucose metabolism and glycemic control is central in physical performance, metabolic health, and daily life for patients with diabetes. For example, exercise endurance is tightly linked to muscle glycogen content (21,22), and the insulin-sensitizing effect following a single bout of exercise secures glucose partitioning toward glycogen in the prior contracted muscles, allowing glycogen levels to become supercompensated if glucose is available (3,11). These adaptations ensure faster recovery and increase endurance during subsequent exercise. The exercise–insulin interaction also constitutes a challenge for subjects treated with insulin due to the significant risk of hypoglycemia during exercise as well as in the postexercise period (23). In fact, fear of hypoglycemia and loss of glycemic control remain a barrier for performing physical activity in insulin-treated subjects (24).

Over the past 25 years, we have performed 13 studies including more than 100 individuals using almost identical study protocols and experimental conditions. We anticipated that a meta-analysis of these studies would allow us to clarify associations of proposed regulators of insulin action both at rest and in the period following exercise. In fact, our analyses allow us to challenge the idea that glycogen breakdown is the primary determinant for the potentiation of muscle glucose disposal after exercise. The analysis also reveals a surprisingly large interindividual variation in in vivo muscle glucose uptake in response to insulin and potentiation of the same with exercise.

Detailed information for the design of each of the 13 studies is shown in Table 1. All subjects (n = 106) were healthy young (26 ± 7 years [mean ± SD]), lean (BMI 23 ± 1.5 kg · m⁻²), recreationally active (oxygen volume [VO₂] peak 50 ± 6.3 mL O₂ · kg body mass⁻¹ · min⁻¹) males recruited from the Copenhagen area from 1996 to 2020. All the studies followed the same design as that illustrated in Fig. 2A. Knee extensor endurance exercise was performed by one leg, while the other leg served as a rested control (within-subject control). A recovery period was followed by a hyperinsulinemic-euglycemic clamp (HIEC) (Table 1). Femoral veins of both legs and one femoral artery were catheterized, allowing measures of substrate arterio-venous (a-v) differences. Blood flow was measured in both legs. Biopsy samples were obtained from the vastus lateralis muscle of both legs just before and at the end of the HIEC in all studies except one (6). Muscle glycogen concentration was measured in all studies (n = 96), and in nine studies, muscle glycogen synthase (GS) activity was determined (n = 68).

Glucose uptake was calculated by multiplying the a-v difference in blood glucose concentration by blood flow (Fick principle). Leg glucose uptake (LGU) (in µmol · min⁻¹) was expressed relative to lean leg mass (LLM), determined by DXA scanning. In four studies (25–28), LGU was originally expressed relative to thigh muscle mass. Glucose uptake measurements for these studies were converted relative to LLM under the assumption that thigh muscle mass represents 74% of total leg muscle mass and that total muscle mass represents 83% of LLM (30).

In the equation above, ex. leg is exercised leg, rest leg is rested leg, and ex. quad is exercised quadriceps muscle. The muscle mass for the contracted quadriceps muscle was calculated with the assumptions that 83% of lean thigh mass determined by DXA scanning represents muscle mass (30), and quadriceps muscle mass constitutes 40% of all thigh muscle. For the rested muscle, muscle mass responsible for insulin-stimulated glucose uptake was given relative to 83% of LLM (30) with the assumption that all muscles in the rested control leg behaved similarly and contribute equally.

The relative contribution of total muscle glucose uptake to whole-body glucose disposal was estimated during steady state. The following assumptions were made for this calculation: 1) hepatic glucose production was fully suppressed by insulin (30), and glucose infusion rate (GIR) reflects rate of glucose disposal; and 2) all muscle mass constitutes 43% of body mass (31), and all muscles, except the prior contracted quadriceps muscle, behave similarly to the muscle of the rested leg. Total muscle glucose uptake was calculated as the sum of the uptake in the prior contracted quadriceps muscle and the uptake in the remaining rested muscle mass.

As no biopsy specimens were obtained immediately following exercise, glycogen utilization was calculated as the difference between glycogen content in the biopsy specimens obtained in the rested and the prior exercised muscle following exercise recovery. This estimate is valid, as the rate of glycogen resynthesis in recovery from exercise in the rested and fasted state (low insulin and euglycemia) is minimal in human muscle (5,25).

The half-activation time was calculated from the velocity constant (k) obtained by fitting the group mean data using the least-squares method to the equation y = α[l − exp(−αk)] + β, where α and β are unknown constants as previously described (26).

All data are expressed as means ± SD, unless otherwise stated. All data were tested for normality by quantile-quantile plots and histograms. Repeated measurements over time were analyzed by two-way repeated-measures ANOVA. The Tukey test was used as a post hoc test when the ANOVA revealed significant main effects or interactions. Correlations were analyzed by Pearson correlation analysis. All statistical evaluations were processed in GraphPad Prism (version 8.4.3), and results were considered significant at P < 0.05. Variation between studies (denoted T) for each variable was estimated as the study variance of covariance parameters in mixed-effects models with “study” as a random factor in SPSS 28. Data were visualized in Figs. 2–5 by color-coding when possible, enabling discrimination of data between studies.

GS activity in muscle was measured as the %I-form (enzyme activity assayed in the presence of 0.02 mmol/L glucose-6-phosphate [G6P] given relative to enzyme activity in the presence of saturating conditions [8 mmol/L G6P]). In two studies, the %I-form was measured in the total absence of G6P, whereas in all other studies it was measured in the presence of 0.02 mmol/L G6P. This methodological difference increased variation for the %I-form data, and we have thus only included data from the studies using 0.02 mmol/L G6P (7 studies, n = 55).

The data sets generated and analyzed during the current study are available from the corresponding author upon reasonable request.

During the HIEC, which was initiated 3–5 h after exercise, arterial insulin concentration increased immediately after the bolus injection and reached, on average, a plateau of 100 ± 22 µU · mL⁻¹ within 15 min (Fig. 2B and C) (coefficient of variation [CV] 22%, 26% related to between-study variation and 74% to between-subject variation). Plasma glucose concentration was clamped at individual fasting levels (5.1 ± 0.7 mmol/L), and the GIR was increased over time during the first 60 min of the HIEC to maintain euglycemia before reaching a steady-state level (37 ± 12 µmol · kg body weight⁻¹ · min⁻¹) during the last 30–40 min of the HIEC (Fig. 2B). A large interindividual variation in GIR of fourfold during steady state was observed (16–69 µmol · kg body weight⁻¹ · min⁻¹; CV 32%, 13% related to between-study variation and 87% to between-subject variation) (Fig. 2D).

Prior to the initiation of the HIEC, glucose uptake across the prior exercised leg had returned to resting levels; thus, glucose uptake was similar to that measured in the rested leg (Fig. 3A). Glucose uptake across the rested leg increased during the HIEC, reaching a plateau at 37 ± 19 µmol · kg LLM⁻¹ · min⁻¹. A 12-fold variation in individual responses was observed (6–72 µmol · kg LLM⁻¹ · min⁻¹; CV 47%, 20% related to between-study variation and 80% to between-subject variation) (Fig. 3B). When accounting for plasma insulin concentration a similar variation was observed (Fig. 3B). Time to reach half-maximal insulin-stimulated glucose uptake was ∼38 min for the rested leg. A strong relationship between GIR and LGU was evident (r² = 0.33, P < 0.0001) (Fig. 3C), which was further strengthened when adjusting for plasma insulin concentration (r² = 0.49, P < 0.0001) (Fig. 3D).

From the data shown in Fig. 3A, it becomes clear that prior exercise potentiated insulin action on LGU by nearly 50%, reaching a level of 58 ± 23 µmol · kg LLM⁻¹ · min⁻¹ across the exercised leg at steady state. The insulin-stimulated glucose uptake across the prior exercised leg also showed a large interindividual variation (sevenfold, ranging from 18 to 127 µmol · kg LLM⁻¹ · min⁻¹; CV 41%, 33% related to between-study variation and 67% to between-subject variation), and this remained even when glucose uptake was normalized to the plasma insulin concentration (Fig. 3B). However, it is even more remarkable that a similarly high variation between subjects was evident in the insulin-sensitizing effect of exercise. Thus, in some individuals the potentiation was as low as 6% above values in the rested leg, whereas in others it was as high as 80% (CV 56%, 15% related to between-study variation and 85% related to between-subject variation) (Fig. 3B). Another important observation is that prior exercise also altered the kinetic profile of LGU in response to insulin (Fig. 3A). Time to reach half-maximal glucose uptake was reduced by 50%, to ∼19 min for the prior exercised leg.

In Fig. 3A, glucose uptake across the exercised leg is presented relative to LLM of the entire leg, which also includes glucose uptake in noncontracted muscle of that leg. Given that the quadriceps muscle is the only muscle activated during knee extension exercise (29), we assumed that this muscle accounts for the difference in insulin-stimulated glucose uptake between the exercised and rested leg. Based on this assumption, the glucose uptake in the prior contracted muscle (quadriceps muscle) was substantially higher (approximately threefold), reaching 113 ± 58 µmol · kg QMM⁻¹ · min⁻¹ compared with the muscle of the rested leg (45 ± 17 µmol · kg LMM⁻¹ · min⁻¹) during steady state of the HIEC (Fig. 3E). When glucose uptake per kilogram of QMM was analyzed for the kinetic profile, time to reach half-maximal glucose uptake was ∼5 min for the prior contracted quadriceps muscle and ∼38 min for the rested muscle (Fig. 3E).

Leg blood flow was ∼10% higher in the exercised leg than in the rested leg at all time points during the HIEC (Fig. 3F). Leg blood flow also increased ∼15% during the HIEC, which is a rather small increase compared with the ∼12-fold increase in LGU in response to insulin. Thus, differences in bulk glucose delivery can only account for a small part of the difference in glucose uptake between the two legs. In line with this finding, it was the ability of the legs to extract glucose (reflected in the a-v difference) that changed, and it was the difference in glucose extraction that mostly accounted for the differences in glucose uptake between the two legs (Fig. 3G).

These observations support the concept that enhanced insulin action following exercise is predominantly caused by intramuscular events localized to the prior contracted muscle rather than exercise-induced systemic alterations.

Recently, we observed that the insulin-sensitizing effect of a single exercise bout was lowered following a period of exercise training (30). As exercise training per se increases muscle insulin sensitivity, we hypothesized that there are one or more mechanisms that limited the maximal obtainable sensitivity. Based on this observation, we hypothesized that individuals with the highest insulin-stimulated glucose uptake in the rested leg would show less of an increase by exercise, resulting in a negative association. However, this was not apparent from the present analysis (Fig. 3H).

GS is a key regulator of glycogen synthesis, and most of the glucose taken up during insulin stimulation is partitioned toward muscle glycogen synthesis (32) (Fig. 1). Both exercise and insulin stimulation induce GS activation (33), and several human and rodent studies have reported that the GS activity is elevated in the glycogen-depleted state (11,34,35). In accordance with this, our meta-analysis demonstrates that 3–5 h after exercise, GS activity was elevated by 100% in prior contracted compared with rested muscle (Fig. 4A). A variation was observed in muscle GS activity at this time point (CV 55% and 51% for rested and contracted muscle, respectively; 46% related to between-study variation and 54% to between-subject variation for the exercised leg, 63% related to between-study variation and 37% to between-subject variation for the rested leg) (Fig. 4A).

It has been suggested that elevated GS activity in the prior contracted muscle is dependent on muscle glycogen content and/or breakdown (36). In accordance with this, before HIEC, GS activity was negatively correlated with glycogen content in the prior contracted muscle (r² = 0.35, P < 0.0001) (Fig. 4B) but markedly less so with glycogen utilization during exercise (r² = 0.10, P = 0.02) (Fig. 4C). Thus, the absolute muscle glycogen concentration rather than the degree of glycogen utilization during exercise seems to be a key determinant of GS activity during recovery from exercise.

Insulin increased GS activity to a similar extent in the rested and contracted muscle such that the ∼100% higher GS activity remained present during insulin stimulation in the contracted muscle (Fig. 4A). This parallel increase in GS activity in the muscle of the two legs is in discordance with the exercise–insulin interaction seen on glucose uptake in the two legs. Correspondingly, neither GS activity in the contracted muscle before initiation (r² = 0.03 and P = 0.17) (Fig. 4D), or by the end of the HIEC (r² = 0.006 and P = 0.61), nor differences in GS activity between muscles of the two legs (r² = 0.002 and P = 0.93) correlated with the increase in insulin sensitivity of glucose uptake in the contracted muscle. Collectively, these data provide correlative evidence against GS activity as a controller of insulin sensitization following exercise.

Observations in rodents suggest that insulin-stimulated glucose uptake can be regulated by muscle glycogen content, particularly when glycogen content is manipulated to lower levels than those under normal resting conditions (37,38). Surprisingly, the meta-analysis revealed only a weak association between rested muscle glycogen concentration and insulin action on GIR (r² = 0.13 and P = 0.006) (Fig. 5A) as well as on rested LGU (r² = 0.07 and P = 0.01) (Fig. 5B). These correlations remained largely unchanged when GIR and muscle glucose uptake were normalized to individual insulin concentration (r² = 0.06 and P = 0.02; r² = 0.04 and P = 0.05, respectively). In addition, these associations were all positive and in stark contrast to the inverse relationship seen in rodents (39). Thus, the present meta-analysis suggests that muscle glycogen content in humans is not directly associated with insulin-stimulated muscle glucose uptake or whole-body insulin action. It could be argued that insulin action is only modified when muscle glycogen concentration drops below a certain threshold. However, this is not supported by our meta-analysis, as glycogen content in prior contracted muscle (as low as 55 mmol · kg dry weight muscle⁻¹) was not associated with glucose uptake (Fig. 5C). Thus, in neither the rested muscle nor contracted muscle could we find an association between glucose uptake and glycogen, even at low glycogen levels.

A muscle in recovery from exercise “recovers” within minutes with respect to the various metabolic perturbations occurring during exercise, e.g., energy status as reflected in adenosine nucleotide concentrations (40). However, fuel storage, e.g., glycogen content, is restored more slowly (in days) depending on, for instance, the diet. Our data revealed that muscle glycogen (re)synthesis rate increased significantly (38 mmol $⋅$ kg dry weight muscle⁻¹) in the contracted muscle (P < 0.001) during the HIEC, while this effect was not evident in the rested control leg (P = 0.672) (Fig. 5D). This difference is most likely linked to the enhanced insulin sensitivity to stimulate glucose uptake in the contracted muscle in combination with the elevated GS activity. During insulin stimulation, rested and contracted muscles differed with regard to glucose uptake (higher in the contracted muscle) and glycogen content (lower in the contracted muscle). Thus, it could be speculated that the magnitude of the difference in glucose uptake between the muscles of the two legs (the insulin-sensitizing effect) is related to either the absolute level of glycogen in the contracted muscle or to the amount of glycogen utilized during contraction. Such a regulatory role of muscle glycogen utilization has been suggested previously based on observations from a small (n = 14) cohort of individuals (41). However, our meta-analysis revealed that neither the absolute muscle glycogen content in contracted muscle (r² = 0.01, P = 0.32) nor muscle glycogen utilization during contraction (r² = 0.004, P = 0.83) was associated with insulin-stimulated glucose uptake of the previously contracted muscle (Fig. 5C and E). Furthermore, glycogen utilization during the exercise bout was not correlated to the insulin-sensitizing effect on glucose uptake (r² = 0.005, P = 0.78) (Fig. 5F). These observations remained when correlated to LGU expressed relative to LLM (r² = 0.0014 and P = 0.74 for exercised LGU versus contracted leg muscle glycogen concentration; r² = 0.0005 and P = 0.84 for delta LGU versus contracted leg muscle glycogen concentration). Thus, we conclude that neither glycogen content nor utilization during contraction determines the magnitude of the insulin-sensitizing effect of exercise in human muscle.

A single bout of exercise induces insulin sensitization of glucose uptake in the prior contracted muscle. This includes a more potent response and a temporal shift toward a shorter half-activation time. These changes are mainly accounted for by more efficient muscle glucose extraction. This points toward local myocellular factors that are activated during exercise and exert their action in response to subsequent insulin stimulation, including improved microvascular glucose delivery, muscle membrane glucose permeability, and glucose partitioning toward glycogen (Fig. 1). Our analyses suggest that neither glycogen content nor the degree of its utilization determines the magnitude of this effect in human muscle. This challenges previous viewpoints primarily based on observations in animal models. Interestingly, in rodent muscle, the AMPK–TBC1D4 signaling axis is necessary for contraction/exercise to induce insulin sensitization. Thus, against a necessary role of glycogen utilization and in support of our meta-analysis are observations in rodents that pharmacological activation of the AMPK–TBC1D4 signaling axis sensitizes muscle to insulin in the absence of lowered glycogen content (10,42).

The large interindividual variations in response to insulin and in the effect of prior exercise are remarkable, as these subjects constitute a homogeneous group of healthy young males. Previous studies report intraindividual day-to-day variation of ∼10–15% for whole-body glucose infusion/disposal under HIEC conditions (43). This aligns well with our own observations (CV ∼14%) for both whole-body GIR and LGU (n = 7 subjects studied 3 times) (unpublished observations). Thus, intraindividual variation explains only a fraction of the interindividual variation observed and reveals heterogeneity in the response to insulin at the muscular level. This expands previous observations of whole-body glucose disposal (44). Our statistical mixed-effects model approach allowed us to estimate the part of the total variation attributable to between-study variation. For the large majority of measures in this study, the between-study variation accounted for only 10–35% of the total variation. Our data therefore suggest that most of the variation can be ascribed to factors intrinsic to the individuals and align with the increasing evidence of individual responses to physiological perturbations. Based on observations from studies of twins, we estimated that heterogeneity in peripheral insulin action (whole-body and muscle parameters) can be ascribed 1:1 between genetic and environmental factors (45,46). We do not know whether the insulin sensitization effect of an exercise bout is under similar influences. Possible environmental factors could be long-term dietary and physical activity and long-term habits that may not be evident in inclusion criteria such as BMI and VO₂ peak.

Recently, we revealed that global analyses of posttranslational protein modifications (phosphorylation) similarly display a highly individual pattern in response to exercise and insulin. Intriguingly, our analyses uncovered that this variance was indeed linked to the variation in physiological phenotype in vivo and allowed us to illuminate new potential phosphorylation events and patterns regulating the action of insulin on muscle glucose uptake and the potentiation thereof by exercise (47). The new candidates included proteins like VAMP, mTORC1, and AMPK.

We estimate that glucose uptake into skeletal muscle accounts for ∼68% of whole-body glucose disposal during steady state. When estimates are performed on the increment caused by insulin, this number rises to ∼72%. These numbers are 15–20% lower than previous estimates by others (48). We believe the discrepancy primarily relates to the estimation of total leg mass, which is based on water replacement leg volume measures in the study from DeFronzo et al. (48) and by DXA measures in most of our studies. Nevertheless, our data suggest a larger role of other peripheral tissues like fat in insulin-induced glucose disposal in lean healthy individuals than previously thought (48).

Published (15,49) and in-house unpublished observations make us confident that whole-body exercise also elicits insulin sensitization of the contracted muscles. However, it is not always translated to improved insulin sensitivity at the whole-body level, as reported by Steenberg et al. (30) and in multiple references therein. This may relate to the observation that exercise under some conditions induces insulin resistance in nonexercised muscle and perhaps other organs as well (30).

It has been debated whether elevated GS activity is sufficient to pull glucose into muscle (50). Based on the elevated GS activity but similar glucose uptake immediately before the HIEC in the contracted compared with the rested muscle, our analyses reveal that GS activation under these conditions is not sufficient to drive an increase in glucose uptake. In support of this, there is no correlation between GS activity and muscle glucose uptake before the HIEC in the exercised leg or in the rested leg (r² = 0.05, P = 0.10 for exercised leg and r² = 0.005, P = 0.59 for rested leg).

The marked changes in glucose uptake kinetics in the prior contracted muscle are noteworthy. From previous studies, we have not been able to detect a faster insulin-induced activation of the insulin receptor or downstream signaling in the contracted compared with rested muscle (25,26). Thus, either further downstream events, like GLUT4 translocation/localization or glucose delivery, may be responsible for these kinetic changes.

Improved insulin sensitivity following a single bout of exercise has been observed in insulin-resistant individuals and patients with type 2 diabetes (51–53) as well as in experimentally induced insulin resistance following prolonged bedrest and intralipid infusion (13). Therefore, mechanistic insight to the insulin-sensitizing effect of exercise may also reveal novel pharmacological targets for improving insulin sensitivity in insulin-resistant patients. Such insight may become beneficial for patients unable to perform physical activity as even very short periods of physical inactivity markedly reduce muscular insulin sensitivity (54).

Our data suggest that insulin sensitization to stimulate muscle glucose uptake following exercise is not regulated in a homeostatic feedback manner involving GS activity, glycogen content, or glycogen utilization. Prior exercise is a prerequisite for glycogen supercompensation, and it has been postulated that the enhanced insulin sensitivity following exercise is necessary for this to occur. However, the apparent lack of negative feedback control from glycogen to insulin sensitivity does not support this hypothesis. Thus, we propose a new regulatory model whereby stability of the cellular milieu of the muscle fiber following exercise is achieved primarily by elevating the set point for glycogen storage, thus allowing for a higher storage capacity. First, this improves the ability of the exercised muscle to act as a metabolic sink for glucose (as glycogen) and will per se accelerate resynthesis up to and beyond normal glycogen levels by decreasing the inhibition by biochemical end product. Second, an increased glycogen content set point will limit the cellular glucose load, which otherwise might have enhanced the conversion of glucose into metabolites detrimental for maintaining a high insulin sensitivity. A term used to describe control/stability through adaptation of set points is allostasis (55–57). Thus, we propose that cellular allostatic regulation of glycogen storage allows for enhanced insulin sensitivity to promote optimal glycogen resynthesis. Our model also implies that the regulatory mechanisms for the enhanced insulin sensitivity and the elevated glycogen set point are separated, although both are elevated by prior exercise. Whereas recent research has provided some novel hints as to the regulators of muscle insulin sensitivity, almost nothing is known about the regulation of glycogen set point (11). Nevertheless, it can be speculated that the change in set point for glycogen storage level imparts an evolutionary advantage. Muscle glycogen content is a key determinant of exercise endurance (21,22), and augmented glycogen levels would effectively improve the chances of a successful hunt or escape in primeval settings. If true, this hypothesis underlines the need for future therapeutics aiming to restore glucose homeostasis by manipulating muscle glucose uptake via improving insulin sensitivity and enhancing the set point for glycogen.

A strength of our analyses is that all the studies included were performed in one laboratory involving the same scientific and technical staff. However, some technology advancements and minor modifications have been implemented over the years. Thus, 1) body composition (anthropometric vs. DXA) and blood flow (thermodilution vs. Doppler) have been estimated by different methods; 2) the studies were performed using the same general experimental design, yet some parameters varied somewhat between studies (Table 1); and 3) we did not standardize the diet in the period leading up to these studies. We acknowledge that such factors may introduce some of the variation observed. The study design, using within-individual leg/muscle comparisons, makes the influence of these factors less for some of the reported measures. All participants included in the studies were healthy, lean, male, White Europeans with a narrow range in age, BMI, and fitness level (VO₂ peak). No studies of female subjects using this design are available yet. However, some of our ongoing studies make us confident that this phenomenon is also present in female, elderly, and obese insulin-resistant subjects.

The experimental settings applied favor mechanistic insight, and data analyzed are collected only by using the one-legged knee extensor exercise model. Thus, translation to real-world whole-body exercise and oral intake of nutrients should be made with caution. While our data allow us to conclude that local factors are necessary for the enhancement of insulin sensitivity following exercise, the data do not allow us to exclude that a humoral factor is also required in this process in humans, as suggested in studies of rodents (58,59).

We acknowledge that insights into the duration of the insulin sensitization following exercise is understudied in humans, and we are open to the possibility that glycogen content or utilization is an important factor in determining the length of the time window in which muscle is sensitized to insulin after exercise.

Previous human studies have demonstrated that prior exercise increased the affinity (K_m) of GS for its substrate, UDP-glucose (34,60). Whether this regulation is further potentiated in the insulin-sensitized state following exercise remains to be investigated. Thus, it must be kept in mind that cellular GS activity in vivo may not be reflected in the activity measured in vitro.

Acknowledgments. The authors thank Betina Bolmgren, Irene B. Nielsen, Nicoline R. Andersen, and Jesper B. Birk, all from The August Krogh Section for Molecular Physiology, Department of Nutrition, Exercise and Sports, Faculty of Science, University of Copenhagen, Copenhagen, Denmark, for skilled technical assistance. The authors also thank all the subjects for their participation in the studies.

Funding. J.D.O., S.H., R.K., D.E.S., M.R.L., G.M., and J.J. have all been supported by a research grant from the Danish Diabetes Academy, funded by the Novo Nordisk Foundation, grant number NNF17SA0031406. K.K. received research grants from The Japan Society for the Promotion of Science, Grants-in-Aid for Scientific Research (no. 19K20007), and the EFSD/JDS Fellowship Program. This study was supported by grants given to J.F.P.W. by the Danish Council for Independent Research (FSS 6110-00498B).

Duality of Interest. B.K., E.A.R., and J.F.P.W. have ongoing collaborations with Pfizer and Novo Nordisk. J.R.H. (from 1 February 2021) and C.P. (from 1 January 2021) are currently working at Novo Nordisk a/s. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. J.R.H., C.F., and J.F.P.W. conceptualized the study. J.R.H., J.O., S.H., C.F., and J.F.P.W. contributed to data collection and analyses. J.R.H., J.D.O., S.H., and J.F.P.W. contributed to writing first draft of manuscript. All authors contributed to original data collection and execution of studies, data interpretation, and editing and approval of final manuscript. J.F.P.W. 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.