Interaction of Insulin and Prior Exercise in Control of Hepatic Metabolism of a Glucose Load

The subjects of this study were 24 mongrel dogs of either sex with a mean weight of 23 ± 1 kg. The animals were housed in a facility that met the American Association for the Accreditation of Laboratory Animals Care guidelines. All procedures were approved by the Vanderbilt University Animal Care and Use Committee. The animals were fed a standard diet (34% protein, 14.5% fat, 46% carbohydrate, and 5.5% fiber based on dry weight). At least 16 days before each experiment, a laparotomy was performed under general anesthesia. Three silastic catheters (0.03 mm ID) were inserted into the inferior vena cava for indocyanine green (ICG), somatostatin (SRIF), glucose tracer, and glucose infusions. Silastic catheters (0.04 mm ID) were inserted into the portal vein and left common hepatic vein for blood sampling. Silastic catheters (0.03 mm ID) were inserted into splenic and jejunal veins and advanced into the portal vein for the infusion of insulin, glucagon, l-[¹⁴C]glucose, and glucose. A silastic catheter (0.03 mm ID) was inserted into the left femoral artery for blood sampling. After insertion, the vascular catheters were filled with saline containing heparin and the free ends were knotted.

Transonic flow probes (Transonic Systems, Ithaca, NY) were used to measure portal vein and hepatic artery blood flows. A section of the portal vein upstream from the gastroduodenal vein was cleared of tissue and fitted with a 6.0-mm ID flow cuff. A section of the hepatic artery was fitted with a 3.0-mm ID flow cuff. The flow probe leads and knotted catheter ends were placed in a subcutaneous pocket made in the abdomen. The femoral artery catheter was placed in a pocket in the inguinal region. Animals meeting laboratory inclusion criteria (13) were fasted 18 h before the beginning of the study to ensure all animals were in the postabsorptive state.

The experimental protocol is shown in Fig. 1. On the day of the experiment, the catheters and flow probes were freed from the subcutaneous pockets using ∼2-cm incisions made after local administration of 2% lidocaine. Saline was infused into the arterial sampling catheter throughout the duration of the study. Animals were subjected to either a period of rest or moderate treadmill exercise (t = −190 to −40 min). Treadmill exercise was performed at 4 mph on a 12% grade and was followed by a 10-min transition period, during which time dogs were positioned in a Pavlov harness. At t = −120 min, peripheral infusions of ICG, [3-³H]glucose, and l-[¹⁴C]glucose were initiated. The ICG and [3-³H]glucose infusions continued for the remainder of the experiment. At t = 0 min, the peripheral l-[¹⁴C]glucose infusion was discontinued and a peripheral infusion of SRIF (0.8 μg · kg⁻¹ · min⁻¹) was initiated to suppress pancreatic insulin and glucagon secretion. Basal arterial plasma glucagon levels were maintained with a portal glucagon infusion (0.5 ng · kg⁻¹ · min⁻¹), and glucose was infused into the portal vein at a rate of 18 μmol · kg⁻¹ · min⁻¹. The portal glucose infusion contained trace quantities of l-[¹⁴C]glucose. The amount of l-[¹⁴C]glucose was chosen to match the rate of the peripheral l-[¹⁴C]glucose infusion used from t = −120 to 0 min. A variable peripheral glucose infusion was used to clamp arterial blood glucose concentrations at ∼8.3 mmol/l. Insulin was infused into the portal vein at a rate that simulated either basal insulin release (0.2 mU · kg⁻¹ · min⁻¹) or mild physiological hyperinsulinemia (1.2 mU · kg⁻¹ · min⁻¹). Arterial, portal vein, and hepatic vein blood samples were taken from t = −30 to 0 min for the assessment of baseline substrate concentrations and hepatic substrate balance. A 100-min interval between the baseline and experimental sampling periods was allotted to establish constant hormone levels and the hyperglycemic clamp. Once a steady state had been achieved, arterial, portal vein, and hepatic vein blood samples were taken every 10 min for the remainder of the experiment (t = 100–150 min). Portal vein and hepatic artery flows were measured at the same time that blood samples were collected. At t = 150 min, the dogs were killed, the liver samples were excised and frozen in liquid nitrogen, and the placement of catheters was confirmed.

Plasma glucose levels were determined on the day of the experiment by the glucose oxidase method using a Beckman Glucose Analyzer II (Beckman Instruments, Fullerton, CA). Plasma and tissue samples that were not immediately analyzed were stored at −70°C for later analysis. Whole blood samples were deproteinized with barium hydroxide and zinc sulfate to assess levels of radioactivity. After centrifugation (3,000g, 30 min), the supernatant was dried and reconstituted in 1 ml of water and 10 ml Ultima Gold scintillent (Packard, Meriden, CT). Radioactivity was determined using a Packard TRI-CARB 2900TR liquid scintillation counter. Plasma insulin and glucagon were measured using previously described radioimmunoassay techniques (1). Liver glycogen concentration and radioactivity were assessed as previously described (14). Liver glycogen phosphorylase and synthase activities were determined as previously described (15). For the determination of hepatic glucose-6-phosphate (G6P) and fructose-6-phosphate (F6P) levels, liver samples where homogenized in 8% perchloric acid and centrifuged (3,000g, 10 min). Blood glucose and hepatic concentrations of glycogen, G6P, and F6P were measured using enzymatic methods on a Technicon auto-analyzer (16).

Net hepatic balance (NHB) was calculated as NHB = [(H − A) × HAF] + [(H − P) × PVF], where H, A, and P are the hepatic vein, arterial, and portal vein blood substrate concentrations, respectively, and HAF and PVF represent the hepatic artery and portal vein blood flows, respectively. For the purposes of this study, net hepatic glucose output (NHGO) and uptake (NHGU) are presented as positive values. Hepatic glucose load (HGL) was calculated as HGL = (A × HAF) + (P × PVF). Net hepatic glucose fractional extraction (NHGFE) was calculated as the ratio of NHGU to HGL.

To assess the mixing of infused glucose in the portal vein, l-[¹⁴C]glucose was used in a manner similar to that previously described for p-aminohippuric acid (PAH) (17). As is the case with PAH, l-[¹⁴C]glucose is not extracted by the liver. Once a baseline level of radioactivity was established via the peripheral l-[¹⁴C]glucose infusion, l-[¹⁴C]glucose was infused with glucose into the portal vein. The appearance of l-[¹⁴C]glucose was calculated as (P −A) × PVF. The rate determined using this equation was compared with the actual portal infusion rate of l-[¹⁴C]glucose. Time points indicating gut outputs that were within 30% of the actual l-[¹⁴C]glucose infusion rate were considered mixed; furthermore, successful portal vein glucose mixing had to occur in at least half of the experimental sampling time points for a given experiment to be included herein.

Incorporation of glucose into hepatic glycogen stores was calculated as the [3-³H]glucose radioactivity in hepatic glycogen per mass tissue divided by the specific activity of inflowing glucose. The specific activity of inflowing glucose was calculated as [(SA_A × HAF) + (SA_P × PVF)]/(PVF + HAF), where SA_A and SA_P are the specific activities of glucose in the artery and portal vein. It should be noted that this calculation includes only the glucose incorporated into glycogen from the direct pathway (glucose → UDP-glucose → glycogen) and assumes that minimal [3-³H]glucose is stored as glycogen before t = 0 min (18).

All data are presented as means ± SE. ANOVA was performed to assess differences among groups and between baseline and experimental periods for all presented variables. Differences were considered significant at P < 0.05.

During the baseline sampling period (t = −30 to 0 min), arterial plasma insulin levels were significantly lower in both exercise groups compared with sedentary group (Fig. 2A). During the experimental period (t = 100–150 min), arterial plasma insulin concentrations were similar in both exercised and sedentary dogs that received a basal portal insulin infusion (4.5 ± 0.3 vs. 4.3 ± 0.3 μU/ml). Exercised and sedentary dogs that were infused with portal insulin to induce mild hyperinsulinemia had similar ∼5-fold increases in arterial plasma insulin levels compared with dogs receiving basal insulin infusion (22.5 ± 3.8 vs. 22.9 ± 3.2 μU/ml). Arterial plasma glucagon levels were not different among the groups during the baseline period, whereas arterial plasma cortisol levels were elevated by prior exercise (Table 1). Glucagon and cortisol were the same in all groups during the experimental period. Arterial plasma epinephrine and norepinephrine concentrations were comparable in all groups during baseline and experimental periods.

During the baseline period, arterial blood glucose and HGL were comparable in all groups (Fig. 2B and C). During the experimental period, arterial blood glucose concentrations were clamped at the same level (∼8.3 mmol/l) in all groups. HGL was also similar in exercised and sedentary dogs with basal (235 ± 17 vs. 249 ± 22 μmol · kg⁻¹ · min⁻¹) and elevated insulin (225 ± 18 vs. 240 ± 27 μmol · kg⁻¹ · min⁻¹).

The glucose infusion rate (GIR; variable GIR + portal glucose infusion) was not significantly increased by prior exercise with basal insulin (25.5 ± 1.5 vs. 20.3 ± 1.7 μmol · kg⁻¹ · min⁻¹) (Fig. 3). In contrast, there was marked enhancement in the GIR in the presence of elevated insulin (97.2 ± 15.1 vs. 64.4 ± 7.4 μmol · kg⁻¹ · min⁻¹). The GIR required to clamp glucose in the presence of hyperinsulinemia in exercised dogs increased to a significantly greater extent than in sedentary dogs that received elevated insulin.

NHGO tended to be higher in the baseline period in exercised compared with sedentary dogs, regardless of whether they were to have basal (11.2 ± 2.2 vs. 10.8 ± 4.3 μmol · kg⁻¹ · min⁻¹) or elevated insulin (12.6 ± 1.0 vs. 7.4 ± 1.4 μmol · kg⁻¹ · min⁻¹) in the experimental period, but this difference was not significant. All groups switched from net hepatic output to net hepatic uptake of glucose during the experimental period. Prior exercise increased NHGU (Fig. 4A) in the presence of both basal (7.5 ± 1.2 vs. 2.9 ± 2.4 μmol · kg⁻¹ · min⁻¹) and elevated insulin (22.0 ± 3.5 vs. 11.5 ± 1.8 μmol · kg⁻¹ · min⁻¹). The insulin-stimulated increase in NHGU (Fig. 4A, inset), calculated as the difference between NHGU with basal insulin and NHGU with elevated insulin, also tended to be more after exercise (14.5 ± 3.5 vs. 8.6 ± 1.8 μmol · kg⁻¹ · min⁻¹), but differences were narrowly insignificant (P = 0.06). NHGFE (Fig. 4B) was increased by prior exercise by both basal (0.04 ± 0.01 vs. 0.01 ± 0.01) and elevated insulin (0.10 ± 0.01 vs. 0.05 ± 0.01). Interestingly, NHGFE was significantly higher in exercised than in sedentary dogs that received basal insulin. The insulin-stimulated increase in NHGFE (Fig. 4B, inset) was twofold more after exercise compared with an equivalent duration of rest (0.06 ± 0.01 vs. 0.03 ± 0.01).

Arterial lactate during the baseline sampling period was similar in all groups (Table 2). The arterial lactate concentration was increased during the experimental compared with the baseline period in all groups. Arterial alanine levels (Table 2) were significantly reduced in exercised compared with sedentary dogs during the baseline period. During the experimental period, arterial alanine levels were significantly lower in sedentary and exercised dogs with hyperinsulinemia compared with dogs that received basal insulin infusion.

Arterial glycerol and nonesterified fatty acid (NEFA) concentrations were elevated by prior exercise in the baseline period (Table 2). During the hyperglycemic clamp, arterial, NEFA, and glycerol concentrations fell to comparable levels in sedentary and exercised dogs with basal insulin. In the presence of elevated insulin, arterial NEFA and glycerol levels in sedentary and exercised dogs were significantly lower than during the baseline period and than in dogs that had received basal insulin during the experimental period.

In a net sense, the liver consumed lactate during the baseline period in all groups except in the sedentary group receiving basal insulin; however, the difference among groups was not significant. During the experimental period, the liver was a net lactate producer during hyperglycemic clamps in all groups (Table 3). Net hepatic alanine uptake during the baseline period was similar to that during the experimental periods in all groups, and there were no significant differences among groups. Net hepatic uptake of glycerol fell from baseline during hyperglycemic clamps because of a decrease in circulating glycerol; there were no significant differences among groups. Net hepatic NEFA uptake was suppressed by hyperglycemic clamps in all groups except in the sedentary group receiving basal insulin during the hyperglycemic clamp. The suppression was greatest in those hyperglycemic clamps in which insulin was also high.

Hepatic glycogen content after the experimental period was similar in the two sedentary groups (Table 4), but was significantly higher than that seen in either exercised group. This indicated that, despite 150 min of increased hepatic glucose load, postexercise glycogen repletion was incomplete. Hepatic G6P and F6P concentrations, as well as glycogen synthase and phosphorylase activity ratios, were unaffected by prior exercise or mild hyperinsulinemia. Net hepatic glycogen synthesis was increased by mild hyperinsulinemia and prior exercise (Fig. 5). The effect of the two were additive, and there was no potentiation of insulin’s action on net hepatic glycogen synthesis (21.5 ± 4.5 vs. 20.9 ± 4.4 mg/g liver) (Fig. 5, inset).

Previous work by Galassetti et al. (6) showed that under conditions of simulated feeding (i.e., hyperglycemia and mild hyperinsulinemia), NHGU is enhanced by prior exercise when variables that control hepatic glucose balance (e.g., HGL, arterial glucose concentration, hormonal environment, and arterial-portal venous glucose gradient) are tightly regulated. This finding is consistent with work done in anesthetized rabbits that showed that after a period of hindlimb contraction, liver uptake of 3-fluoro-3-deoxyglucose was increased (4). In addition, recent work done in the conscious dog has shown that during euglycemia, insulin suppresses NHGO to a greater extent after exercise compared with after an equivalent sedentary period (13). In the present study, we showed that enhanced insulin action is a cause of the increased hepatic extraction of glucose during a glucose load after prolonged exercise.

In sedentary dogs, mild hyperinsulinemia increased NHGU and NHGFE. Exercised dogs that received basal insulin showed an increase in NHGU compared with sedentary dogs with the same level of insulin. In fact, the levels of NHGU and NHGFE in exercised dogs with basal insulin were as high as values seen in sedentary dogs with elevated insulin. Mild hyperinsulinemia in exercised dogs increased NHGU and NHGFE to values that were significantly higher than those seen under any other condition. This enhancement in liver glucose extraction elicited by prior exercise was, at least in part, attributable to an increase in the insulin-stimulated component of liver glucose extraction. Insulin-stimulated NHGFE was ∼2-fold greater in exercised dogs compared with in sedentary dogs (Fig. 4B). The arterial insulin levels present in this study were achieved with a portal insulin infusion. It is likely that had we infused insulin at the same rate peripherally, the stimulation of NHGU would have been reduced and peripheral glucose disposal would be elevated. Portal venous insulin infusion causes a marked increase in hepatic sinusoidal insulin concentration in comparison with peripheral insulin infusion, given that ∼80% of hepatic blood flow is derived from the portal vein. Satake et al. (19) showed that under simulated feeding conditions, it is the direct actions of portal vein insulin at the liver that control NHGU. This is an important distinction because insulin is delivered directly into the portal circulation in healthy subjects, many type 2 diabetic subjects, and in patients using some insulin-delivery strategies. However, in most type 1 diabetic patients, insulin is delivered peripherally.

Under the steady-state conditions of the present study, the peripheral glucose infusion rate required to maintain the glucose clamp, when added to the constant portal glucose infusion, was equal to whole-body glucose uptake (i.e., endogenous glucose production was zero). Hyperinsulinemia increased whole-body glucose uptake to a greater extent in exercised than in sedentary dogs (72 vs. 44 μmol · kg⁻¹ · min⁻¹), primarily because of an increase in the sensitivity of muscle to insulin stimulation (20). NHGU was ∼20% of whole-body glucose uptake in the presence of mild hyperinsulinemia in sedentary animals and remained at that level during mild hyperinsulinemia after exercise. Thus the hepatic sensitization to insulin was proportional to the whole-body enhancement in insulin action observed in the postexercise state. Although there was a slight increase in whole-body glucose uptake in exercised animals receiving basal insulin compared with sedentary animals, the difference was not significant. However, NHGU was significantly increased after exercise with basal insulin. The increase in NHGU in the absence of an increase in whole-body glucose uptake with basal insulin levels could have been attributable to either the greater insulin sensitivity of the liver compared with peripheral tissue (21) or an increase in insulin-independent hepatic glucose uptake. It may be that with basal insulin replacement, only the liver is affected in the postexercise state. When the insulin level is increased, the effect on muscle is readily apparent.

There are two notable points regarding the effects of prior exercise on net hepatic glycogen synthesis. First, this study showed that prior exercise increases hepatic glycogen synthesis even in the presence of basal insulin levels. In fact, prior exercise was nearly as potent in stimulating glycogen synthesis as was hyperinsulinemia. Second, the incremental change in net hepatic glycogen synthesis with elevated insulin was similar in exercised and sedentary dogs. These observations together suggest that enhanced insulin action is not the mechanism for increased channeling of glucose into glycogen after exercise. One could speculate that the concentration of hepatic glycogen participates in directing the fate of ingested glucose so that glycogen synthesis is, to an extent, autoregulatory (22). This notion is consistent with studies in long-term fasted dog (23) and human models (24) showing enhanced deposition of hepatic glycogen in the absence of differences in insulin concentrations. Dogs were fasted overnight in the present study; a shorter fast might have resulted in higher glycogen levels and possibly reduced the drive to synthesize glycogen even after exercise. In these studies, glucose was infused to provide substrate for hepatic glycogen synthesis. In the absence of exogenous glucose, net hepatic glycogen resynthesis does not occur and the liver is a net glucose producer (18). This contrasts with skeletal muscle, which can resynthesize glycogen after exercise even in the absence of exogenous substrate (18). Therefore, hepatic glycogen synthesis is critically dependent on circulating glucose.

The specific mechanism by which prior exercise enhances insulin stimulation of hepatic glucose extraction remains to be elucidated. During exercise, glucagon rises while insulin decreases (25). These hormonal changes result in a rapid depletion of hepatic glycogen stores. During prolonged moderate-intensity exercise, these hormonal changes persist for extended periods of time. Changes in pancreatic hormone levels and the resulting depletion of hepatic glycogen stores could potentially facilitate hepatic glucose uptake and glucose incorporation into glycogen during feeding after a sustained bout of glycogen-depleting exercise. Previous work in rats has shown that hepatic glycogen synthesis is increased after glucocorticoid treatment (26–28). It is possible that the glucocorticoid release that occurs during exercise (29) could potentiate an increase in hepatic glucose uptake and glycogen synthesis during feeding after the cessation of exercise. In addition, work in dogs has shown that the adenosine receptor agonism inhibits (30) and antagonism increases insulin action at the liver (31). Hypoxanthine breakdown products of adenosine act as competitive adenosine receptor antagonists, and their hepatic uptake is increased after exercise (32). Thus it is conceivable that hypoxanthines may improve insulin action after exercise. There is evidence that glycerol and free fatty acids play a role in the regulation of NHGU (33,34). Circulating glycerol and NEFAs did not appear to contribute to the differences in NHGU, as arterial glycerol and NEFA levels were similar in sedentary and exercised dogs with basal insulin and were suppressed to a similar extent with hyperinsulinemia. The enhanced hepatic uptake and glycogen storage occurred despite a lack of detectable changes to various hepatic substrate and enzyme variables. Hepatic concentrations of G6P and F6P as well as glycogen synthase/phosphorylase activity were similar in all groups. Nevertheless, we cannot rule out compartmentalized changes in substrate level or enzyme activities.

The results of the present study define the interaction of prior exercise and insulin in controlling the hepatic handling of an increased glucose load. They demonstrated that prior exercise 1) leads to increases in hepatic extraction of glucose and synthesis of glycogen, even in the absence of elevated insulin levels, 2) enhances the ability of mild hyperinsulinemia to stimulate hepatic glucose extraction by ∼2-fold, 3) does not potentiate insulin’s effect on net hepatic glycogen synthesis, and 4) does not alter the percent of whole-body glucose uptake that can be ascribed to hepatic uptake, suggesting that the enhanced effectiveness of insulin at the liver is roughly proportional to the enhanced effectiveness on previously working muscle. In conclusion, these findings provide support for advocating exercise as a treatment for insulin resistance, not only for its effect on muscle tissue but also as a means of reducing hepatic insulin resistance.

This work was funded by National Institute of Diabetes and Digestive and Kidney Diseases Grant RO1 DK-50277, Diabetes Center Grant DK-20593, and Training Grant 5-T32-DK-7563-08.

We would like to thank Deanna Bracy for her valued assistance with the completion of this work as well as the Vanderbilt University Diabetes Center Hormone Assay Core.