Insulin Sensitively Controls the Glucagon Response to Mild Hypoglycemia in the Dog

Experiments were performed on 20 overnight-fasted mongrel dogs (17.4–29.0 kg, mean 23.1 ± 1.1 kg) of either sex that had been fed a standard meat and chow diet once daily (34% protein, 46% carbohydrate, 14% fat, and 6% fiber, based on dry weight) (Kal Kan, Vernon, CA, and Purina Lab Canine Diet No. 5006; Purina Mills, St. Louis, MO). The dogs were housed in a facility that met American Association for the Accreditation of Laboratory Animal Care guidelines, and the protocols were approved by the Vanderbilt University Medical Center Animal Care Committee. At least 16 days before an experiment, a laparotomy was performed under general endotrachial anesthesia (15 mg/kg body wt pentothal sodium presurgery and 1.0% isoflurane as an inhalation anesthetic during surgery), and catheters for blood sampling were placed into a femoral artery, the portal, an hepatic, a jejunal, and a splenic vein as previously described (8–11,13,16). On the day of the experiment, the catheters were exteriorized under local anesthesia (2% lidocaine; Abbott, Chicago), their contents were aspirated, and they were flushed with saline. Angiocaths (20 gauge; Abbott) were inserted into the right and left cephalic veins for infusion of glucose.

On the day before the experiment, the leukocyte count and hematocrit were determined. Dogs were used for an experiment only if they had 1) a leukocyte count <18,000/mm³, 2) a hematocrit > 35%, 3) a good appetite, and 4) normal stools.

Each experiment consisted of a 100-min equilibration period (−140 to −40 min), a 40-min control period (−40 to 0 min), and a 180-min test period (0–180 min). Four experimental protocols were used. During the test period, a 0.5% methyl cellulose/saline solution (50 ml) with BAY R 3401 (10 mg/kg) was given by mouth at 0 min in each of the four groups. The plasma glucose level was then monitored every 5 min. In protocol 1, saline was infused, and the glucose level was allowed to drop. In protocol 2, insulin was infused peripherally at 3.6 pmol · kg⁻¹ · min⁻¹, and glucose was infused peripherally to clamp the arterial plasma glucose level to that in protocol 1. In protocol 3, glucose was infused peripherally to maintain euglycemia. In protocol 4, insulin was infused peripherally at 3.6 pmol · kg⁻¹ · min⁻¹, and glucose was infused so that the plasma glucose level was maintained at a euglycemic value.

Data from protocols 1 and 3 are contained as part of the database in another article that addresses a separate topic (13). The representation of the data facilitates demonstration of the effect of hyperinsulinemia on glucagon secretion in response to mild hypoglycemia, which is the focus of the present article. The period when protocols 1 and 3 were performed overlapped with the period when protocols 2 and 4 were carried out.

Plasma glucose concentrations were determined using the glucose oxidase method with a Beckman glucose analyzer (Fullerton, CA) (20). Blood concentrations of glucose, lactate, glycerol, ketones, and alanine were determined according to the method of Lloyd et al. (21), as adapted to the Monarch 2000 centrifugal analyzer (Lexington, MA) in samples deproteinized with perchloric acid (3%). The plasma free fatty acid (FFA) concentrations were determined using the Wako nonesterified fatty acid C test (Wako, Osaka, Japan). Plasma insulin, glucagon, cortisol, epinephrine, and norepinephrine levels were determined as previously described (14), with interassay coefficients of variation of 8, 9, 8, 10, and 5%, respectively.

Hepatic arterial and portal vein blood flows were assessed using Transonic flow probes. Net hepatic substrate balance was calculated using the formula H × (F_a + F_p) − [(A × F_a) + (P × F_p)], where A, P, and H are the arterial, portal vein, and hepatic vein substrate concentrations, respectively, and F_a and F_p are the hepatic arterial and hepatic portal blood or plasma flows, respectively. Gluconeogenic flux is equal to the sum of the net hepatic uptake rates of lactate, glycerol, and the gluconeogenic amino acids (alanine, glycine, serine, threonine, glutamine, and glutamate) converted to glucose equivalents (22). The net glycogenolytic rate was obtained by subtracting the gluconeogenic flux rate from the sum of net hepatic glucose output (NHGO), net hepatic lactate output, and glucose oxidation. The glucose oxidation rate was assumed to be 1.7 μmol · kg⁻¹ · min⁻¹ based on previous data (23). Glucagon release was calculated using the formula F_p × [P_g − (0.85 × A_g)], where A_g and P_g are the arterial and hepatic portal plasma glucagon concentrations, respectively, and F_p is hepatic portal plasma flow. A_g is multiplied by 0.85 because the gut fractional extraction of glucagon is 15% (24). Insulin release was calculated using the formula F_p × [P_I − (0.70 × A_I)], where A_I and P_I are the arterial and hepatic portal plasma insulin concentrations, respectively, and F_p is hepatic portal plasma flow. A_I is multiplied by 0.7 because the gut fractional extraction of insulin is 30% (25).

Data are expressed as means ± SE. A one-way ANOVA for repeated measures was used to analyze changes over time. A two-way ANOVA for repeated measures was used to analyze the difference between time course. When significant changes were obtained over time, post hoc comparisons were made using a paired t test (26).

The hepatic arterial blood flow (5.4 ± 0.6, 6.1 ± 0.8, 4.7 ± 0.3, and 4.2 ± 0.5 ml · kg⁻¹ · min⁻¹ in the hypoinsulinemic-hypoglycemic, hyperinsulinemic-hypoglycemic, euinsulinemic-euglycemic, and hyperinsulinemic-euglycemic groups, respectively) and portal blood flow (20.6 ± 2.2, 25.4 ± 2.4, 19.2 ± 1.2, and 20.4 ± 2.7 ml · kg⁻¹ · min⁻¹ in the four groups, respectively) did not change over time or with treatment (data not shown).

BAY R 3401 administration decreased NHGO (μmol · kg⁻¹ · min⁻¹) in the hypoinsulinemic-hypoglycemic group from 8.9 ± 1.7 to 5.0 ± 1.1 by 60 min (P < 0.05) (Fig. 1), after which it rose slightly. As a result, the arterial plasma glucose level (mmol/l) fell from 5.8 ± 0.2 to a minimum of 4.4 ± 0.3 (P < 0.01). In the hyperinsulinemic-hypoglycemic group, NHGO decreased from 10.9 ± 2.2 to zero within 2 h (P < 0.01). The plasma glucose level in the hyperinsulinemic-hypoglycemic group was clamped to the level observed in the hypoglycemic group, falling from 6.1 ± 0.1 (basal) to 4.6 ± 0.1 mmol/l. When the plasma glucose level was maintained at its basal value (5.9 ± 0.1 mmol/l) by exogenous glucose infusion in the presence of drug and basal insulin, the liver took up glucose at −3.3 ± 0.6 μmol · kg⁻¹ · min⁻¹ by the end of the experiment. When the plasma glucose level was maintained at a euglycemic value (6.1 ± 0.1 mmol/l) by glucose infusion in the presence of hyperinsulinemia and drugs, the liver took up glucose at −4.7 ± 1.7 μmol · kg⁻¹ · min⁻¹ by 60 min (P < 0.01), after which it remained unchanged. Thus, the presence of mild hyperinsulinemia blocked the hypoglycemia-induced stimulation of NHGO and hastened the switch to net hepatic glucose uptake in the presence of euglycemia.

As shown in Fig. 2, and as was reported previously (13), arterial and hepatic portal plasma insulin levels (pmol/l) in the hypoinsulinemic-hypoglycemic group decreased from 78 ± 18 and 168 ± 36 to 24 ± 6 and 36 ± 12 during the first 30 min (P < 0.01) and to 18 ± 6 and 24 ± 6 by 90 min (P < 0.01), respectively. The hepatic portal-arterial difference in plasma insulin (pmol/l) decreased from 90 ± 24 to 12 ± 6 by 30 min (P < 0.05) and to 6 ± 3 by 90 min (P < 0.05). The insulin secretion rates (pmol · kg⁻¹ · min⁻¹) decreased from 1.5 ± 0.4 to 0.3 ± 0.1 by 30 min (P < 0.05) and 0.2 ± 0.1 by 90 min (P < 0.05). The arterial and hepatic portal glucagon levels (ng/l) rose from 43 ± 5 and 48 ± 6 to 51 ± 5 and 61 ± 7 by 30 min (P < 0.05) and to 79 ± 16 and 109 ± 30 by 90 min (P < 0.05), after which they drifted down slightly (Fig. 3). The hepatic portal-arterial difference in plasma glucagon increased from 5 ± 2 (basal) to 10 ± 5 by 30 min (P < 0.05) and to 30 ± 15 by 90 min (P < 0.05), after which it remained at 20 ± 13. The glucagon secretion rate (pg · kg⁻¹ · min⁻¹) increased from 112 ± 57 (basal) to 230 ± 113 by 30 min (P < 0.05) and to 556 ± 273 by 90 min (P < 0.05), after which it decreased to 466 ± 232 by 180 min.

During hyperinsulinemic hypoglycemia, as shown in Fig. 2, the arterial plasma insulin level (pmol/l) increased from 48 ± 6 to 156 ± 6 during the first 30 min (P < 0.01), after which it remained constant (156 ± 24). The portal plasma insulin levels (pmol/l) remained at a basal value (138 ± 36 pmol/l). Therefore, the hepatic portal-arterial difference in plasma insulin (pmol/l) decreased from 90 ± 36 to −30 ± 18 by 30 min (P < 0.01) and remained nonsignificantly different from zero (−24 ± 18) thereafter, as did insulin secretion. The arterial and hepatic portal vein glucagon levels (ng/l) did not rise in response to the decline in plasma glucose (40 ± 5 and 46 ± 7 to 45 ± 3 and 54 ± 4 at 30 min and 44 ± 8 and 51 ± 9 at 180 min, respectively) (Fig. 3). The hepatic portal-arterial difference in plasma glucagon (ng/l) and glucagon secretion (pg · kg⁻¹ · min⁻¹) also did not change significantly (6 ± 4 and 216 ± 126 [basal] to 9 ± 2 and 283 ± 71 by 30 min and 7 ± 3 and 252 ± 116 at 180 min, respectively).

When euglycemia was maintained in the presence of the drug, the arterial plasma insulin and glucagon levels did not change significantly; likewise, the portal hormone levels, their hepatic portal-arterial differences, and their secretion rates also remained unchanged (Figs. 2 and 3).

When insulin was raised in the presence of euglycemia, the arterial plasma insulin levels (pmol/l) rose from 54 ± 6 to 126 ± 6 by 30 min (P < 0.01) and averaged 138 ± 6 during the rest of the test period. However, hepatic portal plasma insulin levels did not rise from basal (150 ± 42 pmol/l). The hepatic portal-arterial difference in plasma insulin (pmol/l) decreased rapidly from 102 ± 36 to 18 ± 42 by 10 min (P < 0.05) and thereafter averaged −6 ± 24 (Fig. 2). Thus, insulin secretion was completely inhibited. The arterial and hepatic portal vein glucagon levels and the hepatic portal-arterial difference in plasma glucagon (ng/l) did not change (43 ± 2, 52 ± 2, and 9 ± 1 to 37 ± 3, 44 ± 3, and 7 ± 2 at 180 min during the test period, respectively). The glucagon secretion rate (pg · kg⁻¹ · min⁻¹) also remained unchanged (basal of 150 ± 17 to 150 ± 52 at 180 min) during euglycemic hyperinsulinemia. These data indicate that the rise in arterial insulin blocked the hypoglycemia induced α-cell response and did not suppress basal glucagon secretion.

As shown in Table 1, the arterial norepinephrine, epinephrine, and cortisol levels did not change significantly in the absence of hyperinsulinemia regardless of whether euglycemia or hypoglycemia were present. In the presence of hyperinsulinemia, on the other hand, regardless of the blood glucose levels, arterial plasma norepinephrine, epinephrine, and cortisol levels tended to increase, but only cortisol changed significantly.

As shown in Table 2, when hypoinsulinemic hypoglycemia occurred, the arterial levels and net hepatic uptakes of nonesterified fatty acids (NEFAs), 3-hydroxybutyrate, and glycerol significantly increased. When hypoglycemia occurred in the presence of hyperinsulinemia and when euglycemia was maintained in the absence of elevated insulin, the arterial levels and net hepatic uptake of these metabolites did not change significantly. On the other hand, when plasma glucose levels were maintained in the presence of hyperinsulinemia and drugs, the arterial levels and net hepatic uptake of these metabolites fell significantly. In both cases, therefore, hyperinsulinemia reduced lipolysis and ketogenesis.

In the presence of hypoinsulinemic hypoglycemia, hyperinsulinemic hypoglycemia, euinsulinemic euglycemia, and hyperinsulinemic euglycemia, arterial lactate levels fell and net hepatic lactate balance shifted in favor of increased uptake. The arterial blood alanine levels decreased in all four groups. Net hepatic alanine uptake increased during hypoinsulinemic hypoglycemia, but did not change during hyperinsulinemic hypoglycemia, euinsulinemic euglycemia, or hyperinsulinemic euglycemia.

BAY R 3401 caused net hepatic glycogenolysis to decrease from 7.39 ± 0.78 to 1.83 ± 1.83 μmol glucose equivalent · kg⁻¹ · min⁻¹ by 30 min in the hypoglycemic group, after which it was effectively zero (Table 3). As a result, both the plasma glucose level and the plasma insulin level fell. On the other hand, gluconeogenic flux started to rise at 30 min and, by the end of the experiment, had reached 8.7 ± 1.2 μmol · kg^{− 1} · min⁻¹. From 60 to 180 min, the gluconeogenic flux was almost equivalent to NHGO, indicating that gluconeogenesis (the conversion of gluconeogenically derived glucose-6-phosphate to plasma glucose) was responsible for almost all glucose production. When hyperinsulinemia was added to BAY R 3401 delivery, net glycogenolysis was suppressed more rapidly and to a greater extent, eventually switching to net glycogen synthesis (−6.9 ± 0.9 μmol glucose equivalent · kg⁻¹ · min⁻¹ at 180 min). The glucose level was clamped to the hypoglycemia seen in the former group. On the other hand, gluconeogenic flux rose from 4.9 ± 1.4 to 8.8 ± 1.9 μmol glucose equivalent · kg⁻¹ · min⁻¹ by the end of the study. Thus, although the presence of insulin caused more net glycogen synthesis, it failed to decrease gluconeogenic flux. Since NHGO was zero, however, gluconeogenesis (the conversion of gluconeogenic precursors into plasma glucose) was inhibited, and the gluconeogenic-derived carbon in the glucose-6-phosphate pool was directed to glycogen. In the presence of euinsulinemic euglycemia, BAY R 3401 caused net glycogenolysis to decline from 8.2 ± 1.5 to −7.3 ± 1.4 μmol glucose equivalent · kg⁻¹ · min⁻¹. When hyperinsulinemia accompanied euglycemia and drugs, net glycogen flux switched rapidly from breakdown to synthesis, reaching 6.5 ± 1.7 μmol glucose equivalent · kg⁻¹ · min⁻¹ by the end of the study. Gluconeogenic flux did not change significantly in the presence of euglycemia, regardless of the presence or absence of insulin infusion. Once again, however, the gluconeogenic rate was zero in both groups, and the gluconeogenically derived carbon was stored in glycogen.

Glucagon secretion was increased threefold in response to the non- insulin-induced fall in plasma glucose from 5.8 to 5.2 mmol/l. As such, the plasma glucose threshold was well above the well-recognized concentration of 3.8 mmol/l observed in the presence of insulin-induced hypoglycemia in humans (3–7) and dogs (8–11). In a previous study (10), we showed that decreasing plasma glucose from 5.9 to 4.3 mmol/l in the presence of hyperinsulinemia (480 pmol/l) failed to trigger glucagon secretion. Thus, in the presence of hyperinsulinemia, a fall of 1.6 mmol/l in plasma glucose level failed to cause the plasma glucagon levels to rise, whereas, as we have also previously shown, in the presence of hypoinsulinemia, a fall of 1.4 mmol/l (from 5.8 to 4.4 mmol/l) caused a sixfold rise in glucagon secretion. These data therefore led to the hypothesis that was tested in the present study, namely, that glucagon secretion is sensitively inhibited by insulin such that the α-cell response to hypoglycemia is markedly reduced by the presence of insulin. Because it is known that glucagon secretion increases markedly in response to deep insulin-induced hypoglycemia, our data suggest that a deeper hypoglycemia results in an increase in neural input to the α-cell or some other change that then overcomes insulin’s inhibitory action. Our data also suggest that if a deep hypoglycemia could be produced in the absence of elevated insulin, the rise in plasma glucagon would be much larger than that observed in the presence of equivalent insulin-induced hypoglycemia.

It has been shown that insulin has direct effects on the α-cell to suppress glucagon secretion (18,27–29). Anatomical and physiological studies suggest that blood in the islet flows from β-cells to α-cells in dogs (30) as well as in rats (31,32). Therefore, α-cells are exposed to high concentrations of intra-islet insulin in the presence of euglycemia. In the presence of mild hypoglycemia (Fig. 2), on the other hand, insulin secretion falls markedly, thereby lowering the insulin concentration to which α-cells are exposed. It has been shown that the administration of anti-insulin serum to perfused rat pancreas markedly increased glucagon secretion and abolished the ability of an increase in the glucose concentration to decrease glucagon secretion (27,33,34). Acute insulin deficiency in humans (35,36), as well as alloxan- or streptozotocin-induced diabetes in animals, are associated with hyperglucagonemia (37). It is therefore likely that the decrease in intra-islet insulin contributed in some way to the increased glucagon secretion that we observed in response to mild hypoglycemia.

In the present study, modest arterial hyperinsulinemia almost completely suppressed the increment of glucagon secretion that occurred in response to hypoinsulinemic hypoglycemia (Fig. 3) even though the hypoglycemia was identical in the two groups. It is also clear that this was due to an inhibition of the α-cell response to hypoglycemia because when insulin was infused in the presence of euglycemia, it failed to decrease basal glucagons secretion. Because ∼ 30% of insulin in the arterial blood is cleared by the intestine when blood passes through it (25), it is clear that endogenous insulin secretion was completely suppressed in this protocol. Accordingly, the α-cells were probably exposed to a concentration of insulin that was equivalent to the arterial insulin level. Therefore, the rise in plasma insulin levels caused by exogenous insulin infusion would have induced only a small increase in the intra-islet insulin concentration relative to that present in the hypoglycemic-hyperinsulinemic setting. Nevertheless, it appears to have been enough to inhibit glucagon secretion in response to mild hypoglycemia. Alternatively, the arterial or portal insulin levels themselves may be the key determinants of glucagon secretion. Thus, when each or both fall, a signal (neural or humoral) is generated that then controls the α-cells.

In the presence of euglycemia in the present study, mild hyperinsulinemia (∼150 pmol/l) did not alter basal glucagon secretion. In contrast, we (10,16) and others (8) have shown previously that larger increases in the arterial plasma insulin level (to ∼400–1,440 pmol/l) can decrease plasma glucagon levels by 30–40% in the presence of euglycemia in conscious dogs. It is possible that the rise in the plasma insulin concentration that we brought about in the present study was not large enough to cause a detectable inhibition of basal glucagon secretion under euglycemic conditions. Further, it is unclear whether the inhibition of glucagon secretion by hyperinsulinemia that we observed in the presence of euglycemia in our previous studies was due to a direct action of insulin on the α-cell (8,10,16). Several in vitro studies have shown that high concentrations of insulin (0.01–3.3 units/ml) do not suppress glucagon secretion in the presence of 5.5 mmol/l glucose in perifused islets isolated from normal rats (18, 34,38).

Electrical stimulation of the pancreatic sympathetic nerves (39), local infusion of the classic sympathetic neurotransmitter norepinephrine (40), and a pancreatic sympathetic neuropeptide such as galanin (41) all increase glucagon secretion. Similarly, electrical stimulation of parasympathetic nerves (42,43), local infusion of the classic parasympathetic neurotransmitter acetylcholine (44), and a pancreatic parasympathetic neuropeptide such as vasoactive intestinal peptide (45,46) all stimulate glucagon release from the α-cell. Each of these autonomic inputs is capable of stimulating glucagon secretion. We have no evidence, however, that increased central nervous system (CNS) input can explain the rise in glucagon secretion seen during hypoinsulinemic hypoglycemia in the present study. In fact, we did not observe any change in the level of circulatory norepinephrine. Further, a fall of only 1.4 mmol/l in plasma glucose level would not have activated the autonomic nervous system. In addition, in another recent study, we showed that non-insulin-induced hypoglycemia caused glucagon secretion even in dogs with denervated pancreata (47). Therefore, it is unlikely that the rise in arterial insulin level would have blocked α-cell function by inhibiting CNS input, because if anything, it would have had the opposite effect.

It is known that amino acids have stimulatory effects and that FFAs and ketones have inhibitory effects on glucagon secretion from the α-cell (34,48,49). As shown in Table 2, the blood alanine concentration decreased in the hypoinsulinemic-hypoglycemic group, and the increase in glucagon secretion could not be explained. Likewise, the NEFA and ketone levels rose again, precluding them from explaining the rise in the glucagon secretion seen in response to non-insulin-induced hypoglycemia. Furthermore, raising insulin failed to change the plasma alanine level and decreased FFA and ketones levels, meaning that alterations in the levels of these metabolites cannot explain the insulin-induced inhibition of glucagon secretion.

In summary, the increase in glucagon secretion seen in response to a small decline of blood glucose occurs only when the arterial, portal vein, and intra-islet concentration of insulin are very low. Even maintenance of an arterial concentration (and presumably intra-islet insulin concentrations) of ∼150 pmol/l suppresses the increase in glucagon secretion seen in response to a small fall in blood glucose. An effect of insulin on the CNS or blood metabolite levels cannot explain its effect on glucagon release. It is therefore likely to have resulted from a direct effect of insulin on the α-cell or local neural reflexes.

When the arterial plasma glucose level fell by 0.3 mmol/l as a result of administration of glycogen phosphorylase inhibition, insulin secretion decreased by 80%. This result confirms that insulin secretion is very sensitively suppressed in response to a declining plasma glucose level, in agreement with earlier data (2,3, 6). When euglycemia was maintained, plasma insulin levels and insulin secretion did not change in response to the drug, indicating that it had no direct effect on the β -cell. When insulin was infused through a leg vein to create mild hyperinsulinemia in the presence of euglycemia, the resulting rise in arterial plasma insulin was not accompanied by a rise in the hepatic portal vein plasma insulin level. The hepatic portal vein-arterial difference in plasma insulin (pmol/l) decreased from 102 ± 36 to 18 ± 42 by 10 min and thereafter averaged −6 ± 24 (Fig. 2). Thus, endogenous insulin secretion was decreased markedly and rapidly by a selective rise in arterial insulin per se. In the dog, the insulin concentration in the pancreatic vein is four to five times higher than that in the portal vein (24). Because the mass of the islets represents only a small part of the total mass of the pancreatic tissue, the insulin concentration in the interstitial fluid surrounding the islets and in the venules draining the islets should be higher than that in the pancreatic vein. When endogenous insulin secretion is suppressed completely, the insulin concentration in the interstitial fluid surrounding the islets would be expected to be similar to the concentration of the hormone in arterial plasma. If so, the insulin concentration in islet interstitial fluid during the hyperinsulinemic-euglycemic period was probably lower than that during the control period of the same protocol. This finding raises the question of how insulin inhibited its own secretion.

It has been suggested that insulin can act directly on the brain. Davis et al. (9) showed that increased insulin levels can amplify the sympathetic nervous system response to hypoglycemia in the dog, normal human (5,50), and insulin-dependent diabetic human (51,52). Furthermore, studies in which microneurography was used have shown that insulin can increase sympathetic nerve activity, even under euglycemic conditions (53,54). Epinephrine and norepinephrine are known to inhibit insulin secretion through an α₂-adrenergic mechanism (55,56). Epinephrine is unlikely to have contributed to the suppression of insulin secretion in the present study, however, because its concentration in plasma did not change. On the other hand, it is known that activation of the sympathetic nerves by the islets of Langerhans induces the release of norepinephrine (57) and galanin (58), both of which inhibit insulin release (59). Stagner et al. (60) examined the effects of systemic infusion of insulin (1 mU · kg⁻¹ · min⁻¹) on endogenous insulin and C-peptide secretion from in situ perfused pancreata isolated vascularly but yet innervated in anesthetized dogs. They showed that the systemic infusion of insulin suppressed insulin and C-peptide secretion but that no changes in the pancreatic hormone secretion occurred after the sympathetic nerves were sectioned. It is not clear in the present study, however, whether the small rise in arterial insulin could have amplified the sympathetic discharge to the β-cells so as to decrease insulin secretion. If so, it was not reflected by a change in epinephrine and norepinephrine in plasma.

Amino acids, FFAs, and ketones have been shown to stimulate insulin secretion (48). Because significant decreases in their concentrations occurred in the presence of euglycemic hyperinsulinemia, it is possible that changes in the plasma concentrations of these substances contributed in some way to the decrease in insulin secretion that was observed in response to euglycemic hyperinsulinemia.

In summary, a physiological increase in the arterial insulin level strongly suppressed endogenous insulin secretion even in the presence of euglycemia. The effect of increased insulin levels on insulin secretion may occur via their action on the brain, secondary to their effects on metabolite levels in the blood, or because of a direct action of arterial insulin on β-cells.

In our previous study (14), inhibition of phosphorylase by BAY R 3401, in the presence of a pancreatic clamp to keep plasma insulin and glucagon at the basal levels and glucose infusion to maintain euglycemia, had little effect on the net hepatic uptake of gluconeogenic precursors. The latter provides a good estimate of gluconeogenic flux rate (22), thus suggesting that inhibition of glycogen breakdown per se did not activate the gluconeogenic pathway. In the present study, when euglycemia was maintained (in the presence of basal insulin and glucagon), inhibition of phosphorylase by BAY R 3401 shifted net glycogen flux from breakdown to synthesis but, consistent with the data in our previous study (14), did not affect hepatic gluconeogenic flux. The selective rise in arterial insulin in the presence of euglycemia did not affect the hepatic gluconeogenic rate. However, the suppression of net glycogenolysis occurred somewhat more rapidly when the insulin level was increased. When mild hypoglycemia was allowed to occur in the presence of the phosphorylase inhibitor, the hepatic gluconeogenic rate doubled, regardless of the presence or absence of insulin infusion and the associated changes in plasma insulin and glucagon. Therefore, these results suggest that the changes in hepatic sinusoidal insulin (from 135 to 29 pmol/l) and glucagon (from 48 to 98 ng/l) seen in response to mild hypoglycemia had little effect on the acute increase in net hepatic gluconeogenic flux seen in response to hypoglycemia, at least when glycogenolysis was impaired. This finding is in agreement with our previous suggestions (61,62) and indicates that hypoglycemia per se may be the major factor increasing gluconeogenic flux under our experimental conditions. Further, it is clear that almost all of the glucose being produced by the liver in response to non-insulin-induced hypoglycemia was gluconeogenic in origin.

In euglycemic-euinsulinemic animals, the inhibition of phosphorylase suppressed NHGO without changing gluconeogenic flux. Because net glycogen flux is determined by the balance between glycogen phosphorylase activity and glycogen synthase activity (63), the inhibition of phosphorylase must have shifted the net flux toward synthesis, and, as a result, the gluconeogenic precursors taken up by the liver were incorporated into glycogen. The shift of net hepatic glucose balance from output to uptake, and of net glycogen metabolism from glycogenolysis to glycogen synthesis, occurred in the absence of changes in the plasma levels of regulatory hormones or substrates, suggesting an important role for the dynamic balance between glycogen phosphorylase activity and glycogen synthase activity in determining net hepatic glucose balance. As noted above, the presence of mild hyperinsulinemia in combination with euglycemia and phosphorylase inhibition accelerated the shift of net hepatic glucose and glycogen flux from output and breakdown to uptake and synthesis, respectively, but did not alter gluconeogenic flux (Fig. 1 and Table 3). Therefore, the effect of hyperinsulinemia in this context was to more rapidly activate glycogen synthesis.

In the hypoglycemic-hypoinsulinemic animals, inhibition of phosphorylase decreased NHGO only transiently (Fig. 1). Although net hepatic glycogenolysis was almost completely suppressed, glycogen synthesis did not occur. During hypoglycemia, the plasma concentrations of glucose and insulin, which activate glycogen synthase, fell (64), and the plasma concentration of glucagon, which inhibits glycogen synthase, rose (64). It is likely, therefore, that glycogen synthase activity decreased, and, as a result, the inhibition of phosphorylase activity did not result in net hepatic glycogen synthesis. Indeed, when mild hepatic euinsulinemia was maintained in the presence of hypoglycemia and basal hepatic sinusoidal glucagon levels, net glycogen flux shifted from breakdown to synthesis even in the presence of hypoglycemia. Therefore, glycogen synthase activity appears to have been sensitively regulated by small changes in plasma insulin and/or glucagon levels.

The present study shows that 1) insulin secretion from β -cells in vivo decreases sensitively in response to mild hyperinsulinemia induced by exogenous insulin infusion, even in the presence of euglycemia; 2) mild hyperinsulinemia completely inhibits the increase in glucagon secretion normally caused in response to mild non- insulin-induced hypoglycemia; 3) the decrease in plasma insulin and increase in plasma glucagon levels that occur in response to mild hypoglycemia exert their regulatory effect on glycogen metabolism but not on gluconeogenic flux; 4) mild hypoglycemia per se plays a major role in acutely increasing gluconeogenic flux in response to mild non- insulin-induced hypoglycemia; and 5) coincident inhibition of glycogen deposition (by hormonal changes) and a stimulation of gluconeogenic flux (by hypoglycemia) results in gluconeogenesis, which prevents serious hypoglycemia in response to hepatic phosphorylase inhibition.

This research was supported by grants from the National Institutes of Health (DK43706 and DK20593).

We thank Jon Hastings and the members of the Vanderbilt Diabetes Research and Training Center Core Labs (Wanda Snead, Eric Allen, and Angelina Penaloza) for technical support.

Part of this work was presented at the 57th Annual Meeting of the American Diabetes Association, Boston, Massachusetts, 8–11 June 1997, and 58th Annual Meeting of the American Diabetes Association, Chicago, Illinois, 13–16 June 1998.