Evidence That Extrapancreatic GLUT2-Dependent Glucose Sensors Control Glucagon Secretion

The 10- to 14-week-old RIPGLUT1 × GLUT2^−/−, C57Bl/6, or C57Bl/6 × SV129 male mice from our colonies were used in these studies and housed with an inverted light cycle with the dark phase from 8:00 a.m. to 8:00 p.m. RIPGLUT1 × GLUT2^−/− mice were derived from GLUT2-null mice (29) by transgenic reexpression of GLUT1 specifically in pancreatic β-cells under the control of the rat insulin promoter (30). The RIPGLUT1 × GLUT2^−/− mice have restored normal glucose-stimulated insulin secretion as assessed in perifusion experiments and by hyperglycemic clamps (30).

An indwelling catheter was implanted in mice into the vena cava through the left femoral vein as described (31,32). The other extremity of this catheter was slid under the back skin, exteriorized, and glued at the back of the neck. The mice were then housed individually. After 2 days, they showed normal feeding behavior and motor activity.

At 3–5 days after catheter implantations, food was removed at 8:00 a.m. for 6 h. In awake, freely moving mice, human insulin (Actrapid; Novo Nordisk, Bagsvaerd, Denmark) was infused at the rate of 18 mU · kg^–1 · min^–1 for 60 or 180 min through the femoral vein catheter. A glucose solution (33% wt/vol) was co-infused with insulin to maintain blood glucose levels at the constant values of 2.5, 5.5, 10, or 20 mmol/l during the experimental period. Blood glucose concentration was monitored using a glucose meter (Roche Diagnostics, Rotkreuz, Switzerland) from 3.5 μl blood sampled approximately every 10 min. At completion of the infusions, blood was collected for plasma glucagon and insulin determinations.

To block the ANS, fed mice were injected with 5 mg/kg of the ganglionic blockers hexamethonium (Sigma, St. Louis, MO) or chlorisondamine (Novartis, Summit, NJ). Blood was collected 30 min after the injections for plasma glucagon, insulin, and glucose determinations. To induce a rapid and profound hypoglycemia, 6-h fasted mice were injected with 0.7 U/kg human insulin. Blood was collected 30 or 60 min after the injection for plasma glucagon and glucose determinations. Blood glucose was monitored as described above.

Blood was collected in the presence of 1 μg/ml aprotinine and 1 mmol/l EDTA and immediately centrifuged, and the plasma was stored at –70°C until assayed. Plasma glucagon and insulin concentrations were assessed by radioimmunoassay (Linco, St. Louis, MO). Plasma glucose concentration was assessed using a glucose oxidase method (Sigma).

Table 1 shows the plasma glucose, glucagon, and insulin levels of control and RIPGLUT1 × GLUT2^−/− mice in the fed and fasted states. The data show that glycemia of fed mutant mice was similar to that of control mice. In the fasted state, the mutant mice were slightly hypoglycemic, which could probably be explained by impaired renal glucose reabsorption in the absence of GLUT2. Plasma glucagon levels of the mutant mice were twice those of control mice in the fed state and were similar to those of control mice in fasted animals. Plasma insulin levels were lower in both fed and fasted states in the mutant mice.

To evaluate whether there was a defect in the control by glucose of glucagon secretion in the absence of GLUT2, we first infused mice with insulin and glucose to reach glycemic levels of 2.5, 5, or 20 mmol/l and determined the plasma glucagon levels after 60 min. Figure 1A shows the glycemic profiles established, and Figs. 1B and C show the absolute and relative plasma glucagon concentrations determined at the end of the infusion experiments. Hypoglycemia reached a value of ∼2.0–2.5 mmol/l within ∼40 min and stayed at this level until 60 min. In control mice, this hypoglycemia induced an approximately sevenfold increase in plasma glucagon levels, whereas the increase was only ∼50% in the mutant mice.

Hyperglycemia developed linearly over the 60-min time course of this experiment to reach a value of 20 mmol/l, and in this condition, no significant suppression of glucagon levels could be observed in either type of mice. The lack of glucagon suppression in these conditions may be due to the relatively slow kinetics of establishment of hyperglycemia.

We next performed much longer hyperinsulinemic clamps with glucose maintained at 2.5, 5.5, 10, or 20 mmol/l for 180 min and determined plasma glucagon levels at the end of the experiments. Figure 2A shows the glycemic profiles that were established during the experiments. Figure 2B and C show that, in control mice, hypoglycemia increased plasma glucagon by ∼1.7-fold and hyperglycemia at 10 or 20 mmol/l decreased glucagonemia by half. In contrast, in RIPGLUT1 × GLUT2^−/− mice, no significant changes in glucagon levels occurred in response to either hypo- or hyperglycemia.

Maintaining normoglycemia during the clamp was associated with an increase in plasma glucagon levels compared with nonclamped 6 h–fasted mice (compare Fig. 2B and Table 1). This increase could be explained by the stress due to the experimental protocol, which requires frequent blood sampling of awake animals. A higher stress level was confirmed by elevated plasma catecholamine levels at the end of the clamp (data not shown).

In the experiments of Figs. 1 and 2, glycemic levels were maintained in the presence of continuous exogenous insulin infusion. To evaluate the role of endogenously produced insulin, we performed hyperglycemic clamps by infusing glucose in control and mutant mice to reach a glycemia of 20 mmol/l. Figure 3A shows the glycemic profiles established during glucose or saline infusion for 3 h, and Figs. 3B and C present the glucagon and insulin levels measured at the end of the experiment. These data show that the insulin secretory response is the same in control and mutant mice but that in these clamp conditions, there is still no suppression of glucagon secretion by hyperglycemia, even though insulin secretion is normal.

We next evaluated whether the impaired glucagon secretory activity in response to hypoglycemia was due to a complete or a relative impairment of the effect of low glucose. We thus injected control and mutant mice intraperitoneally with insulin to rapidly induce a profound hypoglycemia. As shown in Fig. 4A, the glycemic levels rapidly reached a value of ≤1 mmol/l. The plasma glucagon levels were then measured at 30 and 60 min and are shown in Figs. 4B and C. Under these severe hypoglycemic conditions, strong elevations in plasma glucagon levels were detected both in control (∼8-fold and ∼2.5-fold over basal at 30 and 60 min, respectively) and mutant mice (∼5-fold and ∼2-fold over basal at 30 and 60 min, respectively).

To assess whether the high plasma glucagon level found in the fed mutant mice may be due to an uncontrolled permanent stimulation of α-cells by the ANS, we injected mutant and control mice with a saline solution or a solution containing hexamethonium or chlorisondamine, two different ganglionic blockers. Plasma glucagon levels were then measured 30 min later. Figure 5A shows that the plasma glucagon levels were not modified in the control mice, but in RIPGLUT1 × GLUT2^−/− mice, they were decreased to a level similar to that of control mice. Figure 5B and C show the plasma insulin and glycemic levels measured in the same mice. Injection of ganglionic blockers had a similar effect in decreasing insulin and glucose levels in control and mutant mice.

This study provides evidence that GLUT2 is required for the normal control of glucagon secretion in response to both hypo- and hyperglycemia. Because GLUT2 is not normally expressed in pancreatic α-cells, it is required for the function of glucose-sensing units controlling glucagon secretion likely located outside of the islets. Furthermore, in the absence of GLUT2, an increased ANS tone appears to be responsible for the elevated glucagon levels found in the fed state.

The GLUT2 glucose transporter is not expressed by pancreatic α-cells of normal mice (9,10) and is therefore not involved in direct glucose sensing by these cells. However, GLUT2 is required for glucose sensing by mouse pancreatic β-cells because its absence leads to loss of first-phase and diminished second-phase glucose-stimulated insulin secretion (29). Re-expression by transgenesis of GLUT1 or GLUT2 in pancreatic β-cells rescues the GLUT2-null mice from early death and normalizes both the first and second phases of glucose-induced insulin secretion, as assessed in vitro in islet perifusion experiments or in vivo during hyperglycemic clamps (30) (Fig. 3). The endocrine pancreas of these mice should therefore function normally. In this context, the low insulin levels of the RIPGLUT1 × GLUT2^−/− mice studied here cannot be explained by a β-cell defect. It can more probably be explained by the fact that in the absence of GLUT2 from the kidney tubule, there is a strong impairment in glucose reabsorption leading to a severe glycosuria (29) and, thus, an efficient insulin-independent glucose disposal. These mice are therefore in a situation just opposite to that found in mild insulin-resistant states in which normoglycemia is maintained because of relatively high insulin levels. It is however similar to the situation found, for instance, in transgenic mice overexpressing GLUT4 in muscles where increased glucose disposal due to glucose transporter overexpression is associated with normoglycemia and lower insulinemic levels in the fasted state (33).

Here, we demonstrate that glucagon secretion was only poorly stimulated by moderate hypoglycemia and not suppressed by hyperglycemia in the mutant mice, whereas strong regulation could be observed in the control animals. However, decreasing glycemia to ≤1 mmol/l markedly stimulated glucagon secretion in the mutant mice.

The impaired stimulation of glucagon secretion in response to moderate hypoglycemia suggests that GLUT2 expression is required for the proper detection of glucose by the sensing units controlling α-cell secretory activity. Furthermore, because normal α-cells do not express GLUT2 and the β-cells of these mice express an alternate glucose transporter, these sensing units are likely located outside of the endocrine pancreas. There is however a glucagon response to severe hypoglycemia. This response suggests a GLUT2-independent mechanism for stimulating the glucagon response to extreme hypoglycemia and indicates that the α-cells have a preserved capability to increase their secretory activity.

The preservation of a robust secretory response to severe hypoglycemia, but of only a mild one to moderate hypoglycemia, can be explained in different ways. First, there may be a single type of glucose-sensing unit controlling glucagon secretion during hypoglycemia, which is dependent on GLUT2 expression. In the absence of the transporter, a response can be stimulated only by extreme glycemic conditions. This would be similar to the secretion of insulin by GLUT2-null β-cells, which is mostly suppressed in the presence of moderate hyperglycemia but still detected with extreme glycemic levels (29,34). Second, there may be separate glucose sensors, activated at different hypoglycemic levels, that rely on different glucose transporter isoforms. It is indeed known that the counterregulatory response to hypoglycemia is graded, with the glucagon response being activated at milder hypoglycemia than the sympathoadrenal response, which, in turn, also stimulates glucagon secretion (12). In such a situation, GLUT2-dependent sensors would be activated only by moderate hypoglycemia, and severe hypoglycemia would activate another type of sensor. Third, it is possible that the increased autonomic tone to the α-cells potentiates a direct hypoglycemic stimulation of the α-cell sensor, which is prominent only at the lower glycemic levels.

In contrast to the glucagon response to hypoglycemia, suppression of glucagon secretion by hyperglycemia appears totally lost. This indicates that hyperglycemia sensing units suppressing glucagon secretion are also GLUT2-dependent. These detection units cannot be the pancreatic β-cells because their function is normal due to the transgenic reexpression of GLUT1 (30). We indeed demonstrated that high glucose normally stimulated β-cell secretory activity in hyperglycemic clamps, yet these experimental conditions did not suppress glucagon secretion.

Where are the glucose detection units located? Many studies have presented evidence for a major role of the ANS in the control of glucagon secretion during hypoglycemic episodes (35). It could therefore be hypothesized that the sensing units are located in brain regions controlling activation of the ANS (in particular the ventromedial and lateral hypothalamus). This is in agreement with the fact that GLUT2 is expressed in the hypothalamus and the brain stem where glucose-sensitive neurons are present and where many of the proteins involved in glucose signaling in pancreatic β-cells are found. This is the case for glucokinase (36,37,38), the K_ATP channel (39,40,41), or the glucagon-like peptide 1 receptor (37,42,43). However, the cellular localization of GLUT2 in the hypothalamus and brain stem is not yet unambiguously established. Some reports described GLUT2 to be present in glial cells (44), whereas others found the transporters in neurons or in ependymal cells lining the third or fourth ventricle (37,38,45). Further mapping of GLUT2 expression in the central nervous system is therefore warranted.

In addition to central glucose detection units, other peripheral glucose sensors, such as those present in the hepatoportal vein, which we recently showed to be GLUT2-dependent (46,47), may also be involved in controlling ANS activity.

Although we do not yet have evidence for the location of these GLUT2-sensing units, we have shown that the hyperglucagonemia of fed mutant mice is the result of increased ANS activity. Indeed, blocking ganglionic transmission with hexamethonium or chlorisondamine rapidly led glucagon levels of mutant mice to decrease to those of fed control mice. Because the same treatment in control mice did not change their glucagon levels, this strongly suggests that in the absence of GLUT2, there is an increased tone of the ANS-stimulating glucagon secretion. Because high plasma glucagon levels are seen only in the fed state, this suggests that the ANS tone, rather than directly stimulating secretion, may sensitize α-cells to the secretagogue activity of other nutrients (e.g., amino acids). This would explain why glucagonemia returns to control levels in fasted mutant mice when the concentration of nutrients in plasma decreases. Whatever the role of the ANS tone, i.e., sensitizing or directly stimulating the α-cell secretory activity, our data suggest that GLUT2 participates in a glucose detection unit that normally controls this autonomic tone.

To explain that both hypoglycemia and hyperglycemia responses are lost in the absence of GLUT2, we propose the following hypothesis. Activity of the ANS controlling glucagon secretion may be modulated by two separate glucose sensors—one sensitive to hyperglycemia, the other sensitive to hypoglycemia, and both GLUT2-dependent. The hypoglycemia sensor continuously activates the autonomic tone, but its stimulatory activity can be suppressed by euglycemia. In the absence of GLUT2, the constant activity of the sensor can no longer be suppressed by euglycemia. In contrast, the hyperglycemia sensor, which may suppress autonomic tone, is mostly silent at euglycemia and may become activated when glucose concentrations rise. Again, in the absence of GLUT2, there is a lack of glucose sensing and therefore a lack of inhibition of the autonomic tone. Although this interpretation requires direct experimental verification, we believe that it is compatible with the glucose sensitivity of neurons found in the hypothalamus, which can respond to low- or high-glucose concentrations by increasing or decreasing their firing rates (17,18,19,20,21). For example, the majority of glucose-sensitive neurons of the lateral hypothalamic area decreased their firing rates in response to small increases in local glucose concentrations. In contrast, in the ventromedial hypothalamus, the majority of glucose-sensitive neurons increased their firing rates in response to small increases in local glucose concentrations (21).

Together, our data present strong evidence that extrapancreatic GLUT2-dependent glucose sensors are involved both in the stimulation of glucagon secretion in hypoglycemic conditions and in the suppression of secretion during hyperglycemia. Identification of these glucose-sensing units and the molecules they involve is of great importance not only in understanding the complexity of glucose sensing in the central nervous system, which controls diverse functions such as feeding and energy expenditure, but also because hypoglycemia unawareness and defective counterregulation are major dysfunctions associated with type 1 diabetes. Our present data point to a key role of GLUT2 in these functions.

This work was supported by grants 31-46958.96 from the Swiss National Science Foundation and 198243 from the Juvenile Diabetes Foundation International.