The pancreatic β-cell paradigm for glucose sensing has been proposed to apply to brain glucose sensors controlling counterregulation to hypoglycemia and feeding behavior. Over recent years, we tested this model in mice by first showing that inactivation of the GLUT2 gene suppressed glucose sensing and correctly regulated insulin secretion by pancreatic β-cells. Then, we restored the function of the β-cell in GLUT2-null mice by transgenic expression of a glucose transporter under the control of the rat insulin promoter. Using these rescued mice, we showed that GLUT2-dependent sensors are present in several anatomical sites, including the hepatoportal vein and the central nervous system. When these extrapancreatic glucose sensors are inactivated, the mice display loss of first-phase insulin secretion and hyperglucagonemia in the fed state, and they eat more than control mice—defects characteristic of developing obesity/diabetes. By gene complementation experiments, we further showed that glucose sensors controlling glucagon secretion require GLUT2 expression in glial cells. However, transgenic expression of GLUT2 in astrocytes or neurons failed to restore the normal control of feeding, indicating that different classes of glucose sensors control the response to hypoglycemia and food intake.
GLUCOSE SENSING BY THE PANCREATIC β-CELLS
Glucose-stimulated insulin secretion requires glucose uptake by the β-cells, which is facilitated by the glucose transporter GLUT2, the phosphorylation of glucose by glucokinase, and the generation of ATP by oxidative phosphorylation. The rise in intracellular ATP-to-ADP ratio then leads to closure of the ATP-sensitive K+ channel and plasma membrane depolarization. This opens the voltage-dependent calcium channel, and the entry of Ca2+ stimulates insulin secretion. Glucokinase catalyzes the rate-controlling step in this signaling pathway (1), and GLUT2, a high Km glucose transporter, allows unrestricted access of glucose to glucokinase (2,3).
Inactivation of the GLUT2 gene in the mouse induces early death of the animals around the time of weaning. This is caused by the suppression of glucose-stimulated insulin secretion (4). However, these GLUT2-null mice can survive in two conditions. First, when injected daily with insulin, they grow normally and well beyond the time of weaning. Second, the transgenic expression of GLUT1 or GLUT2 specifically in their β-cells restores normal glucose-stimulated insulin secretion and allows the mice to survive and breed normally (5). The mice rescued by GLUT1 expression in their β-cells (ripglut1;glut2−/−) have been extensively studied to evaluate the role of extra-pancreatic GLUT2-dependent glucose sensors in the control of glucose homeostasis and feeding behavior.
THE HEPATOPORTAL GLUCOSE SENSOR
The hepatoportal glucose sensor, which is an as yet incompletely defined structure, is activated by a glucose gradient established between the portal vein and the periphery (6). It is linked to afferent branches of the hepatic vagal nerve, which projects to the brainstem and the hypothalamus (7) and then controls several physiological functions, such as inhibition of counterregulation, stimulation of liver glucose uptake, and termination of feeding. We have provided evidence that this sensor, activated by intraportal glucose infusion in conscious mice, increases the rate of glucose clearance by a subset of peripheral tissues: brown fat, soleus muscle, and the heart (8). We showed that this sensor required the presence of GLUT2 but that hepatocytes were not involved in this sensing process (9), in agreement with previous studies showing this sensor to be located upstream of the hepatic hylus (10). We also demonstrated that the function of this sensor required expression of the glucagon-like peptide 1 (GLP-1) receptor to properly function (11), an observation compatible with the role of GLP-1 in regulating the firing activity of hepatic vagal afferents (12) and with the presence of the GLP-1 receptor mRNA in the nodose ganglion (13). A similar role for GLP-1 in the function of the dog hepatoportal sensor has been reported (14).
The signal sent to peripheral tissues to increase glucose utilization has not yet been identified, although a role for insulin could be excluded. This was demonstrated in experiments using mice with muscle-selective inactivation of the insulin receptor (MIRKO ). When glucose was infused in the portal vein of these mice, hypoglycemia and glucose clearance were stimulated to the same extent as in control mice. However, GLUT4 expression in muscle was required, since stimulation of the portal vein sensor in mice with muscle-specific inactivation of this transporter (MGLUT4KO ) did not increase glucose clearance. Similarly, no effect of portal vein glucose infusion could be observed in mice expressing a dominant-negative form of the AMP-activated kinase (17) specifically in muscle. Thus, the portal signal induces hypoglycemia and increases glucose clearance mostly by stimulating glucose uptake in muscle, by a mechanism that does not require the presence of the insulin receptor but that requires the expression of GLUT4 and the activation of the AMP-activated kinase (18).
The hepatoportal sensor is also involved in the control of insulin secretion via an autonomic nervous connection (19,20) and is critical for triggering the first phase of insulin secretion. This was demonstrated in mice with inactivation of the portal sensor due to knockout of the GLUT2 or the GLP-1 receptor genes or caused by injection in the portal vein of the GLP-1 receptor antagonist exendin(9-39) (21). In these experimental conditions, which inactivate the portal sensor, intraperitoneal injections of glucose, which is rapidly drained into the portal vein, failed to induce the expected first phase of insulin secretion (measured as the peak of plasma insulin 2 min after injection). In parallel in vitro experiments, perifused islets isolated from the same GLUT2-null or GLP-1 receptor–null mice showed normal induction of first-phase secretion upon stimulation with high glucose concentrations. These observations therefore suggest that glucose activation of the portal sensor may be the physiological regulator of first-phase insulin secretion, rather than glucose directly stimulating the β-cells. A further consequence is that the loss of first-phase insulin secretion in the pathogenesis of type 2 diabetes may be caused by an initial deregulation of the hepatoportal sensor, even before β-cells become dysfunctional.
EXTRAPANCREATIC GLUCOSE SENSING AND GLUCAGON SECRETION
Glucagon secretion is an immediate response to hypoglycemia that is critical to restore normal blood glucose levels by triggering hepatic glycogenolysis. The secretory activity of the α-cells is regulated by an α-cell glucose sensor (22), by intra-islet insulin secretion (23), by the sympathetic and parasympathetic autonomic nervous systems, and by the sympathoadrenal axis (24). Thus, we investigated whether extra-pancreatic GLUT2-dependent glucose sensors could be involved in regulating the autonomic tone to α-cells and lead to abnormal glucagon secretion in ripglut1;glut2−/− mice. We showed that these mice had hyperglucagonemia in the fed state, which could be rapidly corrected by ganglionic blockade, indicating that in the absence of GLUT2 expression (which is not expressed by α-cells), there was indeed an increased autonomic tone to α-cells. We further showed that in the absence of GLUT2, glucagon secretion was no longer stimulated during hypoglycemic or inhibited by hyperglycemic clamps (25).
More recently, we addressed the question of whether these glucose sensors controlling counterregulation were centrally located and whether GLUT2 needed to be expressed in glial cells or in neurons (26). To answer the first question, we used intraperitoneal or intracerebroventricular injections of 2-deoxy-d-glucose (2-DG) in ripglut1;glut2−/− mice. This glucoprivic signal induces glucagon secretion when injected by both routes in control mice but failed to induce secretion in mutant mice. This indicated that central GLUT2-dependent glucose sensors were involved in the control of glucagon secretion. By performing c-fos labeling studies, we could demonstrate that intraperitoneal 2-DG injections induced an increase in the number of c-FLI+ cells in the nucleus of the solitary tract and the dorsal motor nucleus of the vagus in control mice. In the GLUT2-null mice, no increase in c-FLI+ cells could be observed in these structures. In the ventromedial hypothalamus, however, increase in c-FLI+ cells was the same in control and GLUT2-null mice. Because the ventromedial hypothalamus has been described to be a potentially important site controlling glucagon secretion (27–29), our data suggest that the glucose sensor involved in this structure may be GLUT2 independent.
To evaluate in which cell type GLUT2 was required for the glucagon response, we expressed GLUT2 in glial cells or neurons of ripglut1;glut2−/− mice by transgenesis using the glial acidic fibrillary protein or synapsin promoters, respectively. We demonstrated that glial cell expression of GLUT2 in the GLUT2-null mice restored glucagon secretion in response to 2-DG injections and to hypoglycemic clamps (26). At the same time, this restored the c-FLI labeling in the nucleus of the solitary tract and dorsal motor nucleus of the vagus neurons. No such recovery of glucagon secretion could be observed in mice expressing GLUT2 in neurons. Together, these data demonstrated that 1) central glucose sensors were involved in the control of glucagon secretion in response to both hypoglycemia and the glucoprivic signal induced by 2-DG, 2) these sensors were GLUT2 dependent, and 3) GLUT2 needed to be expressed in glial cells, indicating the existence of a metabolic coupling between astrocytes and neurons.
The above data were obtained in mice in a mixed genetic background, consisting of C57 Bl/6, Sv129, and DBA2 origin. When backcrossed in C57Bl/6 mice, the ripglut1;glut2−/− mice showed the same fed hyperglucagonemia that could be reversed by ganglionic blockade, thereby confirming the role of central GLUT2-dependent sensors in the control of basal glucagon levels and of autonomic tone to the α-cells. However, after intraperitoneal or intracerebroventricular 2-DG injections or hypoglycemic clamps, glucagon secretion was stimulated as in control mice. Thus, these observations provided evidence for the involvement of a complex network of glucose-sensing systems controlling fed glucagon levels and the response to hypoglycemia. One controls fed glucagon levels and is GLUT2 dependent. For the response to hypoglycemia or glucoprivation, there may be two different sensors: one GLUT2 dependent, which predominates in the mixed genetic background mice, and another that predominates in the C57Bl/6 mice. The GLUT2-dependent sensors in the mixed background mice require GLUT2 expression in glial cells.
CENTRAL GLUCOSE SENSING AND THE CONTROL OF FEEDING
A glucostatic theory for the control of feeding had been proposed about 50 years ago (30). However, the exact physiological role of glucose in the regulation of feeding has so far been uncertain (31). We therefore evaluated the feeding behavior of the ripglut1;glut2−/− mice and whether they were normally responsive to glucose and 2-DG intraperitoneal or intracerebroventricular injections (32). Our data demonstrated that ad libitum–fed GLUT2-null mice ate ∼25% more than their controls. Furthermore, when presented with food after a 24-h fast, they initiated refeeding very slowly, even sometimes waiting for up to 3 h before eating. However, after 48 h of refeeding, their cumulative food intake was greater than the controls. This indicated a defect both in feeding initiation and feeding termination. To get further evidence that absence of GLUT2 expression caused a defect in the control of feeding by glucose, we injected fasted mice with glucose and measured refeeding over a 4-h period. Either intraperitoneal or intracerebroventricular glucose administration reduced the cumulative food intake over the period of observation in control mice but not in the GLUT2-null mice. As a second approach, we injected fed mice with 2-DG. Either intraperitoneal or intracerebroventricular 2-DG injections induced a strong feeding response in control mice but not in the GLUT2-null mice. Therefore, in the absence of GLUT2, central sensors sensitive to both glucose and cellular glucoprivation are inactivated.
To evaluate whether the abnormal feeding behavior was correlated with a defect in the regulated expression of the hypothalamic orexigenic and anorexigenic neuropeptides, we measured by quantitative RT-PCR the mRNA levels of neuropeptide Y, agouti gene–related peptide, preopiomelanocortin, and cocaine- and amphetamine-regulated transcript, as well as the level of expression of the neuropeptides thyrotropin-releasing hormone (from the paraventricular nucleus) and orexin (from the lateral hypothalamus) in the fasted state and after 6 h of refeeding. In control mice, fasting was associated with an increase in the mRNA level of the orexigenic peptides and a decrease in that of the anorexigenic peptides. In the GLUT2-null mice, this regulation was lost and the level of expression of either type of peptides remained at a low level. Thyrotropin-releasing hormone expression level was also no longer activated by fasting and the orexin mRNA levels were not regulated, but maintained at a level higher than that found in the control mice.
Because these hypothalamic neuronal circuits are regulated by circulating levels of leptin and insulin (33,34), we assessed the plasma levels of these hormones and the blood glucose concentrations during the fast-to-refed transition. These parameters were normally regulated, implying that the loss of glucose sensing, rather than a dysregulation of these hormones or of glucose circulating levels, was preventing the normal change in expression of these neuropeptides. As a direct test of this hypothesis, glucose was injected intracerebroventricularly in fasted mice. This induced suppression of neuropeptide Y and an increase in the preopiomelanocortin mRNA levels in the control mice, but no regulation in the ripglut1;glut2−/− mice. These data therefore indicate that the loss of central glucose sensing probably caused the deregulated expression of these hypothalamic neuropeptides.
These observations (32) therefore led to several important conclusions: 1) in the absence of GLUT2, glucose-sensing units controlling feeding in response to glucose or glucoprivation are lost; 2) these GLUT2-dependent sensors are, at least in part, located centrally; 3) their inactivation leads to altered feeding behavior in a physiological situation, i.e., during the fast-to-refed transition; 4) abnormal feeding behavior is probably caused by the loss of regulated expression of hypothalamic neuropeptides; and 5) since plasma leptin and insulin levels are normally regulated during the fast-to-refed transition, glucose sensing is therefore critical for the normal action of these two hormones on the melanocortin pathway.
EVIDENCE FOR SEPARATE SENSORS CONTROLLING FEEDING AND COUNTERREGULATION
To evaluate whether the glucose sensors controlling glucagon secretion and food intake were the same or distinct entities, we performed feeding tests on ripglut1;glut2−/− mice in a mixed genetic background and with complementation of GLUT2 expression in astrocytes or neurons. As shown in Fig. 1A, ad libitum–fed ripglut1;glut2−/− mice ate more than control mice and, in contrast to control mice, they failed to respond to intracerebroventricular 2-DG injection by an increase in feeding (Fig. 1B). These mice also failed to increase their glucagon secretion after 2-DG administration (Fig. 1C; 26). Figure 2B and C show that fasted GLUT2-null mice injected intraperitoneally with glucose also failed to reduce the amount of food eaten after a fast, in contrast to the significant reduction in refeeding observed in control mice (Fig. 2A). When GLUT2 was expressed in glial cells of ripglut1;glut2−/− mice by transgenesis, glucagon secretion in response to 2-DG administration (Fig. 1C) or hypoglycemic clamp (26) was restored. However, no restoration of the feeding behavior was observed compared with the GLUT2-null mice (Figs. 1A–C and 2C). Therefore, even though glial cell expression restored regulated glucagon secretion, no rescue of the feeding pattern could be observed. This observation suggests that separate GLUT2-dependent glucose sensors control counterregulation and feeding.
To evaluate whether GLUT2 expression in neurons could restore the control of feeding, we generated transgenic mice expressing GLUT2 under the control of the synapsin promoter and crossed them with the GLUT2-null mice (26). These mice (synapsinglut2,ripglut1;glut2−/−) did not show any correction of the glucagon response to intracerebroventricular 2-DG injections (Fig. 3C for mouse line 2 and  for mouse line 1). Ad libitum feeding for 24 h was as elevated in these mice as in the GLUT2-null mice and higher than in the control mice (Fig. 3A). Similarly, the synapsinglut2,ripglut1;glut2−/− mice did not increase feeding when injected intracerebroventricularly with 2-DG (Fig. 3B). Thus, the glucose sensors controlling feeding behavior were not restored in two lines of transgenic mice expressing GLUT2 in their neurons.
Together, the above data indicated that the sensors controlling counterregulation and feeding consist of distinct entities. For counterregulation, GLUT2-dependent sensors are required for the control of fed glucagonemia and, in mixed genetic background, also for the response to hypoglycemia or cellular glucoprivation. In the C57Bl/6 background, only fed glucagonemia is abnormal because of the elevated tone of the autonomic nervous system to the α-cells. However, the response to glucoprivation or hypoglycemic clamps is normal indicating that other, non–GLUT2-dependent sensors are active in these conditions. Our data had shown that c-fos activation in the ventromedial hypothalamus was GLUT2-independent, whereas it was GLUT2-dependent in the nucleus of the solitary tract and dorsal motor nucleus of the vagus. Thus, hypothalamic ventromedial hypothalamus neurons may be responsive to hypoglycemia by a mechanism that does not involve GLUT2.
For the control of feeding, absence of GLUT2 leads to similar defects, irrespective of the genetic background. These defects were not rescued by glial cell expression of GLUT2 or by neuronal expression of GLUT2 in two lines of transgenic mice. Because expression of GLUT2 under the synapsin promoter was relatively high in both the hypothalamus and brainstem, as judged by Northern blot analysis (26), the failure to restore feeding control may indicate that 1) GLUT2 expression in neurons is not required for control of feeding; 2) GLUT2 is required in neurons but the transgenic lines studied did not express the transporter in the correct neurons; 3) GLUT2 expression is required in another cell type, for instance, in ependymal cells or tanycytes (35,36); or 4) a combination of GLUT2 expression in different cell types is required, for instance, in ependymal cells and neurons. Addressing these questions will require more precise gene deletion approaches, for instance, using the Cre-lox system.
The studies described above have allowed the exploration of the potential role of extrapancreatic glucose sensors presenting some functional characteristics similar to that of the pancreatic β-cells. In particular, based on the initial observation that GLUT2 was required for normal glucose-stimulated insulin secretion, we could demonstrate in physiological approaches in the mouse that this transporter was present in glucose-sensing units present at other anatomical sites. These sensors control several physiological functions, probably mainly by activating the autonomic nervous systems. Of particular interest is the role of the hepatoportal sensor in the control of first-phase insulin secretion and of central sensors in the control of glucagon secretion and feeding. Indeed, when these sensing mechanisms are inactivated, this leads to suppression of first-phase insulin secretion, to fed hyperglucagonemia, and to overfeeding. These are critical deregulations of glucose and energy homeostasis that are associated with development of obesity and type 2 diabetes. These observations therefore strongly suggest that glucose plays a critical role as a signal, not only in the control of glucose homeostasis, but also of feeding behavior. Importantly, they suggest that identification of the cells and molecules forming these extrapancreatic GLUT2-dependent sensors may provide critical novel information on the physiology and pathophysiology of the regulation of energy metabolism.
This article is based on a presentation at a symposium. The symposium and the publication of this article were made possible by an unrestricted educational grant from Servier.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by grants 3100-065219-01 from the Swiss National Science Foundation and 1-2002-366 from the Juvenile Diabetes Research Foundation International.
We thank Martine Emery, David Tarussio, and Joël Gyger for excellent technical assistance.