Minireview: The Value of Looking Backward: The Essential Role of the Hindbrain in Counterregulatory Responses to Glucose Deficit

Abstract

This review focuses on evidence indicating a key role for the hindbrain in mobilizing behavioral, autonomic and endocrine counterregulatory responses to acute and profound glucose deficit, and identifies hindbrain norepinephrine (NE) and epinephrine (E) neurons as essential mediators of some of these responses. It has become clear that hindbrain NE/E neurons are functionally diverse. However, considerable progress has been made in identifying the particular NE/E neurons important for particular glucoregulatory responses. Although it is not yet known whether NE/E neurons are directly activated by glucose deficit, compelling evidence indicates that if they are not, the primary glucoreceptor cells must be located in the immediate vicinity these neurons. Hindbrain studies identifying cellular markers associated with glucose-sensing functions in other brain regions are discussed, as are studies examining the relationship of these markers to counterregulatory responses of NE/E neurons. Further investigations to identify glucose-sensing cells (neurons, ependymocytes, or glia) controlling counterregulatory responses are crucial, as are studies to determine the specific functions of glucose-sensing cells throughout the brain. Likewise, examination of the roles (if any) of hindbrain counterregulatory systems in managing glucose homeostasis under basal, nonglucoprivic conditions will also be important for a full understanding of energy homeostasis. Nevertheless, the accumulated evidence demonstrates that hindbrain glucose sensors and NE/E neurons are essential players in triggering counterregulatory responses to emergencies of glucose deficit.

Glucose is used by virtually every cell in the body and is the essential metabolic fuel for the brain. Therefore, it is important to understand the mechanisms by which glucose levels are monitored and maintained and by which the brain is protected from precipitous declines in glucose availability. Several recent reviews have focused on the variety of cellular glucose-sensing mechanisms in the brain and peripheral tissues, the diversity of cellular phenotypes that appear to engage in glucose sensing, their distribution within the brain and the range of anatomical connections that might integrate responses of various glucose-sensing cells (1–9). These reviews contribute to an appreciation of the complexity of glucoregulatory processes. They also highlight the need to begin asking more specific questions about how particular glucose-sensing cells contribute to systemic glucoregulation under specific physiological or pathological conditions. For many years, our own work has focused on hindbrain glucoregulatory mechanisms that mobilize protective and restorative systemic responses to acute cerebral glucose deficit. Therefore, we will approach this review from the vantage point of the hindbrain. We will discuss evidence indicating a major role for the hindbrain in glucoregulation and the nature of that role. We also will speculate from current data on how the role of the hindbrain may differ from the role of the hypothalamus in this process.

Elicitation of Protective Systemic Responses to Central Glucose Deficit is a Crucial Aspect of Glucoregulation

In 1855, a central control of glycemia was proposed by Claude Bernard (10), who found that a probe “puncturing” the floor of the fourth ventricle (4V) stimulated glucose output from the liver into the blood. Walter B. Cannon, in 1924, identified the contribution of the adrenal medulla to hepatic glucose mobilization during hypoglycemia (11). Subsequently, Smith and Epstein (12) and Miselis and Epstein (13) used the antiglycolytic glucose analog, 2-deoxy-D-glucose (2DG), to reveal that a brain glucose deficit was sufficient to stimulate hepatic glucose mobilization as well as a rapid increase in food intake. An important contribution of this work was the introduction of a new experimental tool. Because it interferes with glycolysis directly at the cellular level (14), 2DG, unlike insulin, can be applied locally in the brain to cause local cytoglucopenia or glucoprivation. Hence, 2DG was the first in a line of antimetabolites that have enabled localized production of cellular glucose deficiency and that have been critical for dissecting central mechanisms for responding to glucose deficit.

It now is clear that a deficiency in glucose, the brain's essential metabolic fuel, triggers rapid and multifaceted behavioral, endocrine, and autonomic defenses that protect and restore the brain's glucose supply (4, 15, 16). These defenses often are referred to collectively as counterregulatory responses to glucose deficit. They include a rapid increase in food intake, which is the ultimate mechanism for restoration of depleted glucose reserves. Other responses, however, extend the brain's operating time, enabling the search for obtaining and ingestion of food. These include autonomic and endocrine responses to mobilize stored glucose and fat and to facilitate redistribution of these nutrients, thereby conserving glucose for use by the brain. Finally, glucoprivation-induced suppression of reproductive function avoids costly metabolic investment in reproductive activity in the face of energy shortage.

These systemic counterregulatory responses to glucose deficit, which are the primary focus of this review, clearly are crucial to survival of acute glucose deficit. Nevertheless, it is important to be clear at the outset that there may be glucose monitoring cells, and even glucoregulatory systems, that do not contribute to counterregulatory responses to glucose deficit. At the cellular level, glucose sensing may involve diverse cellular phenotypes, different glucose transporters (GLUT), ion channels, enzymes, etc. In fact, some glucose-sensing cells appear to be responsive to other nutrients as well as glucose. In most cases, the observation of apparent cellular glucose sensing has not yet been causally associated with any sort of systemic response, counterregulatory or otherwise. Some glucose-sensing cells may in fact function to balance energy influx and expenditure at the local level. Some might be involved in caloric regulation, nutrient balance, food preference, adjustment of metabolic rate, circadian rhythms, or other functions. The point is that we should not be surprised that glucose may be sensed by a multiplicity of cells and mechanisms, in the service of a variety of physiological and cellular processes. Therefore, it would be naïve to think that all glucose-sensing cells participate in counterregulatory responses to glucose deficit, when in fact many (or most) do not. We simply must acknowledge that there is still much to learn and that rigorous experimental effort eventually will reveal the particular functions of various glucose-sensing cells and the contexts under which they contribute cellular, local, or systemic energy homeostasis.

A Role for Hindbrain Glucoreceptors in Counterregulatory Responses

For the past three decades, evidence has steadily accumulated that the hindbrain contains neural circuitry essential for increasing food intake in response to glucose deficit (glucoprivation). Even the earliest attempts to identify brain sites that mediate feeding responses to glucose deficit revealed that glucoprivation of the hypothalamus and other forebrain sites, produced by local intraparenchymal injection of 2DG, did not evoke increased feeding, whereas lateral ventricular 2DG administration did (13, 17). Subsequently, Grill and co-workers demonstrated that feeding (18) and adrenomedullary (19) responses to systemic glucoprivation could be elicited in decerebrate rats in which all connections between forebrain and hindbrain are severed, demonstrating that the hindbrain and, perhaps, its inputs from the periphery are adequate to detect glucose deficit and trigger at least two important counterregulatory responses. Other work, using the glucose antimetabolite 5-thioglucose (5TG) (20), revealed that cerebral aqueduct blockade, which prevents lateral and third ventricular fluid from entering the 4V, eliminates the feeding and hyperglycemic responses to lateral ventricular 5TG injection but does not eliminate the ability of 4V 5TG injection to evoke these same responses (21). Injections into the forebrain ventricles were effective only when the injected 5TG was able to flow through the ventricular system to hindbrain sites that detected and triggered responses to glucose deficit. More recent studies have clearly revealed that 200 nanoliter volumes of 5TG (12–24 μg) injected unilaterally into discrete hindbrain locations increase food intake, adrenal medullary hyperglycemia, corticosterone, and glucagon secretion, whereas the same or higher 5TG doses (24–48 μg) injected into sites throughout the hypothalamus or even into the 4V fail to evoke these counterregulatory responses (22, 23). The inefficacy of local forebrain 5TG injections rules out the possibility that rostral diffusion of the 5TG from the hindbrain to hypothalamic or other forebrain sites is responsible for the elicitation of counterregulatory responses to hindbrain 5TG injection. Therefore, the cumulative evidence is now compelling: sensors in the hindbrain are primary mediators of increased food intake and adrenomedullary responses to acute glucose deficit. Nevertheless, elicitation of counterregulatory responses from hypothalamic injections of antimetabolites has also been reported (24, 25). We will discuss these results later, but first we want to review additional evidence, indicating 1) that receptor cells located within the hindbrain are the primary initiators of key counterregulatory responses to glucose deficit, and 2) that subsets of hindbrain norepinephrine (NE) and epinephrine (E) neurons [the “hindbrain catecholamine (CA) neurons”] are essential participants in triggering these responses.

Organization of Hindbrain Catecholamine Neurons

Hindbrain CA neurons are organized into groups designated as A for those expressing NE and C for those expressing E (26). All uniquely express the biosynthetic enzyme, dopamine β-hydroxylase (DBH). Cell groups A2 and C2 are located within the nucleus of the solitary tract (NTS) in the dorsomedial medulla. Cell groups A1 and C1 are located in the ventrolateral medulla, and C3 is diffusely scattered along the midline rostral and medial to C2. There is a degree of overlap at the transitions between A2 and C2 and between A1 and C1. Nevertheless, the locations and projections of these cell groups have been well described. The most rostral NE cell groups are A5, A6, and A7. Because experimental probing has not detected a role for A5–A7 in glucoregulation, these groups will not be discussed further here. Because the roles of NE vs. E neurons have not been distinguished in many cases, these phenotypes will often be designated as NE/E neurons.

The relatively diffuse distribution of NE and E cell bodies and their proximity and contribution to hindbrain life-supporting neural systems necessary for ventilatory, circulatory, and gastrointestinal controls is an obstacle to anatomical and functional analysis of these neurons using conventional lesioning approaches. However, with the use of the selective NE/E immunotoxin, anti-DBH monoclonal antibody conjugated to the ribosomal toxin, saporin (SAP) (27–29), it has been possible to overcome this obstacle. Anti-DBH-saporin (DSAP) is internalized selectively by DBH-containing NE and E neurons. Subsequent to internalization by terminals, it is retrogradely transported to their cell bodies of origin, where it disrupts ribosomal function, resulting in cell death. The retrograde transport of DSAP is an especially useful property of this toxin, because it enables selective destruction of NE/E subgroups without damage to adjacent hindbrain neurons. Retrograde destruction of neurons to a particular target area by DSAP also enables rigorous assessment of the involvement of specific NE/E neurons in particular glucoregulatory responses and to distinguish them from other intermingled neurons of the same CA phenotype. However, because NE and E neurons appear to be highly collateralized, it must be kept in mind that the site of DSAP injection may not be the site responsible for a particular deficit. A recurrent theme in our results is that DSAP lesions produce deficits in the response of particular systems to glucoprivic challenge without disrupting their basal functions.

Counterregulatory Responses Requiring Hindbrain CA Neurons

Glucoprivic feeding

Pharmacological evidence and lesion data generated over many years have implicated CA neurons in control of food intake (30, 31) and, specifically, in glucoprivic feeding. NE and E are potently orexigenic (32). Glucoprivation increases NE release from nerve terminals in the hypothalamus (33–35) and increases Fos expression in specific subpopulations of hindbrain NE and E neurons, primarily in A1 and C1 and C2 and a small percentage of A2 (36). The distribution of NE and E cell groups activated by systemic glucoprivation overlaps sites where feeding, adrenal medullary, glucagon, and corticosterone responses are elicited by localized nano-injection of 5TG (22, 23).

We injected DSAP into the paraventricular nucleus of the hypothalamus (PVH) to test further the hypothesis that NE or E neurons with projections to the PVH mediate glucoprivic feeding and to begin to identify the subgroups critical for this function (37). Retrograde transport of the toxin selectively destroyed neurons in multiple hindbrain NE and E cell groups that project to the PVH and surrounding medial hypothalamus. Although PVH DSAP injection destroyed some neurons in cell groups A2 and C2 in the dorsomedial medulla, PVH DSAP virtually eliminated NE/E neurons from the rostral portion of the A1 cell group and the caudal 2/3 of the C1 group in the ventrolateral medulla. The DSAP-induced destruction of these PVH-projecting hindbrain neurons selectively abolished glucoprivic feeding, whereas feeding in response to blockade of fat oxidation and overnight food deprivation were not impaired. Rats treated in this way with DSAP did not even increase their intake in response to 2DG when a palatable liquid was infused directly into their mouths, obviating the need for appetitive behavioral responses (38).

Follow-up studies confirmed the essential role of specific of hindbrain NE and E neurons in glucoprivic feeding and, in addition, indicate that the CA subpopulation critical for feeding coexpresses neuropeptide Y (NPY). Hindbrain expression of genes for NPY and DBH is increased after glucoprivation (39, 40). Furthermore, glucoprivic feeding is attenuated by simultaneous selective silencing of Npy and Dbh in A1/C1 cell groups, whereas silencing of each gene separately is less effective or ineffective (41). The unique importance of NPY expression specifically by hindbrain E and NE neurons is reinforced by the fact that global Npy knockout significantly impairs glucoprivic feeding in mice (19, 20, 42, 43), and silencing of Npy/Dbh in the ventrolateral medulla impairs the feeding response in rats, but lesion of the hypothalamic NPY/Agouti-related peptide (AGRP) neurons (44, 45) or deletion of NPY/AGRP in hypothalamic neurons (44, 45) fails to affect glucoprivic feeding. Thus, although hypothalamic NPY/AGRP neurons are doubtless of importance in day to day control of food intake they are not necessary for counterregulatory feeding in response to glucose deficit.

The importance of both DBH and NPY for elicitation of glucoprivic feeding supports the hypothesis that ventral hindbrain NE or E neurons are most critical for mediating this response. However, a conflicting and well-known finding is that lesion of the area postrema in the dorsomedial hindbrain also impairs the feeding response to both systemic (46) and central glucoprivation (47). This lesion requires further analysis due to two factors. One is that not all investigators have found deficits in glucoprivic feeding in area postrema-lesioned rats (48). A second but related factor is that the NTS is a highly complex structure, containing afferent terminals from the entire visceral compartment and is known for its involvement in multiple aspects of food and water intake and gastrointestinal function (49–51). Therefore, the glucoprivic feeding deficit in area postrema-lesioned rats might well be a nonspecific deficit. Variability of the lesion effects may be due to variability in the amount and locus of damage to the underlying NTS, which always accompanies area postrema lesions. Nonetheless, this body of work suggests that cell groups A2 and C2 should not be ruled out as contributors to the glucoprivic control. Indeed, because sites sensitive to glucoprivation appear to be coextensive with CA cell groups throughout the medulla (22, 23), it is possible that multiple CA cell groups are glucose sensing or that non-CA glucose sensors, including those from visceral organs and/or ventricular lining, form a diffuse network for downstream activation of the critical CA neurons (52).

Adrenal medullary secretion

The adrenal medullary response to glucoprivation is triggered when sympathetic preganglionic neurons innervating the adrenal medulla are activated by descending projections from the brain. Adrenal medullary E secretion elicits robust hepatic glycogenolysis that rapidly elevates blood glucose concentration. It also mobilizes the gluconeogenic substrates and stimulates lipolysis. Adrenal medullary E facilitates lipid uptake by tissue-specific regulation of lipoprotein lipase. It suppresses insulin secretion by α2 adrenergic receptor activation and stimulates glucagon secretion by β2 receptor activation. E also increases cardiac output, facilitating delivery of blood (and glucose) to the brain.

Intraspinal DSAP injection results in retrograde destruction of spinally projecting hindbrain E and NE neurons in the rostral 1/3 of cell group C1 and results in loss of adrenal medullary secretion and hyperglycemic responses to glucoprivation. However, feeding in response to glucose deficit is not impaired by intraspinal DSAP (27, 37). Conversely, PVH DSAP injection, which abolishes glucoprivic feeding, destroys NE/E neurons in the caudal portion of A1/C1 and does not impair the adrenal medullary hyperglycemic response. Thus, the glucoprivic control of these two key responses is mediated by different subgroups of hindbrain NE/E neurons.

Although adrenomedullary secretion is often thought of as being controlled in a rather nonspecific manner, that assumption now appears to be incorrect. Ventral hindbrain neurons that control release of adrenal NE are different from those controlling release of E (53), as shown by differences in their conduction velocities. In addition, E, as opposed to NE, is preferentially released from the adrenal medulla in response to glucoprivation (54). Furthermore, electrophysiological examination of the C1 area has identified neurons (of undetermined phenotype) that respond differentially to hypoglycemia but are not baroreceptive. These neurons have slowly conducting action potentials characteristic of C1 neurons (55) and of neurons that control adrenal medullary E secretion in response to 2DG (53). Finally, counterregulatory functions are impaired by direct injection of DSAP into the rostral ventrolateral medulla in the area of C1 (56), as they are by intraspinal DSAP injection (37). Together, these findings strongly support the hypothesis that neurons in the rostral ventrolateral medulla (probably C1 neurons) control the counterregulatory response of the adrenal medulla to glucose deficit and that the critical group is distinct from the C1 neurons involved in cardiovascular control and from NE/E neurons controlling glucoprivic feeding.

Corticosterone secretion

The corticosterone response to glucose deficit shifts metabolism of nonneural tissues away from glucose utilization by mobilizing fatty acids and protein from tissue sites and promoting fatty acid oxidation, gluconeogenesis, and, over the longer term, ketogenesis. Evidence that NE/E neurons stimulate CRH secretion is substantial and compelling. Both NE and E neurons innervate the parvocellular area of the PVH (57, 58), which contains CRH neurons. Injections of NE and E into the PVH increase corticosterone levels (32). Secretion of CRH and CRH gene expression are both increased by NE or E injections into the PVH in vivo or by addition of NE to preparations of cultured CRH neurons (59–61).

Corticosterone levels are elevated by systemic glucoprivation (62), which also elevates NE turnover in the hypothalamus (33, 35). Recent work suggests that the glucoprivation-induced NE-evoked stimulation of CRH secretion requires activation of MAPK signaling cascades (59). Corticosterone levels are increased by localized injection of the glucoprivic agent, 5TG, into hindbrain sites contiguous with or overlapping NE and E cell groups (22). Finally, CRH secretion also is increased by NPY (32, 63), which is coexpressed by nearly all ventral medullary E neurons and some NE neurons projecting to the PVH (64, 65).

Corticosterone secretion can be evoked by many and diverse stimuli. Previous work has suggested that neurons controlling corticosterone secretion are “category specific” (66). Thus, a pertinent question is whether CA neurons that control CRH secretion are selective in their response to glucoprivation or respond indiscriminately to many stress modalities. DSAP lesion experiments clearly indicate that NE/E neurons exert a stimulus-specific control of corticosterone secretion in response glucose deficit (67). PVH DSAP injections profoundly reduce the corticosterone response to glucoprivation. They also eliminate glucoprivation-induced increases in expression level of CRH heteronuclear RNA and c-fos mRNA in the PVH, without damaging CRH neurons themselves and without impairing the corticosterone response to swim stress or the circadian rhythm of corticosterone secretion. These results demonstrate that glucoprivation-induced corticosterone secretion is mediated by hindbrain NE/E innervation of the PVH and that circadian cycles and swim stress influence corticosterone secretion through separate pathways (67). Different subpopulations of NE/E neurons with different afferent and efferent projections, as well as non-CA neurons, also elicit CRH secretion in response to different stress modalities. For example, some data suggest that NE or E projections to the amygdala, but not the PVH, mediate corticosterone secretion in response to restraint stress (68).

Understanding of the control of corticosterone secretion in glucose counterregulation is important, in part because of evidence suggesting a role for this corticosterone in hypoglycemia-induced autonomic failure (HAAF) (69–71), a potentially fatal syndrome in diabetic patients on insulin therapy (9). HAAF occurs after a previous glucoprivic episode and comprises suppression or desensitization of counterregulatory responses to glucose deficit. The pathogenesis of HAAF may be multifactorial, but significantly, this syndrome can be reproduced by systemic or central administration of high corticosterone doses. Therefore, one hypothesis is that high levels of corticosterone secretion during hypoglycemia densensitize or suppress counterregulatory responses to a subsequent glucoprivic episode. HAAF is a limiting side effect of successful insulin therapy but in other situations may be part of a balancing act to conserve strategic resources, maintain brain function, and to allow the development of more slowly adapting responses to long term glucose shortage, such as ketogenesis and gluconeogenesis.

Reproductive hormone secretion

NE is a major neurotransmitter controlling release of GnRH and the resulting released of LH (72). NE neurons provide afferent input to adrenoreceptor-expressing GnRH neurons (58, 73, 74). In addition, the cell bodies of NE neurons in caudal A1 and in A2 cell groups express ovarian steroid receptors (75, 76). Moreover, increased NE activity in the preoptic area increases LH secretion, and this NE effect can also be obtained in cultured cells (77). Correlational studies reveal increased NE turnover and release in preoptic area at the time of the preovulatory LH surge (78, 79). Reproduction is behaviorally and metabolically costly. Previous reports indicate that glucoprivation suppresses reproductive function: frequency of LH secretory pulses is reduced (80–82), estrous cycle length is extended, and reproductive behavior is decreased by glucoprivation (83–85). 2DG also induces estrogen receptor-α expression in cell groups A1 and A2 of ovariectomized female rats (86).

The observations described above support a role for NE/E neurons in stimulating secretion of reproductive hormones. Therefore, the apparent importance of hindbrain E/NE neurons for suppression of reproduction by glucoprivation comes as something of a surprise. Nevertheless, a pivotal role for hindbrain E/NE neurons in glucoprivation-induced suppression of estrous cycles has been demonstrated using hypothalamic DSAP injections (87). Chronic systemic glucoprivation significantly increases estrous cycle length and, in fact, appears to arrest the cycles completely in controls. However, DSAP-injected rats have normal estrous cycles that continue in spite of chronic glucoprivation. Thus, DSAP appears to lesion neurons that are required for inhibition of reproductive function during chronic glucose deficit but that are not required for normal estrous cycling. DSAP injections into the arcuate nucleus also alter glucoprivic control of hypothalamic galanin-like peptide mRNA and LH secretion in the adult male rat but do not influence male sex behavior (88).

Although these results strengthen the link between NE neurons, GnRH/LH secretion, and hindbrain effects of glucoprivation, they also suggest that the NE or E neurons involved in stimulatory control of GnRH secretion under basal conditions may be distinct from those that suppress reproductive hormone secretion and estrous behavior under glucoprivic conditions. Possibly the glucoprivic control is exerted by an indirect NE pathway. Along these lines, it has been shown that the various components of the hypothalamopituitary adrenal axis, including CRH and corticosterone, suppress GnRH secretion (89, 90), suggesting that NE may act through the hypothalamopituitary adrenal axis to suppress reproductive function during glucoprivation. Interestingly, the suppression of LH pulses by stress and hypoglycemia, which can stimulate prostaglandin E2 secretion in the brain, is blocked by inhibition of prostaglandins synthesis (91, 92). Similarly, NE stimulates prostaglandin synthesis in brain tissue, both in vivo and in vitro (93, 94, for reviews see Ref. 95). Thus, prostaglandins also may be a common mediator of LH pulse inhibition and could be downstream of NE/E activation by glucoprivation. Other mechanisms have also been suggested, such as increased opioid signaling and the involvement of γ-aminobutyric acid release in the septopreoptic region of the brain resulting from hindbrain glucose deficit (96, 97).

Glucagon secretion

Glucagon is a difficult hormone to study, because it is under autocrine/paracrine pancreatic control, gut endocrine control, and is also controlled by both sympathetic and parasympathetic pancreatic innervations (98–100). Loss of one control, therefore, may not disable the response to glucose deficit. Our own investigations indicate that plasma glucagon levels are elevated by localized glucoprivation of specific hindbrain sites (22). The distribution of effective sites is very similar, if not identical, to those at which 5TG injections elicit feeding and corticosterone responses. However, PVH DSAP injection does not reduce the glucagon secretion in response to systemic glucoprivation (Ritter, S. and T. T. Dinh, unpublished data). The glucagon response in rats given intraspinal DSAP has not been examined. Glucagon secretion is a crucial counterregulatory response (9, 100). Therefore, more work is required to determine hindbrain mechanisms that contribute to glucagon secretion during glucose deficit and whether these mechanisms interface with reported control of glucagon secretion by the hypothalamus (24).

Are Hindbrain NE/E Neurons Themselves Glucoreceptive?

A number of cell markers implicated in glucose-sensing functions in other brain regions or tissues has been identified in the hindbrain, including glucokinase, AMP kinase (AMPK), K_ATP channel markers, GLUT1–GLUT4, and the monocarboxylate transporter. Electrophysiological studies in dorsal and ventral hindbrain have found neurons that respond to changes in glucose, although the neurochemical phenotype of such neurons is not known. Unfortunately, it is not been clearly established whether any NE or E neurons are directly glucose sensing or whether they contain any of these glucose-sensing mechanisms. However, preliminary findings and convergent circumstantial evidence are promising and consistent with the possibility that some hindbrain NE/E neurons may be glucose sensing. These findings will be reviewed briefly here.

Electrophysiological studies

Glucose-sensing neurons fall into two major categories: those that are excited by high glucose [glucose excited (GE) neurons] and those that are inhibited by high glucose [glucose inhibited (GI) neurons] (101). Glucose-sensing neurons of both types, but of unknown neurochemical phenotype, have been detected electrophysiologically in the dorsal vagal complex (102, 103). In a recent study of 123 neurons in the dorsal vagal complex (104), approximately 12% were GI type, 9% were GE, and 80% were NR. Based on Fos data (36), which indicates that a small percentage of A2 neurons express Fos in response to glucose deficit, and these electrophysiological results, only a few A2 neurons could be expected to be activated by glucose reduction. Nevertheless, it is conceivable that some neurons excited by glucose deficit include those few A2 neurons that express Fos after glucoprivation or that A2 neurons are GE type. Glucose-responsive neurons have also been detected in the ventral hindbrain (105), where many NE/E neurons involved in counterregulatory responses to glucose deficit reside (37, 41, 56, 105). As discussed previously in this review (see Adrenal medullary secretion), convergent evidence is highly suggestive that some C1 neurons may in fact be glucose sensing, but a sufficiently rigorous analysis of this possibility is not yet available.

The K_ATP channel

This ATP-gated potassium channel is implicated in the function of some glucose-sensing cells and in hypoglycemia counterregulation (106, 107) and is thought to operate in conjunction with glucokinase (108, 109). One study, using single-cell RT-PCR of laser microdissected tyrosine hydroxylase-immunoreactive neurons, has reported the presence of gene transcripts for sulfonylurea 1, a K_ATP channel subunit, in A2 neurons (110). With this exception, K_ATP channel has not been implicated in hindbrain counterregulatory functions, and there is reason to be cautious about assuming that the K_ATP channel is a reliable indicator of glucose-sensing cells. Electrophysiological results suggest that only a small percentage of cells in the dorsal vagal complex are glucose sensing, although results indicate a widespread distribution of K_ATP channels in this area (104). Furthermore, the K_ATP channel has been identified in some GE neurons, but it is also present in some glucose-nonresponsive neurons and so far appears to be absent in GI neurons (104).

Glucokinase

Glucokinase is an inducible enzyme, sensitive to metabolic state. Like other hexokinases, it phosphorylates glucose at the initial step in glycolysis but only in certain tissues, such as hepatocytes, pancreatic β-cells, and some brain cells (104, 109, 111–113). Electrophysiological experiments combined with single-cell RT-PCR of dorsal hindbrain neurons have found glucokinase in both GE and GI neurons but not in neurons nonresponsive to glucose or in cells expressing the glial cell marker, glial fibrillary acidic protein (104). Its distribution in the dorsal vagal complex is described as “widespread but sparse” (104). Immunohistochemical analysis has revealed that glucokinase is also present in ependymocytes lining all ventricles, in endothelial cells, and in serotonergic neurons in the hindbrain (114). In these areas, it is coexpressed with GLUT1 and GLUT2 and microtubule-associated protein 2. In addition, laser microdissected A2 neurons appear to express a glucokinase transcript that is inducible by insulin-induced hypoglycemia (110).

Pharmacological experiments using glucokinase inhibitors have been published sporadically over more than 30 years. Early pharmacological experiments demonstrated that lateral (115, 116) or 4V (117) administration of 40 μg of alloxan, a glucokinase inhibitor and β-cell toxin (118, 119), produced long-term deficits in glucoprivic feeding induced by systemic 2DG, centrally administered 5TG, or insulin-induced hypoglycemia. Alloxan's effect was preventable by coadministration of D-glucose but not L-glucose (117, 120). Importantly, the hyperglycemic response to glucoprivation was not altered by alloxan. In contrast to these results, 4V injection of much lower alloxan doses (10, 15, and 20 μg) actually stimulated feeding but, like the higher doses, did not alter blood glucose (117). From these studies, it can be inferred that feeding and adrenal medullary responses to glucoprivation rely on different sensing mechanisms and that glucokinase is of special importance for control of glucoprivic feeding.

Recent work using the glucokinase inhibitor, glucosamine (121), has shown that lateral or 4V injection of glucosamine evokes feeding but, as with alloxan, does not alter blood glucose. In addition, glucosamine-induced enhancement of feeding was abolished by PVH/medial hypothalamic injections of DSAP, suggesting that glucosamine's effect in the hindbrain is mediated by the hypothalamically projecting NE/E neurons. Finally, 4V injection of glucosamine increased Fos expression in NE/E cell populations that mediate key glucoregulatory responses. Other work shows that third ventricular injection stimulates feeding and induces Fos expression in hypothalamic NPY and orexin neurons (122) and that brief refeeding in rats after short-term food restriction lowers glucokinase expression in both the hypothalamus and hindbrain, compared with rats that were not refed (123). Thus, glucokinase is a likely participant in NE/E activation necessary for the counterregulatory feeding response but may also be involved in other hindbrain controls of food intake.

AMP kinase

Another nutrient monitoring mechanism that may participate in hindbrain glucose sensing is AMPK. AMPK is an intriguing candidate for nutrient sensing, because its activity is regulated by cellular ATP/ADP ratio and because its activation suppresses cellular ATP consumption and stimulates ATP production (124). In basomedial hypothalamus, activation of AMPK mimics the effect of decreased glucose on NPY-expressing neurons, which constitute about 40% of GI neurons in that area (125). Both food deprivation (126) and hypoglycemia (127) increase hypothalamic phosphorylated AMPK (pAMPK), the activated form of AMPK. In the NTS, pAMPK is increased by food deprivation and hindbrain administration of compound C, which inhibits AMPK phosphorylation and reduces food intake (128).

We found that 4V administration of compound C also decreases feeding induced by systemic 2DG (129). Systemic 2DG increases phosphorylation of AMPK in tissue punches containing the A1/C1 area of the ventrolateral medulla of control rats but not in tissue from the same area in which A1/C1 neurons were destroyed by PVH DSAP injections. These results are consistent with the possibility that the NE/E neurons damaged by DSAP express pAMPK in response to glucose deficit. Nevertheless, in the absence of a reliable antibody for pAMPK, we were not able to visualize the detected increase in pAMPK in NE/E neurons themselves. In contrast to effects of food deprivation on dorsal hindbrain pAMPK, reported by Hayes et al. (128), glucoprivation did not increase pAMPK in the dorsal hindbrain punches in our experiment. Food deprivation may be a broader spectrum stimulus than glucoprivation, thus activating more cells in the NTS, which is heavily influenced by the status of the gastrointestinal tract. Alternatively, differences in tissue sampling may account for the differing results. Further analysis is needed to determine whether this is a real difference between food deprivation and glucoprivation.

Sodium-glucose cotransporter (SGLT)

Another dimension of glucose sensing that will be important to examine further in the brain is “metabolism-independent sugar sensing,” as has been demonstrated in lateral hypothalamic orexin neurons (130). This mechanism is mediated by SGLT and is responsive to glucose and structural analogs but is independent of glucokinase (130, 131). In cells using this mechanism, 2DG mimics, rather than antagonizes, the effect of glucose. It has been proposed (3) that this mechanism provides a system for responding to extracellular glucose levels without interfering with controls of intracellular fuel utilization. Results of one study in the hindbrain suggest a role for this mechanism in stimulation of food intake (132). In that study, phlorizin, an SGLT antagonist, injected into the 4V evoked feeding but not a hyperglycemic response, whereas 5TG injected via the same route stimulated both feeding and blood glucose responses. Taken together, pharmacological studies using alloxan, glucosamine, and phlorizin suggest that hindbrain glucoregulatory mechanisms may use multiple signaling elements and, further, that different sensing mechanisms control feeding and blood glucose responses.

Are Hindbrain-Mediated Counterregulatory Responses Required for Daily Feeding, Maintenance of Body Weight, and Energy Homeostasis?

Our results clearly indicate that neither PVH DSAP lesions that abolish glucoprivic feeding and corticosterone responses nor intraspinal DSAP injections that impair the sympathoadrenal response are lethal under normal laboratory conditions of ad libitum feeding or during brief periods of food deprivation (e.g. overnight) or glucoprivation (37, 67). Adrenal medullary responses and local islet cell glucagon and insulin responses remain intact in PVH DSAP rats, and these responses continue to ameliorate glucose deficits. Spontaneous daily feeding appears to be normal with respect to both diurnal distribution and amount in DSAP-lesioned rats, consistent with the fact that premeal glucose decline accompanies some, but not all, meals (133) and is not required for meal initiation. Thus, it appears that even if the lesioned NE/E neurons include some that normally participate in control of daily feeding or metabolic homeostasis under stable laboratory conditions, they are not required for this role.

Although NE/E neurons do not appear to be required for control of daily feeding in intact rats, it is important that such a role not be prematurely disregarded. It seems unlikely that such a highly conserved, powerful, and multifaceted system as we see in the CA-ergic glucoregulatory neurons would have evolved solely to deal with rare occurrences of acute and profound glucose deficit, which in a culture of food abundance would seem unlikely to occur except in cases of an overdose of exogenous insulin. It seems more likely that this system exerts more subtle actions to avert hypoglycemia by promoting nutrient intake and/or disposition under nonglucoprivic conditions. One observation along these lines is that PVH DSAP-injected rats gain more weight than SAP controls over time (37). Possibly this effect is associated with the over expression of NPY and AGRP genes in the hypothalamus that has been observed in DSAP-injected rats (134).

How do the Hindbrain CA Neurons Interact with Other Brain Regions Involved in Feeding Behavior and Energy Homeostasis?

Results summarized in this review suggest that hindbrain neurons are essential for key glucose counterregulatory responses. However, other investigators make similar claims for hypothalamic cells (24, 135–138). Thus, it is important to begin to ask how the glucose-sensing mechanisms in the hindbrain interface with hypothalamic mechanisms. A recent review has provided an excellent discussion of the range of anatomical pathways that could potentially provide the anatomical interface between hindbrain, forebrain, and peripheral sites that contribute to glucoregulation (4). Hindbrain CA neurons are unique in their widespread innervation of the central nervous system (26, 139), including various hypothalamic nuclei and the spinal cord. Similarly, the hypothalamus densely innervates hindbrain (2), particularly the dorsal vagal complex, and peripheral signals converge in the dorsal vagal complex as well. Certainly, these anatomical features indicate that the systems are in communication. But what is the nature of the interactions between various glucose-sensing sites? Are there redundant or partially redundant counterregulatory systems in the periphery, hindbrain, and hypothalamus? Or do the systems arising from these sites serve entirely different functions? A few speculations seem appropriate here.

Clearly, glucoregulatory systems must be (and are) capable of employing many and diverse glucoregulatory responses in a coordinated and timely fashion, synchronously in response to an acute and profound glucoprivic challenge and sequentially (or individually) during a more gradual or more limited reduction in glucose availability. Our work has focused on neurons that respond rapidly to acute glucose deficit. However, a system that responds to gradual declines in glucose has been proposed and demonstrated (140). Indeed, recent reports suggest that gradual declines in glucose availability activate responses that are initiated by peripheral glucose sensors and are transmitted to the NTS by spinal and vagal sensory neurons (140). Systems such as these, which presumably remain intact in PVH DSAP-lesioned rats, may partly account for the ability of these rats to maintain normal spontaneous feeding in the absence of the hindbrain CA mediators of glucoprivic feeding (37).

It also seems likely that glucose-sensing neurons in various brain sites may differ in their capacity to integrate multiple endocrine and metabolic signals. Present data suggest that counterregulatory responses to precipitous reduction of glucose availability are not integrative in nature. That is, the sensors that mediate these responses do not seem to be sensitive to or modulated by other metabolic or ingestive signals. At least with response to nutrient sensitivity, the counterregulatory system that responds to glucose deficit does not appear to be integrative with respect to fat, the other major substrate for energy metabolism. For example, feeding responses to systemic or central glucoprivation are attenuated by D-glucose but not lipid infusion (141, 142). DSAP lesions that eliminate the response to glucoprivation do not impair feeding in response to blockade of fatty acid oxidation with mercaptoacetate (37). Lesion of visceral afferent fibers by abdominal vagotomy or capsaicin treatment abolishes mercaptoacetate-induced feeding but does not impair glucoprivic feeding. Mercaptoacetate and 2DG induction of Fos (143) and hypothalamic orexigenic peptide gene expression (144), as well as lesion of sites expressing Fos in response to 2DG or MA (145–150), all indicate that these two antimetabolic drugs activate different neural pathways. Furthermore, high chronic doses of exogenous leptin, which signal replete endogenous fat stores and significantly reduce spontaneous feeding, do not eliminate glucoprivic feeding (151). The advantages of a nonintegrative system is that signal strength is not diminished by competing messages when that system must take charge of autonomic, behavioral, and endocrine controls in a dire emergency.

A glucoregulatory system contributing to control of daily feeding may be more integrative. An advantage of any integrative system is that small reductions in a variety of convergent signals would be additive, rather than requiring a dramatic increase or decrease in a single nutrient to achieve a threshold for responding. Hypothalamic NPY/AGRP neurons may be prototypical of such a system. In contrast to the stimulus specificity of the counterregulatory system, the GI NPY/AGRP neurons in the hypothalamus not only respond to glucose but also to fatty acids, leptin, insulin, and ghrelin (for review, please see Ref. 101). Moreover, NPY/AGRP neurons also are responsive to food deprivation (152–154), a stimulus that remains effective for evoking food intake in DSAP-lesioned rats that do not respond to glucoprivation (37). Although NPY/AGRP neurons may be recruited by hindbrain CA neurons during glucose deficit (134, 144), they are not required for either feeding or hyperglycemic responses to glucose deficit (44, 45). Their major main function may be related to control of daily feeding and adjustment of overall energy intake to match utilization.

A Technical Challenge to Untangling the Diverse Roles of the Brain's Multiple Glucose-Sensitive Cells

It is clear that there are glucose-sensing neurons and potential transduction mechanisms expressed in various forebrain and hindbrain locations as well as in peripheral tissues. It is also becoming clear that not all glucose-sensing cells have the same functions. To understand the complex process of glucoregulation and overall control of energy balance, it will be necessary to decipher the specific roles of the various systems. To do this, one technical problem that will have to be overcome is how to limit the spread (or alternatively, accurately measure the spread) of injected or dialyzed substances within the brain or ventricular system. Failure to adequately account for intraventricular spread has been a point of confusion and a source of contradictory results over the years with respect to injections of glucoprivic agents and glucose into hypothalamic and hindbrain tissue sites and into forebrain and hindbrain ventricles. Two early studies (13, 17) were not able to elicit feeding by injections of 2DG into the hypothalamus or other forebrain tissue sites at doses that were effective in the lateral ventricle. Similarly, in the mapping study discussed above (22), feeding, cort, glucagon, and glucose were elevated by local unilateral injections of 5TG into hindbrain sites (12–24 μg in 200 nl), but the same or higher doses (24–48 μg) injected at sites throughout the hypothalamus were ineffective. For comparison, another study reported that 5TG injected into the ventromedial hypothalamus elicited a feeding response (25). However, in that study, 120 μg of 5TG in 1 μl was injected bilaterally for a total of 240 μg in 2000 nl. Thus, the total dose of 5TG injected into the ventromedial hypothalamus was 10 fold greater, and the volume was 100 fold greater than the intraparenchymal dose and volume found to be effective in the hindbrain mapping experiments, and the dose was more than double effective lateral or 4V 5TG doses (90 μg) (21).

Some studies have used microdialysis of 2DG or glucose into the hypothalamus to localize counterregulatory controls (24). Although this is a reasonable approach, it is particularly difficult to restrict the distribution of an antimetabolite in the brain over the extended time course of the infusion. In addition, the dialysis probes are quite large compared with the dimensions of the hypothalamic nuclei. The obvious question raised by these examples is whether responses associated with these large doses and volumes can legitimately be used to claim localized action of injected antimetabolite, or whether the effects are achieved because the injected substances are accessing the ventricular system and diffusing to more distant sites. This is an old problem but nonetheless a critical one that is rarely addressed when interpreting experimental results. The issue of localization for injected agents must be addressed rigorously if we are to correctly understand the roles of various glucose sensors in glucoregulation.

One approach to the problem of localization is to include and compare multiple injection sites in experimental designs, choosing control sites carefully, for example, with similar proximity to the ventricle. Establishing sites that are negative to the application of glucoprivic agents should be considered of equal importance to identifying positive sites, because negative sites serve to delimit the area over which an agent can be said to act (22, 23). Obviously, this approach becomes progressively more powerful when very small injection volumes and agent concentrations are used. Another approach that has been effective in separating forebrain and hindbrain responses is to occlude the cerebral aqueduct with a silicone plug and test effectiveness of an injected substance downstream and upstream of the blockade (21).

Synopsis

We have reviewed hindbrain counterregulatory mechanisms that protect and restore glucose supply to the brain. Because glucose is essential for brain function and survival, glucose-sensing underlying counterregulatory responses plays a role that is distinctly different from glucose-sensing systems that control overall energy balance in the interest of a variety of physiological needs. In contrast to the latter, it does not require an integrative network for sampling levels of multiple ingestive hormones or circulating nutrient levels, adjusting behavior and metabolic activity to photoperiod, or for finessing level of appetite to match energy output. Rather, the glucose counterregulatory system must respond to and prevent dangerously low brain glucose levels and must do so irrespective of body fat stores or the presence of leptin. We proposed that this system must maintain a rapidly responding, direct line to its effector sites and must be capable of commandeering autonomic, behavioral, and endocrine controls centered in the hypothalamus or other brain regions. We present evidence supporting the hypothesis that hindbrain NE/E neurons fill this important niche. We review data illustrating the functional diversity of this neuronal population and suggesting that specific NE/E subgroups mediate particular counterregulatory functions, whereas some are not involved in glucoregulatory functions at all. We propose areas where additional work is urgently needed to further support or refute these proposed hypotheses and to further understand the complex nature of overall glucoregulation.