Central Mechanisms of Glucose Sensing and Counterregulation in Defense of Hypoglycemia

Abstract

Glucose homeostasis requires an organism to rapidly respond to changes in plasma glucose concentrations. Iatrogenic hypoglycemia as a result of treatment with insulin or sulfonylureas is the most common cause of hypoglycemia in humans and is generally only seen in patients with diabetes who take these medications. The first response to a fall in glucose is the detection of impending hypoglycemia by hypoglycemia-detecting sensors, including glucose-sensing neurons in the hypothalamus and other regions. This detection is then linked to a series of neural and hormonal responses that serve to prevent the fall in blood glucose and restore euglycemia. In this review, we discuss the current state of knowledge about central glucose sensing and how detection of a fall in glucose leads to the stimulation of counterregulatory hormone and behavior responses. We also review how diabetes and recurrent hypoglycemia impact glucose sensing and counterregulation, leading to development of impaired awareness of hypoglycemia in diabetes.

Essential Points

• Glucose-sensing neurons have been identified in the hypothalamus, brain stem, dorsal motor nucleus of the vagus, nucleus accumbens, amygdala, paraventricular thalamus, prefrontal cortex, and the hippocampus

• Mechanisms used by glucose-excited, glucose-sensing neurons to detect changes in glycemia include detection of changes in the ATP/ADP ratio, mitochondrial function, sodium glucose cotransport, and sweet receptors

• Mechanisms used by glucose-inhibited, glucose-sensing neurons to detect changes in glycemia include detection of changes in the ATP/ADP ratio, AMP kinase activity, and the opening of tandem-pore K⁺ channels

• Brain regions with glucose-sensing neurons are highly connected and project to downstream areas that regulate secretion of insulin, catecholamines, glucagon, cortisol, and GH

• Recurrent episodes of hypoglycemia, particularly in patients with diabetes who are treated with insulin or insulin secretagogues, lead to impaired awareness of hypoglycemia

• The mechanisms responsible for the development of impaired awareness of hypoglycemia remain uncertain, but upregulation of brain glucose transport, use of alternative fuels, and altered hypothalamic signaling have been hypothesized

Glucose homeostasis requires an organism to rapidly respond to changes in plasma glucose concentrations. This is particularly true in the face of a fall in plasma glucose, because the difference between normoglycemia and life-threatening hypoglycemia can be as small as 30 mg/dL (1.7 mmol/L). Iatrogenic hypoglycemia as a result of treatment with insulin or sulfonylureas is the most common cause of hypoglycemia in humans and is generally only seen in patients with diabetes who take these medications. Hypoglycemia is extremely common in the life of patients with diabetes. A recent global study that included >27,000 insulin-treated patients found that >80% of the patients with type 1 diabetes and nearly 50% of the patients with type 2 diabetes experienced any hypoglycemia in a month of prospective monitoring (1). In the same study, 14% of patients with type 1 diabetes and 9% of patients with type 2 diabetes experienced severe hypoglycemia during the same month, which by definition only includes episodes associated with sufficient neuroglycopenia to require the assistance of another person to treat the low blood glucose. Hypoglycemia is also a costly complication. Between 2007 and 2011, emergency room visits for hypoglycemia are estimated to have cost >$600 million in the United States alone (2).

The first central response to a fall in glucose is the detection of impending hypoglycemia by glucose-sensing neurons in the hypothalamus and other regions. This detection is then linked to a series of neural and hormonal responses that serve to prevent the fall in blood glucose and restore euglycemia. In this review, we first discuss the current state of knowledge about central glucose sensing and how detection of a fall in glucose leads to the stimulation of counterregulatory hormonal and behavioral responses. We then review how diabetes and recurrent hypoglycemia impact glucose sensing and counterregulation. Finally, we consider the mechanisms that contribute to the development of impaired awareness of hypoglycemia in diabetes, a situation that increases the risk of experiencing severe hypoglycemia by more than sixfold (3). The purpose of this review is to provide scientists and clinicians with an understanding of the complexity of glucose counterregulation.

Glucose-Sensing Neurons

All neurons use glucose as a source of energy and are silenced at very low glucose concentrations, but specialized neural populations also use glucose as a signal (4, 5). These glucose-sensing neurons modulate their firing rate in response to changes in extracellular glucose concentrations (6). Glucose-sensing neural populations can be broadly divided into two groups: glucose-excited (GE) and glucose-inhibited (GI) neurons (7). As the name suggests, GE neurons increase their firing rate in response to increasing extracellular glucose, and their firing rate decreases as glucose concentrations fall. In contrast, the firing rate of GI neurons decreases as glucose levels rise and increases as glucose concentrations fall. Glucose-sensing cells are found outside the central nervous system (CNS), particularly in the pancreas but also in the mouth, gastrointestinal tract, portal vein, and carotid body (8). These peripheral sensors provide signals via polysynaptic pathways involving sensory and vagal afferents to multiple CNS regions, many of which also contain glucose-sensing neurons. Collectively, peripheral and central glucose sensors likely provide a distributed system to detect and respond to altered glucose levels.

Extracellular glucose concentrations in the CNS are much lower than those in the systemic circulation. From microdialysis studies, it is known that interstitial CNS glucose concentrations range from ∼2.5 mM in the fed state to 0.7 mM in fasted animals (9–11). In human studies, brain glucose concentrations have been shown by magnetic resonance spectroscopy to be ∼25% of plasma glucose values under a range of plasma glucose concentrations (12). However, some CNS regions lie outside the blood–brain barrier. For circumventricular organs, such as the median eminence and area postrema (AP), the extracellular glucose concentration is likely to be closer to plasma glucose levels (13). Glucose diffusion from the cerebrospinal fluid into the brain is also variable, leading to differences in extracellular glucose concentrations across brain regions (14). Microdialysis studies suggest that brain interstitial glucose levels may also change with neural activity. For example, extracellular glucose in the hippocampus decreased during a spatial memory task (11). Additionally, there is evidence that glucose transport into the CNS is altered by recurrent hypoglycemia, feeding, fasting, and high glucose (15–17). Finally, CNS glucose levels vary between strains, ages, and species (18, 19). Despite the variability described above, neurons in most brain regions are unlikely to be exposed to glucose concentrations >5 mM, with the exception of cells close to circumventricular organs where glucose levels may be higher.

Locations of and neurotransmitters released by glucose-sensing neurons

Populations of glucose-sensing neurons have been described in multiple areas of the brain [Table 1 (7, 20–35); Fig. 1]. Evidence for the ability of these populations to sense glucose changes comes from several sources, as discussed in subsequent paragraphs. These include expression of early response genes such as c-fos induced by systemic glucose, electrophysiological studies in vivo and in ex vivo slices demonstrating modulation in firing rates with changes in glucose, and expression of putative glucose sensors such as glucose transporters (GLUTs), specific ion channels, and components of the glucose metabolic pathway such as glucokinase. The preclinical studies used to investigate CNS glucose sensing and counterregulation have evolved from electrical stimulation and lesions of anatomical regions to targeted, reversible modulation of genetically defined neural populations. However, many of the neural populations are heterogeneous, comprised of glucose-sensing and nonsensing populations or a combination of GE and GI neurons, so stimulating them may reveal a dominant phenotype but miss the distinct roles of specific subpopulations.

Table 1.

Summary of the Proportion and Markers for Glucose-Sensing Neurons in Specific Brain Regions

Area	Proportion of GE Neurons	Ref.	Marker	Ref.	Proportion of GI Neurons	Ref.	Marker	Ref.
PVH	57%	(20)			42%	(20)
Medial ARC	4%	(21)			14%	(21)	NPY (40%)	(22)
							GHRH (80%)	(23)
Lateral ARC	13%	(21)	POMC (51%)	(24)	1%	(21)
VMH	14%	(7)	Glucokinase (64%)	(7)	3%	(7)	Glucokinase (43%)	(7)
LH	38%	(25)	MCH (83%)	(26)	19%	(25)	Orexin (95%)	(26)
							NPY (70%)	(27)
DMH	11%	(28)	GABA (60%)	(29)	15%	(28)	GABA (30%)	(29)
AP	11%	(30)			12%	(30)
PBN							CCK (35%)	(31)
NTS	35%	(32)	GABA (40%)	(33)	21%	(32)	GABA (33%)	(33)
			TH	(34)			GLUT2	(35)
DVC	22%	(20)			18%	(20)

The proportions of GE and GI neurons in defined brain regions and markers for specific populations are shown. The percentage of each population with glucose-sensing properties is shown in parentheses.

Abbreviation: DVC, dorsal vagal complex.

Figure 1.

Sites of glucose-sensing neurons in mouse brain. GE and GI neurons are found in many brain regions. In the hypothalamus, these include the PVH, ARC, DMH, VMH, and LH. Outside the hypothalamus, glucose-sensing neurons have been reported in the NAc, amygdala, LC, and PBN. In the brain stem, the AP, NTS, and DMV also have glucose-sensing neurons.

The most extensively studied glucose-sensing populations are in the hypothalamus, a region at the base of the forebrain. From studies examining the expression of early immediate genes, particularly c-fos, and ex vivo electrophysiological recordings, glucose-sensing neurons are known to present in many hypothalamic nuclei. These include the arcuate nucleus (ARC) (24, 36), paraventricular hypothalamic nucleus (PVH) (37), dorsomedial hypothalamic nucleus (DMH) (4), ventromedial hypothalamic nucleus (VMH) (38), lateral hypothalamus (LH) (26, 39), supraoptic nucleus (40), and possibly the medial preoptic area (MPOA) (41).

There are several populations of glucose-sensing neurons in the ARC. Both intracarotid glucose administration (42) and insulin-induced hypoglycemia (43) increase c-fos expression in the ARC, suggesting that this region contains neurons activated by both high and low glucose. Several neural populations in the ARC express putative glucose sensors such as glucokinase (7, 23) and GLUT2 (44), as well as components of possible glucose-sensing mechanisms such as ATP-sensitive potassium (K^ATP) channels (24, 45). In keeping with this, ex vivo electrophysiological recordings from brain slices show that subpopulations of neurons in this region are glucose excited and increase their firing rate as glucose levels increase from 0.1 to 10 mM (21). It is possible that these GE neurons may express the neuropeptide pro-opiomelanocortin (POMC), as patch clamp recordings suggest that firing rates of these neurons fell as glucose levels were reduced from 10 to 5 mM (45) or from 5 to 3 mM (24). However, other reports show no effect of glucose levels <5 mM on POMC activity (46). In other studies, c-fos induced by hypoglycemia has been shown to overlap with expression of neuropeptide Y (NPY) (47). Recordings from ARC NPY neurons demonstrate that ∼40% of these neurons increase their activity as glucose falls (5 to 0.5 mM), indicating that they are glucose inhibited in type (46). Additionally, electrophysiological recordings from ARC neurons expressing GH-releasing hormone (GHRH) show that most of these cells are GI neurons. More than 80% of GHRH neurons are activated as glucose levels decrease from 10 to 0.2 mM (23). Taken together, this evidence suggests that the ARC contains populations of GE neurons, possibly expressing POMC, and GI neurons expressing NPY and GHRH.

Similarly, the PVH also appears to have subpopulations of GI and GE neurons because both insulin-induced hypoglycemia (43) and intracarotid glucose administration (48) increase PVH expression of c-fos. The putative glucose sensors glucokinase (7) and GLUT2 (44) are expressed in the PVH, and electrophysiological recordings support the presence of glucose-sensing neurons, particularly in the dorsal parvocellular PVH, where ∼70% of neurons responded to changes in glucose. Overall, in vivo electrophysiological recordings in rats report that 57% of PVH neurons are activated and 42% are inhibited by application of 5 mM glucose (20). The molecular identify of these populations is not clear, but hypoglycemia induces c-fos and FosB in PVH neurons expressing CRH (49, 50).

The DMH also contains glucose-sensing neurons. C-fos expression in the DMH is induced by both hypoglycemia (43) and by glucose administration (42), again suggesting a mixed population, and glucokinase is expressed at high levels in the nucleus (7). In vivo recordings in rats showed that 75 of 279 DMH neurons were responsive to IV glucose administration, and of the glucose-responsive neurons, 41% were excited (GE neurons) and 59% were inhibited (GI neurons) (28). Ex vivo electrophysiological recordings also show that DMH neurons respond to glucose, with approximately a third altering their firing rate as glucose increased from 5.5 to 20 mM (4), and most of these neurons were excited. In keeping with these findings, the activity of γ-aminobutyric acid (GABA)ergic DMH neurons was also modulated by glucose. In ex vivo slices, lowering glucose from 2.5 to 0.5 mM activated 60% and inhibited 30% of GABAergic DMH neurons (29).

One of the most highly examined hypothalamic regions is the VMH. Again, both peripheral glucose administration and insulin-induced hypoglycemia increase expression of c-fos in the VMH (43, 48). VMH neurons express GLUT2 (38), glucokinase (23, 38), AMP-activated protein kinase (AMPK) (51), and K^ATP channels (52, 53), which have all been proposed to be components of glucose-sensing pathways. Electrophysiological recordings confirm the presence of both GE and GI neurons in the VMH. In rats, 14% of neurons are glucose excited, with decreased activity as glucose is reduced from 2.5 to 0.1 mM, and 3% are glucose inhibited and increase their firing rate as glucose is decreased (2.5 to 0.1 mM) (54). Glucokinase is found in 64% of GE neurons and 43% of GI neurons in this region (38). However, specific markers for GE populations or GI populations in the VMH are not known. Interestingly, a recent study done in humans with positron emission tomography (PET) showed there to be a significant relationship between the binding of a highly selective norepinephrine transporter ligand in the hypothalamus and the epinephrine response during hypoglycemia (55), suggesting that the glucose-sensing neurons in the hypothalamus may be norepinephrinergic.

Similar to other hypothalamic regions, the LH also contains subpopulations of GE and GI neurons. Both hypoglycemia and hyperglycemia increase expression of c-fos (43, 56), and subpopulations of LH neurons express glucokinase (23). Extracellular recordings in freely moving rats demonstrated that central administration of the nonmetabolizable glucose analog 2-deoxy-d-glucose (2DG), which mimics hypoglycemia, increased activity in 38% of LH neurons (GI neurons) and inhibited 19% (GE neurons) (25). Extracellular recordings in rhesus monkeys indicated that similar proportions of LH neurons, ∼27%, were glucose inhibited (57). However, in the LH, the identity of several glucose-sensing populations is known. Expression of the neuropeptide orexin marks a population of GI neurons (39) whose activity (in 20 of 21 neurons) was suppressed as glucose increased from 0.2 to 5 mM. Additionally, a proportion (70%) of LH neurons expressing NPY (but not orexin) also appear to be GI neurons, as they are hyperpolarized and reduce their firing rate in response to rising glucose (27). In contrast, neurons expressing melanin-concentrating hormone (MCH) are depolarized by increased glucose (0.2 to 5 mM, 15 of 18 neurons), and although only a small proportion increase their firing rate, they are more likely to fire with the addition of a depolarizing input (26).

Glucose-sensing neurons have also been reported in the supraoptic nucleus of the hypothalamus. These neurons express glucokinase (58) and K^ATP channels (40), and glucose treatment (0.5 to 5 mM) increased intracellular calcium levels, suggesting that they are GE neurons. Glucose also increased vasopressin and oxytocin release, suggesting that the supraoptic GE neurons may express these neuropeptides (40). Other hypothalamic regions may also harbor glucose-sensing neurons. For example, glucokinase is expressed in the preoptic area (23, 58), and hyperglycemia induces c-fos immunoreactivity in the preoptic area in rats (59). Immortalized hypothalamic cell lines expressing GnRH, a neuropeptide abundantly expressed in the preoptic area, increased c-fos and GnRH expression in response to 5 mM glucose (60), and GnRH neurons in vivo have been shown to be glucose sensing (41).

The brain stem also contains a number of glucose-sensing populations in the nucleus of the solitary tract (NTS), AP, and dorsal motor nucleus of the vagus (DMV) as well as the parabrachial nucleus (PBN). Duodenal administration of glucose in rats increased c-fos expression in NTS neurons (61). Glucoprivation with 2DG also induces c-fos immunoreactivity in the NTS (62). The NTS also expresses glucokinase as well as K^ATP channels, and electrophysiological studies show that 35% of NTS neurons are activated by 5 mM glucose whereas glucose inhibits 21% of NTS neurons (32). Studies suggest that glucose-sensing populations in the NTS are marked by several neuropeptides and neurotransmitters. Electrophysiological recordings from NTS GABAergic neurons show that 40% are glucose excited and 33% are glucose inhibited (2.5 to 15 mM glucose) (33). A population of NTS GLUT2-expressing neurons is activated by hypoglycemia. Their firing rate increases as glucose falls from 5 to 0.5 mM (35). Catecholaminergic neurons in the NTS are also reported to be glucose sensing. Extracellular recordings in rats found that a population of glucose-sensing neurons was in the caudal tyrosine hydroxylase–positive region of the NTS (63). Supporting these findings are results from ex vivo slice recordings performed in mice expressing GFP in tyrosine hydroxylase (rate-limiting enzyme in catecholamine synthesis)–positive neurons. Recordings from GFP/tyrosine hydroxylase–positive neurons showed that their firing rate decreased as glucose levels fell from 5 to 1 mM, suggesting that they are GE neurons (34). GE neurons and GI neurons have been described in the rostral ventrolateral medulla (64, 65), and c-fos immunoreactivity increases in catecholaminergic neurons in the locus coeruleus (LC) (66) after hypoglycemia, suggesting that they may also be a glucose-sensing population. Glucose-sensing neurons have also been described in the AP (67) with 21 of 172 neurons inhibited by glucose whereas 20 of 172 neurons were activated by glucose (30).

Neurons in the DMV may also respond to changes in glucose. Glucokinase is expressed in this region (58), and glucose-sensing neurons have been identified by electrophysiological studies (67). Similarly, hypoglycemia induced c-fos immunoreactivity in the lateral PBN of the pons in a subpopulation of neurons expressing cholecystokinin (CCK). In keeping with this, electrophysiological studies confirmed that a subpopulation of CCK LPBN neurons (5 of 14) were depolarized as glucose levels fell from 5 to 0.5 mM (31).

Glucose-sensing populations have been described in areas beyond the hypothalamus and hindbrain, particularly in regions implicated in reward such as the nucleus accumbens (NAc) (68), amygdala (69), paraventricular thalamus (70), prefrontal cortex (71), and hippocampus (72). These areas contribute to dopaminergic pathways involved in reward processing. In the NAc, a fourth of neurons are glucose sensing (68), with GE neurons predominantly in the NAc core and GI populations in the shell. However, these studies used supraphysiological glucose concentrations (500 mM glucose). Both GE and GI neurons are present in the amygdala, with electrophysiological studies in rats (using glucose concentrations from 0.5 to 2.5mM) showing that GE and GI neurons make up 6% and 7.5% of amygdala neurons, respectively (69). Neurons in the hippocampus may also respond to changes in glucose. These neurons express glucokinase (73), GLUT1 (74), K^ATP channels (75), and sweet taste receptors T1R2/T1R3 in the corni ammonis fields and dentate gyrus (76).

Glucose-sensing properties have been reported in CNS populations other than neurons. Tanycytes, which line the third ventricle, express proteins likely to be involved in glucose sensing such as glucokinase, K^ATP channels, and GLUT2. In vitro studies demonstrate that intracellular calcium in tanycytes increases as glucose rises from 0.5 to 8 mM (77). Astrocyte populations may also sense glucose, as GLUT2 is expressed in astrocytes in the hypothalamus (78) and brain stem (79). Glucose taken up by astrocytes largely enters the glycolytic pathway to produce lactate that is released via monocarboxylate transporters into the extracellular space (80, 81) and may provide an alternative energy source for neurons (82, 83).

Many other cell populations are modulated by changing glucose via synaptic inputs from glucose-sensing neurons. Song et al. (54) reported that in the ventromedial hypothalamus ∼17% of cells are directly modulated by changes in extracellular glucose whereas a further 33% of neurons were presynaptically inhibited or activated by altered glucose.

Mechanisms responsible for glucose sensing in GE and GI neurons

Multiple mechanisms have been described to play a role in glucose sensing (Fig. 2). In many GE neurons, the sensing mechanism is thought to be similar to that seen in pancreatic β cells and dependent on glucose metabolism. VMH and ARC GE neurons, similar to the pancreatic β cells, express glucokinase to phosphorylate glucose to glucose-6-phosphate (7, 84). Glucokinase has a low affinity for glucose and there is no end product inhibition (85), so glucokinase activity is proportional to glucose concentrations. Supporting a role for glucokinase, mice and humans with reduced glucokinase activity show an exaggerated hormonal response to hypoglycemia (86). As glucose concentrations increase, so too does glucose metabolism, leading to an increased intracellular ATP/ADP ratio (87) and, through interconversion of adenine nucleotides, to an increased ATP/AMP ratio. In many GE neurons, increased ATP/ADP leads to closure of K^ATP channels, depolarizing the neurons and increasing their firing rate (88). However, K^ATP channels are not the only mechanism involved in GE neurons. Intracellular ATP levels do not increase in hypothalamic neurons as glucose rises from 3 to 15 mM (89), suggesting that K^ATP channels do not play a role in glucose sensing in arcuate neurons that are modulated by glucose levels >5 mM (high GE and high GI). In keeping with this, high GE neurons are still found in K^ATP-deficient mice (90). In these neurons, it is likely that a pathway involving transient receptor potential canonical type 3 (TRPC3) channels is involved in the glucose-sensing mechanism (91). Glucose sensing in GE neurons in the PVH (37) and GnRH neurons (41) is also reported to be independent of K^ATP channels.

Figure 2.

Putative glucose-sensing mechanisms in glucose-sensing neurons. In GE cells (upper panel), glucose can enter cells via GLUTs (usually GLUT2 or GLUT3) and is phosphorylated by glucokinase to glucose-6-phosphate. This, in turn, regulates cytosolic ATP production. In GE neurons, increased ATP closes K^ATP channels, leading to depolarization and calcium influx through voltage-activated calcium channels. Metabolism-independent pathways have also been described using sweet taste receptors and downstream signaling, sodium/glucose cotransporter, or transient receptor potential canonical type 3 (TRPC3) channels. Both sodium/glucose cotransporter and TRPC3 channels lead to influx of cations and depolarization. In GI cells (lower panel), as glucose entry decreases, intracellular ATP falls. Low ATP leads to an increase in AMPK activity. This may reduce activity of chloride channels, possibly via neuronal NO synthase, whereas low ATP decreases activity of Na/K ATPases. Both of these lead to cell depolarization with low glucose. Alternative pathways involving closure of potassium leak channels with low glucose have also been described.

Mitochondrial function can also play a role in glucose sensing in ARC and VMH GE neurons. Glucose sensing in POMC neurons exposed to glucose concentrations ranging from 0.2 to 2.5 mM is disrupted by alterations in the mitochondrial fission regulator, dynamin-related protein 1 (DRP1) (92). In the VMH, mitochondrial function has also been shown to be critical to glucose sensing in GE neurons. Increasing glucose induces mitochondrial fission through phosphorylated DRP1 and uncoupling protein 2. When uncoupling protein 2 expression was increased, the number of VMH GE neurons also increased (93).

“In many GE neurons, the sensing mechanism is thought to be similar to that seen in pancreatic β cells…”

The glucose-sensing mechanisms described above rely on glucose metabolism, but metabolism-independent pathways have also been reported. Sodium-glucose cotransporters (SGLTs) link the inward transport of glucose and sodium, and as glucose levels increase and are transported into the cell, the cotransport of sodium ions would lead to depolarization. SGLT1, SGLT3a, and SGLT3b are expressed in subpopulations of GE neurons (38, 94). An alternative signaling mechanism is via sweet taste receptors. These are heterodimers of T1R2 and T1R3 and are expressed in the CNS, particularly in the ARC as well as in the periphery. The receptors are activated by glucose and also by sweeteners such as sucralose, but signaling does not rely on glucose metabolism. A role for these receptors is supported by studies showing that sucralose application activated a significant proportion of ARC GE neurons whereas the sweet taste receptor inhibitor gurmarin inhibited the glucose responses in >60% of ARC GE neurons (95).

The glucose-sensing mechanisms in GI neurons are less well understood. GI neurons in the ARC and VMH also express glucokinase, and decreased glucose leads to a reduction in intracellular ATP/ADP ratios. Electrophysiological studies and calcium imaging suggest that the activation in VMH GI neurons is mediated by inactivation of an ATP-dependent chloride current (51). The cystic fibrosis transmembrane conductance regulator is activated by ATP and expressed in the hypothalamus. As glucose levels fall, decreased ATP reduces the chloride current through the cystic fibrosis transmembrane conductance regulator, leading to depolarization (51). There is also evidence to support a role for AMPK in glucose sensing. AMPK is activated as the ratio of AMP to ATP rises and, in turn, decreases anabolic processes while increasing catabolic processes to conserve energy (96). In GI neurons on the VMH, activating AMPK with 5-Aminoimidazole-4-carboxamide ribonucleotide mimicked the effects of low glucose (97) and altered neuronal activity (98). The pathway linking low glucose and AMPK to increased cell activity is also likely to involve nitric oxide (NO). Both hypoglycemia and 5-Aminoimidazole-4-carboxamide ribonucleotide increase NO, and this increase is absent when AMPK activity is blocked by compound C (97).

Glucose metabolism–independent pathways are also present in GI neurons. LH orexin neurons are inhibited by both glucose (5 mM) and the nonmetabolizable glucose analog 2DG (1 to 5 mM). The mechanism linking glucose to ion channel activity in these cells is unknown, but glucose and 2DG lead to opening of tandem-pore K⁺ channels, resulting in hyperpolarization and reduced neural activity (39, 99). However, although the tandem-pore K⁺ channels, TASK1 and TASK3, are involved in regulating excitability of orexin neurons, they are not essential for glucose sensing in orexin neurons. Orexin neurons in TASK1 knockout mice, TASK3 knockout mice, or TASK1/3 double knockout mice still responded to altered glucose (0.2 to 5 mM) (100, 101).

Projections from glucose-sensing neurons

Although the specific circuitry of glucose-sensing neurons is not fully defined, it is clear that regions with glucose-sensing neurons are highly interconnected. This provides an anatomical framework to suggest that there may be an integrated network of glucose sensors that fine tune the response to altered glucose concentrations (Fig. 3). Glucose-sensing neurons are present in a number of brain regions and exert their actions through projections to downstream areas, particularly via sympathetic and parasympathetic efferent pathways in the brain stem and spinal cord to metabolically active organs. The cell bodies of vagal preganglionic neurons innervating organs such as the pancreas, liver, and visceral adipose tissue are found in the DMV and express choline acetyltransferase. The DMV receives input from the brain stem and hypothalamic and forebrain regions. In the brain stem, neurons in the NTS are known to project to the DMV. These include GABAergic neurons, of which some are glucose sensing (102). Cells in the A5/rostral ventrolateral medulla also provide input into the DMV. In the hypothalamus, cells in the PVH, DMH, LH, and ARC project to the DMV, and in the forebrain, neurons in the MPOA and bed nucleus of the stria terminalis (BNST) also provide input into the DMV. Additional hypothalamic nuclei then connect to these regions, particularly the VMH (103, 104).

Figure 3.

Connections to sympathetic and parasympathetic efferent pathways in the mouse brain. The DMV is the parasympathetic efferent pathway. The DMV receives input from the NTS and the RVLM as well as from the PVH, DMH, LH, and ARC. The VMH connects indirectly via the PVH. The sympathetic efferent pathway is via the IML of the spinal cord. This receives input from the RVLM, LC, as well as the PVH and LH. Multiple hypothalamic regions, including the ARC, VMH, DMH, project to either the PVH or LH.

Sympathetic pancreatic innervation originates in the cholinergic sympathetic preganglionic neurons in the intermediolateral column (IML) of the spinal cord, and this region also receives input from regions with glucose-sensing neurons. The rostral ventrolateral medulla (RVLM), LC, and several hypothalamic areas, including the PVH and LH, provide direct input into the IML. Multiple additional regions provide indirect projections to the IML via either the PVH or LH, including the ARC, VMH, DMH, MPOA, and BNST (103, 105).

The specific projections of glucose-sensing neurons from across the brain are largely unknown, but studies that examine the projections of neurons expressing defined neuropeptides, a portion of which are known to be glucose sensing, provide some insights into the density and expanse of these projections. In the ARC, a subpopulation of NPY neurons, which also express agouti-related peptide (AgRP), are GI neurons. Within the hypothalamus, ARC AgRP/NPY neurons project to the PVH and LH (106), which would provide indirect pathways to both sympathetic and parasympathetic pathways. Similarly, ARC POMC neurons have also been reported to be glucose sensing and project to the LH and PVH in addition to other hypothalamic regions such as the VMH and DMH, as well as sending dense projections to the BNST (107). VMH neurons expressing steroidogenic factor-1 (SF1), a subpopulation of which are glucose sensing, project widely. SF1 neurons in the dorsomedial VMH project to other hypothalamic nuclei, including the PVH, to lateral regions such as the amygdala, rostral areas such as the BNST, and caudal areas including the RVLM and LC (108). In the LH, GI orexin neurons have multiple projection sites to innervate almost the entire hypothalamus, BNST, and cortex with caudal projections to the LC and the NTS, RVLM, and DMV (109, 110). Similarly, LH MCH neurons, which are largely glucose excited, project to most hypothalamic nuclei, to the amygdala and BNST, and rostral to the RVLM and AP (111). Neurons in the PVH project to the median eminence to regulate pituitary function but also connect to other brain regions. Dense connections are seen within the PVH and to other hypothalamic regions (DMH, LH, MPO), to rostral regions such as the BNST (112) and caudally to the NTS, DMV and IML (113).

Brain stem PBN neurons, which are glucose inhibited, project to thalamic and hypothalamic nuclei (LH, VMH, zona incerta) as well as to the NTS. The specific projections of glucose-sensing CCK neurons in the PBN have been traced using genetically encoded tracers. These neurons project predominantly to the ipsilateral hypothalamus, with a particularly dense input onto the dorsomedial VMH (31).

The Counterregulatory Response to Hypoglycemia in Health and Diabetes

Under normal circumstances, blood glucose is maintained in a narrow range from 4 to 8 mmol/L. There is a delicate balance between glucose influx from ingestion and endogenous glucose production and glucose efflux through utilization and uptake in insulin-sensitive and insulin-insensitive organs. Hormones, neurotransmitters, and behavior all act to maintain this balance, and when blood glucose falls either because of insufficient intake or increased use, a number of counterregulatory measures act to restore blood glucose levels to normal. In healthy individuals, as glucose levels fall, a sequential set of responses occurs. These responses are the result of coordinated effects directly in peripheral organs but also via actions in multiple CNS regions that then influence peripheral organs via their innervation and actions of circulating hormones.

Hypoglycemia-induced reduction in insulin secretion

As glucose falls below the physiological range (∼70 mg/dL or ∼4.0 mmol/L), the initial response is a decrease in insulin secretion. As insulin levels fall, glucose uptake into insulin-sensitive tissues, including liver, muscle, and adipose tissue, also falls and insulin’s inhibition of glycogenolysis and gluconeogenesis is relieved (114). These effects combine to decrease glucose efflux, increase glucose influx, and restore euglycemia.

The fall in insulin release in response to hypoglycemia is partly a consequence of decreased glucose stimulation directly on β cells. However, there is also evidence for neural regulation of insulin release. The pancreas is densely innervated by both parasympathetic and sympathetic fibers (115). Parasympathetic cholinergic inputs may play a role in regulation of insulin and glucagon. In many animal studies across different species, activation of DMV neurons (116), vagal stimulation (117), and acetylcholine administration (118) all increase both insulin and glucagon secretion. Conversely, parasympathetic blockade by atropine inhibits insulin release (119) and blunts glucagon secretory responses to hypoglycemia (120). In human studies, the role of the parasympathetic nervous system is complex. Atropine prevents increased glucose-stimulated insulin release that occurs with the atypical antipsychotic olanzapine (121), but vagotomy increases the periodicity of insulin release (122), so species differences may exist. In contrast, sympathetic activation suppressed insulin release (123), and norepinephrine and epinephrine administration reproduce sympathetic excitation by suppressing insulin and stimulating glucagon release in humans (124, 125).

Both the PVH and LH have direct projections to sympathetic preganglionic neurons. Lesion of the PVH, which projects to sympathetic preganglionic neurons, produces hyperinsulinemia (126). However, the effects of modulating LH activity are contradictory. Several studies reported that electrical stimulation of the LH inhibited glucose-stimulated insulin release (127, 128). However, infusion of norepinephrine into the LH rapidly increased insulin and this effect was reversed by vagal blockade with atropine (129).

Other CNS regions regulate sympathetic activity via indirect projections to the IML and have been implicated in the control of insulin release. There is considerable evidence to support a role for the VMH, which contains both GI and GE neurons, in modulating insulin secretion. VMH lesions result in insulin hypersecretion (130, 131) whereas electrical stimulation of the VMH suppressed insulin secretion (128). These results suggest that VMH activity may have an inhibitory effect on pancreatic insulin release. More recent studies support these findings. Using magnetogenetic neuromodulatory tools, acute stimulation of VMH neurons expressing glucokinase decreased plasma insulin whereas silencing these neurons increased insulin (132). Similarly, optogenetic activation of VMH SF1 neurons increased blood glucose without increasing plasma insulin (133). These findings suggest that VMH GI neurons, which would normally be activated by a decrease in blood glucose, may play a role in suppressing insulin release in hypoglycemia.

ARC neural populations may also regulate insulin release. Studies using targeted overexpression of glucokinase in the ARC to enhance glucose sensing showed blunted glucose-stimulated insulin release (134). CNS administration of melanocortin agonists inhibits basal insulin release (135). This suggests that the melanocortin system, AgRP and POMC neurons of the ARC and melanocortin receptors, may play a role in regulating insulin secretion.

Hindbrain regions are also likely to play a role in control of insulin. Glucose-sensing neurons expressing leptin receptor in the PBN regulate insulin release, as chemogenetic activation of these neurons blunts glucose-stimulated insulin release, resulting in impaired glucose tolerance (136). Although neurons in the RVLM are modulated by glucose and their activation increases blood glucose, the effects on insulin have not been reported.

In patients with insulin-treated diabetes, a fall in glucose occurs when more exogenously administered insulin is present than required for the metabolic condition. Insulin secretion from the β cell is profoundly deficient in such patients, and reductions in serum glucose concentration or changes in neural input to the islet have a minimal effect on reversing hypoglycemia. These patients cannot link brain glucose sensing that occurs in the brain regions detailed in previous paragraphs to a reduction in insulin secretion and must rely on other counterregulatory mechanisms to restore euglycemia.

Hypoglycemia-induced glucagon release

The threshold to elicit the increase in glucagon is in the range of 64 to 74 mg/dL (2.5 to 3.8 mmol/L) (137–140). The major effects of glucagon to increase blood glucose are in the liver, to rapidly increase glycogenolysis and stimulate gluconeogenesis. Glucagon also inhibits hepatic glucose uptake. Collectively, these actions produce a transient increase in hepatic glycogenolysis and a sustained increase in gluconeogenesis (141), leading to increased blood glucose (142). Glucagon can also stimulate lipolysis and ketogenesis when insulin levels are low, for example, following an extended fast (143).

Several mechanisms contribute to glucagon release in hypoglycemia. Perhaps one of the most important is the reduction in insulin secretion that occurs when glucose falls, which in turn removes the inhibition of glucagon release (144). In healthy subjects, this leads to an exuberant response that restores euglycemia. However, this mechanism is not operational in patients with type 1 and advanced type 2 diabetes because they cannot reduce insulin secretion in the setting of a fall in blood sugar. Exogenous insulin cannot be dissipated, so the inhibitory effects of circulating insulin cannot be attenuated and the tonic inhibition of glucagon secretion cannot be released. As a result, these patients generally do not have a glucagon secretory response to a fall in glucose (145)

The sympathetic nervous system and catecholamines also play a critical role in increasing glucagon release. Sympathetic nervous system activation releases norepinephrine locally and increases circulating epinephrine, which together act on B2 adrenergic receptors on the α cell to stimulate glucagon secretion. Blocking the activation of the autonomic nervous system has been reported to eliminate 75% to 90% of the glucagon response to hypoglycemia (146).

“Both cortisol and GH release occur with relatively severe hypoglycemia and provide a more sustained effect to maintain blood glucose.”

Multiple CNS regions have been shown to play a role in hypoglycemia-induced glucagon secretion. In rodents, vagal stimulation increases plasma glucagon, and recent studies showed that optogenetic activation of GLUT2 neurons in the DMV significantly increased vagal activity and glucagon secretion compared with control animals (35). Sympathetic activity also stimulates glucagon release, and many CNS regions with direct and indirect projections to sympathetic outflow regions are implicated in regulating glucagon release. In the brain stem, chemogenetic activation of leptin receptor–positive neurons (136) and CCK neurons (31) in the PBN increased plasma glucagon. Because both the LH and PVH project to sympathetic preganglionic neurons, they may also play a role in sympathetic stimulation of glucagon release. In keeping with this, the glucagon response to hypoglycemia is enhanced in mice lacking the leptin receptor in the LH and premammillary nucleus (147). However, PVH lesions did not alter glucagon responses to hypoglycemia (148).

Although the VMH does not project directly to sympathetic outflow regions, most evidence supports its role in the glucagon response to hypoglycemia. Electrical stimulation of the VMH increases plasma glucagon (128) whereas VMH lesions blunt the glucagon rise following hypoglycemia (130, 131). VMH injection of 2DG to mimic glycopenia induces glucagon secretion (149), and VMH administration of glucose during systemic hypoglycemia blunts the counterregulatory glucagon response (150). More recently, chemogenetic activation of VMH SF1 neurons (133) and magnetogenetic activation of VMH glucokinase neurons (132) have been shown to increase plasma glucagon. Silencing of these populations blunted the counterregulatory response to hypoglycemia. In mice without VMH glucokinase expression, the glucagon response to hypoglycemia was also blunted (151). Similarly, reduced AMPK activity by expression of a dominant negative AMPK in the ARC/VMH of rats reduced the glucagon response to hypoglycemia (152).

Hypoglycemia-induced catecholamine secretion

As glucose levels fall further to a threshold of between 55 and 68 mg/dL (3.1 to 3.8 mmol/L), epinephrine levels begin to increase (137). Increased epinephrine and norepinephrine have several effects to raise glucose. Catecholamines increase hepatic gluconeogenesis and glycogenolysis to increase glucose influx. Additionally, catecholamines also decreases glucose uptake by directly inhibiting insulin secretion from β cells and inhibiting insulin-stimulated glucose uptake (153). Catecholamines also increase lipolysis, providing free fatty acids as an alternative energy source for peripheral tissues (154). Sympathetic activation leading to catecholamine release results in tremor (155), sweating (156), and headache (157).

Several CNS regions have been reported to play a role in this response. Ablation of catecholaminergic neurons in the spinal cord diminishes the catecholaminergic response to insulin-induced hypoglycemia (158). Similarly, neurotoxic destruction of RVLM neurons blunted the catecholaminergic response to systemic hypoglycemia (159). Chemogenetic activation of GI CCK neurons in the PBN significantly increases serum epinephrine concentrations, and silencing these neurons blunted the counterregulatory response to hypoglycemia (31).

Several hypothalamic regions are also likely to play a role in the catecholaminergic response to hypoglycemia. Both the PVH and LH have dense projections to the sympathetic preganglionic neurons and play a role in hypoglycemia-induced catecholamine release. Silencing PVH neurons with lidocaine almost halved the peak epinephrine response to hypoglycemia (148). Overexpression of a dominant negative form of AMPK to reduce glucose sensing in the PVH blunted the catecholaminergic response to hypoglycemia (152). Local injection of thioglucose into the perifornical hypothalamus to elicit glucoprivation rapidly stimulated epinephrine release, whereas depleting serotonin in the perifornical hypothalamus blunted the catecholaminergic response to hypoglycemia (160).

The VMH is also critical to hypoglycemia-induced catecholamine release. VMH lesions blunted epinephrine and norepinephrine release during mild and severe hypoglycemia in rats by 50% to 80% (161). Local VMH injection of 2DG to induce glucoprivation stimulated epinephrine release 30-fold and norepinephrine release more than threefold (149) whereas local VMH glucose infusion blunted the catecholamine response to systemic hypoglycemia (150). A number of signaling mechanisms in the VMH have also been implicated in the full catecholamine counterregulatory response to hypoglycemia, including NO (162), AMPK (163), CRF1 (164), β adrenergic receptors (165), and GABA (166).

Individuals with type 1 and advanced type 2 diabetes must rely on hypoglycemia-induced catecholamine secretion to prevent them from experiencing neuroglycopenia. Unfortunately, each episode of hypoglycemia transiently reduces the glucose level necessary to elicit catecholamine release. When an individual experiences several episodes of hypoglycemia in a short period of time, catecholamine release may not occur before the onset of neuroglycopenia. This situation has been called hypoglycemia-associated autonomic failure (167). It is closely associated clinically with low symptomatic awareness of hypoglycemia, presumably because at least some of the symptoms of hypoglycemia are generated by the physical counterregulatory responses.

Hypoglycemia-induced increases in cortisol and GH

Both cortisol and GH release occur with relatively severe hypoglycemia and provide a more sustained effect to maintain blood glucose. The effects of both cortisol and GH are slow onset (on the order of hours), and some studies suggest that they may not play a significant role in the acute response to hypoglycemia (168). Cortisol activates hepatic gluconeogenesis to increase plasma glucose and ketogenesis to provide alternative energy source for tissues, reduces insulin-dependent glucose uptake (169), and increases fatty acid oxidation. Cortisol also increases lipolysis in some fat depots (170). Cortisol also alters insulin receptor binding and signaling (171) and directly inhibits pancreatic insulin secretion (172). GH induces insulin resistance through a postreceptor mechanism (173), activates gluconeogenesis, and stimulates lipolysis to increase free fatty acids (174).

Cortisol (or corticosterone in rodents) is secreted from the adrenal cortex in response to ACTH produced by the pituitary gland. ACTH release is, in turn, modulated by CRH (also known as corticotropin-releasing factor). CRH is produced by neuroendocrine cells in the PVH and released from terminals in the median eminence into the pituitary portal system to be delivered to the anterior pituitary and to stimulate ACTH release into the systemic circulation. Hypoglycemia increases both CRH (175) and ACTH (176). CRH is produced by PVH neurons and, as expected, this region has been shown to affect the ACTH/corticosterone response to hypoglycemia. Disruption of PVH glucose sensing by expressing a dominant negative AMPK significantly blunted the corticosterone response to hypoglycemia (152). However, silencing PVH neurons with lidocaine suppressed hypoglycemia induced ACTH but not corticosterone secretion (148). Other studies have shown that inputs into the PVH are important to the corticosterone response to hypoglycemia. Chemical lesions of the hindbrain catecholaminergic neurons innervating the PVH blunted the feeding and corticosterone increase with hypoglycemia (177). Other hypothalamic regions also project to the PVH and can modulate the counterregulatory corticosterone release. Silencing DMH neurons with lidocaine blunted ACTH release and delayed corticosterone release with hypoglycemia (178). Optogenetic activation of VMH SF1 neurons markedly increased plasma corticosterone (133). Interestingly, activating SF1 neurons projecting to the PVH produced a small, nonsignificant increase in corticosterone whereas activation of SF1 projections to the BNST significantly increased corticosterone (133). It also appears that the BNST-projecting VMH neurons receive inputs from the PBN, and this region has also been shown to play a role in regulating corticosterone release. Activation of PBN leptin receptor–positive neurons also stimulates plasma corticosterone to mimic the response to hypoglycemia (136). Plasma corticosterone is also increased by chemogenetic stimulation of PBN CCK neurons (31).

GH is also released in response to hypoglycemia. Pituitary GH secretion is stimulated by GHRH and inhibited by somatostatin, both produced in the hypothalamus. Previous studies have shown that arcuate GHRH neurons are activated by hypoglycemia (23) and so may be directly stimulated by low glucose to release GHRH and increase GH secretion. Somatostatin-expressing neurons are found in the periventricular hypothalamus, and there is conflicting evidence as to whether they are glucose sensing. In mice, somatostatin neurons do not express c-fos in response to hypoglycemia (179), but in insulin-induced hypoglycemia induces c-fos immunoreactivity in SST neurons in ewes (180). Other factors also regulate GH release, including GH secretagogues such as ghrelin. Ghrelin knockout mice or mice lacking ghrelin O-acyltransferase, which is required for active ghrelin, become profoundly hypoglycemic with fasting owing to a failure to increase GH (181, 182). However, in humans, studies suggest that ghrelin is acutely suppressed by insulin (with or without hypoglycemia) and does not influence the counterregulatory response to hypoglycemia (183).

Impact of Recurrent Hypoglycemia on Counterregulation

Exposure to repeated episodes of antecedent hypoglycemia results in reduced counterregulatory hormone and symptom response to subsequent hypoglycemia and leads to the development of impaired awareness of hypoglycemia. The mechanisms underlying the development of impaired awareness of hypoglycemia are not clearly understood. Changes in both neurotransmission and adaptations in energy metabolism have been postulated as possible causes that may lead to altered sensing of glucose by the brain. Development of impaired awareness of hypoglycemia could also be related to defects in the coordination of the counterregulatory response to hypoglycemia. In this section we review some of the potential mechanisms (Fig. 4) that may contribute to the development of impaired awareness of hypoglycemia. These proposed mechanisms are not mutually exclusive, it is likely that multiple factors contribute to development of this syndrome, and how patients with diabetes respond to recurrent hypoglycemia may be different than from individuals with normoglycemia.

Figure 4.

Potential mechanisms that may contribute to the development of impaired awareness of hypoglycemia after exposure to recurrent hypoglycemia.

“Exposure to repeated episodes of antecedent hypoglycemia results in reduced counterregulatory hormone and symptom response to subsequent hypoglycemia…”

Upregulation of glucose transport to the brain

Increase glucose uptake across the blood–brain barrier has been proposed as a potential cerebral adaptation to recurrent hypoglycemia, which allows neurons to maintain metabolism during the stress of hypoglycemia and may lead to development of impaired awareness of hypoglycemia. Previous studies in rodents have shown that chronic hypoglycemia resulted in increased expression of GLUT1 transporters in the endothelial cells of the blood–brain barrier along with upregulation of brain glucose transport (184, 185). In a rodent study that used an in vivo microdialysis technique to measure brain glucose concentrations following administration of IP glucose, animals exposed to recurrent hypoglycemia were found to have glucose concentrations threefold to fourfold higher in the hippocampal extracellular fluid as compared with controls (186), whereas the plasma glucose concentration did not differ between the two groups, suggesting upregulated transport of glucose into the brain. However, other studies, especially in humans, have reported conflicting data on whether cerebral glucose transport or metabolism are altered after exposure to recurrent hypoglycemia. Using magnetic resonance spectroscopy, Creigo et al. (187) found glucose steady-state concentrations in the occipital cortex to be higher during a hyperglycemic clamp in subjects with type 1 diabetes and impaired awareness of hypoglycemia compared with healthy controls. Additionally, brain glucose uptake (calculated based on cerebral blood flow and brain arteriovenous glucose difference) was maintained during hypoglycemia in subjects with tightly controlled type 1 diabetes (who were presumably more likely to be exposed to recurrent hypoglycemia) compared with controls and subjects with poorly controlled type 1 diabetes (188). These observations support the hypothesis that brain glucose transport or metabolism are altered after recurrent hypoglycemia. Despite this, measurement of global rates of blood-to-brain glucose transport by PET in humans has not shown evidence of upregulation of glucose uptake. Utilizing PET, glucose transport was shown to be normal in healthy subjects exposed to hypoglycemia (189) and similar in subjects with type 1 diabetes and impaired awareness of hypoglycemia compared with subjects with type 1 diabetes and normal awareness (190). A recent study examined glucose transport kinetics in the hypothalamus of healthy subjects who underwent a protocol of repeated antecedent hypoglycemia to experimentally induce impaired awareness of hypoglycemia (191). In this study, the glucose transport kinetics in the hypothalamus of healthy humans with experimentally induced impaired awareness of hypoglycemia were not different from those measured in the absence of impaired awareness of hypoglycemia. More studies are needed to examine whether there are regional differences in cerebral adaptation to recurrent hypoglycemia.

Utilization of alternate fuels

Brain utilization of nonglucose substrates, including lactate and ketones, has been proposed as a potential mechanism that could maintain energy metabolism during hypoglycemia. Lactate infusion during hypoglycemia has been shown to lower the glucose level, which triggers the hypoglycemia counterregulatory response, reduces the magnitude of epinephrine and hypoglycemia symptom responses, as well as prevents cognitive function deterioration (192). These findings suggest that the brain utilizes lactate as a fuel source when glucose supply is low, leading to alterations in the counterregulatory response to hypoglycemia similar to those seen in people with impaired awareness of hypoglycemia (193). Using in vivo magnetic resonance spectroscopy, brain lactate has been shown to fall significantly during hypoglycemia in subjects with type 1 diabetes and impaired awareness of hypoglycemia and remained stable in subjects with type 1 diabetes and normal awareness of hypoglycemia and in nondiabetic controls. This suggests that lactate was used during hypoglycemia as an alternate fuel source and may contribute to the development of impaired awareness of hypoglycemia (194). Other investigators have examined the effects of physiologically raised plasma lactate concentrations on brain lactate concentration during hypoglycemia in patients with and without impaired awareness of hypoglycemia and in healthy controls (195). In this study, plasma lactate was raised by a single bout of high-intensity interval training. The investigators found that the brain lactate level was increased in all groups at the start of the hypoglycemia that was induced following high-intensity interval training, but the increase was most pronounced in subjects with diabetes and impaired awareness of hypoglycemia. During hypoglycemia, lactate levels decreased below baseline in the group with impaired awareness of hypoglycemia and remained unchanged in the other groups. This finding also suggests that lactate transport and oxidation of lactate may be enhanced in, and contribute to, development of impaired awareness of hypoglycemia. Similar to lactate, infusion of β-hydroxybutyrate has also been shown to reduce the magnitude of counterregulatory hormone responses and symptoms in healthy humans (192, 196).

Acetate transport into the brain has also been shown to be upregulated in subjects with type 1 diabetes and a history of hypoglycemia unawareness compared with healthy controls (197). Acetate is transported into the brain by monocarboxylic acid transporters, which are also responsible for transporting lactate and other ketone bodies. It has been hypothesized that upregulation of these transporters in response to recurrent hypoglycemia can provide the brain with alternate fuel sources during subsequent hypoglycemia, leading to the development of impaired awareness of hypoglycemia. Another study has shown that the increase in lactate levels alone may not be sufficient to offset the energy deficit caused by hypoglycemia (198). In a rodent model, Herzog et al. (17) demonstrated that recurrent hypoglycemia enhances transport of lactate into the brain but the elevated lactate itself was not sufficient to support metabolism as nonglucose fuel source. Instead, increased lactate acted as a ‘metabolic regulator’ to maintain neuronal glucose metabolism during hypoglycemia.

Alterations in brain glutamate metabolism have been reported in subjects with type 1 diabetes and impaired awareness of hypoglycemia (199, 200). Brain glutamate concentration decreases during hypoglycemia, and this reduction has been ascribed to oxidation of glutamate in response to hypoglycemia (199, 201). In human studies, reduction in glutamate during hypoglycemia was seen in heathy controls and subjects with type 1 diabetes without impaired awareness of hypoglycemia but not in subjects with type 1 diabetes and impaired awareness of hypoglycemia (199). These data were interpreted to suggest that in subjects with impaired awareness of hypoglycemia, enhanced glucose and/or alternate fuel transport into the brain eliminated the need to oxidize glutamate.

Increase glycogen storage

Brain glycogen stored primarily in astrocytes can be metabolized to lactate and exported to neurons to be used as an energy source during hypoglycemia (202–204). Following a single episode of acute hypoglycemia in rodents, after plasma and brain glucose concentrations are restored, glycogen concentration has been shown to increase several fold above the baseline condition. This rebound increase in glycogen level above the pre-hypoglycemia level has been termed “supercompensation” (204). As a result of this observation, investigators have hypothesized that supercompensation of glycogen after hypoglycemia can supply extra fuel during subsequent hypoglycemia and contribute to development of impaired awareness of hypoglycemia. Supercompensation of glycogen has also been shown to occur in rodent brains after exposure to recurrent episodes of hypoglycemia (205). However, this increase in glycogen above baseline after single and recurrent hypoglycemia was not seen in other rodent studies (206). Differences in study design, including severity of hypoglycemia, use of anesthetics, and methods for monitoring glycogen content and glucose level during recovery from hypoglycemia, have been noted as possible explanations for the differences in these results (205). However, a recent rodent study found that glycogen supercompensation was independent of blood glucose levels in the post-hypoglycemia period (207). In vivo¹³C magnetic resonance spectroscopy in conjunction with IV infusions of 1-[¹³C]-glucose has been used to measure brain glycogen metabolism in humans. In heathy humans, increased glycogen levels were noted following a single episode of acute hypoglycemia but not after exposure to recurrent hypoglycemia (208). Glycogen levels were also noted to be similar in subjects with type 1 diabetes and impaired awareness of hypoglycemia compared with healthy controls (209). These data from human studies did not support the hypothesis that supercompensation of glycogen contributes to development of impaired awareness of hypoglycemia.

Altered hypothalamic neurosignaling

Studies in rodent models have suggested that hypothalamic GABA signaling plays an important role in regulating the counterregulatory hormone response to hypoglycemia (210). A decrease in the local availability of glucose within the VMH is thought to lower local GABA release, which in turn modulates the magnitude of hypoglycemia-induced hormone secretion (166). Increased GABAergic tone within the VMH has been implicated in the development of impaired awareness of hypoglycemia (211). Rodents exposed to recurrent hypoglycemia have elevated hypothalamic GABA levels at baseline, and they also fail to decrease these GABA levels in response to hypoglycemia (211, 212). These animals show an attenuated counterregulatory hormone response to hypoglycemia, and this response can be restored with local blockage of GABA_A receptors (212). In a human study, pharmacologic activation of GABA_A receptors has also been shown to reduce the counterregulatory response to subsequent hypoglycemia (213).

Hypothalamic AMPK has been shown to play a role in glucose sensing and modulating the counterregulatory response to hypoglycemia (214). Exposure to recurrent hypoglycemia leads to reduced AMPK activity in the VMH and may contribute to development of a defective counterregulatory response to hypoglycemia (215, 216). Pharmacological activation of AMPK within the VMH amplifies the counterregulatory response to acute hypoglycemia and improves the blunted counterregulatory hormone response in both diabetic and normal rats exposed to antecedent recurrent hypoglycemia (216, 217).

Other glucose-sensing mechanisms may also be defective in recurrent hypoglycemia. K^ATP channels open in GE neurons in response to low glucose, leading to hyperpolarization and reduced firing rate, and hypothalamic K^ATP channels are needed for an appropriate response to hypoglycemia. VMH injection of the K^ATP channel opener diazoxide augments the response to hypoglycemia in rodents with recurrent hypoglycemia (218). However, chronic treatment with K^ATP channel openers resulted in channel adaptation and decreased ability to sense hypoglycemia in vitro and in vivo (219, 220).

Increase cerebral oxidative stress

Acute hypoglycemia increases reactive oxygen species (ROS) levels in the hypothalamus (221). NO production by neuronal NO synthase and activation of the NO receptor soluble guanylyl cyclase (sGC) are thought to be critical for activation of VMH GI neurons during hypoglycemia and for the initiation of a counterregulatory response (162). Elevated ROS level leads to nitrosylation of sGC, which desensitizes sGC to NO (222). Hypoglycemia induced increases in hypothalamic ROS, leading to impaired action of NO on its receptor sGC (223), has been implicated in defective counterregulatory responses to hypoglycemia. In nondiabetic rats, enhancing the glutathione antioxidant defense system by pretreatment with N-acetyl-cysteine (NAC) prevented both hypoglycemia-induced VMH ROS production and an impaired counterregulatory hormone response to subsequent hypoglycemia (223). These results suggest that increasing glutathione with NAC could be a potential treatment option for impaired awareness of hypoglycemia. However, in a recent study by the same investigators, pretreatment with NAC did not preserve activation of VMH GI neurons by low glucose in diabetic rats exposed to recurrent hypoglycemia (224). Hyperglycemia associated with diabetes has also been shown to increase ROS levels in the brain (225). Zhou and Routh (224) hypothesize that the NAC-related increase in glutathione maybe not be sufficient to compensate for the combined oxidative stress resulting from both hypoglycemia and hyperglycemia in diabetic rats. They observed that overexpression of VMH thioredoxin (Trx)-1 preserved the counterregulatory response and activation of VMH GI neurons by low glucose in rats with streptozotocin-induced diabetes before but not after exposure to recurrent hypoglycemia. The Trx system, which is composed of the reduced form of NAD phosphate, Trx reductase, and Trx, is a key antioxidant system in defense against oxidative stress (226). These investigators concluded that recruitment of both glutathione and Trx antioxidant systems in the VMH may be needed to prevent impaired awareness of hypoglycemia in diabetes.

Role of adrenal NPY

NPY is cosecreted with catecholamines from adrenal chromaffin cells and has been postulated to be a possible mediator in the development impaired awareness of hypoglycemia. NPY is known to modulate the release and synthesis of adrenal catecholamine through adrenal Y1 receptors (227). In a recent study, investigators demonstrated that recurrent hypoglycemia reduced the epinephrine secretory capacity of mouse adrenal through a mechanism involving NPY activation of the Y1 receptor (228).

Role of opioid signaling

Endogenous opiates have been shown to modulate hormonal responses during hypoglycemia (229). Plasma β-endorphin levels increase in humans after hypoglycemia (230). IV administration of naloxone, an opioid receptor antagonist, during hypoglycemia augments the plasma epinephrine response to hypoglycemia in dogs (231) and humans (232). When infused during antecedent hypoglycemia, naloxone has been shown to prevent the development of the defective counterregulatory hormone response to subsequent hypoglycemia in healthy humans (233) and in subjects with type 1 diabetes (234). These studies suggest that opioid signaling may play a role in the development of impaired awareness of hypoglycemia. The mechanisms by which opioid receptor antagonists may prevent the development of impaired awareness of hypoglycemia are not known. Opioid receptors are expressed in VMH (235), where they could potentially modulate glucose sensing during hypoglycemia (229, 236). Opioid receptor antagonists are being tested as potential pharmaceutical therapy for prevention and treatment of impaired awareness of hypoglycemia (237).

Conclusion

Glucose homeostasis requires an organism to rapidly respond to changes in plasma glucose concentrations. In health, glucose-sensing neurons located in the brain and other regions detect the fall in glucose and trigger the hormonal and neural responses that restore euglycemia through a complicated network of interconnected defenses. In patients with diabetes who are treated with insulin and insulin secretagogues, these counterregulatory mechanisms are often insufficient to overcome the glucose-lowering effects of their medication. As a result, hypoglycemia is a common occurrence in the lives of such patients. Recurrent episodes of hypoglycemia also cause impaired awareness of hypoglycemia where the glucose level that elicits the response falls below the neuroglycopenic threshold. Better understanding of how glucose-sensing neurons respond to a fall in glucose and communicate this event to downstream effectors will significantly improve diabetes care.

Glossary

Abbreviations:

2DG

2-deoxy-d-glucose

AgRP

agouti-related peptide

AMPK

AMP-activated protein kinase

AP

area postrema

ARC

arcuate nucleus

BNST

bed nucleus of the stria terminalis

CCK

cholecystokinin

CNS

central nervous system

DMH

dorsomedial hypothalamic nucleus

DMV

dorsal motor nucleus of the vagus

GABA

γ-aminobutyric acid

GE

glucose-excited

GHRH

GH-releasing hormone

GI

glucose-inhibited

GLUT

glucose transporter

IML

intermediolateral column

K^ATP

ATP-sensitive potassium

LC

locus coeruleus

LH

lateral hypothalamus

MCH

melanin-concentrating hormone

MPOA

medial preoptic area

NAc

nucleus accumbens

NAC

N-acetyl-cysteine

NO

nitric oxide

NPY

neuropeptide Y

NTS

nucleus of the solitary tract

PBN

parabrachial nucleus

PET

positron emission tomography

POMC

pro-opiomelanocortin

PVH

paraventricular hypothalamic nucleus

ROS

reactive oxygen species

RVLM

rostral ventrolateral medulla

sGC

soluble guanylyl cyclase

SF1

steroidogenic factor-1

Trx

thioredoxin

VMH

ventromedial hypothalamic nucleus