Remote control of glucose-sensing neurons to analyze glucose metabolism

Abstract

The central nervous system relies on a continual supply of glucose, and must be able to detect glucose levels and regulate peripheral organ functions to ensure that its energy requirements are met. Specialized glucose-sensing neurons, first described half a century ago, use glucose as a signal and modulate their firing rates as glucose levels change. Glucose-excited neurons are activated by increasing glucose concentrations, while glucose-inhibited neurons increase their firing rate as glucose concentrations fall and decrease their firing rate as glucose concentrations rise. Glucose-sensing neurons are present in multiple brain regions and are highly expressed in hypothalamic regions, where they are involved in functions related to glucose homeostasis. However, the roles of glucose-sensing neurons in healthy and disease states remain poorly understood. Technologies that can rapidly and reversibly activate or inhibit defined neural populations provide invaluable tools to investigate how specific neural populations regulate metabolism and other physiological roles. Optogenetics has high temporal and spatial resolutions, requires implants for neural stimulation, and is suitable for modulating local neural populations. Chemogenetics, which requires injection of a synthetic ligand, can target both local and widespread populations. Radio- and magnetogenetics offer rapid neural activation in localized or widespread neural populations without the need for implants or injections. These tools will allow us to better understand glucose-sensing neurons and their metabolism-regulating circuits.

INTRODUCTION

Studies done over 50 years ago demonstrated that the brain is a primary user of the body’s energy. Despite comprising <2% of human body weight, the brain uses >20% of the body’s daily energy production (69, 144). Except under rare circumstances, such as fasting or in newborns fed maternal milk, where ketones can provide energy (133), the central nervous system (CNS) uses glucose as its fuel and requires a constant supply to maintain its processes (147). To ensure that this glucose supply is maintained, the CNS must be able to detect glucose levels and regulate peripheral organs to provide sufficient energy for bodily processes, including neural function. Glucose sensors, which react to changes in glucose availability, are present in the periphery and in the CNS, where they serve to initiate peripheral responses to maintain glucose homeostasis.

GLUCOSE UPTAKE INTO THE BRAIN

Glucose entry into the CNS appears to be a regulated process (78, 122). The blood-brain barrier, composed of tight junctions between brain microvascular endothelial cells, prevents free diffusion of glucose into the CNS. Glucose enters the CNS through high-affinity glucose transporter (GLUT) type 1 (GLUT1), which is expressed on the luminal and outer membranes of CNS endothelial cells (32, 107). GLUT1 is expressed in glia, the choroid plexus, and most ependymal cells (165). Under physiological conditions, GLUT1 expression in neurons is minimal. Under most conditions, glucose transport does not limit fuel availability to the CNS, but CNS glucose levels are markedly lower than circulating glucose concentrations (20). Microdialysis studies in the hippocampus (67), hypothalamus (5), and other brain regions (102) have demonstrated that interstitial brain glucose levels in awake animals range from 0.7 to 2.5 mM (from fed to fasted state) under normal conditions and from 0.2 to 4.5 mM in pathological states such as hypoglycemia and hyperglycemia, respectively. However, the circumventricular organs (including the median eminence and area postrema), which do not have a blood-brain barrier, are likely to be exposed to glucose levels closer to circulating concentrations (40). In addition, it appears that there is differential diffusion of glucose from cerebrospinal fluid into specific brain regions, such as the arcuate nucleus (ARC) (127). Interestingly, previous work has shown that fasting, hypoglycemia, and hyperglycemia can alter glucose transport into the brain (28, 86).

Once glucose enters the brain, additional transporters move glucose into cells (144). Most neurons express GLUT3, a GLUT with even higher affinity than GLUT1, which transports glucose into the cell (112). Low expression levels of other GLUTs have also been reported (101). GLUT2 is expressed in astrocytes, tanycytes, and neurons, and has been implicated in glucose sensing (90). Intracellular GLUT6 is found in neurons and may transport hexose molecules across organelle membranes. Similarly, low levels of GLUT8 have been described in defined brain regions, including the hippocampus and amygdala (112).

Once within neurons, glucose undergoes oxidative metabolism (3); in glia, however, glucose can undergo anaerobic glycolysis into lactate (55). Recent work indicates that glucose can also undergo glycolysis in activated neurons (33). The primary step in both of these processes is conversion of glucose to glucose-6-phosphate, which is predominantly carried out by hexokinase I, a very high-affinity kinase (26). This sequence of high-affinity transporters and kinases ensures that CNS glucose supplies are maintained over a wide range of physiological glucose concentrations.

GLUCOSE-SENSING POPULATIONS AND MECHANISMS IN THE CNS

Specialized groups of neurons, glucose-sensing neurons, use glucose as a signal as well as a fuel (51, 159). These neurons alter their firing rate with changes in glucose levels (143). Glucose-sensing neurons can be broadly categorized as glucose-inhibited (GI) or glucose-excited (GE) neurons (35). As glucose levels rise, the firing rate of GE neurons also rises and the activity of GI neurons is reduced. As glucose concentrations decrease, GI neurons are excited, whereas GE neurons decrease their activity. Glucose-sensing neurons are present in multiple brain regions but are found at high levels in hypothalamic regions, such as the ARC (25, 124), ventromedial nucleus (VMH) (77), lateral hypothalamus (LH) (16, 17), paraventricular nucleus of the hypothalamus (PVH) (104), and supraoptic nucleus (148). Interestingly, most regions are reported to harbor both GE and GI neurons (35, 177). In some areas, such as the VMH, there may be some anatomic segregation with higher expression of GE neurons in the dorsomedial VMH but an enrichment of GI neurons in the ventrolateral VMH (139, 170). There is also some anatomic division of glucose-sensing neurons in the nucleus accumbens (NAc) (123).

Glucose-sensing neurons have been identified in regions outside the hypothalamus, notably in several nuclei of the reward system, including the NAc (123), amygdala, paraventricular thalamus (89), prefrontal cortex (110), and hippocampus (160). Many of these regions form part of the mesolimbic and mesocortical dopaminergic pathways and are integral for motivation and reward processing, including addiction and different aspects of motivated food intake. Approximately 25% of neurons in the NAc have been reported to be glucose-sensing (123), with the NAc shell predominantly expressing GI neurons and the NAc core predominantly expressing GE neurons. However, these experiments used glucose at 500 mM, which is higher than the expected physiological concentration (123). In the rat, 6% of tested medial amygdala neurons increased their firing with increased glucose (from 0.5 to 2.5 mM), while 7.5% were GI neurons (177). Hippocampal neurons may also be glucose responsive. They express glucokinase (GK) (91), ATP-dependent K⁺ (K_ATP) channels (70), and GLUT1 (23), as well as the metabolism-independent heterodimeric sweet taste receptor T1R2/T1R3, in the CA fields and dentate gyrus (130).

Glucose-sensing neurons are also found in hindbrain regions. The nucleus of the solitary tract (NTS) contains glucose-sensing populations (12, 90), with 35% of NTS neurons excited by 5 mM glucose and 21% inhibited by 5 mM glucose (108). Many of these neurons also respond to melanocortin signaling (108), which can act in the NTS to modulate food intake and body weight (36). In the lateral parabrachial nucleus (PBN) of the pons, cholecystokinin (CCK)-expressing neurons are glucose responsive, showing a significant increase in activity as glucose levels fall from 5 to 0.5 mM (39, 53). Lateral PBN neurons are also known to play a role in appetite suppression (39, 53).

Multiple mechanisms have been reported to contribute to neuronal glucose sensing, as summarized in Fig. 1. The best-characterized glucose-sensing mechanism parallels that in pancreatic beta cells. GE neurons in the ARC and VMH use GK (hexokinase IV), rather than hexokinase I, for the initial step of glucose metabolism, catalyzing phosphorylation of glucose to glucose-6-phosphate (35, 76). Unlike hexokinase I, GK is a low-affinity kinase with a high K_m for glucose and without end-product inhibition (4). GK activity is, therefore, proportional to glucose concentrations. Increasing glucose concentrations increase glucose metabolism and the intracellular ATP-to-ADP ratio (97), which, in some GE neurons, leads to closure of the K_ATP channel, resulting in depolarization and activation (2). Additional K⁺ and cation channels have also been described to mediate glucose sensing in some populations of GE neurons, including GE neurons of the PVH (104) and gonadotropin-releasing hormone neurons (134), but sensing mechanisms remain unknown in other cell types (16, 104, 134). GK is also expressed in GI neurons in the VMH and other regions (149). In VMH GI neurons, reduced glucose levels decreased cystic fibrosis transmembrane conductance regulator Cl⁻ currents, which in turn depolarized and activated the cells (116). Pathways involving AMP-activated protein kinase and nitric oxide are likely to be involved. In addition, AMP-activated protein kinase has been postulated to be a glucose sensor in some ARC and anterior hypothalamic neurons (117, 134).

Fig. 1.

Schematic drawing of proposed neural glucose-sensing mechanisms. Glucose can be transported into the neuron by glucose transporters (GLUTs) or cotransported with Na⁺ by sodium-glucose transport proteins (SGLTs). The latter mechanism of entry would lead to depolarization and increased excitability of the neuron by generation of an inward current. Inside the neuron, glucose phosphorylation by glucokinase (hexokinase IV) acts as a rate-limiting step proportional to extracellular glucose concentrations. Lactate can be taken up by the neuron via monocarboxylate transporters (MCTs) to serve as a glycolytic substrate and metabolized into pyruvate. Mitochondrial uptake and metabolism of pyruvate lead to an increased ATP-to-ADP ratio and inhibition of ATP-dependent K⁺ (K_ATP) channels, resulting in plasma membrane depolarization and neurotransmitter release. Alternatively, an increased ATP-to-ADP ratio could increase Na⁺/K⁺-ATPase activity or trigger cystic fibrosis transmembrane conductance regulator (CTFR_Cl) channel opening, resulting in plasma membrane hyperpolarization. In orexin-expressing glucose-inhibited neurons, both glucose and 2-deoxy-d-glucose (2DG) interact with an unknown protein to trigger K⁺ leak (K_leak) channels, resulting in plasma membrane hyperpolarization. The metabolism-independent sweet taste receptor (STR) T1R2/T1R3 has been implicated in the glucose response of non-proopiomelanocortin (POMC) glucose-excited neurons in the arcuate nucleus (ARC). Transient receptor potential canonical (TRPC) type 3 cation channels are implicated in the glucose response of glucose-excited neurons of the mediobasal hypothalamus via a mechanism dependent on reactive oxygen species (ROS)/H₂O₂ production, rather than K_ATP channels. However, the exact mechanism in which ROS activates TRPC channels remains elusive. Glucose-6-P, glucose-6-phosphate.

Some glucose-sensing neurons do not require glucose metabolism for sensing. Both glucose (5 mM) and 2-deoxy-d-glucose (2DG), a nonmetabolizable glucose analog (1 and 5 mM), can inhibit orexin-expressing GI neurons in the LH (59). Glucose and 2DG interact with an unknown molecular player to initiate events that lead to the opening of a K⁺ leak channel and hyperpolarization. Sodium-glucose cotransporters (SGLTs) are present in some glucose-sensing neurons. SGLT1, -3a, and -3b are present in subpopulations of GE neurons, where expression would link inward transport of Na⁺ with glucose to produce depolarization (77, 119). A small number of VMH GI neurons have also been reported to express SGLT1 (77). A number of hypothalamic glucose-sensing populations also express additional GLUTs, including the lower-affinity GLUT2 and insulin-dependent GLUT4, both of which are expressed in some GE- and GI-sensing neurons (77).

The anorexigenic proopiomelanocortin (POMC) neurons of the ARC have been proposed to be glucose-sensing; however, the glucose-sensing mechanism in this population is unclear (42). The Kir6.2 subunit of the K_ATP channel has been reported to mediate the glucose response of ARC POMC GE neurons at between 3−5 mM and 5−10 mM glucose (71, 124), whereas the mitochondrial fission regulator dynamin-related protein 1 has been implicated at 0.2−2.5 mM glucose (140). However, other studies have found that POMC neurons do not respond to glucose or respond via an indirect, rather than a direct, mechanism (41, 42, 167). On the other hand, metabolism-independent sweet taste receptors have been implicated in the glucose-sensing properties of ARC non-POMC GE neurons at high (10 mM) glucose concentrations (82). The ARC also harbors neuropeptide Y/agouti-related protein-expressing (NPY/AgRP) neurons, a subset of which is inhibited by an increase in glucose from 1 to 10 mM (42, 115). NPY/AgRP GI neurons appear to be directly activated by decreased glucose levels (from 2.5 mM to 0.1 mM) via K⁺ channel closure (66), similar to LH NPY neurons (98). Both ARC POMC neurons and NPY/AgRP neurons have been reported to express GK in rats (35) and mice (149).

Orexin/hypocretin neurons of the LH are inhibited by increasing glucose levels (from 0.2 to 4.5 mM) via tandem-pore K⁺ (K_2P) channels (17, 59). However, while TASK1 and TASK3 channels of the K_2P channel superfamily increase membrane excitability of orexin neurons, they do not appear to be essential for the glucose response of orexin neurons, as revealed by studies using 0.2−5 mM glucose in TASK1 knockout mice, TASK3 knockout mice, or TASK1/3 double-knockout mice (58, 63).

A majority of putative PVH parvocellular neurons are GE and GI neurons, responding to 0.2−10 mM glucose (104). Although the glucose-sensing mechanism is not fully established, the response of PVH GE neurons to glucose does not appear to require K_ATP channels (104).

Transient receptor potential (TRP) canonical type 3 (TRPC3) channels were recently shown to be involved in glucose detection in vivo and required for the glucose response of isolated GE neurons of the mediobasal hypothalamus (including the ARC and VMH) in vitro (identified by a response to an increase from 2.5 to 10 mM glucose) (24). However, it remains to be determined which specific neurons within the mediobasal hypothalamus express TRPC3 (24).

Studies have identified neuronal populations that respond to glucose fluctuations >5 mM, usually referred to as high-GE or high-GI neurons. The firing rate of isolated orexin neurons of the LH was shown to be inhibited at 10−30 mM glucose (18, 171). High-GE and high-GI neurons that respond to >5 mM changes in glucose concentration have also been identified in the mediobasal hypothalamus (24) and, specifically, in the ARC (42, 43, 71), and these high-GE ARC neurons appear to sense glucose via a K_ATP channel-independent mechanism (43, 167).

Glucose sensing is not limited to neurons. Tanycytes lining the third ventricle express several proteins that may be involved in glucose sensing, such as GK, K_ATP channels, and GLUT2. These cells respond to elevated glucose (tested at 0.5−8 mM) by increasing intracellular Ca²⁺ in vitro (37), although these cells also respond to nonmetabolizable 2DG (49). Astrocytes have also been reported to be involved in glucose sensing, and GLUT2 expression has been described in astrocytic populations in the hypothalamus (62) and brain stem (100). Furthermore, astrocytes take up glucose, a large part of which enters the glycolytic pathway and is released as lactate via monocarboxylate transporters into the extracellular space (10, 125). Lactate and glucose exist in similar concentrations in the extracellular space, and there is growing evidence that lactate may be an energy source for neurons (52, 141). Astrocytes also play essential roles in the neurovascular unit, and they may regulate cerebral blood flow based on the metabolic state of neurons (60).

EVIDENCE FOR NEURAL REGULATION OF METABOLISM

The brain not only detects changes in glucose levels, but animal and human studies suggest that the CNS plays a role in glucose regulation. Peripheral and intraventricular injections of 2DG, which mimics hypoglycemia, increase expression of c-fos, an early-immediate gene and a marker of neural activation, in many brain regions expressing glucose-sensing populations, including the ARC (14, 109), VMH (48), and LH (15). Central 2DG administration in rats suppresses pancreatic insulin release (50, 113), increases plasma glucagon, enhances hepatic glucose, induces lipolysis (27), and increases food intake (114). Central injection of glucose suppresses plasma glucose (120). In humans, glucose infusion has been shown to alter CNS activity in hypothalamic regions, including the VMH and LH (121). Additional studies have examined the effects of hypothalamic lesions and electrical stimulation on glucose metabolism. VMH lesioning in mice and rats can modulate pancreatic hormone release (61), peripheral glucose uptake (29), and glucose production by the liver (126). The neural and hormonal responses to restore glucose levels after hypoglycemia are also diminished. ARC lesions with monosodium glutamate prevented hypoglycemia-induced feeding (94), while PVH lesions increased feeding and body weight but only altered insulin secondary to obesity (145). In rodents, VMH, PVH, and anterior hypothalamus stimulation using electrodes increased blood glucose (142). LH stimulation had the opposite effect, decreasing blood glucose in rats (31) and rabbits (142) and increasing glycogenesis (7). Similarly, posterior hypothalamic stimulation reduced blood glucose (154). Lesions and stimulation outside the hypothalamus also regulate blood glucose. Electrical stimulation of the dorsal or median raphe nuclei (75) or the rostral ventrolateral medulla (84) increased blood glucose, while stimulation of the nucleus ambiguus (6), NTS, and dorsal motor nucleus of the vagus increased plasma insulin (73). Lesioning of the medial amygdala was shown to impair the counterregulatory hormonal response to hypoglycemia (177).

More recently, neuron-specific drivers have been used to modulate the expression of specific factors, such as receptors, channels, and neurotransmitters, in distinct cell populations in the hypothalamus and elsewhere to determine how they contribute to glucose regulation. These studies are summarized in Table 1. However, lesioning and electrical stimulation studies affect multiple neural cell types as well as fibers passing through the regions involved. The consequences of transgenic expression or deletion of specific proteins can be masked or blunted by compensatory effects. So technologies that can rapidly and reversibly activate or inhibit defined neural populations provide invaluable tools to investigate how specific neural populations regulate blood glucose.

Table 1.

Summary of studies examining effects of manipulation of anatomically or molecularly defined neural populations on glucose metabolism and appetite

Area	Cell Type	Intervention	Effect on Metabolism	Feeding Effect	Reference
ARC	All	Overexpression of hexokinase I	Increased fasting glucose, impaired glucose tolerance	?	135
	All	Leptin injection	Increased glucose uptake into BAT	?	157
	AgRP + POMC	Leptin receptor KO	Increased insulin, normal glucose tolerance	Increased	164
	AgRP	Ribosomal S6K1 KO	Muscle insulin resistance	Normal	146
	AgRP	Postembryonic deletion	Decreased insulin	Decreased	8
	AgRP	Postembryonic deletion	Normal glucoprivic feeding	Decreased	95
	AgRP	Insulin receptor KO	Loss of insulin suppression of hepatic glucose production	Normal	85
	AgRP	AMPK2α KO	Loss of glucose sensing in subpopulation of AgRP neurons, no effect on insulin	Unchanged	25
	AgRP	Postnatal ablation by diphtheria toxin	Increased lipid utilization, decreased sympathetic outflow	Blunted refeeding	74
	AgRP	TXNIP deletion	Improved fasting glucose and glucose tolerance	Decreased weight	11
	AgRP	Glutamaterigc GluN2B receptor KO	Correction of hyperglycemia in ob/ob mice	Unchanged	162
	AgRP	Leptin receptor replacement in ob/ob mice	Correction of hyperglycemia	Normal	57
	AgRP	GHSR restoration in KO mice	Correction of hypoglycemia with calorie restriction	Normal	166
	AgRP	Carnitine acetyltransferase deletion	Reduced glucose disposal and suppression of glucose production	Decreased	129
	AgRP	CRF receptor 1 KO	Reduced hepatic glucose production in fasting in females	Normal	88
	AgRP	Loss of GABA signaling	Glucose intolerance in young mice	Reduced	106
	AgRP	Antagonism of AP1 transcription factor	Improved glucose clearance and insulin sensitivity	?	72
	POMC	Ribosomal S6K1 KO	Hepatic insulin resistance	?	146
	POMC	AMPK2α KO	Loss of glucose sensing in POMC neurons	Increased	25
	POMC	ATP-insensitive KATP channel	Impaired glucose sensing in POMC neurons, impaired glucose tolerance	?	124
	POMC	Inositol-requiring enzyme 1 deficiency	Impaired glucose and insulin tolerance, reduced hepatic insulin sensitivity	Increased	172
	POMC	Antagonism of AP1 transcription factor	Improved glucose clearance and insulin sensitivity		72
	POMC	Arcuate POMC KO	Improved glucose tolerance		22
PVH	All	MT-II injection	Increased glucose uptake into BAT	?	157
	SIM1	CB1 receptor KO	Increased insulin sensitivity on high-fat diet	Unchanged	19
DMH	All	NPY knockdown	Increased hepatic insulin sensitivity	?	93
	All	CCK receptor replacement in OLETF rats	Improved glucose homeostasis	Decreased	178
LH	All	MC4R reexpression	Improved glucose tolerance	Unchanged	111
	MCH	Insulin receptor KO	Improved glucose tolerance on high-fat diet	Normal	68
	MCH	ATP-insensitive KATP channel	Impaired glucose tolerance	Normal	83
	Orexin	Orexin KO	Loss of diurnal glucose rhythm	Unchanged	161
VMH	All	Leptin injection	Increased glucose uptake into skeletal muscle, BAT	?	157
	All	MT-II	Increased glucose uptake into skeletal muscle, BAT	?	157
	All	LIN28a viral overexpression	Improved glucose tolerance and insulin sensitivity on high-fat diet	Unchanged	79
	SF1	Leptin receptor KO	Glucose intolerance, hyperinsulinemia	Increased	9
	SF1	VGLUT KO	Hypoglycemia during fasting	?	158
	SF1	Insulin receptor KO	Protected against insulin resistance on high-fat diet	Reduced	81
	SF1	FOXO1 KO	Improved glucose tolerance and insulin sensitivity in muscle	Unchanged	80
	SF1	SOCS3 KO	Improved glucose tolerance	Reduced	174
MPO	All	Norepinephrine injection	Increased plasma glucose and insulin	?	47
VLM	CA	CA neuron ablation	Abolished hyperglycemic response to stress	?	176
PBN	CCK	Leptin receptor KO	Enhanced counterregulatory response to hypoglycemia	Unchanged	45
PAG	Leptin receptor	Optogenetic activation	Increased blood glucose	Unchanged	44

AgRP, agouti related peptide; AMPK2α, AMP-activated protein kinase α-subunit; AP1, activator protein 1; ARC, arcuate nucleus of the hypothalamus; BAT, brown adipose tissue; CA, catecholaminergic; CB1, cannabinoid type 1; CCK, cholecystokinin; CRF, corticotropin-releasing factor; DMH, dorsomedial nucleus of the hypothalamus; FOXO1, forkhead box protein O1; GABA, γ-aminobutyric acid; GHSR, growth hormone secretagogue receptor; GluN2B, glutamate ionotropic receptor NMDA type 2B; K_ATP, ATP-dependent K⁺ channel; KO, knockout; LH, lateral hypothalamus; LIN28a, RNA binding protein; MCH, melanin-concentrating hormone; MC4R, melanocortin 4 receptor; MPO, medial preoptic area; MT-II, melanotan II; NPY, neuropeptide Y; OLETF, Otsuka Long Evans Tokushima fatty rat; PAG, periaqueductal gray; PBN, parabrachial nucleus; PVH, paraventricular nucleus of the hypothalamus; POMC, proopiomelanocortin; SF1, steroidogenic factor 1; SIM-1, single-minded 1; SOCS3, suppressor of cytokine signaling 3; S6K1, ribosomal S6 protein kinase 1; TXNIP, thioredoxin-interacting protein; VGLUT, vesicular glutamate transporter; VLM, ventrolateral medulla; VMH, ventromedial hypothalamic nucleus.

TRANSGENIC ANIMAL MODELS FOR STUDIES OF GLUCOSE METABOLISM

Transgenic mouse models have emerged as invaluable tools for targeting specific neuronal populations such as subsets of glucose-sensing neurons, particularly in combination with viral tools for optogenetics, chemogenetics, and magneto-/radiogenetics. However, because of the heterogeneous nature of glucose-sensing neurons and the wide variety and low expression of glucose-sensing markers, specific glucose-sensing populations remain difficult to genetically pinpoint. GK is a marker for some glucose-sensing populations but is not expressed in all glucose-sensing neurons, and some GK-expressing neurons do not respond to glucose. For instance, in the VMH, GK is expressed in 64% of GE neurons, 43% of GI neurons, and 8% of non-glucose-sensing neurons (77). Another marker of some glucose-sensing neurons, the Kir6.2 subunit of the K_ATP channel, is expressed in 42% of GE and GI neurons and 15% of non-glucose-sensing neurons (77). Viral tools can also be targeted to neural populations with cre expression driven by specific neuropeptide promoters such as CCK, POMC, or AgRP. However, often only a portion of these neuropeptide-expressing neurons are glucose-sensing. Targeted mutations in putative glucose-sensing mechanisms can be valuable to address the role of specific neurons. For instance, a mouse model with transgenic expression of mutant Kir6.2 subunits preventing ATP-mediated closure of K_ATP channels was used to disrupt glucose sensing in POMC neurons (124). Moreover, TASK1/3 knockout mice have been used to address the glucose-sensing mechanisms of orexin neurons (58, 63). Advanced Cre-LoxP transgenic mouse models allowing targeting of more defined subtypes of neurons will be invaluable tools for targeting glucose-sensing neurons in the future.

NEUROMODULATORY TOOLS TO ASSESS METABOLIC REGULATION

Several techniques for acute neuromodulation use light, chemicals, or electromagnetic fields to gate ion channels or activate receptors. The most widely used technologies today are optogenetics and chemogenetics. These technologies allow acute activation or silencing of anatomically and/or genetically defined neural populations. When applied to glucose-sensing neural populations, they allow us to examine the effects of modulating these neural populations on metabolic parameters such as blood glucose, hormone release, and sensitivity. Unlike life-long genetic manipulations, compensatory mechanisms that may blunt or hide essential roles of specific populations are unlikely. In addition, the precise temporal control and reversibility may allow better correlation between the activity of specific neurons and physiological changes. These tools, in combination with existing technologies such as transgenic mice, allow unprecedented investigation of the neural populations and circuits regulating glucose metabolism.

Optogenetics

Optogenetic tools use the targeted expression of light-responsive opsins to activate or inhibit specific neural populations. Channelrhodopsins are robustly and rapidly (on the order of milliseconds) opened by light and can be used to deliver a specific pattern of neural activity to the target neural population (13). A wide array of channels, with multiple temporal patterns and other characteristics, can be applied to investigate the roles of neural populations. Because light spread in the CNS is relatively limited (105), optogenetics can be used to modulate cell bodies or specific projection sites with precision (128). However, the limited light spread can be a drawback in some studies. For most optogenetic studies, a fiber-optic implant is needed to deliver light stimuli to the target region, especially for regions deep in the CNS, such as the hypothalamus. Only populations immediately adjacent to the fiber will receive sufficient light delivery to be activated or inhibited. Thus, target populations that are spread throughout a CNS region or structure can be difficult to modulate by optogenetics. In addition, optogenetic studies require light delivery to the implant itself through a head-mounted light source or an optical cable. Animals need to be restrained for attachment of the light source or cable, and this can be stressful and alter glucose metabolism even in well-handled animals. Tethering by the light cable during studies can also be a stressor or interfere with the behaviors being investigated.

Optogenetics has been used extensively for mapping of neuronal networks underlying the control of feeding (see Ref. 136 for review), but far fewer studies have examined the cell populations and circuits controlling glucose metabolism. Photoactivation of VMH steroidogenic factor 1 (SF1) neurons was shown to induce hyperglycemia, whereas photoinhibition had no effect on glucose levels under basal conditions but severely impaired recovery from hypoglycemia (103). Photostimulation of LH neurons expressing melanin-concentrating hormone (MCH) highlighted their role in sensing the nutrient value of sucrose and their potential roles in nutrient preference (34). Optogenetic tools have also been used to modulate GLUT2-expressing neurons. Activation of glucose-inhibited, GLUT2-expressing neurons in the paraventricular thalamus increased sucrose intake (89). Other studies that examined photostimulation of a population of glucose-inhibited, GLUT2-expressing neurons in the NTS showed that activation significantly increased vagal activity and blood glucagon levels (90).

Chemogenetics

Chemogenetics offers an alternative approach for neuromodulation based on engineered molecules and ligands. These systems use G-protein coupled receptors (1, 118) or ion channels (92, 96) engineered to be unresponsive to their natural ligand but sensitive to an inert (or largely inert) reagent. The modified channel or receptor is expressed in the target neural population and stimulated by peripheral injection of its ligand, leading to channel opening or receptor activation and neuromodulation (155). Unlike optogenetics, chemogenetic tools can be used to activate or silence cells in multiple regions or across large structures. The animals are not tethered and can move freely during experiments. However, some of the chemogenetic technologies are relatively slow (of the order of minutes) (137), and animals still need to be handled to administer the ligand, which can act as a stressor to perturb some metabolic studies. Off-target effects from the ligand or its metabolites may also require consideration (56). Chemogenetics has been used to characterize the metabolic effects of activating glucose-sensing neurons (163, 168). The effects of activating VMH SF1 neurons on glucose uptake and insulin sensitivity in skeletal muscle, brown adipose tissue, and heart were tested using the designer receptor exclusively activated by designer drugs (DREADD) hM3Dq (30). Activation of SF1 neurons increased hepatic glucose production and glucose uptake into skeletal muscle. In the ARC, chemogenetic activation of AgRP neurons, but not POMC neurons, impaired glucose tolerance and insulin sensitivity of brown adipose tissue (153). Interestingly, the same study found that optogenetic activation of AgRP neurons only impaired insulin sensitivity, a discrepancy possibly related to methodological differences (153). Chemogenetic stimulation of neurons expressing preproglucagon peptide (GCG neurons) suppressed glucose production without affecting glucose uptake and reduced metabolic rate and food intake in the fed and fasted states of lean GCG-cre mice (54). In GCG-cre mice with diet-induced obesity, the effect of chemogenetic GCG neuronal stimulation on glucose production was abolished, but the effect of reduced food intake persisted (54). Chemogenetic experiments have also been used to assess the effects of activating LH MCH neurons but showed that excitation of MCH neurons had no effect on glucose tolerance or insulin sensitivity (68).

Chemogenetic tools have also been used to examine the roles of glucose-sensing neurons outside the hypothalamus. DREADD modulation of glucose-inhibited CCK neurons in the lateral PBN increased blood glucose, suggesting that these neurons are likely to be involved in the counterregulatory response to hypoglycemia (45). Similarly, DREADD activation of leptin receptor-expressing neurons in the PBN or periaqueductal gray increased blood glucose and adrenal sympathetic nerve activity (44). Chemogenetic activation of an additional glucose-sensing population, catecholaminergic neurons in the ventrolateral medulla, also increased blood glucose (176).

Radio- and Magnetogenetics

Remote neural modulation using electromagnetic fields, which has been termed “magnetogenetic” or “radiogenetic” modulation, allows minimally invasive neural regulation. Low-frequency, alternating magnetic fields are already used clinically to program pacemakers (64) or neural stimulation devices. Unlike tissue, which does not absorb these fields to any significant extent (173), metal oxide nanoparticles absorb energy from electromagnetic fields to a degree that depends on the size, shape, and material of the nanoparticle and the strength of the applied field (46, 65, 99, 132). The energy captured by nanoparticles from electromagnetic fields can be used to open a modified ion channel, which, in turn, regulates neural activity. A number of technologies using either injected or genetically encoded nanoparticles have been developed for this purpose. Iron oxide nanoparticles modified to prevent cell uptake can be injected into the CNS (21) and will then act on target cells that have been altered to express a genetically modified multimodal ion channel, TRP vanilloid 1 (TRPV1). An alternative approach is to express a modified TRPV1 channel with an epitope tag inserted into the first extracellular loop in target cells. This is combined with injection of iron oxide nanoparticles coated with a tag-specific antibody that then bind to the channel (150). When alternating magnetic fields are applied, the nanoparticles absorb this energy and transduce it into TRPV1 channel activation, leading to Na⁺ and Ca²⁺ entry into the cell and depolarization. However, injected nanoparticles can only target a limited area to modulate cell activity. In addition, the particles can be cleared from the injection site by endocytosis (138), so multiple nanoparticle injections may be needed for repeated neural activation.

Modified versions of this technology have been developed to overcome the need for nanoparticle injection. These systems use genetically encoded nanoparticles, rather than extracellular particles, to activate ion channels. Genetically encoded nanoparticles in these technologies are produced intracellularly by a modified form of ferritin. Ferritin is an almost ubiquitous iron storage protein consisting of 24 subunits that self-aggregate to form a protein shell. Once assembled, iron ions from the cytoplasm are taken up into the ferritin shell and form an iron oxide nanoparticle (156, 175). In the first system, outlined in Fig. 2, two components are required: 1) ferritin, which has been modified by addition of green fluorescent protein (GFP), and 2) a modified TRPV1 channel with the addition of a small anti-GFP domain. When these components are expressed in the target cells, the GFP-tagged ferritin self-assembles and iron oxide nanoparticles form within the shell, while the modified TRPV1 channel is expressed at the cell surface and tethers the GFP-ferritin through its anti-GFP domain (151, 152). In the second system, the alternative multimodal ion channel TRPV4 is expressed as a fusion protein with ferritin and targeted to defined cell populations (169). In both cases, when the target cells are treated with alternating or static magnetic fields (radio- or magnetogenetic systems, respectively), energy absorbed by nanoparticles in the ferritin lead to TRP channel activation, depolarization, and neural activation. The mechanism(s) linking energy absorption by the nanoparticles to channel opening remain unknown. One possibility is that the applied magnetic field leads to a change in entropy and a very localized temperature change, which, in turn, activates the associated ion channel.

Fig. 2.

Neuronal modulation using radio- or magnetogenetics. A viral vector can be used to deliver the 2 components required for radio- or magnetogenetics, GFP-ferritin and anti-GFP-TRPV1, to the target cell. GFP-ferritin is a ubiquitous iron storage protein that has been modified by addition of green fluorescent protein (GFP), whereas anti-GFP-TRPV1 is a transient receptor potential vanilloid 1 (TRPV1) cation channel modified by addition of a small anti-GFP domain. Once expressed by the cell, the GFP-ferritin self-assembles into a shell and sequesters iron ions in a nanoparticle. TRPV1 subunits self-assemble on the cell membrane. The GFP tag of GFP-ferritin will bind to the anti-GFP domain of the first extracellular loop of the modified TRPV1 channel, thus tethering GFP-ferritin to the modified TRPV1 channel. Application of a radio frequency or a magnetic field will transfer energy to the iron oxide nanoparticles within the GFP-ferritin and trigger ion channel opening, Na⁺ and Ca²⁺ entry into the cell, and depolarization.

Electromagnetic treatment leads to rapid (on the order of <1 s) neural activation to assess the function of defined neural populations. This technology has been further modified to develop a tool for remote neural inhibition. Mutation of an amino acid residue in the pore region of the TRPV1 channel leads to a change in its ion conductance from cations (predominantly Ca²⁺ and Na⁺) to anions (Cl⁻) (87). When this mutant TRPV1 (TRPV1^mutant) channel is fused to the anti-GFP domain and expressed in cells with GFP-ferritin, electromagnetic field treatment opens the TRPV1^mutant channel, leading to Cl⁻ entry, hyperpolarization, and silencing of neural activity. The components required for radio- or magnetogenetics can be delivered by virus injection into defined anatomic regions and/or targeted to defined cell populations with use of cre-lox technology (151, 169).

Radio- and magnetogenetic tools have a number of advantages. They can be applied to activate or silence populations in multiple areas or across a large CNS region. There is no need to handle the experimental animal to connect the optical cable or inject a ligand before the study, and the animal can move freely during the experiment. The range of tools is currently smaller than those for optogenetics or chemogenetics, but these techniques allow fast, wireless neuromodulation in a freely moving animal without an implant and without the need for injection. These characteristics make these technologies particularly useful for studying physiological functions that can be disrupted by stress, such as glucose regulation.

RADIO- AND MAGNETOGENETIC NEURAL MODULATION TO ASSESS GLUCOSE METABOLISM

The roles of a population of GK-expressing VMH glucose-sensing neurons in glucose metabolism and feeding have been examined using electromagnetic modulation. GK-expressing neurons were targeted by injection of the cre-dependent activating construct (adeno-associated virus containing anti-GFP-TRPV1 and GFP-ferritin) into the VMH in GK-cre and wild-type (WT) mice (151). In fed GK-cre mice, both alternating magnetic field (radio waves, 31 mT, 30 min) and gradient magnetic field treatment used to activate VMH GK neurons significantly increased blood glucose. This effect was proportional to field strength and similar in time course and magnitude to optogenetic stimulation of VMH GK neurons (15-ms pulse width for 30 min, 5 Hz) (151). Plasma glucagon and growth hormone levels were significantly increased, as was expression of hepatic glucose-6-phosphatase, a gluconeogenic enzyme. Magnetic field activation of VMH GK neurons also significantly increased feeding. Magnetogenetic activation in WT mice or in the dorsal striatum, a brain area not known to express glucose-sensing neurons, did not alter blood glucose. The endocrine and behavioral responses generated by activation of VMH GK neurons mirror many of the responses to hypoglycemia, suggesting that GI neurons are the functionally dominant population in this region (151).

The effects of magnetogenetic silencing of VMH GK neurons were also assessed using cre-dependent expression of anti-GFP-TRPV1^mutant/GFP-ferritin in GK-cre and WT mice (151). In contrast to activation, inhibition of GK-expressing neurons in GK-cre mice led to a significant decrease in blood glucose in the fasted state with increased plasma insulin, reduced hepatic expression of glucose-6-phosphatase, and decreased feeding after a 4-h fast compared with WT mice. In addition, silencing VMH GK neurons diminished the counterregulatory response to neuroglycopenia following 2DG injection (151).

In combination, these studies demonstrate the utility of remote neural activation and silencing using electromagnetic fields. In addition, they suggest that GK-expressing VMH neurons may acutely modulate food intake and glucose metabolism and, therefore, may contribute to maintaining blood glucose during fasting and hypoglycemia.

SUMMARY AND FUTURE DIRECTIONS

Studies using targeted neuromodulatory tools such as optogenetics, chemogenetics, and radio-/magnetogenetics have already transformed our knowledge of the cells and circuits underlying many neurological diseases. These tools are now being used to assess how specific glucose-sensing neural populations regulate metabolism, including plasma glucose. When combined with in vivo imaging techniques to assess neural activity, tissue clearing to define circuits in whole brains (38, 131), and technologies to examine gene expression in defined neural populations, these tools may allow us to better understand how defined cells and their projections contribute to metabolic regulation. Future studies will determine if similar cells and circuits play a role in human metabolic regulation and how these are disrupted in metabolic disease.

GRANTS

A. Alvarsson is supported by a postdoctoral fellowship from the Swedish Society for Medical Research. Support for this work was also provided by National Institutes of Health Grants MH-105941 and 1R01 NS-097184, American Diabetes Association Grant 1-17-ACE-31, an Einstein-Mt. Sinai Diabetes Research Center Pilot and Feasibility Award (National Institute of Diabetes and Digestive and Kidney Diseases Grant P30 DK-020541), and an Alexander and Alexandrine Sinsheimer Scholar Award.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

A.A. and S.A.S. prepared figures; A.A. and S.A.S. drafted manuscript; A.A. and S.A.S. edited and revised manuscript; S.A.S. approved final version of manuscript.