Brain Defense of Glycemia in Health and Diabetes

Mammalian survival depends on the biological defense of key metabolic variables against myriad environmental perturbations. These variables (e.g., cardiac output, blood pressure, core temperature, body fat mass, blood glucose) are homeostatically controlled within narrow but adaptive ranges—or biologically defended levels (BDLs)—through cooperation between nervous and endocrine systems. For example, the central nervous system (CNS) and hypothalamic-pituitary-adrenal axis (i.e., cortisol) cooperate to maintain the BDL of blood pressure (1), the CNS and hypothalamic-pituitary-thyroid axis (i.e., thyroxine/triiodothyronine) maintain core temperature (2), and the CNS and adipocytes (i.e., the adipose-derived hormone leptin) maintain body fat mass (3). Similarly, the CNS and islet β-cells (i.e., insulin) cooperate to control the BDL of glycemia (BDL_G) (4). Synchronized surveillance of these highly interdependent metabolic variables by interoception enables the CNS to coordinate simultaneous control of each (Fig. 1).

The environmental challenge posed by cold exposure exemplifies CNS homeostatic control. On cold exposure, increased sympathetic nervous system (SNS) outflow stimulates heat production (5–7) (brown adipose tissue thermogenesis and skeletal muscle shivering) to maintain core temperature. The resulting rapid rise in glucose utilization—facilitated by increased insulin sensitivity in thermogenic tissues (8)—should theoretically plummet blood glucose levels. Instead, parallel SNS outflow to islet β-cells adaptively suppresses insulin secretion such that blood glucose levels remain stable throughout the transition to a cold environment (7,9). The net effect of SNS coupling of insulin secretion to insulin sensitivity is a shunting of glucose from nonthermogenic to thermogenic tissues along with maintenance of stable levels in blood (i.e., to supply the brain). The brain engages yet a third adaptive response to cold exposure: rapid activation of hypothalamic neurons called Agouti-related peptide (AgRP) neurons induces hyperphagia to match food intake to the energy expenditure of heat production (10). The parallel engagement of thermoregulatory, glucoregulatory, and energy homeostasis systems highlights the unique capacity of the brain to drive feed-forward homeostatic control to ensure that key metabolic variables—core temperature, blood glucose, and body fat mass—are unperturbed by the environmental challenge.

Here, we focus on CNS control of the BDL_G. Conceptually, the BDL_G is the expected steady-state level achieved after all competing mechanisms that influence the ambient glucose level, both insulin-dependent and -independent, have been accounted for (4). Accordingly, a single effector mechanism, such as insulin acting on its target tissues, does not change the BDL_G but, rather, influences the ambient level as one of multiple effectors contributing to the BDL_G. When ambient levels rise above the BDL_G, insulin action is among the balancing mechanisms weighted to restore the BDL_G. Conversely, if insulin action results in ambient levels below the BDL_G (e.g., pharmacological insulin-induced hypoglycemia), then other (counterregulatory) mechanisms are engaged and weighted (and insulin mechanisms unweighted) to restore the BDL_G. Although β-cells are glucose sensing and autonomously execute glucose-stimulated insulin secretion (GSIS), the brain via the autonomic nervous system (ANS) can and continuously does modulate GSIS (see below) (11–13). This arrangement is analogous to how the brain coordinates many complex tasks (e.g., locomotion) involving multilevel sensorimotor integration. Locomotion, in its simplest form, is executed by local spinal cord circuits using “central pattern generators” (to provide coordinated, rhythmic motor output controlling muscle activation) and reflex loops that entrain motor output to body mechanics. Layered onto this basic function is descending modulation by the brain that enables task- and context-specific adjustments (14). Likewise, descending modulation of GSIS enables the brain to tune β-cell function according to task and context. The brain integrates interoceptive signals of circulating and stored fuel, exteroceptive signals of food availability, and the anticipated glycemic effects of planned behaviors and environmental challenges to coordinate its descending control of both insulin-dependent (9,11–13) and -independent (4,15,16) glucose effectors.

We highlight evidence from animal and human studies supporting core CNS homeostatic functions that maintain the BDL_G, including 1) afferent input from central and peripheral sensors (Fig. 2, part 1), 2) central integration and processing in the hypothalamus and hindbrain (Fig. 2, parts 2 and 3), and 3) efferent control via ANS innervation of peripheral metabolic tissues (Fig. 2, part 4, and, highlighting key references, Tables 1,–,3). We present these components in “reverse” order, starting where we have the most understanding and proceeding to where we have the most opportunity. In parallel, we discuss how defective CNS homeostatic control may contribute to diabetes pathogenesis.

Pancreatic islets are richly innervated by the SNS and parasympathetic nervous system (PNS) (11,12,17). Sympathetic outflow travels from preganglionic neurons in the spinal cord intermediolateral cell column to the prevertebral celiac and superior mesenteric ganglia, from which postganglionic fibers project through the splanchnic nerves to reach the islets. Parasympathetic outflow is conveyed by the vagus nerve from the dorsal motor nucleus (DMX) to intrapancreatic ganglia, from which postganglionic short-range axons project to islets. Although innervation patterns at the islet are slightly different in rodents and humans (18,19), autonomic activity affects islet secretion similarly across species (11,12,17). Postganglionic SNS neurons release norepinephrine onto β-cell α2-adrenergic receptors to restrict calcium influx and restrain β-cell depolarization, ultimately suppressing GSIS and basal insulin secretion (20). Conversely, postganglionic PNS neurons release acetylcholine onto M3 muscarinic acetylcholine receptors (mAChRs), activating Gq signaling to raise intracellular calcium levels and enhance β-cell depolarization and GSIS (21). Although the primary stimulus for β-cell depolarization is glucose (β-cell glucose sensing reviewed in 22), ANS modulation of the β-cell membrane potential directly influences β-cell glucose sensitivity (i.e., changes the glucose and insulin relationship).

Physiologically, PNS activity drives meal-associated GSIS, as the mAChR antagonist atropine or vagotomy suppresses postprandial insulin secretion (23–25). This activity also underlies the “cephalic phase” of insulin secretion, referring to the early increase in circulating insulin (within minutes of meal ingestion) in anticipation and independent of a rise in blood glucose levels. Consistently, pharmacological inhibition or genetic deletion of M3 mAChRs results in impaired insulin secretion and glucose intolerance in mice (26). SNS modulation of β-cell function also plays fundamental physiological roles, demonstrated by the coupling of insulin secretion and insulin sensitivity during cold exposure (see introduction) (9). Although not discussed here, the brain also indirectly modulates β-cell function through neuroendocrine hormone secretion (e.g., somatostatin, cortisol) (27,28).

Atropine-mediated blockade of mAChRs suppresses GSIS in humans as in other species. Furthermore, human studies have demonstrated that this suppression is observed following an oral, but not intravenous, glucose challenge (29). This suggests an interaction between the PNS and gut-derived incretin peptides known to augment insulin secretion, such as glucagon-like peptide 1 (GLP-1) (30). Following a meal, GLP-1 is secreted by intestinal enteroendocrine L cells onto synapse-like contacts, called “neuropods,” formed with GLP-1 receptor (GLP1R)-expressing vagal afferents (31). This signal ascends to distributed neurocircuits in the hindbrain and hypothalamus and ultimately feeds back onto glucose-effector organs (32–34). Thus, although pharmacological doses of GLP1R agonists may act directly on β-cells (in addition to GLP1R-expressing neurocircuits), low levels of blood-borne GLP-1 following a meal suggest that physiological GLP-1 signaling may rely on the PNS (30,32). Alternatively, atropine’s suppression of GSIS only following an oral glucose challenge may implicate PNS stimulation of cephalic-phase insulin secretion.

SNS modulation of β-cells has also been demonstrated in humans. In healthy subjects, intravenous infusion of α-adrenergic antagonist phentolamine results in a small but significant increase in basal insulin secretion and augments GSIS following an intravenous glucose challenge (35). Basal SNS outflow thereby normally provides mild restraint of β-cell secretory function. The same studies conducted in subjects with type 2 diabetes (T2D) reveal a significantly greater increase in basal insulin secretion than in healthy subjects, and fivefold greater GSIS (35). This suggests that excess α-adrenergic signaling in β-cells contributes to insulin secretory dysfunction characteristic of T2D. Consistently, subjects with diabetes had significantly elevated levels of circulating catecholamines (35).

Three-dimensional histological studies of human pancreata demonstrate robust remodeling of islet innervation in obesity and T2D (18,36,37). Islets from subjects with obesity had increased fatty infiltration associated with BMI (18). Compared with healthy subjects, those with T2D had increased islet nerve density and number of innervated islets, with preserved nerve contacts onto β-cells (37). Although these analyses were not specific to autonomic innervation (i.e., spinal sensory nerves also innervate the islet), more recent work has revealed that islet sympathetic innervation specifically is increased in T2D versus healthy subjects. Islets from subjects with diabetes had threefold higher sympathetic nerve density, and this was positively correlated with β-cell dedifferentiation (36). Furthermore, as samples in this study were derived from patients undergoing pancreatic surgery, hyperglycemic clamps performed preoperatively enabled an assessment of β-cell function. Multiple functional parameters of insulin secretion were shown to be negatively correlated with increased sympathetic innervation (36). Consistently, worsening glucose tolerance, from normal to impaired tolerance to T2D, correlated with increased islet sympathetic fibers. These studies provide an anatomical framework for understanding autonomic dysfunction in T2D pathogenesis.

Beyond autonomic innervation of the β-cell, excess sympathetic tone has been linked to T2D pathogenesis (38). In prospective cohort studies, plasma norepinephrine levels both at baseline and in response to stress (i.e., sympathetic reactivity) predict future T2D risk. Reduced heart rate variability, another biomarker of dysautonomia with excess sympathetic tone, has also been associated prospectively with significantly increased T2D risk (39–41). Cross-sectional studies also corroborate this link: T2D patients exhibit increased resting muscle sympathetic nerve activity compared with those with impaired glucose tolerance (42) or obesity (43) alone. Perhaps providing an etiological clue, human studies suggest that diet-induced obesity is associated with chronically elevated sympathetic nerve activity (44,45).

Like the islet, the liver receives SNS postganglionic fibers via the splanchnic nerves originating in the celiac and superior mesenteric ganglia, and PNS preganglionic fibers from the vagus originating in the DMX. In mammals, this innervation is largely localized to the portal vein, hepatic artery, and bile ducts, with generally less innervation of the parenchyma (17). In the mid-19th century, French physiologist Claude Bernard reported that pricking the floor of the cerebral fourth ventricle in rabbits increased hepatic glucose production (HGP), an effect that was abolished by prior transection of the splanchnic nerves of the SNS but not the vagal hepatic branch of the PNS (46). More than a century later, electrical stimulation of the rabbit splanchnic nerve was shown to increase (within 30 s) enzymatic activity of liver glycogen phosphorylase and glucose-6-phosphatase, thereby mobilizing glycogen stores, in a manner that persisted after adrenalectomy or pancreatectomy (i.e., not mediated by adrenal or islet hormones) (47). Since then, our understanding of autonomic control of HGP and the converse, hepatic glucose uptake (HGU), has advanced significantly.

The liver is a critical effector of glucose homeostasis, capable of producing glucose (via gluconeogenesis and glycogenolysis) in the fasted state and disposing glucose in the fed state. Autonomic modulation of these functions is exemplified in transitions between fasted and fed states, which have been studied extensively in dogs (48–50). In the fasted state, basal sympathetic tone to the liver blocks HGU to maintain the circulating level. Following a meal, glucose levels in the portal vein are much higher than they are in the arterial circulation. This negative arterial-portal glucose gradient (called the portal signal) suppresses sympathetic tone to produce a rapid, robust, and insulin-independent rise in HGU (50). Earlier work in rats showed that portal glucose infusion reduced firing rates in the hepatic splanchnic (sympathetic) nerve and this effect was blocked by sectioning of the hepatic (afferent) branch of the vagus nerve (51). Altogether, this suggests that hepatic vagal afferents mediate the portal signal, resulting in suppressed hepatic sympathetic tone.

Elevated hepatic sympathetic tone has been implicated in T2D pathogenesis. While short-term epinephrine infusion in dogs decreases insulin-dependent glucose disposal, long-term infusions dramatically reduce insulin-independent glucose disposal (with no change in insulin secretion or sensitivity) (52). In dogs chronically fed a high-fat, high-fructose diet, the portal signal is unable to induce HGU, resulting in glucose intolerance (53). Finally, sympathetic denervation along the common hepatic artery in these dogs partially restores HGU, independent of insulin, and improves glucose tolerance (54). These findings suggest that diet-induced obesity elevates hepatic sympathetic tone to produce glucose intolerance.

The Bergman minimal model of glucose kinetics, a mathematical modeling approach to data from a frequently sampled intravenous glucose tolerance test, quantifies insulin-dependent and -independent glucose disposal in human subjects (55). When it was developed, the parameter estimate of insulin-independent disposal, called glucose effectiveness (GE), was a means of capturing glucose’s ability to promote its own (passive) disposal down a concentration gradient after an intravenous load. Studies with use of the minimal model quickly revealed that GE was, in fact, an active and dynamic process that could be manipulated with experimental intervention (56,57) and altered by disease (58,59). Subsequent work showed that the insulin-independent mechanisms quantified by GE were modulated by the CNS (16).

Minimal modeling studies have shown that T2D is characterized not only by reduced insulin sensitivity but also by reduced GE (59). Nearly twofold reductions in GE have also been reported in subjects without diabetes with glucose intolerance in comparisons with control subjects with normal glucose tolerance (60). Most notably, in a 25-year prospective study of 155 normoglycemic offspring of parents who both had T2D, the 25 subjects who developed diabetes had significantly lower GE and insulin sensitivity present at least 10 years prior to the diagnosis (61). Reduced GE imparted the greatest (15×) relative risk of progression to T2D among the variables analyzed. These studies provide compelling evidence that in addition to β-cell dysfunction and insulin resistance (62), T2D pathogenesis involves reduced insulin-independent glucose disposal. As a potent modulator of both insulin-dependent (9) and -independent glucose disposal (16), the ANS (i.e., brain efferent control) plays a critical role in diabetes pathogenesis.

Although activity in multiple and distributed brain areas may influence glycemia, subsets of neurons in “preautonomic” networks of the hypothalamus (4,15,63–65) and hindbrain (66–69) have been the best characterized. Neurons in these networks receive multiple modalities of sensory input conveying the ambient glucose level (see Table 3), integrate this information with other feed-forward cues of planned behavior (70,71), and provide descending modulation through the ANS and neuroendocrine system to maintain the ambient level within the BDL_G (72). We focus on subsets of neurons in the hypothalamic arcuate nucleus (ARC) and ventromedial nucleus (VMN) that support the brain’s critical role in glucose homeostasis and diabetes pathogenesis.

In the ARC, AgRP and proopiomelanocortin (POMC)-expressing neurons comprise antagonistic populations that regulate melanocortin signaling through downstream projections to the paraventricular nucleus (PVN). Both AgRP (an inverse agonist) and the POMC-derived peptide, α-melanocyte–stimulating hormone (an agonist), bind melanocortin-4 receptors in the PVN, but to signal contrasting conditions of either fuel deficiency (AgRP neurons) or excess (POMC neurons). This system has a well-established role in energy homeostasis (73). However, mounting evidence, including experimental activation or inhibition of AgRP or POMC neurons using chemogenetic tools, shows this system’s role in glucose homeostasis and that its dysfunction produces glycemic abnormalities characteristic of T2D (74,75).

AgRP neurons lie at the ventromedial-most corner of the ARC, directly adjacent to the median eminence circumventricular organ and to tanycytes lining the 3rd ventricle, an arrangement that may provide privileged access to humoral and cerebrospinal fluid signals, respectively (76). Diet-induced obesity induces hypothalamic gliosis (77,78) and renders AgRP neuron activity less responsive to several exteroceptive and interoceptive signals of fuel availability (79). AgRP neuron activity is elevated in multiple rodent models of diabetes, including the leptin-deficient ob/ob model of T2D (80,81) and the streptozotocin (STZ)-induced insulin-deficient model of type 1 diabetes (T1D) (82). A key role for AgRP neurons in T2D pathogenesis is supported by BDL_G normalization in rodent models of T2D achieved with a single intracerebroventricular (ICV) administration of fibroblast growth factor-1 (FGF1) (83). This effect 1) is recapitulated by FGF1 microinjection into the ARC (84), 2) requires intact melanocortin signaling (85), 3) depends on restored extracellular matrix specializations around AgRP neurons (86), and 4) elicits lasting suppression of AgRP neuron activity, based on transcriptomic (85) and electrophysiological (87) data. Finally, suppression of AgRP neuron activity is also implicated in the remarkable ability of ICV leptin administration to normalize glycemia in the STZ model of T1D (see next section) (82).

The CNS has well-established roles in mounting coordinated autonomic and neuroendocrine counterregulatory responses (CRRs) to hypoglycemia and fasting to avoid fuel deficiency and starvation (88). CRR activation produces sympathoadrenal epinephrine release, hypercortisolemia, hyperglucagonemia, increased HGP, and suppression of GSIS, serving to restore hypoglycemia to the BDL_G. Over the last decade, the VMN has emerged as a key node in the neurocircuit mediating CRRs—independent laboratories have identified overlapping subsets of VMN neurons required to fully activate CRRs (15,65,89,90). For example, optogenetic-based photoinhibition of VMN neurons expressing steroidogenic factor 1 blocks recovery from insulin-induced hypoglycemia, associated with suppressed glucagon and corticosterone levels (89). Furthermore, experimental activation of these neurons during euglycemia rapidly produces diabetes-range hyperglycemia via the same CRRs, including suppressed GSIS despite profound hyperglycemia (65,89–91). This supports hierarchical control whereby descending neuromodulation overrides what the islet would otherwise do autonomously. This scenario also provides insight into diabetes pathogenesis. CRRs elicited by fasting or hypoglycemia are paradoxically activated in uncontrolled insulin-deficient T1D (and less dramatically in T2D), suggesting that the brain activates CRRs in response to reduced afferent input from leptin (fasting), glucose (hypoglycemia), or insulin (T1D) (13), and these aberrant CRRs may drive diabetic hyperglycemia.

Support for this hypothesis comes from the robust, multiply reproduced (independent laboratories) finding that ICV leptin administration normalizes the BDL_G in rodent T1D models through insulin-independent mechanisms (4,92). Rather than simply lowering the ambient level, ICV leptin reestablishes the BDL_G in the normal range and maintains this level despite perturbations by glucose or insulin administration. This is recapitulated by leptin microinjection into the VMN (93). Inactivation of a subset of VMN neurons expressing the cholecystokinin receptor B (CCKBR)—the activity of which drives CRRs to hypoglycemia (68)—reduces the BDL_G to near normal in STZ T1D mice (in the absence of insulin) (15). Although counterintuitive that hyperglycemia following STZ-induced β-cell loss is partly driven by VMN neurons, this highlights the significance of the brain’s CRR to deficiency of insulin (and leptin, which also plummets in uncontrolled T1D) (13). VMN^CCKBR neuron inactivation in euglycemic mice also reduces the BDL_G in an insulin-independent manner (i.e., reduces HGP) (15), demonstrating the brain’s key role in establishing the BDL_G.

The glycemic threshold for triggering CRRs is elevated by 40% in subjects with well-controlled T2D compared with subjects without diabetes, such that CRRs can be initiated even with normal glucose levels (94). Since the magnitude of CRRs is preserved, however, it follows that the BDL_G in subjects with T2D would be elevated. This aligns with evidence that blood glucose levels continue to be defended within a narrow range in T2D, only at a higher BDL_G (95). Data from subjects with monogenic diabetes due to glucokinase mutations provide insight into an underlying mechanism. CRRs are activated with higher magnitude and at even higher glycemic thresholds in these subjects than in those with common forms of T2D, suggesting that defective glucose sensing mediated by glucokinase plays a role (see next section) (96).

Clinically, elevated glycemic thresholds for engaging CRRs may explain paradoxical outcomes of tight glucose control in T2D patients in the intensive care unit (97,98). Emerging evidence suggests that while tight control results in lower mortality among patients without diabetes, it is associated with higher mortality among those with preexisting diabetes. In T2D patients, tight control may excessively activate CRRs that not only influence glycemia but also produce significant physiological stress that could lead to worse outcomes (97).

Neurosurgical patients undergoing epilepsy monitoring with depth electrodes offer unique access to human brain electrophysiology. A recent study of three patients with depth electrode monitoring and simultaneous time-aligned continuous glucose monitoring revealed significant correlations between high-frequency activity (HFA) (70–170 Hz oscillations) at multiple brain sites and interstitial glucose (IG) variations (99). In one patient with hypothalamic depth electrodes, the strongest lag-corrected HFA-IG correlation represented hypothalamic HFA leading IG by 2.8 h. The brain areas that showed the highest HFA-IG correlation were those with highest functional connectivity with the hypothalamus. Although limited to a few subjects, these data suggest that central processing of glycemia is distributed among multiple brain areas in humans, potentially integrating both interoceptive and exteroceptive information.

Multiple hypothalamic and hindbrain neuron populations, including subsets in the ARC and VMN, express the molecular machinery required for cellular glucose sensing (88,100,101). These cells, denoted glucose-excited (GE) or glucose-inhibited (GI), increase their activity to signal a rise or fall, respectively, in the ambient glucose level. GE neuron glucose sensing occurs through a mechanism shared with the β-cell: following GLUT2-mediated entry, glucose is phosphorylated by glucokinase and metabolized to increase intracellular ATP, ultimately triggering closure of K_ATP channels and depolarization. GI neurons (e.g., both ARC AgRP neurons and VMN neurons whose activity drives CRRs) are activated when the ambient glucose concentration, and thus glucokinase activity, falls. Reduced glucokinase activity increases the intracellular AMP-to-ATP ratio, activates AMPK neuronal nitric oxide synthase, and leads to chloride channel closure and cell depolarization (88). Colocalization of these neurons within glucoregulatory brain areas appears cogent with a putative role in both sensing and responding to changes in glycemia. However, most of these neurons (e.g., VMN) only have access to brain interstitial fluid (ISF). Relative to blood glucose, for brain ISF glucose changes are both dampened (i.e., brain ISF glucose concentration is 20%–30% of blood [102]) and delayed (i.e., peak changes in brain ISF glucose occur 30–60 min following changes in blood [63,103,104]), such that the physiological utility of central neuronal glucose sensing may be limited (105). An exception may be glucose-sensing neurons near circumventricular organs—including ARC neurons by the median eminence and nucleus tractus solitarius (NTS) neurons by the area postrema—where fenestrated capillaries allow exposure to the circulation (106–108). Within the ARC, GI neurons tend to be located medially and GE neurons laterally, both responsive to changes in brain ISF glucose in the physiological range (108).

Humoral signals, such as insulin and leptin, may also provide interoceptive cues that are proxies of the circulating glucose level. In ARC AgRP neurons, these two signaling pathways converge on K_ATP channels that, when activated, hyperpolarize and inhibit these cells (82,109–112). While the respective contributions of insulin and leptin signaling to activate these channels in different physiological contexts is unclear, evidence shows that 1) activating hypothalamic K_ATP channels reduces blood glucose levels through suppression of HGP via hepatic vagal efferents (112,113), 2) hyperglycemia-induced insulin-independent suppression of HGP depends on these K_ATP channels (16), 3) K_ATP channels in AgRP neurons are required for leptin-mediated energy and glucose homeostasis (82), and 4) insulin action in AgRP neurons also activates K_ATP channels and suppresses HGP, but AgRP neuron–specific insulin receptor knockout mice display normal energy and glucose homeostasis (111). Altogether, these data suggest that leptin may provide an insulin-independent signal mediating hyperglycemia-induced suppression of HGP.

Consistent with the animal data above, euglycemic pancreatic clamp studies in human subjects treated with the brain penetrant K_ATP channel activator diazoxide showed insulin-independent suppression of HGP (113). When the same studies were performed in subjects with T2D (and animal models of T2D), diazoxide failed to suppress HGP (a key driver of diabetic hyperglycemia) (114). Finally, in subjects without diabetes, hyperglycemia induced insulin-independent suppression of HGP, which was abrogated by K_ATP channel blocker glyburide (16). This work suggests that hypothalamic K_ATP channels may also regulate HGP in humans. We must be careful to acknowledge that many neurons throughout the nervous system express K_ATP channels and that the response to K_ATP-targeted drugs may represent a combined effect on multiple populations. However, a provocative possibility is that T2D pathogenesis involves defective K_ATP channel–dependent sensory feedback onto AgRP neurons, consistent with the effects of AgRP neuron–specific K_ATP channel deletion (82). This possibility remains to be further investigated.

Just as cutaneous thermo-sensory afferent nerves supply continuous temperature information to the brain’s thermoregulatory system, peripherally positioned glucose sensors are ideally positioned to provide rapid and dynamic glycemic information to the brain’s glucoregulatory system (105). The speed of this communication is a discrete advantage over mechanisms described above. Peripheral glucose-sensing neurons are found in the oral cavity, gastrointestinal tract, carotid body, and portal-mesenteric vein (PMV). We focus on PMV glucose sensors and their well-characterized role in hypoglycemia detection and CRRs (88).

The PMV was identified as a site for hypoglycemia detection through elegant studies that revealed locally “clamping” euglycemia in the portal vein dramatically reduced the sympathoadrenal response to systemic hypoglycemia (115). Earlier work had identified afferent fibers in the hepatic branch of the vagus nerve that innervated the PMV wall and displayed firing rates that were inversely related to portal vein glucose concentration (116). These vagal afferents were a suitable candidate for hypoglycemia detection and CRR initiation but were ultimately shown to be dispensable for this response (88). In contrast, selective ablation of TRPV1-expressing spinal sensory afferents—with topical application of capsaicin to the PMV—resulted in >90% suppression of the sympathoadrenal response to hypoglycemia (117,118). These ascending pathways may provide parallel streams signaling hypo- or hyperglycemia. Vagal afferents innervating the PMV wall terminate in the DMX (119) and exhibit linear suppression of firing rate over a wide range of portal glucose concentration (ranging from euglycemia to hyperglycemia, 5–20 mmol/L) and may be involved primarily in responses to elevated PMV glucose levels (e.g., the portal signal) (88). Spinal sensory afferents innervating the PMV wall travel in the splanchnic nerve, traverse the celiac-superior mesenteric ganglion to their cell bodies in dorsal root ganglia, and ultimately ascend in the spinal cord to the NTS, DMX, and area postrema (88,120). Lesioning at multiple sites along this ascending pathway dramatically diminishes sympathoadrenal responses to hypoglycemia. Finally, from the hindbrain to the hypothalamus, two parallel pathways mediating CRRs have been identified. One projects from the ventrolateral medulla to the PVN, lateral hypothalamus, and ARC (88); a second pathway projects from the NTS to leptin receptor–expressing neurons in the parabrachial nucleus to VMN^CCKBR neurons, previously described (67,68).

This functional connectivity appears tedious, but these networks enable the speed of neurotransmission to facilitate the brain’s capacity to immediately account for challenges to the BDL_G. This speed is evident in recent work showing that within 60 s, interventions that raise circulating glucose levels lead to suppression of VMN^PACAP neuron activity (63), known to drive CRRs (91).

Our understanding of the brain’s defense of glycemia in health and diabetes is in a period of rapid growth—building on a rich existing knowledge of islet-based control—but many questions remain. We outlined the three basic components of the CNS control system, with evidence from studies primarily focused on one component, but functional connectivity between these processes is poorly understood. Furthermore, how the brain senses circulating glucose levels in vivo in a manner that facilitates its rapid efferent modulation of glucose effectors is a fertile area for discovery. This is evidenced by recent work identifying neurons in the dorsomedial hypothalamus (121), VMN (63), and lateral hypothalamus (122) exhibiting rapid in vivo responses (<30–90 s) to changing glycemia (i.e., before changes in brain ISF glucose). A fundamental challenge to overcome in this area is the timescale difference between changes in neuronal activity (milliseconds in sensory networks) and changes in glycemia (seconds to minutes). This dramatic difference complicates the ability to precisely follow the flow of sensory information from the incipient glycemic variation to neuronal activity changes at ascending nodes in peripheral and/or central networks.

Despite these challenges, this work holds the promise of CNS-based therapeutic strategies that may improve diabetes management, highlighted by the success of GLP-1 agonists. Recent in vivo evidence suggests that GLP1R-expressing neurons in the dorsomedial hypothalamus are glucose sensing and their activity reduces blood glucose (121). Minimal modeling experiments in human subjects administered GLP-1 agonists showed that in addition to a 66% augmentation of GSIS, there was a 50% enhancement in insulin-independent glucose disposal (i.e., mechanisms regulated by the CNS) (56). This dual benefit of GLP1 agonists may underlie their incredible efficacy, targeting glycemia both postprandially (i.e., predominantly insulin-dependent mechanisms) and in the basal state (i.e., predominantly insulin-independent mechanisms) (13). In addition to systemic pharmaceutical approaches targeting the brain, there is potential for application of more direct, device-based neuromodulation strategies (4). Proof of this concept comes from the unanticipated result of increased insulin sensitivity and suppression of HGP with deep brain stimulation of the striatum (123). Neurosurgical approaches to diabetes may be distant, but incorporation of a brain-based understanding in therapeutic development is likely to reveal new high-yield targets.

Funding. Z.M. is supported by the National Institutes of Health Director’s New Innovator Award Program/National Institute of Diabetes and Digestive and Kidney Diseases (DP2DK128802), the Department of Defense Peer-Reviewed Medical Research Program (W81XWH2010250), and the Barrow Neurological Foundation (18-0025-30-05).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Z.M. and C.F. wrote and revised the manuscript. C.F. created the figures. Z.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 84th Scientific Sessions of the American Diabetes Association, Orlando, FL, 21–24 June 2024. A video presentation can be found in the online version of the article at https://doi.org/10.2337/dbi24-0001.