Sympathetic circuits regulating hepatic glucose metabolism: where we stand

Abstract

The prevalence of metabolic disorders, including type 2 diabetes mellitus, continues to increase worldwide. Although newer and more advanced therapies are available, current treatments are still inadequate and the search for solutions remains. The regulation of energy homeostasis, including glucose metabolism, involves an exchange of information between the nervous systems and peripheral organs and tissues; therefore, developing treatments to alter central and/or peripheral neural pathways could be an alternative solution to modulate whole body metabolism. Liver glucose production and storage are major mechanisms controlling glycemia, and the autonomic nervous system plays an important role in the regulation of hepatic functions. Autonomic nervous system imbalance contributes to excessive hepatic glucose production and thus to the development and progression of type 2 diabetes mellitus. At cellular levels, change in neuronal activity is one of the underlying mechanisms of autonomic imbalance; therefore, modulation of the excitability of neurons involved in autonomic outflow governance has the potential to improve glycemic status. Tissue-specific subsets of preautonomic neurons differentially control autonomic outflow; therefore, detailed information about neural circuits and properties of liver-related neurons is necessary for the development of strategies to regulate liver functions via the autonomic nerves. This review provides an overview of our current understanding of the hypothalamus-ventral brainstem-liver pathway involved in the sympathetic regulation of the liver, outlines strategies to identify organ-related neurons, and summarizes neuronal plasticity during diabetic conditions with a particular focus on liver-related neurons in the paraventricular nucleus.

CLINICAL HIGHLIGHTS.

Excessive hepatic glucose production is a contributor to hyperglycemia in type 2 diabetes mellitus. Neural regulation of hepatic functions, among others, plays an important role in the etiology of diabetes mellitus. Therefore, elucidating these mechanisms is necessary to better understand the clinical aspects of diabetes mellitus. Increased sympathetic nervous system activity is involved in the development and progression of metabolic disorders; thus establishing neural circuits and properties of neurons in the brain-liver pathway could lead to specific therapeutic interventions.

1. INTRODUCTION

In healthy subjects, regulation of energy metabolism is a complex, yet efficient, mechanism involving multiple organs that participate in the maintenance of constant body weight throughout adult life. A proper balance between energy intake (food) and energy expenditure (basal metabolism and exercise) is crucial and requires precise mechanisms that sense the needs of the organism as well as appropriate future storage once immediate needs are met. Glucose is the primary substrate used by cells to produce energy, and the maintenance of tightly regulated blood glucose levels is critical for survival. In response to increased blood glucose concentration, the pancreas secretes insulin to signal insulin-sensitive tissues that glucose uptake is necessary. Glucose is stored as glycogen mainly in the liver and muscle, and excess glucose contributes to weight gain through lipogenesis (1–3). In the past decades, overconsumption of added sugar, which includes sucrose and high-fructose corn syrup, became an increasingly important problem due to the highly lipogenic nature of fructose. Please, see reviews by Peterson et al. (2) and Ter Horst et al. (4) for more details on mechanisms. In contrast, when blood glucose concentration is decreased and energy is needed, glycogen and noncarbohydrates are converted to glucose and released into the bloodstream to supply cellular needs. These mechanisms, which contribute to the maintenance of tightly controlled glycemia by regulating availability and storage of glucose depending on the energy status of the organism, are defined as glucose homeostasis.

The brain plays a critical role in the regulation of these processes. Energy homeostasis regulation involves cross talk between the central nervous system (CNS) and peripheral organs and tissues. The CNS receives information about the metabolic status of the body through circulating signals detected by specialized neurons located in the CNS and by sensory pathways. Afferent signals arrive from peripheral organs, including the gastrointestinal tract where nutrients are absorbed, or from other tissues such as adipose tissue, liver, and pancreas. One important area for nutrients, including glucose sensing, is the hepatic portal vein, which is innervated by both spinal and vagal afferents (5–7). Sensory signals through the dorsal root ganglia or the left nodose ganglion are relayed to the dorsal vagal complex. A recent study conducted in rats used herpes simplex virus-1 (an anterograde, transsynaptic viral tracer) to define sensory pathways from the hepatic portal and superior mesenteric veins (8). Results confirmed that both spinal and vagal sensory information reaches the dorsal vagal complex. Findings also showed that sensory information from the portal and superior mesenteric veins are part of the vago-vagal and spino-vagal reflexes, which are able to modulate energy homeostasis, including food intake and glycemia (8, 9). Information about the metabolic status of the body is processed in a variety of brain nuclei, and responses are triggered to modulate organ functions. Neural pathways ultimately transmit commands to organs and tissues via the autonomic nervous system (ANS), consisting of sympathetic and parasympathetic divisions. In general, the parasympathetic nervous system is activated in response to energy intake, with the primary purpose of digestion and storage of nutrients for future needs. On the other hand, the sympathetic nervous system (SNS) is activated in response to situations when energy is needed, such as low blood glucose levels.

Metabolic disorders, comprising obesity, type 2 diabetes mellitus (T2DM), hypertriglyceridemia, and nonalcoholic fatty liver disease, are rising worldwide despite treatment advances. Intriguingly, an imbalance of the ANS, primarily increased sympathetic nervous system activity and decreased parasympathetic nervous system activity, is recognized as a contributing factor to the pathogenesis and progression of these diseases (10–12). There are four major theories linking elements of metabolic diseases and changes in β-adrenergic responsiveness with sympathetic activation: all have different viewpoints depending on the order of events (reviewed in Ref. 10), specifically whether demonstrating sympathetic activation as a cause or result of metabolic dysregulation. Ample evidence supports both points of view, but a prospective study convincingly showed that the sympathetic nervous system plays an important role in development of the metabolic syndrome (11). Subjects were followed for 10 years, whereupon investigators found that increased sympathetic nerve activity measured by higher plasma noradrenaline levels preceded hyperinsulinemia, impaired glucose metabolism, and hypertension. The authors concluded that sympathetic nervous system overactivity is the initial event, followed by hyperinsulinemia (11). Here, we focus on the role of the sympathetic nervous system in the regulation of hepatic glucose metabolism in the context of obesity and diabetes.

Type 2 diabetes mellitus, which is 1 of the 10 leading causes of death worldwide, according to the World Health Organization, is characterized by elevated blood glucose levels and is the most common metabolic disorder. Over time, uncontrolled hyperglycemia leads to serious complications, comprising but not limited to retinopathy, nephropathy, and neuropathy, and it is associated with an increased prevalence of cardiovascular disease, cognitive decline, and many more conditions. In addition to glucose uptake, hepatic glucose production (HGP) is one of the principal mechanisms responsible for retaining glucose levels in the appropriate range. In T2DM, excessive HGP, the result of inappropriately elevated glycogenolysis and overactive gluconeogenic pathways, is a major contributor to hyperglycemia. Elevated HGP triggers persistent insulin secretion contributing to the decreased sensitivity of insulin-sensitive tissues and thus to decreased effectiveness of insulin to normalize blood glucose levels (13, 14). Intriguingly, SNS overactivity was proposed as one of the main mechanisms underlying prolonged HGP, indicating the need to understand detailed sympathetic regulation of the liver.

The traditional view of autonomic regulation of the liver is based on studies conducted in the 1960s demonstrating that activation of sympathetic nerves increases HGP and glycogenolysis, whereas parasympathetic nerve activation promotes glucose storage and reduction in HGP (15, 16). Although this view is largely accurate, recent observations suggest the existence of more complex regulatory pathways and recognize our limited knowledge about the organization and mechanisms governing organ-specific circuits, both centrally and peripherally. In this review, we discuss approaches for organ-related circuit identification and highlight neurocircuits involved in the sympathetic regulation of the liver.

2. IDENTIFICATION OF ORGAN-RELATED NEURONS WITH FOCUS ON THE BRAIN-LIVER PATHWAY

The central nervous system regulates autonomic functions through direct and indirect control of the autonomic outflow. Direct control is achieved by monosynaptically connected neurons that regulate the activity of sympathetic and parasympathetic preganglionic neurons in the spinal cord and brainstem. Indirect control is provided via interneurons with monosynaptic connections to preganglionic neurons. These preautonomic cells modulate the activity of preganglionic neurons; however, they are part of a broader network and receive information from a wide variety of nuclei that project to these preautonomic regions. Ultimately, through this network of neurons, the autonomic outflow can be adjusted based on the body's requirements (17).

Rodent sympathetic preganglionic neurons involved in liver innervation were identified in the intermediolateral column (IML) of the spinal cord at the level of T7-T12 (18, 19) with projections to sympathetic postganglionic neurons, mainly located in the celiac and superior mesenteric ganglia (19, 20). Sympathetic fiber distribution at the level of the liver is highly species dependent with distinct differences noted in rats and mice compared with humans, guinea pigs, and cats (21–27). Consistent with these findings, our recent study using tissue clearing and staining with iDISCO (immunolabeling-enabled three-dimensional imaging of solvent-cleared organs) demonstrated that in mice, liver-projecting sympathetic postganglionic neurons are located in the celiac-superior mesenteric complex, aortico-renal, and suprarenal ganglia (20). These liver-projecting postganglionic neurons were tyrosine hydroxylase positive (TH) (20), which is consistent with a recent immunohistochemical analysis of the mouse celiac ganglion (28). Furthermore, differences in the spatial distribution of kidney- and liver-related postganglionic neurons and nerve bundles were revealed, which may suggest segregation of sympathetic postganglionic neurons based on the innervated organ.

Sympathetic preganglionic neurons receive inputs from supraspinal areas and the neurons sending projections to preganglionic neurons are called presympathetic neurons. In the past decades, viral tracers, in addition to, or instead of, conventional tracers became popular choices for identification of neural circuits, and detailed reviews are available about their use, advantages, and disadvantages (29, 30). Although static viral tracers, which remain within the infected cells, are good choices for revealing local circuits, to establish organ-related circuits in the central nervous system, transsynaptic viral tracing is necessary. The attenuated Bartha strain of pseudorabies virus (PRV) (an alpha-herpes virus) is a retrograde transsynaptic virus, which became a common tool for identification of organ-related neurons, including liver-related neurons (31–37). PRV-labeled neurons were identified in the CNS 4 to 5 days following left lobe inoculation of the rat liver. Liver-related neurons were located in the following areas: the nucleus of the solitary tract (NTS), ventrolateral medulla (VLM), C1/C3 adrenaline cell group, A5 noradrenaline cell group, locus coeruleus, gigantocellular reticular nucleus, raphe pallidus, and magnus, as well as the lateral hypothalamus (LH), zona incerta, retrochiasmatic area of the hypothalamus, and the paraventricular nucleus (PVN) (38). The same infection pattern was found in mice 5 to 6 days following left lobe PRV inoculation of the liver (39). In our studies, the median/left lobe of the liver was PRV injected and labeling was observed 4 days after inoculation, with specific expression in the ventral brainstem, the A5 noradrenaline cell group, the lateral paragigantocellular nucleus (LPG), the raphe, the lateral reticular nucleus, and the rostroventrolateral reticular nucleus (FIGURE 1). At this time point, we identified a moderate number of PRV-labeled neurons in the PVN (40). By extending animal survival, higher orders of liver-related neurons were revealed (FIGURE 2). La Fleur and coworkers (38) found an increasing number of liver-related neurons in the LH and PVN 5 days after inoculation of the rat liver, whereas additional hypothalamic areas, the suprachiasmatic nucleus, dorsomedial and ventromedial hypothalamic nucleus, arcuate nucleus, and preoptic area contained labeled neurons (34, 38). Stanley and coworkers (39) demonstrated similar results in mice 6 to 7 days after liver inoculation. Interestingly, at these further time points, other brain areas contained higher order liver-related neurons, such as the periaqueductal gray, the amygdala, and the bed nucleus of the stria terminalis (34, 38, 39). Despite the slight laboratory-related differences in the timeline, which could be due to the species (rat versus mouse) or injection site, the location of preautonomic neurons is comparable between rats and mice and PRV provides a reproducible infection pattern (34, 35, 38, 40).

FIGURE 1.

Location of presympathetic liver-related neurons in the brainstem. Left: schematic illustration of mouse brainstem areas where presympathetic neurons (orange) and liver-related neurons (green filled circles) were previously identified. Distance from bregma −5.34 mm (top), −6.64 mm (middle), −7.64 mm (bottom). Right: liver-related neurons (green) ∼ 96 hours after inoculation of the mouse liver with a retrograde viral tracer, pseudorabies virus 152. Note that parasympathetic liver-related neurons are also labeled in the dorsal brainstem. A1, noradrenaline cells; A5, noradrenaline cells; Amb, nucleus ambiguus; C1, adrenergic cells; DMV, dorsal motor nucleus of vagus; KF, Kolliker-Fuse nucleus; LC, locus coeruleus; LPG, lateral paragigantocellular nucleus; NTS, solitary nucleus; PB, parabrachial nucleus; py, pyramidal tract; RVLM, rostroventrolateral medulla; S5, sensory root of the trigeminal nerve; 7n, facial nerve. Scale bar: 200 µm.

FIGURE 2.

Sympathetic circuits involved in the regulation of the liver. The sympathetic innervation of the liver arrives from the postganglionic neurons mainly located in the celiac-superior mesenteric complex, which receives input from preganglionic neurons in the spinal cord. Sympathetic preganglionic neurons receive inputs from supraspinal areas. Presympathetic neurons were identified in a variety of brain nuclei, including in the ventral brainstem and paraventricular nucleus of the hypothalamus. Higher order liver-related neurons were identified in the dorsomedial hypothalamus (DMH), lateral hypothalamus (LH), ventromedial hypothalamus (VMH), arcuate nucleus (ARC), preoptic area (POA), and suprachiasmatic nucleus (SCN). PVN, paraventricular nucleus of the hypothalamus.

Like most organs, the liver receives both sympathetic and parasympathetic innervation; therefore, PRV labeling alone does not allow a distinction between presympathetic and preparasympathetic liver-related neurons in higher brain areas, including the PVN. In general, presympathetic neurons could be identified based on their projection to the spinal cord, using conventional monosynaptic tracers like fluorogold. In the brainstem, presympathetic neurons were identified in the Kolliker-Fuse nucleus of the parabrachial complex, the ventrolateral A5 cell group, the A1/C1 cell group adjacent to the nucleus ambiguous in the caudal and rostral parts of the ventrolateral medulla (CVLM/RVLM), as well as in the ventromedial medulla (VMM), the medullary raphe, and the caudal NTS (34, 41–44). FIGURE 1 illustrates the location of mouse presympathetic neurons in the brainstem and the location of liver-related neurons identified with PRV.

In higher brain areas, including the hypothalamus, a combination of tracers or denervation of one of the autonomic branches needs to be used to further differentiate liver-related neurons to presympathetic and preparasympathetic neurons. Using denervation in rats, Buijs and coworkers (34) demonstrated a complete separation between presympathetic and preparasympathetic neurons at the level of the hypothalamus, including in the PVN, although interaction between the two autonomic branches is likely (34). Even though hepatic denervation in the mouse is possible (45, 46), it might not be feasible in every laboratory or ideal in every experimental setting; therefore, tracer combinations are good alternatives to separate presympathetic and preparasympathetic liver-related neurons. Here, we provide examples for distinguishing liver-related presympathetic PVN neurons by combining viral tracers. Please, note that we focus on viral tracers and do not discuss double labeling with conventional static tracers (29, 30, 34).

As shown in FIGURE 3, PRVs with distinct fluorescence could be used to differentiate neurons based on projection sites. Tracer injections into the brainstem or spinal cord were used to identify preautonomic neurons in the PVN (47–50), and the same concept could be used to differentiate presympathetic liver-related neurons in the PVN. For instance, inoculation of the liver with PRV-red fluorescence protein (RFP; PRV-614) allows identification of liver-related neurons, whereas additional injection of PRV-green fluorescence protein (GFP; PRV-152) into the VLM/VMM results in identification of presympathetic neurons; therefore, double-labeled PVN neurons are presympathetic liver-related neurons. Successful combination of dual labeling with PRVs was demonstrated earlier, revealing the ability of PRVs to infect the same neurons via different pathways in complex neural circuits. Buijs and coworkers (34) combined the inoculation of the adrenal gland and liver utilizing PRV-GFP and a PRV-expressing β-galactosidase (PRV-Bablu) to distinguish sympathetic- and parasympathetic-related neurons in the PVN and suprachiasmatic nucleus. Cano and coworkers (51) used the same combination of PRVs injected to both kidneys in the rats, revealing the different hierarchical levels of the circuit controlling the sympathetic outflow to the kidneys. Stanley and colleagues (39) revealed a population of neurons in the NTS, A5 region, PVN, LH, and arcuate nucleus that project both to the liver and white adipose tissue using a combination of PRV-GFP and PRV-RFP. Triple labeling with PRVs is also possible by combining PRV-GFP, PRV-RFP, PRV-mTurquoise2 (PRV-290), and PRV-mCerulean (PRV-273); however, the analyses of the fluorescence labeling require the use of narrow band filters specific for the detection of GFP, RFP, mTurquoise2, and mCerulean without bleed through (34, 51–53).

FIGURE 3.

Tracing strategies to identify liver-related presympathetic neurons in the paraventricular nucleus (PVN). Left: identification of presympathetic neurons with PRVs expressing distinct fluorescence. PRV-RFP inoculation of the liver allows identification of liver-related neurons, whereas PRV-GFP injection into the ventral brainstem reveals presympathetic neurons. Presympathetic liver-related PVN neurons will be double labeled. Note: due to the transsynaptic nature of PRV, interneurons, and relay areas might be labeled. Right: identification of presympathetic neurons with a combination of PRV and retrograde AAV. PRV-RFP allows identification of liver-related neurons, whereas injection of a retrograde AAV into the VLM/VMM identifies neurons with direct projections to the ventral brainstem. With this combination of viral tracers, the identified double-labeled neurons are presympathetic liver-related. GFP, green fluorescence protein; PVN, paraventricular nucleus of the hypothalamus; PRV, pseudorabies virus; RFP, red fluorescence protein; rgAAV, retrograde adeno-associated virus; VLM, ventrolateral medulla; VMM, ventromedial medulla.

It must be mentioned that although PRV is a unique tool to unveil complex organ-related neural circuits, it presents certain caveats that must be kept in mind. One is the toxicity of the viral infection, which is still significant even though the Bartha strain PRVs have reduced virulence. Infected rodents exhibit a variety of symptoms: hunched posture, lethargy, spike coat, oronasal excretions, or weight loss after viral inoculation, depending on the strain and injection site (37, 54). Regarding cellular properties of PRV-labeled neurons, studies using slices reported that at the time of recordings, cellular properties were indistinguishable from other neurons (31, 55–59). On the other hand, in cultured rat sympathetic neurons infected with the attenuated PRV-GFP, increased action potential frequency was observed 18 hours postinfection (60), whereas, in auditory brainstem neurons, Porres and coworkers (61) found the opposite: a decrease of action potential frequency following PRV-GFP infection. Nevertheless, to reduce possible confounding effects on the cellular properties of PRV-infected neurons, recordings must be conducted within the proper time frame. Another caveat originates from the transsynaptic nature of the virus, which increases the complexity of the analysis. To avoid misinterpretation or overspeculation, it is crucial to design time-dependent studies with multiple endpoints to establish the nature of the connections (direct versus indirect) or the infection sequence because of the possibility of labeling interneurons, second-order neurons, or neurons in a relay area (FIGURE 3). In the case of inoculation of multiple organs, timing is especially important, because first and second-order neurons could be labeled at different time points depending on how many neurons are in the given circuit. For instance, Kerman and coworkers (62) showed that dual injection of PRVs into the adrenal gland and the muscle needed to be separated by 24 h to achieve temporal matching. Alternatively, Buijs and coworkers (34) found that decreasing the dose of PRV injected in the adrenal gland was sufficient to slow the infection and obtain first-order adrenal sympathetic neurons together with first-order liver parasympathetic neurons, despite that both PRVs were injected during the same surgery.

Nevertheless, to circumvent some of these potential confounders, PRV could be combined with retrograde adeno-associated viral (rgAAV) tracing. Because rgAAV needs to be injected first to allow sufficient time for expression (3–4 weeks), this method requires more time to identify sympathetic liver-related circuits than the injection of PRV into the VLM/VMM (FIGURE 3). On the other hand, it excludes transsynaptic labeling and allows the identification of neurons only with direct projections to the ventral brainstem. Furthermore, use of Cre-dependent AAVs in transgenic Cre mice provides additional information about the circuit of interest. For example, rgAAV injected into the ventral brainstem of Sim1^Cre mice (single-minded homolog 1, mainly expressed in PVN) will identify presympathetic Sim1-expressing PVN neurons. Combined with PRV liver inoculation, not only the autonomic division, but the additional feature (Sim1) of the presympathetic liver-related PVN neurons can be revealed.

As mentioned, establishing the location and the circuit of higher order neurons is challenging. Indeed, some of the PRV studies contained longer timelines and they showed an increasing number of liver-related neurons in brain areas, which already contained third-order neurons (34, 39), making differentiation of third-order from fourth-order neurons in the same nucleus impossible. We propose that this could be partially solved by combining a monosynaptic tracer (conventional or viral), injected into the PVN with PRV inoculation of the liver for a longer time (1 or 2 additional days). In this case, we expect to reveal liver-related neurons in brain areas directly projecting to the PVN. With this type of viral tracer combination, the location of higher order neurons could be better identified. Although these combined approaches have just started to be utilized for the identification of organ-related central circuits in mice, they will allow us to establish liver-related circuits in a more precise manner and provide a platform for functional studies at both cellular and systemic levels.

3. NEUROCIRCUITS AND MECHANISMS INVOLVED IN THE SYMPATHETIC REGULATION OF THE LIVER

3.1. Liver-Related Neurons in the PVN

Paraventricular nucleus neurons are known to project to the brainstem and spinal cord, and it is assumed that sympathetic and parasympathetic outflow will be modulated to normalize glycemia following a response of PVN neurons to change blood glucose concentration (63). The hypothalamus, including the PVN, contains glucose-sensing neurons, specialized neurons in which electrical activity increases/decreases in response to rising extracellular glucose concentration (64). Melnick and coworkers (65) determined that 68% of PVN parvocellular neurons were sensitive to changes in glucose concentration, with 24% excited by and 26% inhibited by glucose, although their connection to the liver remains to be determined. Liver-related PVN neurons express corticotrophin-releasing hormone (CRH; ∼ 6%) and oxytocin (∼ 6%) but do not express arginine-vasopressin or thyrotropin-releasing hormone (39). In addition, it has been shown that the anorexigenic neuropeptide vasoactive intestinal peptide increased the activity of CRH and oxytocin neurons, which in turn elevated blood glucose and corticosterone levels, although this effect was diminished in adrenalectomized rats (66). Interestingly, PVN neurons expressing melanocortin 4 receptor (MC4R) were found to project to the ventral brainstem (LPG, VLM, and C1 neurons), and chemogenetic activation of MC4R-expressing PVN neurons suppressed feeding and increased heart rate and blood pressure, without affecting glucose homeostasis (67). These data suggest that MC4R-expressing neurons in the PVN are likely preautonomic but not liver-related neurons.

Nevertheless, it is well-known that the hypothalamus plays a crucial role in the central regulation of energy homeostasis, including glucose metabolism and numerous hypothalamic areas, which have been identified as part of this central regulatory pathway. For instance, the arcuate nucleus is known for its role in the regulation of food intake and glucose homeostasis; the ventromedial hypothalamic area is involved in the regulation of temperature, hunger, and sexual activity; and the dorsomedial hypothalamus (DMH) is part of the regulatory pathways modulating temperature, body weight, and metabolism (68–74). These nuclei relay information to the PVN, which integrates information and commands neuroendocrine and autonomic responses (63–65, 75, 76). The activity of presympathetic and preparasympathetic PVN neurons via descending projections to sympathetic or parasympathetic preganglionic neurons has a large role in determining the sympathetic and parasympathetic outflow. Despite previous in vivo studies that clearly demonstrated the importance of the PVN in the central regulation of metabolism, there is only limited information about the PVN-liver circuit and its involvement in hepatic carbohydrate metabolism. In general, stimulating PVN neurons with NMDA or blockade of GABA_A receptors increased plasma glucose levels due to SNS activation (77). Norepinephrine injection into the PVN also resulted in hyperglycemia and sympathetic activation, similar to splanchnic nerve stimulation (78). More specific investigations demonstrated that pituitary adenylyl cyclase-activating polypeptide (PACAP) administered into the PVN caused hyperglycemia and increased HGP (79). Although these studies successfully demonstrated the importance of PVN in the regulation of SNS, and thus glucose homeostasis, they were based on nonspecific sympathetic output activation and affected not only liver metabolism, but hormone secretion of the pancreas, adrenal gland, or muscle glucose uptake (16, 80).

3.2. Neurophysiology of Liver-Related PVN Neurons

Whereas in previous studies technical limitations prevented the specific manipulation of organ-related circuits in vivo, at the cellular level, transsynaptic viral tracing combined with electrophysiological approaches has revealed intrinsic and extrinsic properties of neurons that are part of the brain-liver circuit. In these electrophysiological studies, PRVs expressing fluorescence proteins were used to identify liver-related neurons followed by whole cell patch-clamp recordings (35, 36, 81, 82). These studies established the synaptic regulation of liver-related PVN neurons in adult mice and showed that most liver-related PVN neurons receive excitatory inputs expressing transient receptor potential vanilloid type 1 (TRPV1). The increase of frequency of the action potential independent postsynaptic glutamate currents, after application of a TRPV1 agonist, was specific to preautonomic neurons (35). This TRPV1-dependent regulation of excitatory neurotransmission was diminished in the streptozotocin-treated type 1 diabetic mouse, whereas, in vivo or ex vivo, the application of insulin reinstated the increase (35). These findings were aligned with previous studies showing that insulin modulates TRPV1 receptor sensitivity in a phosphatidylinositol 3-kinase/PKC-dependent manner (83). TRPV1-expressing neurons were identified in a variety of nuclei, including those that project to the PVN (84, 85). Furthermore, a population of TRPV1-expressing neurons in the PVN and DMH receive excitatory inputs from TRPV1-expressing neurons, which indicates direct interaction between TRPV1-expressing neurons and suggests the existence of a TRPV1 network. TRPV1 was shown to be expressed in the hypothalamus and plays a role in thermoregulation (86–89); however, the contribution of hypothalamic TRPV1-expressing neurons to the regulation of overall glycemia and liver metabolism remains to be determined.

Because the brain-liver pathway plays an important role in the regulation of liver functions and ANS imbalance is known in obese and diabetic conditions, it is crucial to identify potential changes associated with these pathophysiological states. A study conducted in lean and obese and diabetic db/db mice compared the excitability of liver-related PVN neurons (36). At baseline, liver-related PVN neurons were differentiated into silent and spontaneously firing cells. In lean mice, most liver-related PVN neurons were silent, whereas in db/db mice most of the recorded neurons showed spontaneous activity (36). Furthermore, in the db/db mouse model, liver-related PVN neurons were overactivated compared with the lean mice. Although the excitability of these neurons increased in db/db mice, the synaptic excitation and inhibition were unaltered. In addition to synaptic regulation, extrasynaptic inhibition of liver-related PVN neurons was established; however, this tonic inhibition of liver-related PVN neurons was not different between db/db and lean mice (36). The study also determined the role of TRPV1-dependent neurotransmission and showed diminished TRPV1-dependent excitatory neurotransmission in liver-related PVN neurons of db/db mice. Intriguingly, the overall inhibitory neurotransmission to liver-related PVN neurons was not regulated by TRPV1, which is in contrast to previous observations in neurons of the dorsal motor nucleus of the vagus (90). Altogether, these data suggest that the regulation of inhibitory neurotransmission is nuclei specific (90).

Changes in synaptic regulation of liver-related PVN neurons were revealed in other mouse models, including mice deficient in neuronal lipoprotein lipase (LPL). LPL, a critical enzyme for hydrolysis of triglyceride-rich lipoproteins and fatty acids uptake, was shown to play a role in the autonomic regulation of liver functions (81). Mice with neuronal LPL deficiency showed improved glucose tolerance, reduced hepatic glucose production, and lipid accumulation. Loss of LPL in PVN neurons resulted in altered metabolic flux, and these findings were accompanied by reduced inhibitory neurotransmission to liver-related PVN neurons (82) suggesting plasticity of the PVN-liver pathway.

Ultimately, these findings provide important information about the cellular properties of liver-related PVN neurons; however, due to the mouse models (e.g., db/db), they have limited relevance to human obesity. Nevertheless, a recent study showed that in high-fat diet-fed mice the excitability of liver-related PVN neurons is increased and a high-fat diet modulated the firing rate of neurons within the brain-liver pathway (91). Current clamp recordings of action potential responses were used to determine the firing rate of liver-related PVN neurons, and higher action potential frequencies were observed in high-fat diet-fed mice compared to mice fed with the control diet. These data suggest that a high-fat diet increases the excitability of liver-related PVN neurons. Furthermore, in control mice insulin did not change the excitability of liver-related PVN neurons, whereas in high-fat diet-fed mice, the current-action potential frequency responses showed a reduction in firing rate after insulin application. These findings further indicate that the firing rate of liver-related PVN neurons was altered in response to insulin, which could be due to insulin resistance in neurons (91).

Because the liver is innervated with both sympathetic and parasympathetic nerves, the recordings were likely conducted from both presympathetic and preparasympathetic liver-related PVN neurons, which could be one of the reasons for the observed heterogenic responses (e.g., silent and spontaneously firing neurons). To distinguish neurons with sympathetic involvement, Molinas and coworkers (91) identified liver-related PVN neurons with projections to the ventral brainstem and conducted patch-clamp recordings in mice fed with control and high-fat diet. Despite the overactivity of liver-related PVN neurons observed in high-fat diet-fed mice, the excitability of VLM-projecting PVN neurons and the response to insulin were similar in diet-induced obese conditions. This is somewhat surprising because increased excitability was shown in presympathetic kidney-related PVN neurons in hyperglycemic streptozotocin-treated mice (92) and in VLM-projecting PVN neurons in a model of heart failure (93). Although these studies provide an overall assessment of the electrophysiological properties of liver-related PVN neurons suggesting that presympathetic PVN neurons are heterogenous, detailed investigations are needed to gain more insight into the cellular properties of neurons. We can speculate that a specific neuronal population is responsible for the increased excitability of liver-related PVN neurons: 1) PVN neurons with direct projections to preganglionic neurons in the spinal cord, 2) neurons projecting to the VLM/VMM via a relay area, 3) neurons with a specific phenotype, and/or 4) preparasympathetic PVN neurons (FIGURE 2). On the other hand, these findings highlight the need for further investigations in phenotype- and circuit-dependent manners.

3.3. Presympathetic Ventral Brainstem Neurons as Major Determinants of Sympathetic Tone

A dramatic decrease in blood glucose is a life-threatening situation that requires immediate responses from the organism to avoid substantial consequences; therefore, a multitude of mechanisms are triggered. Food seeking becomes the most important task for the individual, with a drive for high-calorie nutrition (94), whereas the decreased concentration of glucose detected by peripheral organs triggers endocrine responses, involving increased secretion of glucagon from pancreatic alpha cells and corticosterone from the adrenal gland (95). Centrally, autonomic responses are triggered, including the activation of presympathetic neurons sending signals to peripheral organs (96, 97). Whereas neurons in the VLM are clearly associated with the regulation of cardiovascular functions (98, 99), it is well-accepted that catecholaminergic neurons in the VLM are critical for glucose regulation, particularly during hypoglycemia (95, 100–108).

In rats, an immunohistochemical study following glucoprivation induced by an antimetabolic glucose analog, 2-deoxy-d-glucose (2DG), revealed activation of distinct groups of catecholaminergic neurons, specifically A1/C1, C2, and dorsal C3 neurons (109). Furthermore, hindbrain injection of the antiglycolytic 5-d-thioglucose produced localized glucoprivation in areas containing catecholaminergic neurons and led to secretion of glucagon and corticosterone, further supporting the importance of this neuronal population in triggering and regulating autonomic responses during glucose deficiency (108). Similarly, glucoprivation was found to activate C3 neurons and optogenetic stimulation of these cells increased thoracic sympathetic nerve activity with a greater effect when C1 + C3 stimulation was combined, resulting in an additive sympathoexcitation (110). In rats, selective destruction of spinally projecting neurons in the rostral C1 and ventral C3 impaired glucoprivic responses, without affecting feeding behavior after subcutaneous injection of 2DG (103).

Selective chemogenetic activation of rat C1 neurons was shown to be sufficient to activate counterregulatory responses (102, 111, 112). Stimulation of rostral and middle C1 and A1/C1 neurons increased feeding and corticosterone secretion, whereas glucose levels increased only after simultaneous stimulation of rostral and middle C1 neurons. These data suggest that spinally projecting neurons are likely sparsely distributed throughout the C1 region (102). On the other hand, chemogenetic activation of dopamine beta-hydroxylase (DBH) neurons in the VLM of transgenic mice increased secretion of epinephrine, corticosterone, and glucose, as well as resulted in significantly higher hepatic expression of glucose-6-phosphatase (G6Pase) and phosphoenolpyruvate carboxykinase 1 (Pepck), two key genes that regulate glycogenolysis and gluconeogenesis (104). Cre-dependent monosynaptic tracing combined with modified rabies virus injection in transgenic DBH^Cre mice identified two different C1 neuronal populations, one in the rostral VLM projecting to the spinal cord and another located in the caudal area projecting to the hypothalamus (113). These findings were confirmed with a combination of two retrograde tracers, cholera toxin b-subunit conjugated with Alexa 488 injected in the hypothalamus and Texas Red conjugated with dextran amine injected into the spinal cord (104). Findings showed that the spinally projecting neurons were located rostral compared with neurons projecting to the hypothalamus and represented two distinct catecholaminergic populations (104). Interestingly, light-evoked activation of spinally projecting RVLM neurons also resulted in increased glycemia; however, these effects were abolished after bilateral adrenalectomy suggesting that activation of catecholaminergic neurons in the RVLM stimulated the secretion of stress hormones by the adrenal gland, which in turn induced hepatic glucose production (104). In addition, chemogenetic activation of rat rostral and middle C1 neurons resulted in a longer lasting increase in glucose levels than was observed after administration of antiglycolitic 2DG (102). Stimulation of the RVLM after targeted elimination of DBH neurons with antibodies conjugated with the ribosomal toxin saporin showed that non-C1 neurons contribute to the regulation of sympathetic nerve activity (114). Moreover, the existence of glucose-responsive cells and glucoregulatory systems that do not participate in counterregulation is acknowledged (96) and suggests that multiple cell groups are involved in the regulation of glycemia (15, 16). Presympathetic neurons, including RVLM neurons, are heterogenous (100, 115–118), and a subset of RVLM neurons sends collaterals to diverse spinal segments (e.g., T2, T10) and activates functionally dissimilar nerves (119). Ultimately, these findings suggest the existence of both generalized and specialized, organ-related, presympathetic neurons in the ventral brainstem.

It is well-accepted that presympathetic neurons in the RVLM are responsible for the generation of sympathetic tone and use glutamate as a neurotransmitter (99, 120, 121). RVLM neurons with projections to the IML of the spinal cord thoraco-lumbar segments are heterogenous with a population expressing TH and phenylethanolamine N-methyltransferase (PNMT), indicating that they synthesize epinephrine (122–125). This was also demonstrated by Schreihofer and coworkers (123) when they characterized putative presympathetic RVLM neurons by extracellular recordings and identified their phenotype. Recordings were mainly conducted from RVLM neurons with confirmed projections to the thoracic spinal cord. Most recorded neurons showed TH immunoreactivity, and PNMT immunoreactivity was confirmed in some of them. Although these studies convincingly showed the catecholaminergic nature of RVLM neurons, they also confirmed the existence of noncatecholaminergic neurons in the RVLM.

In addition to the catecholaminergic phenotype, a variety of neuropeptides were identified in the RVLM (FIGURE 4). Neuropeptide Y (NPY) immunopositivity was identified in multiple species, including rodents, and it was demonstrated that most of the catecholaminergic and NPY-positive terminals found in the IML originate from the RVLM (126, 127). In situ hybridization studies suggested that ∼20% of TH-positive neurons in the RVLM express substance P or neurokinin A (128), and enkephalins were found in TH-positive and noncatecholaminergic RVLM neurons (129). Similarly, calbindin, a calcium-binding protein was identified and the majority of calbindin positive neurons contained TH, while half showed PNMT positivity (130). Interestingly, almost all adrenergic C1 neurons were shown to express cocaine- and amphetamine-regulated transcript as well as noncatecholaminergic neurons (131, 132). IML-projecting PACAP and TH immunoreactive neurons originate from C1 to C3 neurons (133) and intrathecal injection of PACAP caused a prolonged sympathoexcitation. These catecholaminergic C1 neurons, together with non-C1 neurons projecting to sympathetic preganglionic neurons in the spinal cord, are important cardiovascular response regulators (123, 134, 135). Interestingly, studies in rats have shown that a subset of presympathetic RVLM neurons was inhibited by cholecystokinin (CCK) (136) and 60% of the CCK inhibited neurons were non-C1, suggesting that CCK can inhibit non-C1 neurons with some selectivity (137). These findings also suggest that CCK responsive neurons could be involved in the sympathetic vasomotor outflow to the gastrointestinal tract (137) (FIGURE 4).

FIGURE 4.

Presympathetic neurons in the RVLM set sympathetic tone and thus modulate organ functions. The excitability of RVLM neurons depends on the intrinsic properties and activity of the neuronal network (balance of inhibition and excitation). RVLM neurons are glutamatergic and project to sympathetic preganglionic neurons and thus regulate the sympathetic output and organ functions. In addition to glutamate, numerous neuropeptides were also identified in RVLM neurons. C1, adrenergic neurons; CaBP, calbindin; CART, cocaine- and amphetamine-regulated transcript; CCK, cholecystokinin; CVLM, caudoventrolateral medulla; DBH, dopamine β-hydroxylase; DMH, dorsomedial hypothalamus; ENK, enkephalin; GABA, γ-aminobutyric acid; GABA_AR, GABA receptor; Gly, glycine; GlyR, glycine receptor; Glut, glutamate; GlutR, glutamate receptor; LH, lateral hypothalamus; LPG, lateral paragigantocellular nucleus; NPY, neuropeptide Y; PACAP, pituitary adenylate cyclase-activating polypeptide; PNMT, phenylethanolamine N-methyltransferase; PVN, paraventricular nucleus of the hypothalamus; RVLM, rostroventrolateral medulla; SP, substance P; TH, tyrosine hydroxylase.

The tonic discharge of RVLM neurons corresponds with sympathetic nerve activity. Evidence demonstrates that this tonic discharge can be modulated by the baroreflex (138–140) and central respiratory network (141–144). The discharge reflects the intrinsic properties of neurons, which depends on neuronal network activity, which also determines the balance between the presynaptic release of excitatory and inhibitory neurotransmitters. Although several studies suggest that there is no change in the tonic discharge of RVLM neurons after blockade of fast excitatory synaptic neurotransmission (145, 146), other studies showed the importance of synaptic inputs (99, 114, 147–149) and their control over the discharge of RVLM neurons (150, 151). In our opinion, it is highly likely that synaptic inputs are needed to alter the intrinsic properties of RVLM neurons and thus trigger discharge, although the magnitude of contribution (synaptic versus intrinsic) and environment (e.g., hormones, nutrients, pathophysiological condition) will influence it.

While measuring the activity of a specific sympathetic branch is challenging, identifying the cellular properties of organ-related neurons in the RVLM has been equally challenging due to the high density of fibers making visualization of neurons difficult. Nevertheless, based on findings over the past decades, it is well-accepted that descending projections from presympathetic RVLM neurons represent a major path to preganglionic neurons. Despite that most studies investigated presympathetic RVLM neurons in the context of control of cardiovascular functions, we can speculate that many of these findings are relevant to the regulation of metabolically important organs, including the liver. As stated above, the activity of RVLM neurons depends on the intrinsic properties of neurons and the balance of inhibitory and excitatory inputs. Several studies have established that GABA and glutamate are two major neurotransmitters that control the excitability of presympathetic RVLM neurons. Inhibitory mechanisms are shown to be crucial for the maintenance of proper sympathetic output, and GABAergic inputs from the midbrain, medullary regions, and GABAergic neurons in the caudal VLM were identified as the main source of baroreflex-mediated sympathoinhibition (123, 149, 152–157). Blockade of GABAergic neurotransmission in the RVLM increased sympathetic nerve activity and blood pressure (98, 148, 158–161), suggesting that RVLM neurons are primarily under sustained tonic GABAergic inhibition. Glycine, another major inhibitory neurotransmitter, also plays a role in the regulation of ventral brainstem neurons (147); however, its role in the regulation of sympathetic output is controversial due to observations that blockade of glycine receptors in the RVLM had no effect on sympathetic nerve activity and blood pressure (98). Alternatively, anatomical studies clearly demonstrated that GABA and glycine are often coexpressed in the same presynaptic terminals (162–165). Indeed, patch-clamp recordings from presympathetic RVLM neurons revealed the existence of both GABAergic and glycinergic postsynaptic currents (147, 161, 166). These observations suggest that functional glycinergic receptors are expressed on presympathetic RVLM neurons, but the possible mechanism of glycine release and the role of glycine in the control of sympathetic tone are still debated. To reveal the potential roles of glycine, a series of experiments was performed in anesthetized rats, showing that blockade of glycine receptors in the RVLM shortened the recovery time of renal sympathetic nerve activity following an increase in blood pressure (147). Whole cell patch-clamp recordings from presympathetic RVLM neurons revealed that potentiation of inhibitory inputs with a nonselective, voltage-dependent potassium channel blocker increased glycine release (147). Together, these data suggest that in the RVLM GABA and glycine work together to inhibit RVLM neurons and control sympathetic tone, and it is highly likely that most of these mechanisms will play a role in determining sympathetic discharge to regulate not only cardiovascular functions but also metabolism.

4. TARGETED MODULATION OF ORGAN-SPECIFIC CIRCUITS, IS THIS THE FUTURE?

Whereas the previous investigations clearly pointed to a PVN-ventral brainstem-liver circuit, demonstrated the importance of the SNS in modulating glycemia, and established the cellular properties of liver-related neurons, they have technical limitations: in particular the lack of targeting liver-related circuits in vivo. In a recent study, this limitation was circumvented by using a combination of approaches to only modulate neurons that were part of the brain-liver pathway (46). The authors showed that a subset of arcuate POMC neurons is liver related and that some directly project to cholinergic DMV neurons. Light stimulation of ARC POMC fibers to liver-related DMV neurons resulted in elevated blood glucose levels ∼15 min after the start of stimulation in both male and female mice. Additional assessment of liver tissues showed that the increase in systemic glucose levels was partly due to improved hepatic glucose production, as indicated by higher G6Pase and Pepck gene expression levels. This was further supported by the increase of glycemia during pyruvate tolerance test and hepatic branch vagotomy, which prevented the blood glucose rise caused by light stimulation (46). The increase in glucose concentration was also prevented by a MC4R antagonist and was absent in Mc4r gene knock down mice, suggesting melanocortin system involvement. In addition, reduced activity of liver-projecting cholinergic DMV neurons led to the elevation of blood glucose in an insulin-independent manner through increasing hepatic glucose output (46). Currently, this is the first and only publication that demonstrated that centrally modulating the activity of liver-related neurons alters hepatic functions and overall glycemia. Intriguingly, these findings suggest that the parasympathetic nervous system contributes to the increase of blood glucose levels likely through regulation of hepatic glucose metabolism; however, detailed studies are required to determine the role of the SNS and the interaction between the two autonomic branches, as well as revisit the afferent and efferent innervation of the liver.

Nevertheless, these novel observations demonstrate the feasibility of modulating organ-specific neural pathways and suggest the existence of more complex circuits than previously believed. Therefore, it is crucial to understand the mechanisms involved in the regulation of neuronal excitability, to identify receptors, characterize phenotype, and specify location of presympathetic and preparasympathetic neurons. The findings also clearly underscore the need for carefully designed studies to better understand the autonomic regulation of the liver to develop novel strategies to modulate glycemia via the autonomic nerves.

GRANTS

This work was supported by funding from the National Institutes of Health (NIH) DK-122842 (to A.Z. and A.V.D).

DISCLOSURES

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

AUTHOR CONTRIBUTIONS

A.Z., L.D.D., and A.V.D. prepared figures; A.Z., L.D.D., and A.V.D. drafted manuscript; A.Z., L.D.D., and A.V.D. edited and revised manuscript; A.Z., L.D.D., and A.V.D. approved final version of manuscript.