Liver innervation and hepatic function: new insights

Abstract

The hepatic nervous system has a well-known impact on the regulation of liver function and organism homeostasis. The aim of this review is to summarize the new available data regarding the role of hepatic nerves. In the last decade, studies have shown that hepatic nerves exert subtle but significant modifications on the regulation of glucose and lipid metabolism, food intake, and liver regeneration. They also play a role in liver disease pathogenesis, and hepatic denervation has beneficial results to liver graft ischemia–reperfusion injury. Available data are still limited, and further research toward neural pathways involving the liver that can modify response to disease is required.

Introduction

The role of the hepatic nervous system in the maintenance of normal liver function and overall organism homeostasis has long been minor, since after orthotopic liver transplantation (OLT) and, therefore, denervation of the organ, liver continues to function well without serious consequences. This perspective has changed over the last two decades, after publication of detailed reviews, emphasizing the morphologic and functional aspects of liver innervation [1], [2].

The liver receives afferent [3] and efferent [4] nerve fibers of both sympathetic and parasympathetic origin. Hepatic nerve distribution is highly species dependent, with the human liver exhibiting both intra-acinar and portal tract innervation, whereas in mice and rats hepatic innervation is evident only in portal tracts (Fig. 1) [5]. During embryonic development, the liver is poorly innervated and does not contain intrinsic neurons derived from the neural crest cells, in contrast to the developing gastrointestinal tract [6]. Furthermore, in the first two trimesters of gestation, neural distribution in the human liver predominates in the portal tracts, with intra-acinar innervation appearing only toward term [7]. These data suggest that fetal human liver does not need extensive neural control to exert its functions, mainly hematopoietic, whereas after gestation its role changes as reflected in its more extensive innervation pattern. In the adult liver, therefore, hepatic nerves have been shown to play an important role in the regulation of the hepatic neuroendocrine compartment [8], glucose metabolism [9], circadian rhythm [10], liver cell hydration and osmolyte content [11], liver regeneration [12], and liver repair [13].

The scope of this review is to summarize and critically appraise new available data on the role of hepatic innervation on liver function and organism homeostasis, focusing on experimental studies of surgical and pharmaceutical liver denervation or the administration of neuromodulator agents (Fig. 2). Relevant literature was searched using the PubMed database with emphasis on studies after 2004, when the anatomy and function of liver innervation had been extensively reviewed [2].

Hepatic progenitor cells (HPCs) in human liver and oval cells (OCs) in rodents are two equivalent terms for the same stem cells, differentiating into hepatocytes and cholangiocytes in severe acute and chronic liver injury [14], [15]. These cells are localized in four possible locations as follows: canals of Hering, intralobular bile ducts, periductal “null” mononuclear cells, and peribiliary hepatocytes [16]. The combination of selective hepatic branch vagotomy and galactosamine-induced hepatitis results in significant reduction of OCs in comparison with sham-operated rats exposed only to galactosamine intoxication. Furthermore, in transplanted human liver with hepatitis, the number of HPCs is significantly decreased in comparison with that of control livers with hepatitis and normal innervation. Cassiman et al. have shown that hepatic vagal nerves activate HPCs in the injured liver, probably through acetylcholine release to type-3 muscarinic (M3) receptors present on these cells [17]. On the other hand, denervation of normal rat liver results in increase of the number of OCs between 5 and 14 d after two-thirds partial hepatectomy (PH) in comparison with partially hepatectomized rats that did not undergo denervation [18]. Thus, hepatic nerves may play a role during liver regeneration, although the regenerative response after PH is known not to involve OCs, but rather the other hepatic cellular populations [14], [15]. In addition to M3 receptors, HPCs also express vasoactive intestinal peptide receptor type 2 and receive nerve endings as well [19].

Liver innervation plays a critical role on the regulation of carbohydrate and lipid metabolism. A thorough review on the interaction between liver innervation and metabolism has recently been conducted by Yi et al. [20].

Hepatic parasympathetic denervation, either surgically induced or through hepatic muscarinic receptors or nitric oxide synthase blockade, results in 50% decrease of the glucose disposal effect of insulin from the blood after a meal and may reduce insulin corresponding levels in the normal fasting state. The glucose disposal effect of insulin was attributed to both insulin direct action and also to hepatic insulin sensitizing substance (HISS), released from the liver and acting on skeletal muscles, stimulating storage of glucose as glucogen. A critical factor for the release of HISS is hepatic parasympathetic innervation and, therefore, its blockade may lead to HISS-dependent insulin resistance [21], also found in an animal model of hypertension [22]. Liver parasympathetic innervation is responsible for 45%, 35%, and 67% of skeletal muscle, heart, and kidney postprandial glucose clearance respectively, and hepatic parasympathetic denervation leads to a significant decrease in skeletal muscle glucose clearance and consequent 51% increase in plasma glucose concentration, implicating deregulation of the parasympathetic neural component of the liver in the pathogenesis of type 2 diabetes [23]. Chronic partial sensory denervation, through 2% capsaicin administration around the anterior hepatic plexus, also results in decreased insulin sensitivity and diabetes in rabbits [24]. As far as muscarinic receptors are concerned, Li et al. [25] found that lack of M3 receptors in genetically modified mice do not result in significant metabolic alterations after hepatic vagal stimulation, thus implying an acetylcholine-independent mode of action of the vagus nerve, at least in this species. Furthermore, vagal nerves are not necessary for the leptin-induced upregulation of insulin-like growth factor binding protein-2, a liver-derived protein related to glucose regulation, in genetically modified mice [26].

Hepatic sympathetic denervation may block hepatic insulin resistance induced by dexamethasone administration into the arcuate nucleus in rats, whereas this blocking effect is not altered after parasympathetic denervation or in sham-operated animals. Hepatic insulin resistance is also blocked by the intracerebroventricular infusion of the neuropeptide Y1 receptor antagonist BIBP3226. Taken together, the previously mentioned results imply a dexamethasone-induced inhibitory signaling toward liver insulin sensitivity through neuropeptide Y-containing hepatic sympathetic fibers [27]. Complete liver denervation results in loss of net hepatic glucose uptake (NHGU), whereas sympathetic and nitrergic liver innervation exhibit tonic NHGU repression, which resolves with food intake [28]. DiCostanzo et al. [29] reported that the hepatic parasympathetic system is not critical in NHGU regulation in an experimental canine model of vagal cooling, whereas selective sympathetic denervation leads to an increase in NHGU during hyperglycemia [30].

Glucose regulation and food intake are significantly influenced by the incretin hormone glucagon-like peptide-1 (GLP-1), which is derived by two main sources in the body after food ingestion as follows: L-cells in the small intestine and neurons of the hindbrain [31]. GLP-1 exerts its actions after binding to its GLP-1 receptors (GLP-1r), expressed on vagal afferent fibers in the hepatic portal region, apart from the pancreatic islets, regions of the nervous system, and other sites [31]. Hepatic portal GLP-1r, though, were found not to be key regulators of food intake because only very high doses of GLP-1 direct infusion to the portal vein resulted in decreased food intake in rats, whereas intraportal administration of a GLP-1r antagonist did not cause any food intake alterations [32]. Hayes et al. [33] also reported that the metabolic effects of GLP-1 do not require the common hepatic branch of the vagus nerve, supporting a paracrine-like signaling on GLP-1r of vagal afferent fibers along the gastrointestinal tract. On the other hand, GLP-1 is rapidly degraded by dipeptidyl peptidase-IV, and the intraportal administration of a dipeptidyl peptidase-IV inhibitor results in alterations of both insulin secretion and food intake in an experimental rat model via the hepatic vagal nerves [34]. Glucose metabolism is regulated by the neurotransmitter serotonin (5-hydroxytryptamine [5-HT]), which may exert a dual action on glycogen synthesis in hepatocytes, stimulatory and inhibitory, through distinct receptors, 5-HT1/2A and 5-HT2B, respectively [35].

In an experimental rat model of phenol-induced hepatic denervation, serum triglyceride and cholesterol concentrations were increased in the denervated group, through increment on very low density lipoproteins (VLDL) secretion, and were significantly higher than those in the control group, whereas serum glucose showed a significant decline. On the other hand, in both groups, the contents of liver triglyceride and cholesterol did not show significant differences [36]. Triglyceride-rich VLDL (Tg-VLDL) are secreted in higher amounts in type 2 diabetes. Bruinstroop et al. found that hepatic sympathetic innervation is necessary for maintaining Tg-VLDL secretion during fasting, whereas parasympathetic innervation does not cause significant modifications. Furthermore, sympathetic denervation blocked the stimulatory effect of hypothalamic neuropeptide Y on Tg-VLDL [37]. In accordance with the previously mentioned information, Yue et al. [38] reported that although hepatic vagotomy did not alter the rate of Tg-VLDL secretion, it blocked the effect of glycine infusion into an extrahypothalamic region termed the dorsal vagal complex on lowering Tg-VLDL secretion.

Liver triglyceride content is associated with hepatic steatosis, considered related, among other etiologic factors to hyperphagia, obesity, and other features of the metabolic syndrome. Warne et al. showed that chronic central infusion of leptin results in both decreased hepatic lipogenic gene expression and triglyceride content through hepatic sympathetic stimulation. This leptin function is independent of feeding and body weight and is mediated via phosphatidylinositol 3-kinase (PI3K) signaling. On the other hand, attenuated leptin-induced PI3K signaling, through transgenic expression of phosphatase and tensin homologs, may reverse PI3K kinase activity in leptin receptor neurons and result in decreased hepatic sympathetic activity and liver steatosis without affecting adiposity [39]. Leptin, along with insulin and ghrelin, target the central nervous system (CNS) melanocortin system, a central player in lipid metabolism. Wiedmer et al. [40] showed that although the metabolic effects of the CNS melanocortin system on liver lipid metabolism are not mediated through either the hypothalamo-pituitary adrenal or the hypothalamo-pituitary thyroid axes, an intact hypothalamo-pituitary adrenal axis is required for hepatic lipid accumulation after melanocortin system blockade.

Liver regeneration after PH is a very complex process involving the activation and interaction of multiple cytokines and growth factors that regulate cell growth and proliferation [41], [42]. Hepatic vagal innervation is also implicated in the regenerative response because it has been shown that subdiaphragmatic vagotomy results in impaired liver mass restoration and suppression of DNA synthesis after PH compared with those of sham-operated rats. On the other hand, hepatic vagotomy alone does not have detrimental effects in hepatectomized rats, except for delay in DNA synthesis [43], [44].

Ohtake et al. examined the effect of hepatic sympathetic denervation after PH and showed that, despite increased blood flow in the liver, the regenerative response is not affected. Only a contemporary decrease in norepinephrine (NE) is evident 24–48 h after PH that returns to normal levels at 8–9 d after the operation [45]. This result contradicts the findings of Knopp et al. [46], who showed that the NE levels rise 20 min (min) after PH and return to normal in 1 h, whereas epinephrine levels remain elevated at 4 h compared with those of sham-operated rats and are restored at 24 h after PH. During the regenerative process after PH, the serotonin content is increased, and serotonin receptors are upregulated in both brain stem and cerebral cortex, inducing hepatocyte proliferation via increased sympathetic tone [47]. Hamada et al. showed that complete hepatic denervation amplifies liver regeneration after PH in rats, in terms of increased liver to body weight ratio and increased proliferating cell nuclear antigen expression. This finding, although not coincident to the aforementioned results, may be attributed to the increased liver blood flow in the remaining liver because of loss of neural control [48]. Undoubtedly, though, the autonomic nervous system regulates the expression of liver regeneration-related genes as shown by Xu et al. [49] who examined the patterns of gene expression in the regenerating liver, using experimental models of PH combined with either vagotomy, sympathetic, or complete hepatic denervation.

It has been shown from earlier studies summarized by Colle et al. [12] that the hepatic vagus nerve plays at least a partial role in diet selection, but overall does not seem to have an effect on food intake under normal conditions. Along these lines, hepatic branch vagotomy combined with Roux-en-Y gastric bypass (RYGB) results in decreased body weight, adiposity, and food intake compared with that of sham-operated rats, but does not show any significant differences in these parameters when compared with that of RYGB alone. Thus, the common hepatic branch is not considered a critical factor for the beneficial effects of RYGB [50]. Eisen et al. examined the effect of cholecystokinin-8 (CCK-8) and -33 (CCK-33) administration on food intake in two different rat models of vagotomies, either sparing only the hepatic proper vagal nerves (H) or sparing the common hepatic branch (HGD), which contains the hepatic proper and gastroduodenal nerves. Intraperitoneal administration of CCK-33 in H rats has a more profound inhibitory effect on food intake than CCK-8 and shows no difference in contrast to sham-operated rats, whereas in HGD rats both CCK-33 and -8 resulted in decreased effect compared with that of sham-operated rats. Thus, it is suggested that CCK-33 is a more effective stimulant of hepatic vagal nerves than CCK-8, when hepatic branch proper afferent innervation is intact, whereas the decreased effect of both peptides in HGD rats is attributed to the inhibitory interactions between the afferent fibers of the common hepatic branch [51]. In another experimental rat model, De Jonghe and Horn [52] showed that after hepatic vagotomy the cisplatin-induced pica is reduced by 61% over 10 d after injection of the chemotherapeutic agent and also the chemotherapy-induced suppression of daily food intake is attenuated. Hepatic vagotomy has also been shown to result in the blockade of obesity-related hypertension through the hepatic peroxisome proliferator-activated receptor-γ-fat-specific protein 27 pathway [53].

The effects of liver denervation on the hepatic blood flow have been previously reviewed [12] and are quite controversial, showing either decrease in hepatic microcirculation [54] or increase in the hepatic arterial blood flow [55] or no effect after either selective hepatic sympathectomy or vagotomy [56]. In a later study, intracisternal injection of corticotrophin releasing factor (CRF), which is a neurotransmitter in the CNS, resulted in hepatic surface perfusion decrease and portal pressure elevation in rats in a dose-dependent manner. Two types of CRF receptors exist, CRF1 and CRF 2, and mode of action is probably mediated mainly through CRF2 receptors because administration of a selective CRF2 receptor antagonist, K41498, abolished all CRF effects. Furthermore, in animals subjected to chemical hepatic sympathectomy, the effects of CRF on hepatic microcirculation were blocked, whereas no change is noted after hepatic branch vagotomy or atropine administration, indicating the existence of a sympathetic-noradrenergic route mediating the actions of intracisternal CRF on the liver [57]. Mehrabi et al. studied the effect of epinephrine and NE administration on denervated and transplanted porcine livers. In the denervated liver, vasopressor infusion leads to 10% decrease in the portal vein flow, which is counter-regulated by an increase in hepatic artery flow (hepatic arterial buffer response), resulting in only a slight alteration in transhepatic flow. On the other hand, in the transplanted liver, catecholamine administration results in significant decline in both transhepatic flow and hepatic microcirculation, through an unclear mechanism that cannot be attributed to the denervation after liver transplantation [58]. Biernat et al. [59] have reported that hepatic arterial buffer response is mediated by capsaicin-sensitive nerve fibers, using a model of sensory-denervated rats, with calcitonin gene-related peptide (CGRP) being the most effective vasodilator.

The role of sympathetic nervous system (SNS) on liver repair after injury has been thoroughly reviewed. SNS stimulation induces hepatic fibrosis, through hepatic stellate cells activation, and also attenuates liver regeneration, through OCs accumulation inhibition [13]. Thus, selective SNS inhibitors could represent a novel therapeutic strategy against fibrogenesis, leading to reduced scarring after liver injury [60]. Consistent to the information mentioned previously are the findings by Xia et al., showing that chemical sympathetic denervation results in attenuated carbon tetrachloride (CCl₄)-induced rat liver injury. This was evident from the less decline in arterial ketone body ratio and also the decreased indocyanine green retention rate, in comparison with animals subjected solely to liver injury [61]. Furthermore, the SNS may contribute at least partly to the sensitizing effect of ethanol on lipopolysaccharide-induced liver injury in mice and concurrent administration of the beta-blocker propranolol with ethanol results in attenuation of this effect [62]. Carvedilol, an agent, which can block the SNS completely via β1, β2, and α1 adrenergic receptors, may have a protective role against development of hepatosteatosis in rats with alcohol fatty liver disease [63]. On the other hand, hepatic sympathetic innervation is protective against Fas-mediated fulminant hepatitis. In particular, both mortality and number of apoptotic hepatocytes are higher in sympathectomized mice treated with the anti-Fas antibody Jo-2, than in sham-operated animals. Administration of NE, 30 min before Jo-2 treatment, decreases mortality in the sympathectomized mice, in a dose-dependent manner, revealing, therefore, the possible beneficial role of hepatic sympathetic nerve NE in alleviation of Fas-induced fulminant hepatitis [64]. In an experimental rat model of liver ischemia–reperfusion injury (IRI), Friman et al. [65] reported that sympathetic innervation does not affect the course of the insult in terms of alkaline transaminase and mean arterial measurements.

Lam et al. examined the role of cholinergic denervation, through both atropine administration and hepatic branch vagotomy, on CCl₄-induced liver fibrogenesis. Elimination of cholinergic innervation, evident from decreased number of acetylcholinesterase-positive nerve fibers, results in attenuated liver chemical damage, through downregulation of hepatic stellate cells activation and decline in the fibrogenetic cytokines transforming growth factor-β1 and bone morphogenetic protein-6 levels [66]. Taurocholic acid exerts a protective role against the injury caused to the biliary tree, either after cholinergic [67] or adrenergic denervation of the liver [68]. It is of interest that hepatic parasympathetic denervation in hamsters with amoebic liver abscess leads to an altered inflammatory reaction, increasing the response against infection [69]. Furthermore, functional innervation of hepatic invariant natural killer cells has been found to be immunosuppressive after stroke in mice and blocking this innervation can be beneficial in terms of improving defense against infection, the leading cause of death after stroke, suggesting a possible treatment option [70]. CGRP may prevent inflammatory liver injury. In a model of liver injury after administration of lipopolysaccharide to galactosamine-sensitized mice, CGRP was found to have a protective role through induction of intrahepatic expression of transcriptional repressor inducible cyclic adenosine monophosphate early repressor and not through interleukin (IL)-10 independent inhibition of tumor necrosis factor-alpha (TNF-α) production [71]. The anti-inflammatory effects of CGRP have also been evident in a study by Kamiyoshi et al. [72] on a mouse model of concanavalin A-induced hepatitis, whereas Song et al. [73] showed that pretreatment with CGRP may result in attenuated liver IRI.

In liver cirrhosis, hepatic innervation is significantly decreased compared with that of normal liver and liver with chronic hepatitis. Expression of hepatic nerves was evaluated immunohistochemically using the pan-neuronal marker S-100 protein, and the decline in hepatic nerves' density was attributed to the neural damage caused by the inflammatory process associated with cirrhosis [74]. Furthermore, in patients with liver cirrhosis, bound leptin and soluble leptin receptor levels are significantly elevated in comparison with those of controls and are associated with increased sympathetic activity and higher expression of proinflammatory cytokines [75]. In an experimental rat model of liver cirrhosis, chemical splanchnic sympathectomy increases the intraperitoneal influx of polymorphonuclear cells, prevents bacterial translocation, and decreases incidence and severity of infections with Escherichia coli but not Staphylococcus aureus. It is considered that sympathetic tone hyperactivity in the splanchnic circulation results in host response impairment mainly to gram-negative bacteria. SNS inhibitors could represent a therapeutic option for prevention of this type of infections in advanced cirrhosis [76]. Autonomic dysfunction during liver cirrhosis has been reported as a predisposing factor to variceal bleeding [77].

Many genes in the liver, including clock genes, show a circadian rhythm of expression, reviewed in 2004 by Shibata [10], focusing on their neural regulation. In a later study, Cailotto et al. investigated the effect of sympathetic denervation on the daily variation of plasma glucose concentration in rats. They found that by disrupting hepatic sympathetic innervation in combination with a regular feeding schedule (six meals per day), the circadian rhythm of glucose concentration is lost, but circadian clock-gene messenger RNA expression remains intact [78]. On the other hand, complete denervation of the liver does not result in impaired daily variation of plasma glucose concentration, leading to the conclusion that derangement of autonomic nervous balance is more harmful for liver metabolism than maintenance of only one of its components [79]. Furthermore, although hepatic complete denervation does not affect the basal clock-gene expression, it results to the abolishment of the stimulative effects of nocturnal light on both clock-gene expression and metabolic liver enzymes [80].

In 1994, Kjaer et al. [81] reported that the levels of epinephrine and NE in transplanted livers 1 y after OLT were lower than those in normal subjects. Further evidence supporting these results came by May et al., who performed a clinical study on patients subjected either to OLT or to kidney transplantation analyzing sympathetic activation induced by water drinking. Their findings showed reduced plasma NE levels in the OLT group after drinking water compared with those in the kidney transplant group, implying impaired response to the stimulant due to liver graft denervation [82].

The denervated liver is more susceptible to hypovolemic shock according to Henderson et al. [83], who showed reduction in portal flow of denervated animals compared with that of controls in an experimental swine model. In contrast, no difference in the effect of hypovolemic shock between transplanted and sham-operated animals could be seen, based on the expression of the cytokines TNF-α, IL-6, IL-10, and the protective heat shock protein 70 [84].

Liver transplantation represents the only treatment option for end-stage liver failure, and emphasis is given on the decline of liver graft IRI, which is the leading cause of both initial graft dysfunction and primary hepatic failure posttransplantation. Sinusoidal endothelial cells play an important role in warm IRI [85]. The effect of liver denervation on the expression of the sinusoidal endothelial cell-derived intercellular adhesion molecule-1 and P-selectin, that induce liver graft IRI through hepatic microcirculation impairment, was evaluated [86], [87]. Administration of examethonium to donor rats before OLT leads to decreased intercellular adhesion molecule-1 and P-selectin messenger RNA levels of the liver graft reducing IRI risk. This result is further amplified by the simultaneous elimination of Kuppfer cells through administration of gadolinium chloride before OLT [88]. The finding of Zhu et al. supports the beneficial role of hepatic denervation and Kuppfer cells' elimination on liver graft IRI. Thus, treatment of donor rats with examethonium and/or gadolinium chloride results in decreased graft IRI, based on decreased levels of alkaline transaminase, TNF-α, and interleukin-6, and increased arterial ketone body ratio and the antioxidant enzyme superoxide dismutase [89]. Golling et al. [90] confirmed the potential protective role of surgical hepatic denervation before OLT in porcine brain-dead and living donor models, showing that elimination of liver nerves reduces hepatic oxidative stress, only in brain-dead donors though. Furthermore, in an experimental model of warm liver IRI in rats, Chen et al. [91] showed that lidocaine injection into the hepatoduodenal ligament blocking hepatic innervation before ischemia is as beneficial as ischemic preconditioning in reducing ischemic liver injury.