Anat Embryol(1992) 186:431-442

Anatomyand

Embryology

9 Springer-Verlag1992

Original articles An anterograde tracing study of the vagal innervation of rat liver, portal vein and biliary system Hans-Rudolf Berthoud, Michael Kressel, and Winfried L. Neuhuber Anatomy Institute, Universityof Ztirich, Winterthurerstrasse 190, CH-8057 Zfirich,Switzerland Accepted July 30, 1992

Summary. In order to investigate the distribution and structure of the vagal liver innervation, abdominal vagal afferents and efferents were selectively labeled by injecting WGA-HRP or Dil into the nodose ganglia, and DiA into the dorsal motor nucleus, respectively. Vagal afferent fibers produced characteristic terminal-like structures at three locations in the liver hilus: 1. Fine varicose endings preferentially surrounding, but not entering, the numerous peribiliary glands in the larger intra and extrahepatic bile ducts 2. Large, cup-shaped terminals in almost all paraganglia 3. Fine varicose endings in the portal vein adventitia. No fibers and terminals were found in the hepatic parenchyma. While about two thirds of the vagal afferent fibers that originate in the left nodose ganglion, and are contained in the hepatic branch, bypass the liver hilus area on their way to the gastroduodenal artery, a significant number (approx. 10% of the total) of vagal afferents that do innervate the area, originates from the right nodose ganglion, and projects to the periarterial plexus of the common hepatic artery and liver pedicle most likely through the dorsal celiac branch. Varicose vagal efferent fibers were present within the fascicles of the vagal hepatic branch and fine terminallike structures in a small fraction of the paraganglia. No efferents were found to terminate in the hepatic parenchyma or on the few neurons embedded in nerves or paraganglia. In contrast to the paucity of vagal terminals in the hepatic parenchyma, an abundance of vagal efferent and afferent fibers and terminals with distinctive distribution patterns and structural characteristics was present in esophagus and gastrointestinal tract. It is concluded that vagal intralobular hepatic innervation is largely absent in rats, and that the putative hepatic vagal nutrient, osmo-, and pressure receptors are located either in the portal vein, bile ducts, and/or the hepatic paraganglia. Key words: Hepatic branch of vagus - Vagal paraganglia - Dil - WGA-HRP - Laser scanning confocal microscoPY

Correspondence to: H.-R. Berthoud, Pennington Biomedical Research Center, Louisiana State University, 6400 Perkins Road, Baton Rouge, LA 70808, USA

Introduction Anatomical and functional aspects of the extrinsic innervation of the liver have been discussed in several review articles (Sawchenko and Friedman 1979; Shimazu 1979; Lautt 1983; Jungermann 1989). On the parasympathetic side, relatively few vagal efferent preganglionic fibers that originate ahnost exclusively in the left dorsal motor nucleus (dmnX) project through the hepatic branch of the ventral abdominal vagal trunk (Rogers and Herrmann 1983; Fox and Powley 1985; Carobi etal. 1985; Kohno etal. 1987). On the basis of vagotomy-induced degeneration and disappearance of stained nerves, Alexander (1940) concluded that the vagus contributes to the innervation of the cat gall bladder and biliary pathways; however, Lautt (1983) concludes in his review that vagal stimulation does not have significant effects on bile flow. With respect to the liver itself, stimulation of the parasympathetic supply also led to controversial results. It increased hepatic glycogen storage in rabbits (Shimazu 197]), decreased glucose production in sympathectomized cats (Lautt and Wong 1978), and increased blood flow in some hepatic sinusoids (Koo and Liang 1979), but did not affect basal glucose and lactate metabolism, nor blood flow in portally perfused rat liver (Jungermann 1989). Also, an unknown proportion of the vagal efferent fibers that project through the hepatic branch is likely to bypass the liver and innervate pylorus and proximal duodenum (Stavney et al. 1963; Berthoud et al. 1991 a, b). The acetylcholinesterase method was used to demonstrate cholinergic nerves selectively (Sutherland 1965; Skaaring and Bierring 1977; Cai and Gabella 1983; Mawe and Gershon 1989), but was not able to distinguish intramural cholinergic from vagal cholinergic innervation. In an attempt to label vagal hepatic innervation specifically, the cervical vagi were soaked in cobalt chloride solution in vitro for I day (Azana 1987). A dense network of stained fibers spreading with the connective tissue from the liver capsule into the parenchyma was reported. It is, however, not clear whether the connective tissue itself was stained, and the method does not allow the selective labeling of vagal efferents. Therefore the distribution

432

and structure of vagal efferent terminals in the liver and biliary system are not known. There are approximately ten times more vagal afferents than efferents (Prechtl and Powley 1990), and an impressive number of physiological and behavioral studies suggest the existence of vagal hepatic afferents sensitive to metabolites such as glucose (Niijima 1980, 1982), amino acids (Tanaka et al. 1990), and fatty acids (Langhans and Scharrer 1987a), as well as osmolarity (Dennhardt et al. 1971; Passo et al. 1973; Adachi et al. 1976; Rogers and Novin 1980) and temperature of the portal blood (Adachi and Niijima 1982). It is thought that the fluctuations in hepatic oxidation of different metabolic fuels modulate the activity of the hepatic sodium pump, and affect hepatocyte membrane potential, so that innervated hepatocytes may act as metabolic receptors (Russek 1963; Langhans and Scharrer 1987a, b; Tordoff et al. 1991). In most of these studies vagal afferents have been implicated on the basis of selective hepatic vagotomy. Therefore, vagal afferent endings could be expected in the hepatic parenchyma at the level of the hepatocyte. Similarly, a role for the hepatic osmoreceptors in salt intake has been demonstrated (Rogers and Novin 1980; Contreras and Kosten 1981). Except for experiments by Tsai (1958), using the gold/ silver impregnation method on liver tissue from vagotomized humans, no attempt to label vagal afferents selectively was made. Only a few degenerated nerve fibers at the porta and in hepatic lobules were found in that study. Since we have used the fluorescent carbocyanine dyes as well as WGA-HRP, to anterogradely label vagal fibers and terminals in the digestive tube (Neuhuber 1987; Berthoud et al. 1990, 1991 a; Berthoud and Powley 1992) and pancreas (Neuhuber 1989; Berthoud and Powley 1991), we wanted to apply these tracers to label the vagal liver innervation. The fluorescent dye DiA was injected into the dorsal motor nucleus and DiI or WGAH R P were injected into the nodose ganglia of rats, for a more specific and detailed assessment of vagal efferent and afferent hepatic innervation, respectively.

Materials and methods

Animals. Male (n=22) and female (n=2) rats ( Z U R : S I V strain, Institute for Laboratory Animal Science, University of Zfirich), weighing 150-300 g were held individually in acrylic cages with wood shavings in a climatized colony room (12/12 h L : D cycle, 23 _+3~ C). Normal rat pellets and tap water were available ad libiturn, except the night before surgery.

Labeling of vagal afferents. Animals were anesthetized with a mixture of Sedalande (Delalande, 10 mg/kg) and Fentanyl (Janssen, 0.2 mg/kg, i.m.), and the additional administration of Valium (Roche, 2.5 mg/kg, i.m.). Then, when the animals were fully unresponsive, either the left or right nodose ganglion was exposed by a ventral approach. The ganglion capsule was slit open and a finetipped (20-60 ~tm external diameter) micropipette containing the tracer was guided into the ganglion by means of a micromanipulator. Tracer was then injected in volumes of 0.1-1.0 gl, using either air pressure, or a fluid-filled system with a microliter syringe. Four rats were bilaterally injected with W G A - H R P (2% in saline, Sigma)

and 20 rats with delta-9-DiI (1,1'-diolelyi-3,3,3',3'-tetramethylindocarbocyanine methanosulfonate, 25 mg/ml in methanol or methanol/ethanol solution, Molecular Probes, Eugene, Ore.). DiI was injected into the left (n=14) or right (n=2) nodose ganglion. In two animals a left supranodose vagotomy (Berthoud and Powley 1992) was performed 2 weeks prior to DiI injection into the ipsilateral nodose ganglion, in order to eliminate the efferent vagal fibers. In two more animals the left cervical vagus nerve was transected immediately following the injection, in order to control for possible unspecific, non-vagal, transport.

Labeling of vagal efferents. The anesthetized rats were mounted in a stereotaxic apparatus with a sharply flexed neck position. The dorsal surface of the caudal medulla oblongata was then exposed surgically by removing the atlanto-occipital membrane. Using the obex as a reference point, the left dmnX was then injected at three different rostro-caudal levels (approx. 1 mm apart), in order to blanket the entire nucleus. At each site, 50-200 nl of DiA (4-(4-dihexadecylaminostyryl)-N-methylpyridinium iodide, 1.5% in 50% dimethylformamide/50% EtOH, Molecuiar Probes) was injected slowly. Some animals (n = 7) received both DiA injections into the dmnX and DiI injections into the nodose ganglia, while two animals received only DiA injections into the dmnX.

Fluorogold counterstain. In order to label autonomic and enteric ganglia and the perineurium of the peripheral nerves, Fluorogold (1-3 mg in 1.2 ml sterile saline, Fluorochrome, Engelwood, Colo.) was injected i.p. in food-deprived animals, 5 days before sacrifice. For details of this method see Powley and Berthoud (i991). The Fluorogold injections served additionally to retrogradely label vagal afferent neurons in the nodose ganglia and vagal efferent preganglionic perikarya in the dmnX.

Tissue processing. Following a lethal dose of sodium pentobarbitai (100 mg/kg, i.p.), and administration of 100 units of heparin into the saphenous vein, the animals were transcardially perfused with 300 ml ringer solution and 800 ml of 10% phosphate-buffered formalin (a mixture of 1% paraformaldehyde and 1.25% glutaraldehyde for W G A - H R P injected rats, pH 7.4). The liver and gastrointestinal tract were then extracted and stored in the same fixative in the refrigerator. The hepatic pedicle containing the hepatic artery proper, the portal vein, and the major bile ducts was excised with microscissors under binocular microscope guidance. Pieces of adhering liver parenchyma were scraped off, and the three components were then separated from each other with fine forceps. Small pieces of parenchyma (approx. 500 mm3), from close to the hilus were trimmed with a razor blade, and cut with either a Vibratome at 100 lam or a cryostat at 40-60 gm, and collected in phosphatebuffered saline (pH 7.4). The medulla oblongata was also cut in a cryostat at 40 gin. Sections and whole mounts were then mounted in 100% glycerol to which 50 g/1 N-propylgallate had been added as an antifade agent. For the demonstration of WGA-HRP, the tetramethyl benzidine method of Mesulam (1982) was used. Microscopic analysis. A Zeiss Axiophot epifluorescence microscope was used to detect the fluorescent markers. DiI was best seen through filter block 15, DiA through filter block 10, and Fluorogold through filter block 02. Photomicrographs were taken on b/w film T-max 400, or color negative film EES, both exposed at 800 ASA. A laser scanning confocal microscope (Zeiss), atlowing optical sectioning and 3-D reconstruction of selected objects, was also used. Typically a x 40 objective was used and additional electronic zoom factors of up to • 6. In order to obtain all-in-focus images of thicker structures, up to 25 serial optical sections (z-sectioning), 0.5-2.0 pm apart were taken and electronically superimposed. A helium-neon laser, in combination with appropriate filters, resulted in images of the DiI-labeled profiles with little background detail in the longer wavelengths, hnages created with the argon-ion laser, and its appropriate filters, on the other hand, showed greater detail

433 of the surroundingstructures, causedby either the Fluorogoldlabel or autofluorescence.By successivelygeneratingimagesthrough the two channels, and then subtracting one from the other, the brightest structures of the subtrahend became negative and appeared in black. DiI-labeled structures or surroundingelementslike collagen fibers could thus be demonstrated simultaneouslywith high contrast. The built-in reliefand histogram functionswere also used to improve contrast and clarity. Results

Vagal afferents In general, the same patterns of labeling were observed with the two tracers. For the description of the gross distribution of vagal afferent fibers whole mounts were most valuable, and DiI-labeled specimens were used for this purpose. Since the nerve plexus is intimately attached to the hepatic artery, it remained more or less intact when the artery was separated from the bile ducts and the portal vein. This complex network of diverging and reanastomosing nerve branches became visible through the Fluorogold labaled nuclei of the perineural sheath (Fig. 1 B). Only the very rare Fluorogold-labeled neuron or small group of neurons was found embedded in some of the thicker nerve branches. An average of 9.1 _+1.3 (n = 14) paraganglia containing almost exclusively glomus cells, with only the occasional neuron, were either embedded within nerve bundles or formed bulb-like extensions, a few of them being quite large. Upon switching to the filter block appropriate for DiI, the gross distribution of vagal afferent fibers in the liver pedicle could be easily assessed. In animals with left nodose ganglion injections, DiI-labeled vagal afferents were contained in at least two subbranches of the so called hepatic branch, which is loosely associated with the hepatoesophageal artery (Fig. 1). Near the junction of this latter artery with the hepatic artery proper, the main DiI-labeled subbranch clearly divided into a portion that turned towards the common hepatic and gastroduodenal arteries, and a portion that turned towards the liver hilus. A number of smaller nerve fascicles containing fewer DiI-labeled axons divided from the larger branches, and while some of them entered the common hepatic artery plexus, some turned towards the individual liver lobes (Fig. 1). On average, approximately two thirds of the DiI=labeled fibers contained in the hepatic branches bypassed the liver hilus on their way to the gastroduodenal and common hepatic arterial plexuses. In the two animals with right nodose ganglion injections, DiI-labeled fibers could be seen to enter the liver pedicle from the common hepatic artery, and to enter some of the smaller nerve bundles that accompany the small arteries supplying the individual liver lobes, but not in the hepatic branch.

Hepatic parenchyma. A large number of cryostat and vibratome sections from liver parenchyma of different lobes, particularly near the hilus from DiI- and WGAHRP-injected animals was inspected at high magnification. Some labeled fibers were found to accompany the

larger triads, particularly associated with the bile ducts, and less frequently with portal vessels (see below). However no DiI- or WGA-HRP-labeled fibers were seen to enter the parenchyma. No suspiciously labeled terminal structures could be found within lobules and near hepatocytes. Analysis of WGA-HRP-processed material was somewhat impeded by heavy background staining due to endogenous peroxidase-like activity of hepatocytes. In specimens of DiI-injected rats, DiI label was found in presumed Kuppfer and/or Ito cells. Since these same cells were also labeled in control rats with cervical vagotomy, but not in rats without DiI injections they must have taken up the dye from the circulation.

Bile ducts. The main targets of vagal afferent fibers were the larger bile ducts, particularly the portions around the exit from the individual liver lobes (Fig. 8). In material from Fluorogold-injected rats, all bile ducts were easily recognized in the fluorescence microscope, because the many peribiliary glands were strongly Fluorogoldlabeled (Fig. 2A, B). Only the finest nerve fascicles remained on the biliary duct system when it was separated from the other components of the liver pedicle. Originating from such fascicles, DiI-labeled fibers were seen to distribute on the outer surface of the bile ducts, and to produce, locally very dense, clusters of terminal-like structures that characteristically spared the peribiliary glands (Fig. 2C, D, E). As seen in semithin sections, the rat bile duct wall consists of a columnar epithelium surrounded by a rich connective tissue sheath of fine and coarse collagen fiber bundles. There are very few smooth muscle fibers or none. In cross-sectioned ducts from WGA-HRP material, fine varicose labeled fibers were seen penetrating the connective tissue sheath and approaching the epithelial layer. Terminal-like swellings were found immediately beneath, but not within the columnar epithelium (Fig. 2 F). In the two animals with right nodose ganglion injections, DiI-labeled terminals showed the same distribution and structure (Fig. 2E), but they were much less frequent (Table 1). For the remainder of this paper, the terms "terminal-like structures", "terminals" and "endings" will be used interchangeably, and merely to distinguish them from passing axons. It is implied that these structures make synaptic contacts with surrounding tissues, or are free sensory nerve endings, but ultrastructural analysis will be needed for proof.

Portal vein. In all but two animals with left nodose ganglion injections, the portal vein showed small areas with DiI-labeled vagal afferent fibers that seemed to form fine terminal arborizations (Table 1, Fig. 3 A). From serial optical sections in the confocal microscope, it could be clearly established that some of the terminals ran deep within the collagen fiber-rich adventitia (Fig. 3 B, C). In one of the two rats with right nodose ganglion injections, similar varicose endings were found (Fig. 3 D). In cross-sectioned portal vessels from WGAHRP material, labeled fibers could also occasionally be detected in the adventitia.

median lobe left bbe

V

V

- - - - - ventral trunk

/

/

hepatoesophageal A.

vent. celiac br.

hepatic branch

\ to right lobe

~Y t)~---k~\\ %

vent. gastr, br.

to caudate

gastroduod. _ _ branch

hepatic A proper

,~

gastroduodenal A.

Fig. 1. Gross distribution of hepatic afferent vagal innervation, reconstructed from whole-mount preparation of hepatic artery of animal with DiI injected into left nodose ganglion. The hepatic arterial supply with its complex nerve plexus was traced with the aid of a camera lucida and the support of a photomontage taken with the • 10 objective. The inset at the top is an example of this nerve plexus, which is visible because of the Fluorogold-labeled perineural nuclei. DiI-labeled vagal afferent fibers became visible

hepatic A.

by switching to the appropriate filter, and their reIative density in each of the nerves of the plexus is indicated by the size of the dots. The inset at the lower right is a montage and shows the hepatic branch as it distributes fascicles to the nerves of the plexus. Several paraganglia, that were all innervated, are also shown. Fascicles containing vagal afferents, that were torn by the separation of the arterial plexus from the portal vein and biliary tree are indicated by open triangles

435

Fig. 2A-F. Vagal afferent innervation of bile ducts. A Fluorescent conventional micrograph of whole mounted bile duct with Fluorogold-labeled peribiliary glands. B Higher power, laser scan confocal microscope image (single tangential optical section) of bile duct wall, showing Fluorogold-labeled peribiliary gland and surrounding collagen fibrils. C Same peribiliary gland as in B, optically sectioned at more peripheral level, and encircled by DiI-labeled varicose vagal afferent fibers (subtraction image). D Dense cluster of DiI-labeled vagal afferents sparing peribiliary glands in bile duct

wall. Overlay image of five optical sections, i gm apart. E Higher magnification of DiI-labeled varicose vagal afferent endings in bile duct of animal with right nodose ganglion injection. Overlay image of ten optical sections, 1 gm apart. F Conventional lightfield micrograph from sectioned liver hilus, showing WGA-HRP-labeled vagal afferents (arrows) immediately beneath epithelium (e) of cross cut bile duct. Bar 500 gm for A, 25 gm for B, C, and F, 50 gm for D, 20 gm for E

436

Fig. 3A-D. Vagal afferent innervation of portal vein. A Conventional low power photomicrograph of network of DiI-labeled vagal afferents in wall of portal vein. B, C Laser-scan confocal microscope images of varicose vagal afferent fibers (white) embedded in collagen fibers (black) of portal vein adventitia. Dual laser subtraction images of series of 16 (B) and 12 (C) optical sections,

1 lam apart. The specimen in C was from an animal with supranodose vagotomy prior to DiI injection. D Conventional microscope view of DiI-labeled vagal afferent terminal on portal vein from animal with right nodose ganglion injection. Bar 120 ~tm for A, 20 gm for B, and C, 35 gm for D

Vagalparaganglia. One of the most striking observations in rats injected in the left nodose ganglion was that almost every paraganglion was innervated by DiI-labeled vagal afferents (Table 1, Fig. 4A, B, C). Even at low magnification, the strongly labeled, sometimes dense accumulations of terminal arborizations covering parts of, or entire, paraganglia were easily seen. At higher magnification with the conventional microscope, and particularly from optical sections and 3-D reconstructions with the laser scanning confocal microscope, further details became apparent (Fig. 4 B). Often, small spherical clusters of three to six glomus cells were partially surrounded by large, cup-shaped labeled terminals. In one of the two rats with right nodose ganglion injections, one paraganglion was innervated by DiI-labeled terminals.

(Fig. 5 B) could be easily detected. As can also be seen in Table 1, both animals with right nodose ganglion injections showed the same degree of completeness of label as the left nodosal injections. A third animal with right nodose ganglion injection showed poorer labeling in the stomach wall and was therefore not included in the analysis. Similarly, all results of WGA-HRP-labeled hepatic vagal afferents were obtained from experiments in which heavy labeling was observed in esophagus, stomach, and duodenum (Neuhuber 1987).

Esophagus and gastrointestinal tract. Strongly DiI-labeled vagal afferent fibers and terminals were found throughout the myenteric plexus of the stomach and duodenum (Table 1, Fig. 5A). Additionally, the often very fine vagal afferents in the gastric smooth muscle layers (Fig. 5C), duodenal submucosa, and mucosa

VagaI efferents Only a few very thin DiA-labeled vagal efferent fibers, sometimes forming larger varicosities, were found in the main fascicles of the vagal hepatic branch and, less frequently, in some of the finer fascicles of the liver pedicle. A small fraction of the total number of paraganglia contained a few fine, varicose, DiA-labeled terminals (Table 1, Fig. 4D). In two cases, DiA-labeled fibers were found on the portal vein. DiA-labeled fibers and terminals were not found in the liver parenchyma, on bile

437

Fig. 4A-D. Laser scan confocal microscope images of vagal afferent (A, B, C) and efferent (D) innervation of paraganglia associated with hepatic branch of vagus and liver hilus. A Single optical section through medium sized paraganglion adjacent to distal hepatic branch. Several DiI-labeled afferent endings (bright white) surround small groups of glomus cells. Note the many capillaries (asterisks), and the accompanying nerve (N). B Higher magnification of glomus cell group engulfed by large, cup-shaped afferent terminals

(white). C Example of vagal afferent innervation of small paraganglion from animal with supranodose vagotomy prior to DiI injection into left nodose ganglion. Overlay image of seven optical sections, i gm apart. Inset shows location of paraganglion in nerve (N) containing few labeled fibers. D DiA-labeled vagal efferents in paraganglion of liver hilus. Bar 20 gm for A and D, 8 gm for B, 13 gm for C

ducts, or in the few neurons or groups of neurons embedded in nerve bundles in any of the animals. In contrast to this relative paucity of vagal efferents in the liver area, large beaded as well as finer varicose terminals were very frequent in the myenteric plexus of stomach and duodenum (Table 1, Fig. 5 D).

left cervical vagotomy immediately after injection of the dye into the left nodose ganglion.

Control experiments Supranodose vagotomy. In the two animals that received DiI injections into the left nodose ganglion following ipsilateral supranodose vagotomy, the gross labeling pattern (Table 1), as well as the morphological characteristics of the terminal fields in the bile ducts, portal vein (Fig. 3 C), and paraganglia (Fig. 4C), were similar, although the frequency of labeled structures was slightly less. Control for non-vagal transport. No DiI-labeled fibers and terminals were found in the animals that underwent

Injection sites and labeling in medulla Nodose ganglion injections. Following DiI injections into the nodose ganglia in non-vagotomized animals various degrees of anterograde labeling in the ipsilateral solitary tract and NST, with some crossing over to the contralateral side, were observed (Fig. 6A). In the two animals with prior supranodose vagotomy there was no such label in the NST (for microphotographs see Berthoud and Powley 1992). Successful supranodose (and cervical) vagotomies were further indicated by the absence of Fluorogold-labeled preganglionic neurons in the ipsilateral dmnX (Fig. 7). In the animals without vagotomy, there was a population of Fluorogold-labeled and another population of DiI-labeled dmnX neurons (Fig. 6). No double-labeled neurons were present. The neurons without Fluorogold label must have been damaged by the nodose ganglion injection procedure, and only these

438

Fig. 5A-D. Demonstration of vagal afferent (A, B, C) and efferent (D) innervation of gastrointestinal tract in same animals analyzed for vagal hepatic innervation. A DiI-labeled vagal afferents (white) in myenteric plexus of gastric corpus. Strongly labeled single fiber leaves larger bundle to produce complex, intraganglionic laminar ending. Dual laser subtraction image shows details of longitudinal smooth muscle layer (horizontal direction) in background. B Ex-

ample of afferent innervation in lamina propria of duodenal mucosa. Overlay image of five optical sections, 2 gm apart. C Fine vagal afferent fibers in longitudinal smooth muscle layer of gastric fundus. Overlay image of 21 optical sections, 2 gm apart. D Conventional microscope image of DiA-iabeled vagal efferent terminal in myenteric plexus of gastric fundus. Bar 40 Ixm for A, 60 gm for B and D, 100 gm for C

Table 1. Frequency of observed afferent and efferent vagal innervation of liver hilus N

Bile ducts

Portal vein

Gastric wall

Vagal afferents (DiI) Left nodose ggl. injection with prior supranodose vagotomy with cervical vagotomy Right nodose ggl. injection

14 2 2 2

+ + to + + + + + + to + +

- to + + + - to +

+ + to + + + + + to + + + -+ + to + + +

Vagal efferents (DiA) left dmnX injection

10

-

- to +

+ to + + +

Paraganglia Total

Labeled

9.1_+1.3 7.5+_0.4 10.0_+0.7 10i0+-1.5

8.1_+0.9 7.0+_0.5 0 0.5_+0.5

9.7_+2.1

1.8_+1.0

+ + + abundant; + + more than one area with label; + at least one area with label; -- no label

d a m a g e d n e u r o n s a c c u m u l a t e d DiI in their cell body. T h e i n t a c t efferent fibers o f passage (which were retrogradely F l u o r o g o l d - l a b e l e d ) did n o t t r a n s p o r t D i I (for f u r t h e r p h o t o m i c r o g r a p h i c p r o o f see B e r t h o u d a n d Powley 1992).

D m n X injections. O u r D i A injections into the d m n X were relatively large a n d b l a n k e t e d m o s t o f the left nucleus. These injections also spilled i n t o the n e i g h b o u r i n g N S T , hypoglossal nucleus, a n d the right d m n X . D i A labeled cell bodies in the n o d o s e ganglia were n u m e r o u s ,

439

Fig. 7. Verification of supranodose vagotomy with intraperitoneal Fluorogold injection. Absence of retrogradely Fluorogold-labeled neurons in left dmnX (arrows), demonstrates complete left vagotomy

Fig. 6A, B. Photomicrographs of frontal section of dorsal medulla at level of area postrema (AP), showing DiI and Fluorogold label in nucleus of the solitary tract (NST) and dorsal motor nucleus (dmnX). A Anterograde DiI label in the left solitary tract and NST (arrows), and retrograde DiI label in perikarya of left dmnX (arrowhead). B Photomicrograph taken from the same section as in A, with different filter, to show retrogradely Fluorogold-labeled cells in dmnX as a test for their viability. Note that more than half of the cells in the left dmnX are labeled

but there was no evidence for transganglionic transport into the abdominal afferent terminals. DiA label was never found in intramusuclar and mucosal fibers and terminals. It was difficult to estimate the number of damaged dmnX neurons, because bleedthrough of the high concentration of DiA at the center of the injection precluded detection of Fluorogold. Since in the rats, which received both nodosal and dmnX injections, damage could have been cumulative, two more animals received only dmnX injections. The high frequency of DiA-labeled terminals in the myenteric plexus of the gastric wall suggests that a large percentage of dmnX neurons remained intact.

vagotomy, which eliminated all efferent fibers of passage, prior to the injection, neither abolished labeling in the liver pedicle, nor did it change the distribution and architecture of the labeled terminals. Furthermore, the distribution and architecture of labeled vagal efferent fibers and terminals in the gastric wall following DiA injections into the dmnX was different from the DiIlabeldd afferents. Using the same methodologies, we have earlier shown that neither DiI nor W G A - H R P injected into the nodose ganglia or directly into the cervical vagus nerve, label intact fibers of passage (Neuhuber 1987; Berthoud and Powley 1992).

Bile duct innervation This is the first description of vagal afferent innervation of hepatic bile ducts. In the guinea pig, vagal afferent innervation of the gall bladder was indicated on the basis of retrograde labeling experiments (Mawe and Gershon 1989), but no anterograde tracing study is available. The locally very dense vagal afferent innervation is somewhat surprising, because there has been neither physiological evidence, nor a hypothetical functional role, for such innervation. A rich supply of neuropeptide-containing, mostly sympathetic fibers has been documented (Goeler et al. 1988; Inoue et al. 1989; Mawe and Gershon 1989; Terada and N a k a m u r a 1989; Ding et al. 1991). The most likely role for vagal afferents is a mechanoreceptive one. Since the rat has no gall bladder, the larger hepatic bile ducts may be the functional analog to the gall bladder, and their tension may be monitored by the vagal afferents. A chemoreceptive role in the detection of bile composition and/or osmolarity is also possible.

Discussion

Portal vein innervation We have reason to believe that our tracer injections into the nodose ganglia selectively labeled vagal afferents, and not efferent fibers of passage. For one, supranodose

This is the first structural identification of vagal afferent innervation of the portal vein. Using the retrograde

440 tracer True Blue, Barja and Mathison (1984) found a few labeled neurons in the nodose ganglia, and a large number of substance P-containing neurons in the dorsal root ganglia of T8 T13. The DiI-labeled endings were dearly associated with the wall of the portal vein. Quantitatively, this innervation was, however, less pronounced than the bile duct innervation. We can only speculate about its possible function. Are they mechanoreceptors, chemoreceptors, or both? Physiological evidence for portal vein osmoreceptors (Baertschi and Vallet 1981) and pressoreceptors (Niijima 1977), as well as glucosensitive vagal afferents (Niijima 1980) has been presented, although it is generally believed, that the receptive site for osmo and glucose-sensors is located within the hepatic parenchyma (Dennhardt et al. 1971; Adachi et al. 1976; Rogers and Novin 1980; Niijima 1982).

Paraganglia innervation

Almost every single paraganglion was innervated by vagal afferents, and a few of them also received a vagal efferent innervation. Since there was, however, no label in the paranganglia of the animals with vagotomies, it is highly unlikely that this label was of vascular origin. This could be suspected because of the high degree of vascularization, and preferential accumulation of e.g. Evans Blue (McDonald and Blewitt 1981) by the paraganglia. A neuronal origin of the label is also indicated by labeled axons entering the paraganglia. The first systematic inventory of abdominal vagal paraganglia was presented by Goormachtigh (1936), in the mouse. He found that one of the largest and some of the smaller abdominal paraganglia were regularly located in the liver pedicle, and suggested that this must be of physiological importance. Goormaghtigh believed that these paraganglia were functionally related to the efferent fibers of the vagus. Experiments by Hollinshead (1946) provided evidence that the rat abdominal paraganglia produced chemoreceptive effects on blood pressure, but that the dorsal root ganglia were the major link to the CNS, with the vagus playing a minor role. More recently, ultrastructural studies found the presence of two types of paraganglion cells: chief or Type I cells, and satellite or Type II cells (Chen and Yates 1970; Morgan et al. 1976). Many larger, cup-shaped afferent nerve endings with multiple synaptic contacts, and smaller efferent nerve endings with fewer synaptic contacts on chief cells were observed (Morgan et al. 1976). Furthermore, the large afferent nerve endings degenerated upon infranodose vagotomy, but persisted following supranodose vagotomy (Kummer and Neuhuber 1989), and vagal afferent origin of terminals in thoracic and gastric paraganglia was directly demonstrated by WGA-HRP anterograde tracing from the nodose ganglia (Neuhuber 1987; Kummer and Neuhuber 1989). Our own observation, that paraganglia of the hepatic pedicle are innervated by vagal afferent fibers suggests that this is a common feature of paraganglia, and that they serve a chemoreceptive function.

Fig. 8. Schematic summary of results, showing pathways, distribution, and termination of vagal afferents in hepatic pedicle. Most of the vagal afferents that innervate bile ducts, paraganglia, and portal vein, originate in the left nodose ganglion and project through the hepatic branch. A smaller contribution to the innervation of these targets originates in the right nodose ganglion and projects through the dorsal celiac branch and the periarterial plexus of the common hepatic artery

Vagal access routes to the liver

The left nodose ganglion and its projection through the so-called hepatic branch is not the only source of vagal afferents to the liver pedicle. Approximately 10% of the total innervation comes from the right nodose ganglion and reaches the periarterial plexus of the common hepatic artery most likely through the dorsal celiac branch (Fig. 8). This confirms earlier electrophysiological (Niijima 1983; Rogers et al. 1984), and retrograde tracing studies (Carobi 1990). Furthermore, approximately two thirds of the vagal afferents contained in the hepatic branch seem not to terminate in the liver area, but follow the gastroduodenal and/or common hepatic arteries. It has earlier been demonstrated that a portion of the vagal efferents contained in the hepatic branch also bypasses the liver hilus and innervates the area of the pyloric sphincter, proximal duodenum, and pancreas (Stavney et al. 1963; Berthoud et al. 1991 a, b). The term "hepatic branch" is therefore misleading and should be changed to "common hepatic branch", in analogy to the hepatic arterial supply. The term" hepatic branch proper" could then be used for the fascicles that indeed terminate in the liver hilus, while "gastroduodenal branch" would designate the descending portion.

441 These anatomical configurations should be considered when behavioral results of experiments involving hepatic branch vagotomies are interpreted.

Absence of hepatic parenchyma innervation In spite of the considerable evidence for vagal afferent innervation of the liver p a r e n c h y m a from behavioral and physiological studies, we did not find any significant n u m b e r of labeled fibers and terminals in this compartment. It is unlikely that our method failed to make such vagal fibers and terminals visible, unless they are much finer than all the other vagal afferents described in this study, including the very fine fibers in the gastric smooth muscle layers, and duodenal mucosa. There was also no problem with masking Dil labeled fibers by high background and/or autofluorescence. Labeled axons traveling with obliquely cut small bile ducts could easily be detected in sectioned material. It could also be argued that we only labeled a small fraction of vagal fibers, because they were either damaged, or did not take up the tracer. The p r o p o r t i o n of Fluorogold-labeled d m n X neurons is an indicator for the number of damaged efferent fibers. In the best cases, only approximately 25~40% of these efferent fibers were damaged. If the damage to afferents and efferents is proportional, and if we assume (a valid assumption, considering the relatively large injections, and the abundant label in the gastric wall), that the great majority of the intact fibers t o o k up the label, we can conclude that on each side at least half of all vagal afferents and efferents were anterogradely labeled. Furthermore, the results with the very different tracer W G A - H R P were essentially the same. Sparse or absent innervation of the hepatic p a r e n c h y m a of the rat has been reported earlier (Alexander 1940; Metz and Forssmann 1980). The absence of vagal afferent innervation in the hepatic p a r e n c h y m a is in contrast to the considerable evidence for vagal hepatic nutrient and osmoreceptors. The first explanation of this dilemma, that comes to mind, is to ascribe this receptor function to the vagal afferent terminals that we found outside the parenchyma, in the walls of bile ducts and portal vein, and in the hepatic paraganglia. The suggestion of Forssmann and Ito (1977) that there is an inverse relationship between the number of gap junctions and neural elements within the hepatic p a r e n c h y m a is most interesting in this respect. In the rat, which has a large n u m b e r of gap junctions, the hepatocytes m a y thus be maximally electrotonically coupled, and do not need to be individually innervated. Rather, a c o m p o u n d m e m b r a n e potential of m a n y hepatocytes could be detected by relatively few receptors located at the periphery of the parenchyma. We have not seen any suspicious terminal-like structures at the periphery, but we have seen m a n y such structures on intrahepatic bile ducts, in close proximity to hepatocytes. Therefore, it would have to be further suggested that the gap junctions between parenchymal cells continue into the bile duct epithelium, thus allowing electrotonic coupling of the hepatocyte potential with vagal afferents.

Alternatively, some of the receptive functions could be accomplished by terminals in the wall of the portal vein. Most interestingly, it has been noted by Niijima (1980) that changes in hepatic branch nerve discharge with glucose could be obtained in isolated perfused portal vein preparations after removal of the liver lobes. Similarly, Baertschi and Vallet (1981) found evidence for portal vein osmoreceptors in the rat. Finally, some of the putative vagal sensory functions of the liver could be performed by the terminals found in hepatic paraganglia. I f these glomus-like bodies are chemosensors that m o n i t o r blood gases, it is difficult to see their role in specific hepatic sensation, unless the paraganglia are irrigated by portal or hepatic venous blood. There is no evidence for such portal circuits.

Acknowledgments.We thank Ruth Russi, Sabine Richter, and Claudia Meyer-Gresele for technical assistance, Margrit Miiller and Helga Weber for photography, and Jos6 Perez and Curt Bernet for animal care. We would like to dedicate this article to Wolfgang Zenker, who is retiring after many years of dedicated service and guidance as director of the Anatomy Institute, University of Zfirich. We thank him for his general supportive attitude, and his suggestions on an earlier draft of the manuscript. The project was also supported by grants from the Roche Foundation and the Hartmann-Miiller Stiftung, Z/irich (No. 439), to W.L.N.

References Adachi A, Niijima A (1982) Thermosensitive afferent fibers in the hepatic branch of the vagus nerve in the guinea pig. J Auton Nerv Syst 5:101-109 Adachi A, Niijima A, Jacobs HL (1976) An hepatic osmoreceptor mechanism in the rat: electrophysiological and behavioral studies. Am J Physiol 231 : 1043-1049 Alexander WF (1940) The innervation of the biliary system. J Comp Neurol 72:357-370 Azana MJ (1987) The vagal contribution to the rat liver innervation: a demonstration with the cobalt impregnation method. Comp Biochem Physiol 86A:275-279 Baertschi AJ, Vallet PG (1981) Osmosensitivity of the hepatic portal vein area and vasopressin release in rats. J Physiol (Lond) 315:217-230 Barja F, Mathison R (1984) Sensory innervation of the rat portal vein and the hepatic artery. J Auton Nerv Syst 10:117-125 Berthoud HR, Powley TL (1991) Morphology and distribution of efferent vagal innervation of rat pancreas as revealed with anterograde transport of DiI. Brain Res 553:336-341 Berthoud HR, Powley TL (1992) Vagal afferent innervation of the rat fundic stomach: morphological characterization of the gastrointestinal tension receptor. J Comp Neurol 319:261-276 Berthoud HR, Jedrzejewska A, Powley TL (1990) Simultaneous innervation of vagal innervation of gut and afferent projections from the visceral forebrain with DiI injected into the dorsal vagal compiex in the rat. J Comp Neurol 301:65-79 Berthoud HR, Carlson NR, Powley TL (1991a) Topography of efferent vagal innervation of rat gastrointestinal tract. Am J Physiol 260 (Regulatory Integrative Comp Physiol 29):R200 R207 Berthoud HR, Fox EA, Powley TL (1991 b) Abdominal pathways and central origin of rat vagal fibers that stimulate gastric acid. Gastroenterology 100:627 637 Cai WQ, Gabella G (1983) Innervation of the gall bladder and biliary pathways in the guinea pig. J Anat 136:97-109 Carobi C (1990) Vagal afferent innervation in regenerated rat liver. Experientia 46:50 53

442 Carobi C, Della Torre G, Magni F (1985) Differential distribution of vagal afferent neurons from the rat liver. Neurosci Lett 62:255 260 Chen IL, Yates RD (1970) Ultrastructural studies of vagal paraganglia in syrian hamsters. Z Zellforsch Mikrosk Anat 108: 309-323 Contreras R J, Kosten T (1981) Changes in salt intake after abdominal vagotomy: evidence for hepatic sodium receptors. Physiol Behav 26:575 582 Dennhardt R, Ohm WN, Haberich FJ (1971) Die Ausschaltung der Leber/iste des N. vagus an der wachen Ratte und ihr Einfluss auf die hepatogene Diurese - indirekter Beweis ffir die afferente Leitung der Leber Osmorezeptoren fiber den N. vagus. Pflfigers Arch 328:51-56 Ding WG, Fujimura M, Mori A, Toojama I, Kimura H (1991) Light and electron microscopy of neuropeptide Y-containing nerves in human liver, gallbladder and pancreas. Gastroenterology 101 : 1054-1059 Forssmann WG, Ito S (1977) Hepatocyte innervation in primates. J Cell Biol 74:299-313 Fox EA, Powley TL (1985) Longitudinal columnar organization within the dorsal motor nucleus represents separate branches of the abdominal vagus. Brain Res 341:269-282 Goehler LE, Sternini C, Brecha NC (1988) Calcitonin gene-related peptide immunoreactivity in the biliary pathway and liver of guinea pig: distribution and colocalization with substance P. Cell Tissue Res 253:145-150 Goormaghtigh N (1936) On the existence of abdominal vagal paraganglia in the adult mouse. J Anat 71:77-90 Hollinshead WH (1946) The function of abdominal chemoreceptors of the rat and mouse. Am J Physiol 147:654-660 Inoue N, Magari S, Sakanaka M (1989) Distribution, possible origins and fine structure of neuropeptide Y-containing fibers in the rat River. Brain Res 493 : 87-96 Jungermann K (1989) Regulation von Leberfunktion durch autonome hepatische Nerven. Naturwissenschaften 76 : 547-559 Kohno T, Mori S, Mito M (1987) Cells of origin innervating the liver and their axonal projections with synaptic terminals into the liver parenchyma in rats. Hokkaido J Med Sci 62: 93-946 Koo A, Liang IYS (1979) Microvascular filling pattern in rat sinusoids during vagal stimulation. J Physiot (Lond) 295:191 199 Kummer W, Neuhuber WL (1989) Vagal paraganglia of the rat. J Electron Microsc Techn 12:343-355 Langhans W, Scharrer E (1987a) Evidence for a vagally mediated satiety signal derived from hepatic fatty acid oxidation. J Auton Nerv Syst 18:13-18 Langhans W, Scharrer E (1987b) Evidence for a role of the sodium pump of hepatocytes in the control of food intake. J Auton Nerv Syst 20:199-205 Lautt W (1983) Afferent and efferent neural roles in liver function. Progr Neurobiol 21 : 323-348 Lautt W, Wong C (1978) Hepatic parasympathetic neural effect on glucose balance in the intact liver. Can J Physiol Pharmacol 56 : 679-682 Mawe GM, Gershon MD (1989) Structure, afferent innervation, and transmitter content of ganglia of the guinea pig gall bladder: relationship to the enteric nervous system. J Comp Neurol 283 : 374-390 McDonald D, Blewitt RW (1981) Location and size of carotid body-like organs (paraganglia) revealed in rats by the permeability of blood vessels to Evans blue dye. J Neurocytol 10: 607643 Mesulam MM (1982) Tracing neuronal connections with horseradish peroxidase. Wiley, New York, 251 p Metz W, Forssmann WG (1980) Innervation of the liver in guinea pig and rat. Anat Embryol 160:239 252

Morgan M, Pack RJ, Howe A (1976) Structure of cells and nerve endings in abdominal vagal paraganglia of the rat. Cell Tissue Res 169 :467-484 Neuhuber WL (1987) Sensory vagal innervation of the rat esophagus and cardia: a light and electron microscopic anterograde tracing study. J Auton Nerv Syst 20:243-255 Neuhuber WL (1989) Vagal afferent fibers almost exclusively innervate islets in the rat pancreas as demonstrated by anterograde tracing. J Auton Nerv Syst 29:13-18 Niijima A (1977) Afferent discharges from venous pressoreceptors in liver. Am J Physiol 232 (Cell Physiol 1):C76-C81 Niijima A (1980) Studies on the nervous regulatory mechanisms of blood sugar levels. Pharmacol Biochem Behav 3 [SuppI 1]:139-143 Niijima A (1982) Glucose-sensitive afferent nerve fibers in the hepatic branch of the vagus nerve in the guinea pig. J Physiol (Lond) 332:315-323 Niijima A (1983) Electrophysiological study on nervous pathway from splanchnic nerve to vagus nerve in rat. Am J Physiol 244:R888-R890 Passo SS, Thornborough JR, Rothballer AB (1973) Hepatic receptors in control of sodium excretion in anesthetized cats. Am J Physiol 224:373-375 Powley TL, Berthoud HR (1991) A fluorescent labeling strategy for staining the enteric nervous system. J Neurosci Methods 36:9-15 PrechtI JC, Powley TL (1990) The fiber composition of the abdominal vagus of the rat. Anat Embryol 181 : 101-115 Rogers RC, Hermann GE (1983) Central connections of the hepatic branch of the vagus nerve : a horseradish peroxidase histochemical study. J Auton Nerv Syst 7:165-174 Rogers RC, Novin D (1980) The neurobiology of hepatic osmoregulation and its implications for studies of ingestive behavior. Brain Res Ball 5 [Suppl 4] : 18%194 Rogers RC, Karilas PJ, Herman GE (1984) Projection of the hepatic branch of the splanchnic nerve to the brainstem of the rat. J Auton Nerv Syst 11:223-225 Russek M (1963) An hypothesis on the participation of hepatic glucoreceptors in the control of food intake. Nature 197:79-80 Sawchenko PE, Friedmann MI (1979) Sensory functions of the liver- a review. Am J Physio1236 (Regulatory integrative Comp Physiol 5): R5-R20 Shimazu T (1971) Regulation of glycogen metabolism in liver by the autonomic nervous system V. Activation of glycogen synthetase by vagal stimulation. Biochim Biophys Acta 252:28-38 Shimazu T (1979) Nervous control of peripheral metabolism. Acta Physiol Polonica [Suppl 18]:1-18 Skaaring P, Bierring F (1977) Further evidence for the existence of intralobular nerves in the rat liver. Cell Tissue Res 177:287290 Stavney LS, Kato T, Griffith CA, Nyhus LM, Harkins HN (1963) A physiologic study of motility changes following selective gastric vagotomy. J Surg Res 3 : 390-394 Sutherland SD (1965) The intrinsic innervation of the liver. Rev Internat Hepatol 15 : 569-578 Tanaka K, Inoue S, Nagase H, Takamura Y, Niijima A (1990) Amino acid sensors sensitive to alanine and leucine exist in the hepato-portal system in the rat. J Auton Nerv Syst 31:41~46 Terada T, Nakamura Y (1989) Innervation of intrahepatic bile ducts and peribiliary glands in normal human livers, extrahepatic biliary obstruction and hepatolithiasis. An immunohistochemical study. J Hepatol 9:141 148 Tordoff MG, Rawson N, Friedman MI (1991) 2,5-Anhydro-Dmannitol acts in liver to initiate feeding. Am J Physiol 261 (Regulatory integrative Comp Physiol 30):R283-R288 Tsai TL (1958) A histological study of sensory nerves in the liver. Acta Neurovegetativa 17:354-385