The human phrenic nerve serves as a morphological conduit for autonomic nerves and innervates the caval body of the diaphragm

Communicating fibres between the phrenic nerve and sympathetic nervous system may exist, but have not been characterized histologically and immunohistochemically, even though increased sympathetic activity due to phrenic nerve stimulation for central sleep apnoea may entail morbidity and mortality. We, therefore, conducted a histological study of the phrenic nerve to establish the presence of catecholaminergic fibres throughout their course. The entire phrenic nerves of 35 formalin-fixed human cadavers were analysed morphometrically and immunohistochemically. Furthermore, the right abdominal phrenic nerve was serially sectioned and reconstructed. The phrenic nerve contained 3 ± 2 fascicles in the neck that merged to form a single fascicle in the thorax and split again into 3 ± 3 fascicles above the diaphragm. All phrenic nerves contained catecholaminergic fibres, which were distributed homogenously or present as distinct areas within a fascicle or as separate fascicles. The phrenicoabdominal branch of the right phrenic nerve is a branch of the celiac plexus and, therefore, better termed the “phrenic branch of the celiac plexus”. The wall of the inferior caval vein in the diaphragm contained longitudinal strands of myocardium and atrial natriuretic peptide-positive paraganglia (“caval bodies”) that where innervated by the right phrenic nerve.

SCIENTIFIC REPORTS | (2018) 8:11697 | DOI: 10.1038/s41598-018-30145-x fibres are present in such communications. We, therefore, conducted a detailed histological study to establish the presence of catecholaminergic fibres in the phrenic nerve throughout its course from its cervical roots to its connection with the celiac plexus. Since phrenic nerve stimulation can be performed on both sides and at different levels of the nerve, we also looked at morphological differences along the course of the phrenic nerve and compared left and right phrenic nerves at different levels.

Methods
Nerve tissue was harvested from thirty-five (16 female, 19 male) formalin-fixed cadavers between 58 and 101 (x = 84 ± 11) years of age from the body donation program of the Department of Anatomy and Embryology, Maastricht University. The tissue donors gave their informed and written consent to the donation of their body for teaching and research purposes as regulated by the Dutch law for the use of human remains for scientific research and education (Wet op de Lijkbezorging, 1991). Accordingly, a handwritten and signed codicil from the donor posed when still alive and well, is kept at the Department of Anatomy and Embryology Faculty of Health, Medicine and Life Sciences, Maastricht University, Maastricht, The Netherlands. The bodies were preserved by intra-arterial infusion with 10 L fixative (composition (v/v): 96% ethanol (21%), glycerin (21%), 36% formaldehyde (2%), water (56%), and 2.4 g/L thymol), followed by 4 weeks of fixation in 96% ethanol (20%), 36% formaldehyde (2%) and water (78%). Samples were only taken from bodies without signs of previous surgical interventions on neck, thorax or abdomen.
Phrenic nerve sampling. The cervical and thoracic portions of the phrenic nerve with surrounding connective tissue and accompanying pericardiophrenic vessels were collected and subdivided into levels A to I (−J on the left side), as depicted in Fig. 1. To investigate whether an abdominal branch of the phrenic nerve extends to the celiac plexus (inset Fig. 1), the abdominal phrenic nerves, including the part that traversed the diaphragm, were collected and embedded 'en bloc' for further histological processing. Furthermore, periarterial tissue  accompanying left and right inferior phrenic arteries was collected to establish whether or not the phrenic nerve contributes to the nerve plexus accompanying the inferior phrenic arteries.
Histological processing. All cervical and thoracic nerve samples were cut transversally into two parts.
The upper parts were post-fixed overnight in 1% osmium tetroxide (OsO 4 )/phosphate-buffered saline (PBS) and then embedded in paraffin. The lower parts, the abdominal samples, and the 'en bloc' samples of diaphragm and abdominal phrenic nerve were subjected to standard paraffin embedding. Five micrometre-thick sections were prepared of all samples with a Leica 2245 microtome. Mounted sections were used for haematoxylin and eosin (HE) staining or immunohistochemistry. To determine the intradiaphragmatic course of the phrenic nerves, the 'en bloc' samples were sectioned every 250 µm (4 sections per mm). For reconstruction, one abdominal phrenic nerve was sectioned completely into 5 µm consecutive transverse sections resulting in 7,500 slides.
Complete images of selected large slides were digitized with an Olympus BX61 scanning microscope and the DOTSLIDE program (Olympus, Zoeterwoude, The Netherlands). AMIRA software (version 6.0; base package; FEI Visualization Sciences Group Europe, Mérignac Cédex, France) was used to generate 3D reconstructions after image loading, alignment and segmentation 19 . Morphometric analysis. Slides were photographed with a Leica (type DMRD) photomicroscope 20 . Surface areas of myelinated axons, specifically stained axons, and the entire surface of the nerve within the perineurium, including its supporting tissue, were measured with Leica Qwin v.3.5.1 analysis software at 10x magnification. Furthermore, the number of nerve fascicles was counted. Two persons independently determined whether staining exceeded background levels. The average of these values was used as threshold. Statistical analysis was performed with Graphpad Prism v6.0 software. Data were tested for normality with the Shapiro-Wilk normality test. Comparisons were made by Student's t-tests and one-way ANOVA followed by Bonferroni post-hoc tests. Data are presented as means ± SD. Individually presented data are displayed after Savitzky-Golay filtering. P-values < 0.05 were considered as statistically significant.

Results
Cervical and thoracic findings. Surface area measurements. Nineteen cadavers were analysed. The mean surface area of the right and left nerves excluding epi-and perineurium was 0.35 ± 0.02 and 0.29 ± 0.01 mm 2 , yielding diameters of 1.2 and 1.1 mm, respectively. No significant differences in surface area along the proximo-distal course of the nerve were observed for either the right or left nerve (P ≥ 0.71). Furthermore, no significant differences in surface area between corresponding levels of the left and right nerves were detected (all P ≥ 0.09).
Myelination. Myelinated neurons occupied 26 ± 1% and 29 ± 1% of the total surface areas of the right and left nerves, respectively. No significant differences in myelination were observed along the proximo-distal course of the right and left nerve (P ≥ 0.10), or between corresponding levels of the right and left nerves (all P ≥ 0.35).
Fascicles. In the cervical area, the phrenic nerves contained 3 ± 2 fascicles (as defined by having a distinct perineurium; range: 1-10 fascicles; Figs 2A,B and 3A,B) that merged to form the single fascicle usually seen in the thorax (range: 1-4; Figs 2C and 3A). Just cranial to the diaphragm the number of fascicles increased to 3 ± 3 (range 1-9; Fig. 3A,B). No significant differences were found between left and right phrenic nerves (all P ≥ 0.08).
Expression of Tyrosine Hydroxylase. Tyrosine hydroxylase (TH)-positive fibres were found in all phrenic nerves studied (Fig. 2D), occupied 1.6 ± 0.5 and 2.0 ± 0.6% of the total nerve surface area of the right and left nerves, (D) TH-positive fibres (yellow) presented as distinct fascicles or as distinct areas within the phrenic perineurium or epineurium, respectively (most often seen between levels A and E and on the right side only at level I) as opposed to a homogenous distribution of TH-positive fibres (most often seen between levels F and J except for level I on the right side; for levels confer respectively, and were present in the nonmyelinated areas (Fig. 2C). The size of the TH-positive areas was similar along the proximo-distal course of the nerve for both the right (P = 0.63) and left nerves (P ≥ 0.63), and between corresponding levels of the left and right nerves (all P ≥ 0.09).
TH-positive fibres and fascicles. TH-positive fibres were either distributed homogeneously or presented as distinct areas or fascicles within the phrenic perineurium or epineurium, respectively (Fig. 3D). The distinct areas were seen most frequently in the left and right cervical and the right thoracic region just above the diaphragm (Fig. 3C).
Findings near the diaphragm. Intradiaphragmatic course of the phrenic nerves. The course of the right phrenic nerve along the foramen of the inferior caval vein was analysed in 4 cadavers on transverse sections of ~5 cm diameter (Fig. 4A,B). The branching pattern of the phrenic nerve was irregular, with TH-positive branches accompanying the motor nerve. The intradiaphragmatic course of the left phrenic nerve was also analysed in 4  (Fig. 4D-F) and was similar to that of the right phrenic nerve, except that TH-positive fibres were completely absent (Fig. 4E,F).

Caval bodies.
In the wall of the inferior caval vein of the 4 cadavers studied, we identified a total of eight tangled structures (Fig. 4C, lower left inset and 5) that contained an extensive venous plexus surrounding large cell bodies with granular inclusions (Fig. 5 inset) and a network of nerve fibres (Figs 5 and 6A,B) originating from the right phrenic nerve on the opposite side (Fig. 4C). Although the large cells morphologically resembled neural cell bodies, only a fraction stained positive for PGP9.5 (Fig. 6C) or TH (Fig. 6D). Neither VAChT-or ChAT-positive cells were observed (Fig. 6F). Many of the large cells bordering the veins (Fig. 5 inset) contained atrial natriuretic peptide (NPPA)-positive granules (Fig. 6E). The phenotypic characteristics of these structures resemble those of paraganglia 21 . For this reason, we have named them "caval bodies".
Myocardial muscle sleeves. In 3 of the 4 cadavers, the wall of the inferior caval vein inside the diaphragm contained longitudinal strands of myocardium that stained positive for α-smooth muscle actin and SERCA2a (supplemental Fig. 1).
Findings within the abdomen. Abdominal course and characteristics of the right phrenic nerve. The abdominal course of the right phrenic nerve was studied in 5 cadavers. Textbooks indicate that the nerve extends all the way to the phrenic ganglia in the periphery of the celiac plexus 7 . We tested this assumption by sectioning the abdominal portion of the nerve of one cadaver from the diaphragm to the caudal-most of the phrenic ganglia, (7,500 sections of 5 µm each; Fig. 7A-H). A 3D reconstruction of these sections (Fig. 7I) revealed that the phrenic nerve (Fig. 7A,I (brown)) split into two main, myelinated branches that innervated the diaphragm (Fig. 7B). A TH-positive autonomic fascicle presented as a third, separate branch distal to this level and extended all the way to the celiac ganglion (Fig. 7C-H). We, therefore, named this branch the phrenic branch of the celiac plexus. Of note, the diameter of the branch excluding the ganglia increased towards the celiac ganglion (Fig. 7J) and was interspersed with neuronal cell bodies (Fig. 7E,F,H) that stained positive for PGP9.5, TH and dopamine β-hydroxylase DBH (supplemental Fig. 2).
Left and right peri-arterial nerve plexuses of the inferior phrenic artery. Nerve plexuses that accompany the left and right inferior phrenic arteries were studied in 5 cadavers and found to consist of small, purely TH-positive branches in the tunica adventitia. These plexuses, therefore, do not resemble the phrenic branch of the celiac plexus (supplemental Fig. 3).

Discussion
In this study, we have characterized both phrenic nerves in man with respect to fascicles, fibre composition and myelinisation. We demonstrate that the phrenic nerve also serves as a conduit for catecholaminergic fibres and that the phrenicoabdominal branch of the right phrenic nerve is, instead, a branch of the celiac plexus. The "phrenic branch of the celiac plexus" is, therefore, a more appropriate name for this nerve. In addition, we report

Morphological characterization of the left and right phrenic nerve.
Our data show that the thoracic phrenic nerve generally consists of a single fascicle, which arises from the fusion of several cervical fascicles, and diverges into several fascicles again near the diaphragm. In agreement with an earlier ultrasound study at the cervical level 22 , we found comparable diameters for the left and right human phrenic nerves throughout their course. A small (N = 2) electron-microscopic (EM) study of the phrenic nerve of the rat at the level of the entrance of the inferior caval vein into the right atrium revealed that the right phrenic nerve contained ~30% more axons than its left counterpart 23 .
This study histologically validates the presence and composition of TH positive communicating nerve fibres between the right phrenic nerve and celiac plexus (see further). In the supradiaphragmatic part of the right phrenic nerve, the TH positive fibres are present as distinct fascicles or as distinct areas within the phrenic perineurium or epineurium, respectively. Such distinct TH positive areas were only seen in the right phrenic nerve just above the diaphragm, but also in both phrenic nerves in the cervical region. Therefore, we hypothesize that TH-positive fascicles form communications with nearby nerves or organs in this area too.
Abdominal course and characteristics of the right phrenic nerve. Previous gross-anatomy reports suggest that the phrenic nerve continues on the abdominal side of the diaphragm to the phrenic ganglia 7,9,10 . This branch is classically described as the 'phrenicoabdominal branch of the right phrenic nerve' 34 . Our 3D reconstruction of the right phrenic nerve and analysis of the composition of the left phrenic nerve revealed that the phrenic nerve motor branches do not continue beyond the diaphragm. Instead, the right phrenic nerve continues as a completely catecholaminergic nerve branch that, based on its increasing diameter, arises from the celiac ganglia and is, therefore, more appropriately termed the 'phrenic branch of celiac plexus' . TH-and DBH-positive cell bodies were encountered throughout the phrenic branch of celiac plexus, that is, also outside the two macroscopically visible ganglia that are classically described 9,11 . Based on these findings, we conclude that the phrenic motor nerve innervates the diaphragm, but also serves as a conduit for the peripheral autonomic nervous system. The absence of catecholaminergic fibres in the intradiaphragmatic part of the left phrenic nerve emphasizes the asymmetry of the distribution of the autonomic fibres. We hypothesize that this asymmetry corresponds with the presence of paraganglia in the wall of the (right-sided) inferior caval vein (see next paragraph). Paraganglia at the level of the diaphragm. An unexpected finding in this study was the identification of paraganglia in the wall of the inferior caval vein where it passed the diaphragm. Based on their morphological appearance, the extensive venous network around NPPA-positive cells and the rich innervation by fibres arising from the phrenic nerve, we hypothesize that these structures have a neuroendocrine function like paraganglia elsewhere. Fibres of the phrenic nerve encircling the inferior caval vein have also been described in the foetus 35 . The presence of NPPA-positive granules further suggests that these cells have a role in regulating plasma volume in a similar manner as NPPA-containing cardiomyocytes. By analogy to the vagal B-type atrial receptors 36 , which monitor central venous pressure as stretch of the atrial wall, the phrenic or autonomic nerve endings could act as low-pressure receptors for the central venous pressure. As elsewhere 37 , the NPPA-positive cells could be under efferent catecholaminergic neural control. It would be interesting to investigate whether right phrenic nerve stimulation affects plasma levels of NPPA (fragments) and influences volume homeostasis. Blood pressure management is a key element in the treatment of patients suffering from HF and many other conditions. Further characterization of these neuroendocrine structures is desirable, since neuroendocrine, chemosensory and neuroimmunomodulatory functions exist in other paraganglia like the carotid body 21 .
Myocardial muscle strands. Another unexpected finding was the presence of longitudinal cardiac muscle strands in the wall of the inferior caval vein. A caval sphincter supplied by the right phrenic nerve is a well-known feature of diving mammals 38 . However, this sphincter is usually described as consisting of striated skeletal muscle that is continuous with the diaphragm 38 . The expression of α-smooth muscle actin indicates that the myocardium in these strands is poorly differentiated 39 . Myocardial 'sleeves' with such properties have also been described in pulmonary veins and at the base of the pulmonary trunk, where their presence can establish extranodal pacemaker activity 40 .
Clinical implications. We observed myelinated and non-myelinated nerve fibres in both phrenic nerves without differences between left and right or along the proximo-distal course of the nerves. Such information is important for nerve stimulation, because myelinated nerve fibres have a much lower amplitude-duration threshold upon nerve stimulation than non-myelinated fibres 41,42 . Typical stimulation protocols for (transvenous) phrenic nerve stimulation can vary up to a hundred-fold in intensity (0.1-10 mA), 5-fold in duration (60-300 µs) and 2-fold in frequency   13 . If applied for central sleep apnoea, the stimulation should target the myelinated fibres and should, therefore, be accomplished with the lowest possible amplitude-duration thresholds that result in the intended rhythmic activation of the diaphragm. This is necessary to prevent any undesired stimulation of the nonmyelinated catecholaminergic fibres that are also present within the phrenic nerve. This concern is relevant, because (direct) electric stimulation of the right subclavian ansa did elevate noradrenaline and cAMP concentrations in plasma harvested in the coronary sinus of dogs 43 . Such a catecholaminergic stimulation of the heart may, therefore, further increase the chronic upregulation of sympathetic activity that is already present in patients with central sleep apnoea 44,45 and that is associated with increased mortality these patients [15][16][17] .