Distribution and morphology of sensory and autonomic fibres in the subendocardial plexus of the rat heart

Abstract Cardiac reflexes originating from sensory receptors in the heart ensure blood supply to vital tissues and organs in the face of constantly changing demands. Atrial volume receptors are mechanically sensitive vagal afferents which relay to the medulla and hypothalamus, affecting vasopressin release and renal sympathetic activity. To date, two anatomically distinct sensory endings have been identified which may subserve cardiac mechanosensation: end‐nets and flower‐spray endings. To map the distribution of atrial receptors in the subendocardial space, we have double‐labelled rat right atrial whole mounts for neurofilament heavy chain (NFH) and synaptic vesicle protein 2 (SV2) and generated high‐resolution maps of the rat subendocardial neural plexus at the cavo‐atrial region. In order to elucidate the nature of these fibres, double labelling with synaptophysin (SYN) and either NFH, calcitonin gene‐related peptide (CGRP), choline acetyltransferase (ChAT) or tyrosine hydroxylase (TH) was performed. The findings show that subendocardial nerve nets are denser at the superior cavo‐atrial junction than the mid‐atrial region. Adluminal plexuses had the finest diameters and stained positively for synaptic vesicles (SV2 and SYN), CGRP and TH. These plexuses may represent sympathetic post‐ganglionic fibres and/or sensory afferents. The latter are candidate substrates for type B volume receptors which are excited by stretch during atrial filling. Deeper nerve fibres appeared coarser and may be cholinergic (positive staining for ChAT). Flower‐spray endings were never observed using immunohistochemistry but were delineated clearly with the intravital stain methylene blue. We suggest that differing nerve fibre structures form the basis by which atrial deformation and hence atrial filling is reflected to the brain.


| INTRODUC TI ON
Important cardiac reflexes help to ensure that the blood supply to all tissues and organs is maintained in the face of constantly changing demands. Venous return to the right atrium influences the measured variable of the atrial volume reflex. Returning blood stimulates cardiac mechanoreceptors located at the cavo-atrial junction which in turn signal to cardiac control centres in the brain and moderate sympathetic drive to the heart and kidney. Atrial filling and contraction produce structural deformation in three dimensions, and therefore, cardiac mechanoreceptors must accurately reflect the deformation. Atrial receptors form a mechanically sensitive family of afferent parasympathetic neurons whose sensory endings are primarily concentrated within the endocardium at the cavo-atrial junctions of the heart (Woollard, 1926;Coleridge et al., 1957;Holmes, 1957a;Tranum-Jensen, 1975).
However, the morphological nature of the mechanoreceptor/sensory nerve ending type underpinning this important function has not been unequivocally determined.
Despite detailed histological studies of sensory nerve endings in the mammalian heart, there still remains some disagreement concerning the types and distribution of these structures.
Silver impregnation (Nonidez, 1937) and methylene blue staining (Holmes, 1957a;1957b;Miller and Kasahara, 1964) indicate that the atrial endocardium contains cardiac mechanoreceptors (see Linden and Kappagoda, 1982 for review). Two anatomically distinct nerve endings have been described as candidate sensory specialisations for atrial receptors: end-nets and complex unencapsulated endings (diffuse and compact) or flower-sprays. End-nets are described as large arrays of anastomosed dendrites and flower-sprays as arborised complex unencapsulated endings. The morphology of end-nets and flower-sprays has been well described with methylene blue labelling in dogs, cats, monkeys, rabbits and guinea pigs (Coleridge et al., 1957;Miller and Kasahara, 1964;Linden and Kappagoda, 1982).
A third type, complex encapsulated ending has also been described but is thought to be rare in mammals (see Hainsworth, 1979 for review). Cheng and colleagues, in the rat, labelled vagal afferents with DiI (1,1′-dioleyl-3,3,3′3′-tetramethylindocarbocyanine methanesulfonate) and identified "flower-spray" or complex unencapsulated endings and "end-net" terminals (Cheng et al., 1997). It has been suggested that the morphological differences that appear to exist between unencapsulated endings and end-nets are quantitative rather than qualitative and that end-nets should in fact be considered as extensive unencapsulated endings (Hainsworth et al., 1979).
For complex unencapsulated endings, it is reported the atria contain more of these than the endocardium of any other part of the heart and only the atria contain these endings and end-nets.
The end-net structure has been described as distinct from that of the complex unencapsulated end organs (Coleridge et al., 1957;Holmes, 1957a;1957b;1958) and suggests this structure is present in both atrial and ventricular endocardium but more abundant in the atria (Miller and Kasahara, 1964). Notwithstanding the discrepancies about receptor identity, it is apparent that end-nets and complex unencapsulated end-organs form challenging three-dimensional structures.
The observation that end-net distribution is more extensive throughout the atrial endocardium than the complex unencapsulated nerve endings (Coleridge et al., 1957;Holmes, 1957b) may be important in terms of signalling atrial function to the brain. One suggestion is that receptor discharge is related to the force exerted on the atrial wall due to ventricular contraction (Langrehr, 1960).
Another theory argues that the activation of the receptors is a function of atrial wall tension itself and is best described by the Law of Laplace (Kidd et al. 1966(Kidd et al. , 1978. Nevertheless, it is apparent that during atrial contraction parts of the atria undergo extensive mechanical distortion. The question arises: what is the optimal type of network to convey this information faithfully to the brain? Therefore, to understand the anatomical basis of receptor morphology and density we mapped the distribution of atrial receptors in the subendocardial space of the rat heart. Immunohistochemistry was carried out on right atrial whole mounts for neurotransmitters/proteins/markers to enable discrimination between sensory and motor (autonomic) axons to reveal putative cardiac receptor morphology.
Early findings relating to NFH/SV2 double labelling have been presented in preliminary form at the Physiological Society annual conference -Proc Physiol Soc 43 (2019) C102.

| Atrial tissue preparation
Atrial tissue was prepared using different methodologies at two institutions (Dublin and Durham). For method one (Dublin), Wistar rats (n = 15, 140-400 g) were euthanised by cervical dislocation whilst under inhalational anaesthesia (5% isoflurane). Thereafter, a median sternotomy was performed, heparinised saline was injected into the cavity of the right ventricle, and exposed mediastinal tissue was briefly perfused with oxygenated Tyrode's solution.
The heart was removed and placed in oxygenated Tyrode's solution (20°C). Coronary circulation flushing with Tyrode's solution through the ascending aorta removed blood. The right atrium and great veins were isolated, and the right atrium was oriented such that the lumen of the cavity was exposed. An incision was made along the long axis of the inferior and right superior vena cava. The tissue was pinned flat to expose the endocardial surface.

| Immunohistochemistry
Whole mounts obtained by method one were blocked with 1% (w/v) bovine serum albumin (BSA; Sigma-Aldrich, A2153), 5% (v/v) normal goat serum (Sigma-Aldrich, G9023) and 1X phosphate-buffered saline (PBS, Gibco). Primary antibody diluent was prepared in blocking solution, and subsequently, incubation was performed overnight in a humidity chamber at 4°C. The primary antibodies were anti-SV2 and anti-NFH (see Table 1). Following For details of all primary and secondary antibodies, see Table 1.
After washing (×3 in PB), secondary antibodies were applied for two hours at room temperature in the dark. Finally, whole mounts were washed as before and spread onto charged slides (Superfrost Plus; Fisher, 10149870). After air drying overnight, whole mounts were dehydrated through a series of alcohols (2 min in 50%, 70%, 95% and 100%), cleared in xylene (4 min), mounted in DPX and cover slipped. All of the described secondary antibodies were fluorescent conjugates. These protocols are presented as flow diagrams in the Appendix S1 section. Different vesicular markers (SV2 or SYN) were used at each institution because each research team had prior experience with vesicle labelling (SV2 at UCD and SYN at Durham University), and both markers were expected to produce similar labelling.

| Stereological analysis of subendocardial neural plexus density
Subendocardial neural plexus density measurement was performed using atrial whole mounts prepared by method one. A digital stereological probe was constructed and consisted of six  Table 1 details primary and secondary antibody dilutions and suppliers. In terms of specificity, these are all commercial antibodies subject to routine quality assurance. Where positive results were obtained, the pattern of reactivity was characteristic of that particular antibody with distinct cell populations consistently labelled by that antibody on repeat assays. There was an absence of labelling with secondary antibodies alone.

| Methylene blue labelling and imaging
Isolated whole mounts of Wistar rat right atrium (n = 11) were prepared as per method one, that is pinned flat (

| Neurofilament density mapping in right atrial whole mount
For the atrial whole mounts, profound axonal neurofilament labelling was noted in all samples. The right atrium was observed to be richly innervated with a neural plexus present just below the level of the endocardium. This subendocardial neural plexus was observed to be most dense at the superior cavo-atrial junction (Figure 1). Ten measurements of L A were required to ensure a coefficient of error <5% (CE 1.5%, n = 10, Figure 1c). Length area density was calculated 10 times for the superior cavo-atrial junction and the mid-atrial region. This was performed in four neurofilament whole mounts, totalling 80 measurements. Mean L A for each region was calculated and compared by a paired Student's ttest. Mean L A was significantly greater at the superior cavo-atrial region than the mid-atrial region (L A 20.8 ± 1.7 vs. 14.6 ± 0.5 mm −1 , p = 0.007, Figure 1d). The coefficient of error for the superior cavo-atrial region was 8.2% and 3.6% for mid-atrial region. Thus, greater biological variation was observed at the superior junction than the mid-atrial region

| Neurofilament and synaptic vesicle protein 2 double labelling
In atrial whole mounts, it was observed that NFH antibodies labelled large fibres whilst SV2 antibodies labelled both large and fine axons. Nets of SV2 labelled endings were coincident with NFH labelled nets particularly at the mid-atrial region and at the superior and inferior cavo-atrial junctions ( Figure 3). Fine SV2 labelled fibres appeared to extend from end-net formations and most of these SV2 labelled fibres did not co-label for NFH.
With manipulation of the fine focus, it was discernible that SV2 labelled fibres were closer to the endocardial surface than neurofilament labelled fibres. Across all atrial whole mounts, no flower-spray endings were observed. Also, no obvious motor varicosities were observed. In the teased lumbrical preparations, differential labelling of axons was also present with SV2 clearly labelling the vesicles of terminal boutons whilst NFH labelled axonal projections leading to end plates (Figure 3d,e).
The walls of blood vessels in the endocardium were densely innervated with SYN-positive endings (Figures 5-7). On the adluminal surface (superficial), there was a distinctive loose network of single SYNlabelled fibres, most apparent at the caval entrances to the atrium.
Synaptophysin immunoreactivity was also evident on distinctive, "wavy" single fibres running parallel to each other and perpendicular to the direction of blood flow. Beneath the adluminal network, fibres were in bundles of different sizes as well as occurring singly ( Figure 6).

| Calcitonin gene-related peptide and neurofilament labelling
There was less abundant labelling throughout the entire wholemount with either CGRP or NFH antibodies. Neurofilament heavy

| Tyrosine hydroxylase labelling
Tyrosine hydroxylase co-labelled SYN reactive axon bundles and single fibres in all locations, including those on blood vessels ( Figure 6).
The single fibres were fine and varicose, and this was especially apparent both within the IVC, where they ran in parallel lines as described above for the SYN reactive fibres; and in the loose, open network evident within the superficial adluminal endocardial layer.
However, the TH and SYN labelling did not precisely co-localise.
Tyrosine hydroxylase positive structures were observed in amongst SYN-positive structures within axon bundles, usually being fewer in number (Figure 6). Even on single fibres, as in the loose superficial endocardial network, the labelling appeared compartmentalised ( Figure 6).

| Choline acetyltransferase labelling
Choline acetyltransferase labelling was far less prevalent than SYN/TH labelling. It was not present in the superficial endocardial layer but was present on large bundles of axons below

| Innervation of the inferior cavo-atrial junction
Over the inferior IVC there were distinctive, "wavy", single fibres running in parallel with one another. These fibres coursed

| Methylene blue labelling
Twenty to thirty discrete areas of intense methylene blue uptake (consistent with flower-spray endings) were observed in each atrial whole  Figure 9a). End-net formations were observed across the entire whole mount and appeared most dense at the superior cavo-atrial junction. Flower-spray endings were difficult to image clearly which could reflect that they are deeper than the end-net. One flower-spray ending was suspended in a leaflet of the Eustachian valve and thus was more easily imaged than other endings. A contrast-enhanced view of this ending revealed that it had a multilobulated structure (Figure 9b).

| D ISCUSS I ON
This study has examined the architecture of the cardiac mechanoreceptors present in endocardium of the right atria of the rat. All the labelling techniques adopted consistently revealed end-net structures. The structures labelled by SV2 antibodies were consistent with those elicited by SYN. These structures were found within the subendocardium around the entrances of the great veins to the atria.
We propose that some end-nets are sensory fibres and argue that because deformation of the atrial wall during atrial filling and contraction occurs three dimensionally, these structures can accurately reflect this to the brain.

| Complex unencapsulated nerve endings/ flower-sprays
The flower-spray atrial mechanoreceptor type was only revealed following methylene blue staining. Transillumination was adopted to obtain clear images of these endings. The protocol consistently labelled 20-30 small structures primarily located between the hourglass and pectinate muscles. The structures identified by methylene blue uptake are putatively identified as flower-sprays because they are irregularly shaped and approximately 100 μm in diameter.
This size is consistent with the flower-spray endings reported in rodent anterograde tracing experiments (Cheng et al., 1997) and with those described (50-350 μm) using methylene blue labelling in other species such as the dog, cat, monkey, lamb, rabbit and guinea pig (Coleridge et al., 1957;Miller and Kasahara, 1964;Linden and Kappagoda, 1982). It was difficult to image an afferent axon emerging from these endings; however, it is reasonable to assume that our imaging may be unable to capture these as they are reportedly 4-6 μm in diameter (Linden and Kappagoda, 1982).
The combined number of flower-spray endings found in both the left and right atria has been studied in dogs, cats, monkeys and lambs and varies from 150 to 300 (Holmes, 1957a;Miller and Kasahara, 1964) with approximately one third being present in the right atrium (Linden and Kappagoda, 1982). Our observations pertaining to the number of flower-spray endings in the right atrium of the rat are in keeping with these estimates given its smaller size. It is possible that the number we report is an underestimation related to the methodological approach adopted.
Nearly all descriptions since Berkley (1895) and Smirnow (1895) using heavy metal silver and methylene blue staining report extensive complex unencapsulated terminals and end-net formations in the atrial endocardium. Methylene blue staining and DiI appear to consistently localise complex end-organs and end-net formations (Linden and Kappagoda, 1982;Cheng et al., 1997). Whilst the antibody markers SYN, CGRP, SV2 and NFH failed to localise flower-spray endings, they consistently revealed the end-net formations. Therefore, the ability to detect these structures may depend on staining/methodological regime employed. However, manipulation of antibody concentration/ incubation duration, alteration of permeabilisation agent conditions and adjusting tissue clearance techniques did not label flower-sprays.
It is possible that flower-spray endings are located deeper than the end-net and if this is the case then perhaps despite the adjustments the antibodies did not penetrate the tissue sufficiently. Irrespective of F I G U R E 8 Distinctive "wavy" single fibres at the inferior cavo-atrial junction. Shown are endocardial views of tissues from different animals. Immunohistochemical labelling revealed distinctive "wavy" single fibres running parallel to each other and perpendicular to the direction of blood flow. These fibres (white arrows) were observed for CGRP (a), TH (b), SV2 and SYN (not shown). A cell soma is present (yellow arrow) and could be that of a SIF cell. CGRP, calcitonin gene-related peptide. SIF, small intensely fluorescent; SV2, synaptic vesicle protein 2; SYN, synaptophysin; TH, tyrosine hydroxylase A B whether flower-spray endings are deeper than the end-net, it should also be considered that flower-spray endings may not express NFH, SYN, CGRP or SV2 and thus were not labelled with the utilised immunohistochemical techniques.

| End-net structure
Synaptophysin CGRP, SV2 and NFH, whether as single or in combination consistently revealed a similar innervation pattern of the end-net structure within the endocardium. Synaptophysin is a protein found in synaptic vesicles and as such has been frequently used to identify synapses. However, synaptic-like vesicles have been commonly noted in mechanosensitive nerve terminals for many decades (Katz, 1966) and these vesicles also express synaptophy- Neurofilament is utilised mainly for identifying myelinated endings (Lawson and Waddell, 1991;Molliver et al., 1995;Henry et al., 2012).
Neurofilament labelling revealed a rich neural plexus just below the level of the adluminal endocardium, a pattern consistent with those previously observed with anterograde labelling techniques (Cheng et al., 1997). Interestingly, varicosities were not observed in these deeper structures, which suggest that the observed neurofilament labelling may be sensory in nature.
The end-net mesh was most dense at the superior and inferior cavo-atrial regions. Stereological quantification further confirmed and qualified this by suggesting the subendocardial neural plexus innervation density is greater for the superior cavo-atrial F I G U R E 9 Methylene blue labelled structures in the right atrium of the rat. (a) View of the endocardial surface of the mid-atrial region. In each whole mount approximately 20-30 discrete areas of intense methylene blue labelling were observed. Inset is a high-power view of four such structures which have a diameter of approximately 100 μm consistent with flower-spray endings.
(b) Endocardial surface of the right atrial whole mount. An end-net is visible on the left which straddles two flower-spray endings. Inset is a contrast-enhanced view of a flower-spray ending the morphology of which is consistent with terminals labelled with anterograde tracing techniques. Flower-spray endings were difficult to image but were observed most clearly visualised with transillumination. As (b) was imaged using non-calibrated imaging apparatus, scale bar is not included. IVC, inferior vena cava; R-SVC, right superior vena cava junction compared to the mid-atrial region. This corresponds with the early descriptions of atrial receptors elicited from functional data arising from the cat and dog and other species including humans (Johnston, 1968) indicating a rich distribution around the vein-atrial junctions (Coleridge et al., 1973;Linden and Kappagoda, 1982).
With both SYN and SV2, over some areas of the inner endocardial surface, the net consisted of single fibres. Small beaded SYN/ SV2 labelled fibres and coarser NFH labelled fibres consistently demonstrated the end-net to repeatedly divide and coalesce, forming bundles of axons which presumably project to the vagal nerve and towards the brain. Also, with NFH the consistent hourglass arrangement of axons in the posterior atrial wall was revealed. The inter-caval region of the rabbit right atrium has been studied with cholinesterase/silver staining and similarly an hourglass arrangement of axons has been described (Roberts, 1991). In the rat right atrium, we have observed that an upper hourglass limb (anatomical left) consistently tracks towards the root of the cardiac vagal branch and as such we hypothesise that this hourglass structure could be vagal in origin. Whether the end-net is sensory in nature has not been confirmed. However, one study (Coleridge et al., 1957) identified the location of receptors following physiological stimulation and recording of subsequent action potentials. The location of the receptors was confirmed by methylene blue staining and revealed end-nets as well as complex unencapsulated endings. This suggest the potential for the end-net being a sensory structure.

| Synaptophysin & calcitonin generelated peptide
Calcitonin gene-related peptide, commonly used as a biomarker for chemosensitive afferents and peptidergic pain-sensing C-fibres (Kopp et al., 2001;Russell et al. 2014), was far less prevalent. We found CGRP expressed in fine varicose endings which appeared singly or within larger bundles of SYN-labelled fibres. It was also most frequently in the endocardial layer nearest the myocardium rather than the adluminal surface, and this is in agreement with our previous findings (Shenton and Pyner, 2014). Although CGRP is often considered a marker of unmyelinated nerves (Kakudo et al., 1988;Alvarez et al., 1991;Bickel et al., 1999;Ishikawa et al., 2005), the distinction is not always categoric. For example, whilst investigating the rat vagal nerves, Kakudo et al. (1988) found CGRP immunoreactivity in 5% of myelinated axons, whereas 50% of unmyelinated axons expressed CGRP. Although fewer in number, some of the CGRP positive myelinated fibres found in the vagus may have a role in stretch sensitivity and control of ingestive behaviour of the oesophagus (Andrew, 1956;Rodrigo et al., 1985).

| Neurofilament & synaptic vesicle protein 2
Synaptic vesicle protein 2 was used to try and identify vesicle-rich mechanotransductive regions within the right atrium. The fourth lumbrical muscle served as a positive control and no annulospiral endings were labelled. However, axons in motor end plates were differentially labelled with NFH whilst SV2 labelled discrete vesicles of the terminal bouton. We attribute the lack of annulospiral labelling to inadequate disruption and permeabilisation of the spindle capsule during the tissue preparation process. In the atrial whole mounts, extensive co-localisation of NFH and SV2 was noted in larger fibres however fine fibres which extended from end-net formations labelled for SV2 almost exclusively. Synaptic vesicle protein 2 labelled fibres were observed across the entirety of whole mount samples but were noted to be most dense at the cavo-atrial junctions. No flower-spray endings were observed.
The SV2 labelling approach may also have labelled post-ganglionic motor fibres. It should be noted that SV2 labelled fibres were not observed to possess varicosities and were located closer to the endocardial surface than NFH labelled end-nets which is at odds with the supposition that these fibres are motor in nature. To better elucidate areas of vesicle recycling and identify mechanically sensitive atrial receptors, future studies could use vesicular glutamate transporter (vGluT) family which has been implicated in mechanotransduction. Merkel cells have been shown to express vGluT1, vGluT2 and vGluT3 (Nunzi, Pisarek, and Mugnaini, 2004), and recently, vGluT1 has been used to label proprioceptive sensory terminals in the muscle spindle and Golgi tendon organ of the mouse (de Nooij et al., 2015). It has also been described that incubation with glutamate increases discharge rate from the muscle spindle (Banks et al., 2002). With this evidence it is reasonable to investigate the role of the vGluT family in mechanically sensitive atrial receptors.

| Distinguishing between sensory and motor fibres
Whilst SYN, CGRP, NFH and SV2 have labelled fibres, it cannot be stated with certainty that the observed neuronal structures are truly sensory even if they resemble the end-net as described in the literature. The presence of motor varicosities (or lack thereof) is also an inadequate means of differentiation. Therefore, the additional targets of ChAT and TH in combination with SYN were examined. Choline acetyltransferase is expressed in acetylcholinereleasing motor neurons (Kou et al., 1995) and anti-ChAT antibodies would label parasympathetic efferent neurons present in the atria.
Tyrosine hydroxylase is expressed in sympathetic motor fibres (Cheah and Geffen, 1973) and anti-TH antibodies would label catecholamine-releasing sympathetic motor fibres present in the atria.

| Synaptophysin & tyrosine hydroxylase
Within the fine, superficial end-nets labelled with SYN and TH, it is probable that the same fibres were labelled and that the internal concentration of SYN and TH is variable and compartmentalised. In large bundles of fibres, TH and SYN also appeared compartmentalised and may have been on separate fibres however higher magnification confocal imaging would be required to precisely discriminate between fibres. Electron microscopy from mini-pigs indicates efferent sympathetic nerve fibres course with the afferent fibre to end near the receptor ending (Tranum-Jensen, 1975, 1979. In zebra fish larvae, TH fibres described as efferents have been shown to play a role in modulating sensory nerve endings (Haehnel-Taguchi et al., 2018). The exclusive TH-positive fibre terminals we describe projecting in the endocardium may be sympathetic efferents and may release monoamines locally for yet unknown purposes. Tyrosine hydroxylase is a marker of catecholaminergic neurones, where the neurotransmitter is adrenaline, noradrenaline or dopamine and is commonly used to identify sympathetic efferents (Foss et al., 2015) although there is evidence for TH expression on mechanosensitive afferents (Vyas et al., 2017;Wu et al., 2018).
The finding that the majority of SYN-labelled endings also expressed TH was a surprise since TH is usually considered as a marker of sympathetic efferents. Tyrosine hydroxylase/SYN-labelled endings have also been reported in human right atrial material (Marron et al., 1995;Bohlender et al., 2018). With SYN or TH as co-markers, Bohlender described abundant angiotensin fibre innervation of human right atrium. The TH and SYN fibres were varicose with a comparable anatomical distribution. The TH fibres were mostly SYN-positive.
Bohlender assumed these fibres to be efferent, but as alluded to above this may not be the case. In the right cavo-atrial junction we have shown that the SYN-labelled terminals in the endocardium (Shenton and Pyner, 2014) have a morphology characteristic of sensory endings (Drummond et al., 1998;Maeda et al., 1999). Tyrosine hydroxylase has been described in sensory afferents, both chemosensory (Katz et al., 1987;Czyzyk-Krzeska et al., 1991;Finley et al., 1992) and, more recently, mechanosensory (Vyas et al., 2017;Wu et al., 2018). Tyrosine hydroxylase fibres could be dopaminergic and dopaminergic mechanosensory endings have been reported in C. elegans (Han et al., 2017) and these endings also expressed the mechanosensory TRP-4 channel.
Furthermore, the nematode TRP-4 channel is very similar to mammalian transient receptor potential vanilloid 4 (TRPV4), which we have found previously in cavo-atrial endocardium of the rat heart coinciding with SYN labelling (Shenton and Pyner, 2014).

| Synaptophysin & choline acetyltransferase
A minority of SYN-labelled nerves were co-labelled with ChAT (a marker of parasympathetic efferents) and these were large bundles of axons below the adluminal level of the endocardium, suggesting the preponderance of SYN immunoreactivity was indeed on afferents.

| Intrinsic innervation of the atria
Our investigations in whole mounts have focussed on the subendocardial nerve plexus, whereas previous wholemount studies, in both rats (Richardson et al., 2003) and mice (Rysevaite et al., 2011;Li et al., 2014) have examined the epicardial surface. Intracardiac ganglia (ICG) are found within the epicardial tissue and in rats these form a ring-like plexus around the entry of the pulmonary veins (Richardson et al., 2003). We imaged the endocardial aspect of the right atrium and hence the ICG were not readily visible through the relatively thick myocardium. Nevertheless, there remains a possibility that we have imaged some components of the intrinsic innervation of the atria, especially with regard to the ChAT labelling. Richardson et al. (2003) found all intracardiac neurons contained immunoreactivity to ChAT and neuropeptide Y (NPY). By contrast, whilst nerve fibres around the ganglia occasionally expressed TH, the neuronal somata themselves were never positive for this marker; although it did strongly label small intensely fluorescent cells (Huber, 2006) associated with the ganglia. Around the ganglia TH labelled nerve fibres did not express SYN, whereas broadly in agreement with our own observations, TH-positive fibres present in the myocardium and blood vessel walls always co-expressed SYN. This could be suggestive of a distinct functional role for these dual labelled endings. Similarly, none of these three studies (Richardson et al., 2003;Rysevaite et al., 2011;Li et al., 2014) found CGRP in the cell bodies of the ICG.

| Functional considerations
The exact nature of the stimulus which activates cardiac mechanoreceptors remains unknown. It is likely to be a combination of chemical and mechanical stimuli. In rodent muscle spindles stretch initiates the opening of stretch sensitive channels and the sensitivity of the mechanosensitive endings is modified by glutamate released from the endings themselves (Bewick et al., 2005;Bewick & Banks, 2015). Other mechanoreceptors are likely to share similar features, although the detail may vary. Whilst this may describe the transduction of a stimulus, the precise nature of these stimuli remains elusive. However, three-dimensional deformation during atrial filling and contraction changes the curvature and thus tension of the wall and may offer a mechanism to initiate the transduction process. Our study revealed a dense network of putative sensory fibres arranged in an end-net mesh in the region where the caval veins enter the atria. This region undergoes large changes in dimensions and tension and this end-net mesh may be the mechanism by which these dynamic variables are communicated to the brain.
The presence of SYN/CGRP/TH axons at the adluminal side of the endothelium may indicate a chemosensing role for these fibres. It stands to reason that because of their close proximity to the blood-filled atrial lumen, they could be receptive to gases and molecules that diffuse into the paracellular space. Some authors report most atrial afferent fibres to be chemosensitive (Waldmann et al., 2006). Fine endocardial terminals protruding between endothelial cells into the atrial lumen, which are likely to be chemosensitive, have been reported in human cadaver endothelium (Bohlender et al., 2018). Furthermore, these were either angiotensin II/SYN or TH/SYN positive.

| SUMMARY
The subendocardial nerve nets were found to be denser at the superior cavo-atrial junction than the mid-atrial region. The rat heart has a bicuspid Eustachian valve which is richly invested with nerve fibres. Adluminal networks had the finest diameters and stained positively for synaptic vesicles (SV2 and SYN), CGRP and TH. These plexuses may represent sympathetic post-ganglionic fibres and/or sensory afferents. The latter are candidate substrates for volume receptors which are depolarised by myocyte contraction in atrial systole or excited by stretch during atrial filling. Deeper nerve fibres appeared coarser and may be cholinergic (positive staining for ChAT). Flower-spray endings were never observed using immunohistochemistry but were delineated clearly with the intravital stain methylene blue. This suggests that either flower-spray endings do not express NFH, SYN, SV2, TH, CGRP or ChAT or that antibody access to these endings was restricted. Thus, future directions need to address the anatomic relationship of the nerve fibres to the cardiac muscle with the discharge characteristics of atrial receptors reflecting differing receptor type.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
The authors take responsibility for the integrity of the data and accuracy. All authors contributed to conceptualisation and experimental design, data analysis, manuscript drafting and approval of the submitted manuscript. SYN, NFH, TH, ChAT and CGRP labelling, imaging and analysis was performed by Dr. Fiona Shenton and Dr. Susan Pyner. NFH, SV2 and methylene blue imaging and analysis was performed by Dr. Thomas Campbell and Prof. James Jones.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.