Animal experiments were conducted in accordance with United States Federal Regulations as set forth in the Animal Welfare Act, the 2011 Guide for the Care and Use of Laboratory Animals (1), the Public Health Service Policy for the Humane Care and Use of Laboratory Animals, the policies and procedures established by the University of California - Los Angeles (UCLA) Animal Research Committee (ARC) (UCLA ARC protocol #s 2016–033, 2016–084, 2016–085), These studies also conform to the principles and regulations as described in the editorial by Grundy(Grundy, 2015). Animals were provided ad libitum access to food and water, kept under a 12-hr light/dark cycle, and were housed in an Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) accredited facility.

Experiments using human tissue conformed to the standards set by the latest revision of the Declaration of Helsinki, except for registration in a database, and were approved by the UCLA Institutional Review Board (IRB). Tissues were procured consistent with organ donation preferences indicated by the donor and/or after signed consent was obtained from family members in compliance with the bioethical laws of the United States of America, the Uniform Anatomical Gift Act (2006), as well as UCLA policies and procedures for use of human tissue.

Adult C57BL/6J (Jackson Labs) mice (n = 18; 10 male, 8 female, age 10 ± 2 wks) were anesthetized deeply with isoflurane (5%) and exsanguinated. The thoracic cavity, including ribs, heart, and lungs, was removed and placed in ice cold physiologic salt solution (PSS) containing (in mM): 121 NaCl, 5.9 KCl, 1.2 NaH₂PO₄, 1.2 MgCl₂, 25 NaHCO₃, 2 CaCl₂, 8 D-glucose; pH 7.4 maintained by 95% O₂ - 5% CO₂ aeration. The heart was excised from the thorax and observed beneath a stereomicroscope to isolate epicardial ganglia by fine dissection. Connective tissue enveloping isolated ganglia, which were devoid of contracting myocardium, was pinned to the SylGard (Dow Corning) floor of a glass bottom petri dish.

Adult Yucatan minipigs (Premier Biosource; n = 15; 9 castrated male, 6 female; age 6 to 8 mo) were sedated with intramuscular telazol (4–6 mg/kg), intubated, and mechanically ventilated. General anesthesia was maintained with inhaled isoflurane (1.5–2.5%) and intravenous boluses of fentanyl (total: 10–30 μg/kg) during surgical preparation by way of median sternotomy to expose the heart. Major vessels were clamped, the heart was removed, rinsed in PSS, and the dorsal right atrium, at the junction of the superior vena cava (SVC) and the inferior vena cava (IVC), was dissected for subsequent dissection of epicardial ganglia beneath a stereomicroscope.

Human hearts (n = 21; 15 male, 6 female; aged 31 to 72yrs) were procured from deceased organ donors following a formal declaration of brain death. Hearts received were designated for biomedical research after being declined for cardiac transplantation. Deidentified organ donor demographic data is provided in Table 1. The hearts were transported in ice cold crystalloid cardioplegia (University of Wisconsin solution) and were received within six hours after aortic cross clamp. Upon receipt, a portion of the posterior right atrium containing the RAGP (Figure 1C₁) was immediately excised and the epicardial ganglia were isolated under a stereomicroscope by fine dissection. The isolated ganglia were pinned to the SylGard floor of a glass bottom petri dishes using methods identical to those for pig and mouse tissue.

Isolated ganglion preparations were transferred to the stage of an upright compound microscope (Zeiss AxioExaminer) equipped with differential interference contrast optics, an Axiocam camera, and a 5X air and a 40X water-immersion objective. Ganglia were superfused (6–7 ml/min; 5 ml chamber volume) continuously with PSS maintained at 34 ± 2 °C.

Visually identified neurons were impaled with borosilicate glass microelectrodes filled with 2% neurobiotin (Vector Labs, Burlingame, CA) in 2.0 M KCl. Electrode tip resistances measured between 60 and 120 MΩ. Recordings of membrane potential were made with a Multiclamp 700B amplifier (Molecular Devices, Sunnyvale, CA) in current clamp configuration. Data was acquired using a Digidata 1550B (Molecular Devices) analog-to-digital converter at a sampling rate of 5 or 20 kHz. The amplifier and data acquisition were controlled with pCLAMP software (Version 10, Molecular Devices).

Depolarizing and hyperpolarizing currents were injected through the recording electrode to determine responsiveness of the neuronal membrane. Small amplitude hyperpolarizing current steps (500 ms duration; −50 to −100 pA) were used to measure whole cell input resistance and the membrane time constant (tau). Tau was measured by single exponential fit to the membrane potential change in response to the hyperpolarizing current (−50 or −100 pA). The instantaneous and steady-state membrane potential changes in response to large hyperpolarizing steps (500 ms duration; −200 to −500 pA) were also measured. Depolarizing current steps (500 ms duration; 50 to 600 pA)) of increasing intensity were used to assess membrane excitability. Excitability curves were produced by plotting the number of action potentials evoked by depolarizing currents(Tompkins et al., 2018). While the data denotes a continuous variable (spikes/time), they are presented as spike counts for both clarity and simplicity. Neurons were classified as either phasic (firing ≤4 spikes) or tonic (firing >4 spikes continuously) based on the number of action potentials evoked by long-duration, supramaximal depolarizing currents (500 ms). The amplitude and duration of the action potential were measured from spontaneous or nerve evoked action potentials. The amplitude and duration of the after-hyperpolarizing potential was measured from action potentials evoked by brief intracellular current injection (5 ms duration). Cumulatively, the following membrane properties were assessed: resting membrane potential, rheobase, whole cell input resistance, tau, AP 1/2 width, AP duration, AHP amplitude, 2/3^rds AHP duration. Neurons with RMPs less negative than −40 mV or neurons that did not produce action potentials were excluded from the analysis.

Synaptic potentials were evoked within intact ganglion preparations by focal stimulation of interganglionic nerves using extracellular, concentric bipolar electrodes (FHC; Bowdoin, ME). Graded shocks (50–800 uA; 100 μs) were delivered by an AMPI Master 8 and IsoFlex optical Stimulus Isolation Unit. Stimulus recruitment curves were generated by plotting the latency of the evoked potential against the stimulus current. Five to 20 stimuli were delivered at each stimulus intensity, at an interval of three seconds between stimuli. Latency was calculated from the onset of the command pulse to the upstroke of the evoked potential. Trains of stimuli were given at 5, 10, or 20 Hz to assess spike following frequency at the post-synaptic neuron, with a two- to three-minute interval between trains. The amplitudes of spontaneous excitatory post-synaptic potentials (sEPSPs) were quantified using MiniAnalysis software (Synaptosoft, Inc.). The following synaptic properties were assessed: latency, spike following frequency, sEPSP amplitude.

After intracellular recording, ganglia were fixed by overnight immersion in 4% paraformaldehyde at 4 °C. The fixed tissue was rinsed in phosphate buffered saline (PBS) and stored in PBS + 0.02% sodium azide at 4 °C until subsequent immunohistochemical processing. For immunostaining of ganglion whole mounts, tissue was blocked in 0.01M PBS, 0.02% sodium azide, 0.1% Triton X-100, and horse serum for four hours at room temperature with agitation. The tissue was then incubated in primary antibodies, in a solution of 0.01M PBS, 0.02% sodium azide, and 0.1% Triton X-100, with agitation over two nights. Tissues were rinsed in a solution of 0.01M PBS, 0.02% sodium azide (1hr rinse, 3 rinses) and incubated in secondary antibodies, diluted in 0.01M PBS with 0.1% Triton X-100 and 0.02% sodium azide, over 48 hours at room temperature with agitation. Stained tissue received a final rinse in 0.01M PBS, 0.02% sodium azide (1hr rinse, 3 rinses) and was mounted on glass slides in anti-fade media (Citifluor, Electron Microscopy Sciences) and coverslipped. The source and concentration of primary and secondary antibodies is given in Table 2.

Tissue blocks of the RAGP fat pad were frozen on dry ice and stored at −80 °C until sectioning. Blocks were attached to specimen plates, cut at −20 to −25°C using a Leica CM3050S cryostat (Leica Microsystems Inc.), and 30 μm thick sections collected on adhesion slides. Tissues were sectioned in a plane parallel to the epicardial surface, and sections were collected in a sequence that yielded eight sets of sections that each spanned the entire thickness of the specimen. Each set of tissue sections was stored at −20 °C until further processing. Slide-mounted tissue sections were immunostained at room temperature for markers using standard methods of fluorescence IHC, as described previously(Kaestner et al., 2019). Briefly, slides were rinsed with 0.01 mol/L PBS (pH 7.3), incubated for 10 minutes in 0.01 mol/L PBS containing 0.4% Triton X-100 and 0.5% bovine serum albumin (BSA), and blocked for 2 hours in 0.01 mol/L PBS containing 1% BSA, 0.4% Triton X-100, and 10% normal donkey serum (Jackson ImmunoResearch Laboratories). Each section was then incubated overnight in blocking buffer containing one or more of the primary antibodies listed in Table 2. Sections were washed again and incubated in blocking buffer before application of secondary antibodies. Species-specific donkey secondary antibodies conjugated to Alexa Fluor 488, 555, 594 or 647 (Jackson ImmunoResearch Laboratories) were applied at a 1:200 dilution in 0.01 mol/L PBS containing 0.4% Triton X-100 and 1% BSA, and sections were incubated for 2 hours before final washing in 0.01 mol/L PBS. After final washes with 0.01 mol/L PBS, cover glasses were applied with Citifluor (Ted Pella, Inc.) or SlowFade Gold antifade reagent (Life Technologies Corporation). Specific staining did not occur in negative control sections processed without the addition of the primary antibodies.

Slides were viewed under fluorescence illumination with an Olympus BX41 microscope equipped with an Olympus DP74 digital camera and cellSens software (Olympus America Inc., Center Valley, PA, RRID:SCR_016238). After localizing ganglia within sections, they were evaluated with a Leica TCS SP8 Confocal Microscope with 10x, 20x and 40x objective lenses (Leica Microsystems Inc.). Confocal images were collected at a resolution of 1024 × 1024 using 488 and 552nm laser lines. Stacks spanned tissue thicknesses of 25–33μm unless otherwise noted. Figures were created using maximum intensity projection (MIP) images for individual channels and merged images.

Whole ganglia were imaged with a confocal microscope (Zeiss LSM880) by tile-scanning in X, Y, and Z planes using a 63X objective (Zeiss Plan-Apo, 1.4NA). Z-stack images were acquired at step sizes consistent with Nyquist sampling in relation to the numerical aperture of the objective. Three-dimensional scans of whole ganglion neuroanatomy were acquired with 10X objectives. Images were quantified using ImageJ to measure neuronal cross-sectional areas (measured at Z-plane containing nucleus), and total cell counts per ganglia were quantified with ImageJ (FIJI distribution).

Neurobiotin filled neurons were serially imaged to visualize cell structure. For cases in which a neurobiotin labeled soma could not be located, or there were not enough dendrites labeled in the volume, the cell was omitted from further analysis. Neuronal structures were segmented, and quantified, using Neurolucida 360 and Neurolucida Explorer (MBF Bioscience). The file format of the traced structures was converted from .dat to .swc using HBP Neuron Morphology Viewer(Bakker et al., 2017). SWC files were further quantified using the ImageJ plugin SNT(Arshadi et al., 2021). Quantified values include the following: cell soma and dendrite area in X, Y and Z planes, soma volume, number of dendrite stems (originating branches from soma), number of dendrite bifurcations, number of dendrite branches, number of dendrite tips, total dendrite length, mean dendrite length, total dendrite surface area, total dendrite volume.

Samples of whole-mount right atrial ganglion plexi (RAGPs) intended for electron microscopy (EM) analysis were carefully isolated from the heart using a dissection microscope and then submerged in a solution containing 2% paraformaldehyde (PFA) and 2.5% glutaraldehyde overnight at 4°C in a 0.12 mol/L Millonigs buffer (MB). After thorough washing with Millonigs buffer and double-distilled water (ddH₂O), the tissues underwent fixation in a 1% osmium solution diluted in ddH₂O, followed by a series of ethanol (EtOH) and propylene oxide dehydration steps. Subsequently, the tissues were embedded in a plastic resin (Epon, Ted Pella). Semi-thin cross-sections (0.5 μm thick) were prepared and stained with a 1% toluidine blue solution diluted in ddH₂O to enable a general overview of RAGP organization under a light microscope (Nikon Eclipse E600 microscope equipped with a Nikon DS-Fi3 camera). Ultra-thin sections (70 nm thickness) were then mounted onto formvar-coated grids (Ted Pella), stained with uranyl acetate and lead citrate for contrast enhancement, and examined in a Tecnai G2 Spirit Twin transmission electron microscope (FEI, ThermoFisher Scientific) operating at 80 kV. Detailed characterization of epicardial ganglion organization was performed, and images were captured using a Gatan Orius SC 1000B digital camera (Gatan) and SerialEM software(Schorb et al., 2019).

Statistical analysis was performed using GraphPad Prism statistical software (version 9.2; La Jolla, CA). Data are presented as mean ± SD. Means of two independent groups were compared with an unpaired t test. Grouped data with two categorical variables and one continuous dependent variable were compared by two-way ANOVA and Šidák’s multiple comparisons test. Means between three groups were compared with one-way ANOVA and Tukey’s post hoc analysis. For datasets with missing values a mixed-effects analysis model was used. Values were considered statistically significant at P ≤ 0.05.

Data associated with this study were collected as part of the Stimulating Peripheral Activity to Relieve Conditions (SPARC) program and are available through the SPARC Portal (https://sparc.science/) (RRID: SCR_017041) under a CC-BY 4.0 license.

In mice, epicardial ganglia are found on the epicardial surface beneath a layer of connective tissue on the dorsal atrial surface near the pulmonary veins (Fig 1A). Approximately 4–7 large ganglia were identified following fixation and incubation with an antibody to the pan-neuronal protein PGP9.5 (Fig 1A). Counted ganglia (8 ganglia from 8 mice) contained 332 ± 142 neurons, with a mean ganglion area of 110 ± 74 mm², and neuron area of 347 ± 60 μm², giving a total neuronal density of 37 ± 21 neurons per 100 μm². From this data, we estimate the population of mouse ICNs to be between 1000 and 2400 neurons. This is the first quantification of ganglion cell numbers using whole isolated ganglia and three-dimensional imaging with confocal microscopy. Representative quantified ganglia are shown in Figure 1A. For single cell electrophysiology and morphology, we focused on neurons within the ganglion located at the merger of the superior and inferior vena cava, as indicated in Figure 1A.

In adult pig and human cardiac tissues, epicardial ganglia were more challenging to visualize given a thick layer of enveloping subepicardial fat. We focused on clusters of epicardial ganglia referred to commonly as the RAGP. The localization of the RAGP in pig and human hearts is shown in Figures 1B and 1C. Neurons from these ganglia primarily project to the right atrial myocardium and sinoatrial node(Bluemel et al., 1990; Randall et al., 1998; Hanna et al., 2021a). Macroscopically, in pig and human, the RAGP appeared as an interconnecting web of nerve bundles with interspersed ganglia of varying size, consistent with the nomenclature of a ‘ganglionated plexus’ (Fig 1B₃ & C₂). This neuroanatomy, while conserved between animals, showed significant interspecies, and more subtlety, inter-animal variability. The areas of individual ganglia, 168 ± 200 mm² in pig (30 ganglia from 14 pigs) and 188 ± 207 mm² in human (26 ganglia from 11 humans), were similar between pig and human; however, the cross-sectional area of the neuronal soma (1140 ± 207 μm² vs 1689 ± 242 μm²) were smaller in pigs (298 ICNs in 14 ganglia from 9 pigs) compared to humans (521 ICNs in 7 ganglia from 7 humans). The density of neurons in pig (6 ganglia from 6 pigs) and human (7 ganglia from 7 humans) ganglia were similar (6.7 ± 1.2 neurons/100 μm² vs 3.0 ± 1.5 neurons/100 μm²). While ganglion size was not significantly different between all three species, the number of ICNs per unit area was greater in mouse compared to pig (P = 0.0034) or human (P = 0.0054).

Cellular neurochemical expression was probed with rigorously validated antibodies (Table 2), in both ganglion sections and whole mounts. In all three species, intense punctate labeling of the vesicular acetylcholine transporter (VAChT) was observed surrounding most neuronal perikarya, supporting a ubiquitous preganglionic cholinergic innervation to these neurons. Intracellular VAChT+ labeling was observed in most, but not all neurons (Fig 2), indicative of a large population of post-ganglionic cholinergic neurons in epicardial ganglia from mice, pigs and humans.

Immunolabeling with an antibody to tyrosine hydroxylase (TH), a marker for noradrenergic neurons, revealed moderate to strong staining of the neuronal cytoplasm for some ICNs (Fig 2). While cell bodies were strongly positive for TH in each species, only within human ganglia did we clearly observe TH-immunoreactive (IR) nerve terminals encircling the postganglionic neurons (Fig 2).

In addition to adrenergic and cholinergic neurons, nitrergic neurons have also been identified within intracardiac ganglia(Parsons & Dun, 2004). Using an antibody to the neuronal isoform of nitric oxide synthase (nNOS)(Table 2), sparse cytoplasmic immunolabeling of nNOS was observed in several neurons from mouse and pig, but not human (Fig 2). No intraganglionic nNOS-IR fibers were observed in sections or whole mounts of cardiac ganglia from mice, pigs or humans.

Peptidergic neuromodulators are also present in peripheral ganglia of multiple species, including human(Hoover et al., 2009a). Immunostaining for the sensory neuropeptides substance P (SP) and calcitonin gene-related peptide (CGRP) exposed striking differences across species. In mice, SP and CGRP are colocalized within nerve fibers surrounding neurons as well as within intracardiac nerve bundles; however, no CGRP, or SP-IR cell bodies were identified (Fig 3). Pig intracardiac ganglia showed little to no staining for CGRP or SP containing perineuronal fibers or cell bodies. We only identified a few dually positive fibers within the intracardiac nerve bundles. Human intracardiac ganglia showed colocalization of CGRP and SP, like mouse, and displayed a dense meshwork of perineuronal fibers, also as observed in mice. Two human ICNs out of approximately 100 appeared to exhibit immunoreactivity (IR) to these peptides, though a definitive confirmation of sensory neurons within intracardiac ganglia is lacking.

Neuropeptide Y (NPY) and vasoactive intestinal polypeptide (VIP) are two peptidergic co-transmitters commonly identified within sympathetic and parasympathetic neurons, respectively(Dalsgaard et al., 1986; Pardini et al., 1992; Richardson et al., 2003). In the mouse, NPY-IR was observed in cell bodies, a few fibers, and was often colocalized with TH in nerve fibers surrounding blood vessels (Fig 3). In the pig, few NPY-IR nerve fibers were identified. NPY was principally localized to cell bodies and some co-labeling with TH was observed (Fig 3). In human, a greater abundance of NPY-IR nerve fibers surrounding postganglionic neuronal somata was identified. Perineuronal VIP-IR nerve fibers were also significantly more abundant in human ganglia compared to pig or mouse. Within pig RAGP, only a few VIP-IR cell bodies were identified. In mouse, a few varicose VIP-IR fibers were observed.

Importantly, immunoreactivity for the peptides showed variability across ganglia. The above observations were inferred from imaging of 18 sections and 2 whole mounts from 3 mice, 27 sections and 2 whole mounts from 5 pigs, and 32 sections and 2 whole mounts from 4 humans.

Intracellular microelectrode recordings from intact ganglion preparations were used to assess basic membrane properties of ICNs from each species (Fig 4). Recordings were obtained in identical extracellular solutions made at 34 ± 2 °C. The ganglion preparations rested for at least 1 hr in PSS before commencement of recording. Representative action potentials from each species are overlaid in Fig 4D and show similar initial rates of depolarization and repolarization. The mouse neurons had low resting membrane potentials (−51 ± 7 mV, n = 31), relative to pig (−57 ± 7 mV, n = 43) or human (−56 ± 5 mV, n = 19)(Fig 4E). The action potentials recorded from mouse ICNs were also smaller in amplitude (59 ± 9 mV, n = 24) than the action potentials recorded from pig (66 ± 8 mV, n = 24) or human (68 ± 12 mV, n = 17) ICNs (Fig 4F). Action potential half-widths were similar between groups (Fig 4G). The overlaid after-hyperpolarizing (AHP) potentials illustrate the markedly larger AHP of human ICNs, which differed in amplitude but not duration between species (Fig 4H,I,J). The Human AHP amplitudes (−14.2 ± 4.0 mV, n = 22) were greater than those of mouse (−9.9 ± 2.7 mV, n = 28; P< 0.0001) or pig (−11.4 ± 2.6, n = 33; P = 0.0034). The smaller mouse ICNs had higher whole cell input resistance (166 ± 88 MΩ, n = 31), in comparison to the larger pig (90 ± 58 MΩ, n = 44) and human (87 ± 35 MΩ, n = 20) neurons (Fig 4K). The threshold current for action potential general (rheobase) was less in mouse (60 ± 26 pA, n = 20 ICNs from 9 mice) compared to human (179 ± 70 pA, 14 ICNs from 10 humans; P = 0.0124) or pig (435 ± 161 pA, 29 ICNs from 10 pigs; P < 0.0001) neurons. The pig neurons required significantly more current to elicit an action potential compared to human (P <0.0001).

Excitability of the post-synaptic membrane was assessed in response to long duration, depolarizing current steps (+50 to +800 pA, 500ms). Autonomic neurons are characterized based on the number of action potentials elicited by depolarizing stimuli(Cassell et al., 1986). Phasic neurons fire short bursts of action potentials in response to long depolarizing current steps, while tonic neurons fire steadily throughout the depolarization. In mice, both phasic and tonic neurons were identified (Fig 5A). In pig, no tonic neurons were observed (0/40), and in the human, mostly tonic neurons (19/20) were observed (Fig 5B,C). Tonic human ICNs reached a peak firing frequency of 23 spikes/500ms (46 Hz) while mouse neurons peaked at twice that frequency (45 spikes/500ms; 90Hz). Excitability curves showing the number of action potentials elicited by each depolarizing current step are shown for phasic and tonic neurons in each species (Fig 5D,E,F).

The membrane time constant, tau, was calculated from −100pA hyperpolarizing steps. Mouse neurons had the shortest charge time (2.6 ± 1.2 ms, n = 34 ICNs from 12 mice) relative to human (15.6 ± 11.7 ms, n = 19 ICNs from 10 humans; P < 0.0001) or pig (9.7 ± 8 ms, 33 ICNs from 11 pigs; P = 0.0011). Despite being similar in size (based on neuron volume), pig neurons charged faster than human neurons in response to −100pA steps (P = 0.0121). Larger amplitude, long duration (500ms), hyperpolarizing steps (−100 to −600 pA) were given to assess responsiveness to membrane hyperpolarization. A difference between the instantaneous and steady-state voltage response (Fig 5D–F) indicated the presence of a hyperpolarization-activated non-selective cationic current, known to be mediated by hyperpolarization-activated cyclic nucleotide-gated (HCN) channels(Cuevas et al., 1997), which opposes the hyperpolarization. This change in the hyperpolarization is commonly termed ‘sag’. A ‘sag’ or difference potential of greater than 2 mV was observed in 75% human ICNs, 33% of porcine ICNs, and 85% of mouse ICNs.

Both the A-type voltage-gated potassium channels and small conductance Ca²⁺-activated potassium (SK) channels have been shown to affect intrinsic excitability of mammalian ICNs(Xi‐Moy & Dun, 1995; Rimmer & Harper, 2006). 4-AP, an inhibitor of A-type channels, increased excitability of pig but not human or mouse ICNs (Fig 6A,B,C). 4-AP (1 mM), slowed repolarization of the action potential which was measured as an increased action potential half-width in mouse and human but not pig ICNs (Fig 6A,B,C). Apamin, an inhibitor of SK channels, significantly increased the number of action potentials elicited with depolarizing current steps in all three species (Fig 6D,E,F). In mouse and pig, phasic neurons were converted to tonically firing ones. After 5 mins, apamin (100 nM) significantly reduced the duration of the AHP in human (130 ± 34 ms before, 51 ± 21 ms after, P = 0.03, n = 3 ICNs from 3 humans, paired t test) and mouse (81 ± 20 ms before, 12± 2 ms after; P = 0.02, n = 3 ICNs from 3 mice, paired t test), but not pig ICNs (62 ± 8 ms before, 53 ± 12 ms after; P = 0.1, n = 3 ICNs from 3 pigs, paired t test).

Brief (100 μs) stimulation of interganglionic nerves elicited orthodromic and antidromic potentials at ICNs from mice, pigs, and humans (Fig 7). Single stimuli elicited fast, excitatory potentials which, at mouse ICNs, were always suprathreshold for the generation of an action potential. These neurons exhibited a large excitatory post synaptic potential (EPSP)(Fig 7A). At pig and human ICNs, both subthreshold and suprathreshold EPSPs were observed (Fig 7B,C). At all inputs tested with the ganglion nicotinic blocker hexamethonium (500uM), the potentials were blocked, almost completely, confirming activation of ganglion nicotinic receptors with acetylcholine as the principal mechanism for fast excitatory neurotransmission (data not shown). Antidromic potentials could be discerned from synaptic potentials in that they were all-or-none and were unaffected by hexamethonium.

To determine if the ganglionic neurons received multiple inputs, we used either graded stimuli at a single input, or, where possible, multiple ganglionic inputs were stimulated with independent extracellular electrodes. At pig ICNs, stimulation of ganglionic nerves evoked fast synaptic potentials. Most of the nerve evoked responses (32/35) were orthodromic, with only three antidromic spikes. 8 of 28 synapses tested received two discrete inputs based on latency alone. As an example, the pig stimulus recruitment curve shown in figure 7B shows a short latency input in response to a 400 μA current and a long latency input was revealed with a 100 μA pulse. 9 of 15 neurons tested for dual inputs with multiple electrodes had them. At human ICNs, nerve evoked potentials were obtained in 8/20 ganglion preparations. 5/7 of the potentials were identified as antidromic. In 4/8 preparations, the synaptically mediated EPSPs were graded in nature, showed variability in synaptic latency, and were sensitive to inhibition by hexamethonium (500 μM). Similar to mouse, a large suprathreshold EPSP was observed, which occasionally, due to the excitability of the post-synaptic membrane, elicited multiple spikes. In one cell, two synaptically-evoked responses were observed after stimulation of a single nerve (Fig 7C), with unique latency values that were both inhibited by hexamethonium, supporting the concept of convergent innervation, as observed for pig but not mouse ICNs. The frequency distributions of latency values are shown in the bottom panels of figure 7 A, B and C. Mouse latencies were tightly distributed around a mean value of 2.1 ± 0.7 ms, while pig and human latencies were more broadly distributed around means of 5.1 ± 2.2 ms and 24.5 ± 23.76 ms respectively. Upon stimulation of interganglionic nerves, latency values exceeding 20ms were observed in 3 human ganglion preparations indicating the possibility of multisynaptic pathways within the ganglion.

In addition to single stimuli, trains of stimuli (5, 10, and 20 Hz) were given to test for presence of slow potentials and to measure synaptic efficacy. At pig ICNs, of 18 neurons tested for slow potentials, 8/18 exhibited a slow hyperpolarization. No slow depolarizations were observed out of the 18 neurons tested. Synaptic efficacy tended to be very low at porcine ICNs, with fewer than 40% of the evoked potentials triggering spikes. In contrast, at mouse ICNs, the evoked potentials elicited spikes nearly 100% of the time, and at human ICNs multiple spikes were occasionally observed giving greater than 100% synaptic efficacy (Fig 7D).

Spontaneous EPSPs (sEPSPs) were observed in mouse, pig and human ICNs. sEPSPs are mediated by spontaneous vesicle fusion, release of transmitter, and quantal-like activation of post-synaptic receptors. Representative sEPSPs from mouse, pig and human ICNs are shown in Figure 7E. The sEPSPs observed at pig ICNs (2.4 ± 0.6 mV, n = 8) were smaller in amplitude than mouse (4.5 ± 1.1 mV, n = 8; P < 0.001) but were comparable to human sEPSPs (3.3 ± 1.1 mV, n = 7; P = 0.16)(Fig 7F). The durations, measured from 90% to 10% peak amplitude, of mouse sEPSPs (16.2 ± 3.4 ms, n = 8) were shorter than those from pig (38.3 ± 15.3 ms, n = 8) or human (64.3 ± 25.6 ms, n = 7), and the human sEPSPs lasted longer than pig (Fig 7G).

The segmented cellular morphologies of representative neurons from each species are shown in Figure 8. Mouse neurons (15 ICNs from 10 mice) were unique from pig or human ICNs in that few or no dendritic projections were observed with light microscopy. In only 2/15 neurons were any small dendrites observed. The small diameter neuronal cell bodies projected a single axon which typically, though not always, exited the ganglion (Fig 8A₁–₄). In contrast, the manually segmented neurons from pig (13 ICNs from 8 pigs) and human (11 ICNs from 9 humans) exhibited elaborate dendritic arbors consisting of multiple branches projecting from the soma (Fig 8B,C). Typically, these projections extended from the inner (epicardial facing) membrane. A few neurons located deeper within the ganglion had projections extending from both inner and outer membranes. Ganglion depth varied from 1 to 2 layers of neurons to as many as 7 layers of neurons with the greatest neuronal mass concentrated in the ganglion center. Axons were identified based on larger cross-sectional diameter of the initial segment, and as the longest projections which exited the ganglion to join a nerve bundle. These projections are labeled ‘Ax’ on the shown reconstructions (Fig 8). Axons appeared to originate near dendrites on the soma. Given that the ganglia were isolated from the whole heart, the eventual targets of the axons could not be determined. Some pig and human ICN axons showed small bifurcations near the soma, giving rise to one or two short collateral neurite projections (Fig 8B,C). The 3D confocal images and 3D cellular reconstructions are available for download (see Data Availability statement).

The measured morphological properties are presented in Figure 9. The height (Y axis), width (X axis), and depth (Z axis) of the dendritic arbors, including cell somas, was similar between pig and humans ICNs (Fig 9A,B,C). Both dendrite area and ICN soma volume were significantly greater in pig and human than mouse (Fig 9A,B,C,D). Soma shapes, in all three species, are best described as oblate spheroids, having a sorter minor axis (Z axis) relative to the major axis (X,Y axis).The number of dendrite stems extending from the human ICNs was greater than for pig ICNs (Fig 9E). In mice, only 2 of 15 manually segmented ICNs had dendrites, each with three dendrite stems and a total dendrite length of less than 100 μm. The number of bifurcations and branches of human ICNs were greater than pig ICNs, but only the number of dendrite tips was significantly different from the pig neurons. Total dendrite length and were similar between the pig and human ICNs, both greater than mice. Despite the similar dendritic arbor size, human ICNs had more dendrite length close to the cell body, as shown with Sholl analysis of pig and human dendrite length (Fig 9k).

Across the population of manually segmented mouse, pig and human ICNs, some variability in neuronal morphology was observed. In mouse, we found that not all axons were fully matured and showed growth cones at the distal end of the axon (Fig 8A₁). From most neurons we observed only a single axon according to the classification given above. In pig, a couple of neurons appeared to lack any axon (Fig 8B₂), and there was one human cell in which an axon could not be identified. In human, two neurons lacked long dendritic projections and had only short dendritic projections constrained within an envelope around the soma (Fig 8C₂). Every effort was made to completely delineate neuronal morphology, although, truncation due to either incomplete filling or imaging cannot be excluded. Whether these unique morphologies relate to a difference in function have not been determined. Also, whether the long intraganglionic projections were definitively dendrites or might function as collateral axons cannot be determined at this time.

Ultrastructural organization of the intracardiac ganglia increased with complexity from mouse to human. The differences observed exceed a simple increase in cell size but include a reorganization of synaptic contact and a greater partitioning of the membrane in human and pig epicardial neurons. At mouse ICNs, synaptic specializations formed upon the somatic membrane forming axosomatic connections (Fig 10B). The synaptic boutons were filled with small clear vesicles (SCVs) and mitochondria (Fig 10B1,B2). However, large dense core vesicles (LDCVs) were also observed among the SCVs (Fig 10B1). The mouse did not show any dendritic projections from the ICN somata, validating the results from light microscopy. In stark contrast, at pig ICNs, no axosomatic specializations were identified out of 19 ICNs inspected with TEM (Fig 10C). Instead, multiple protrusions of the somatic cytoplasm into dendritic processes were observed (Fig 10E). Synaptic densities were challenging to locate but could be found at appositions between axonal and dendritic segments in areas adjacent to the neuronal cell bodies (Fig 10D,D1,F). These terminals were filled with mitochondria, SCVs, and a few LDCVs.

TEM of the somata of human ICNs revealed a nucleus with a prominent nucleolus and surrounding cytoplasm filled with varied quantities of vacuoles and lipofuscin granules (Fig 11). In total 54 neurons were examined. All images available on the online database. Some of the somata showed dendritic processes (Fig 11B1,D). As in pig, axosomatic synapses were not encountered in the human ICN samples. Instead, axodendritic connections were present (Fig 11E2). These axon terminals contained mitochondria, SCVs, and some LDCVs; and showed synaptic specialization with an electron dense active zone on the pre-terminal side of the synaptic cleft (Fig 11 E2), similar to axodendritic terminals in pigs.

The cross-sectional areas of the SCVs in ICNs of mice, pigs, and humans were quantified from high resolution TEM images. All tissues had been processed for ultrastructural studies using identical fixation and staining protocols. The mean area of mouse SCVs (1437 ± 89 nm², n = 3 terminals on 3 ICNs) was significantly smaller than SCVs at minipig (2174 ± 320 nm^2, n = 3 terminals on 3 ICNs) and human ICNs (2136 ± 115 nm², n = 3 terminals on 3 ICNs) (Fig 11 F). There was no difference in the size of minipig or human SCVs.