M-current downregulation leads to stellate ganglia sympathetic hyperactivity associated with hypertension

Cardiac sympathetic nerves are hyperactive in many cardiovascular diseases, however, the mechanisms underlying this are unknown. In humans, this phenotype is known to precede the development of hypertension and contribute to the severity of the disease. We highlight an electrophysiological phenotype in post-ganglionic sympathetic stellate ganglia neurons from prehypertensive spontaneously hypertensive rats (SHR), and use single cell RNA-sequencing, molecular biology, perforated patch-clamp and computational modelling to uncover the underlying mechanism. IM appears to be transcriptionally downregulated and pharmacological inhibition of IM in control neurons can recapitulate the SHR phenotype. We also show IM expression in human stellate ganglia. We demonstrate the contribution of a plethora of INa, IK, ICa and ICl channels in stellate ganglia neuronal firing, alongside a thorough characterization of this physiologically important ganglia. This is the first evidence of cellular electrophysiological hyperactivity in cardiac sympathetic neurons in hypertension and highlights INa and IM as targets of interest.


Introduction
Sympathetic hyperactivity is a well-documented co-morbidity that contributes to the etiology of many cardiovascular diseases (Herring, Kalla and Paterson, 2019). For example, in hypertension, increased cardiac sympathetic drive is linked to the development of left ventricular hypertrophy (Schlaich et al., 2003) and subsequent heart failure, which are independent predictors of mortality (Levy et al., 1990). A significant component of the dysautonomia associated with hypertension resides at the level of the post-ganglionic sympathetic neuron. For example, neurons from the stellate ganglia, which predominately innervate the heart Schmid, 1989, 1990), have enhanced Ca 2+ driven (Li et al., 2012;Larsen et al., 2016), norepinephrine release (Shanks, Manou-Stathopoulou, et al., 2013) and impaired re-uptake via NET (Shanks, Mane, et al., 2013). However, both animal models and patients with hypertension also have increased sympathetic nerve firing rate as measured by muscle and renal sympathetic nerve activity (Grassi, 2009a;Manolis et al., 2014). Whether this is centrally driven or results from changes in the excitability of post-ganglionic neurons before the onset of hypertension is unknown.
We therefore characterized the electrophysiological behavior of sympathetic neurons from the stellate ganglia of prehypertensive rats that develop high blood pressure over time (Minami et al., 1989). We undertook single cell RNA-sequencing to identify channel subunit expression in control and disease conditions and used this to guide a comprehensive electrophysiological exploration of their contribution to firing rate. Further to this, we highlighted that key targets were also present in the human stellate ganglia. In particular, we investigated the role of M-current (IM), sodium current (INa), T-type calcium current (ICaT), 7 mM KCl, 10 mM HEPES, 10 mM Na + -Phosphocreatine, 4 mM MgATP, 0.3 mM Na2GTP. Internal pH was adjusted to 7.3 with KOH.
To study KNa, NaCl was substituted for LiCl to allow an inward gLi via NaV channels but prevent activation of slick and slack channels by intracellular sodium (Silvana et al., 2003;Kaczmarek et al., 2013). For Ca 2+ free experiments, Ca 2+ was removed from the external solution and 5 mM EGTA, a Ca 2+ buffer, was added.
qRT-PCR Total RNA from whole flash frozen stellate ganglia was isolated using a RNeasy minikit (Qiagen, US) and immediately stored on dry ice before cDNA library preparation. For cDNA synthesis, Superscript IV VILO with ezDNase genomic DNA depletion (Thermofisher, US) was used, cDNA was then stored at -80 o C until required. Taqman PCR primers were used for the transcript identification of KCNQ2 (Rn00591249_m1), KCNQ3 ( Rn00580995_m1), KCNQ5 (Rn01512013_m1), SCN10A (Rn00568393_m1), where either GAPDH (Rn01775763_g1) or B2M (Rn00560865_m1) were used to normalize values via the ΔΔCT method (Livak and Schmittgen, 2001). Samples were measured on an ABI Prism 7000 (Thermofisher, US) as per the standard protocol for taqman.
Cryosectioning and Immunohistochemistry Freshly isolated stellate ganglia were immediately transferred to 4% paraformaldehyde for 1-2 hours, after which the tissue was incubated overnight in 20% sucrose-PBS at 4 o C, before embedding in OCT compound (Tissue-Tek). Tissue was then frozen and stored at -80 o C until cryosectioning the tissue as 12 µm sections. Slides were then permeabilized in 0.3% triton-X for 30 minutes at room temperature, before blocking for 2 hours in 1% BSA, 5% donkey serum. Sections were then incubated for 24 hours with primary antibodies at 4 o C, followed by five 5 minute washes in PBS and 2 hours incubation with the relevant secondary antibodies. Sections were author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2019.12.11.872580 doi: bioRxiv preprint 8 subsequently washed 3 times in PBS, and incubated with DAPI/PBS for 5 minutes, before a final 2 washes in PBS. Slides were then mounted with 50% glycerol in PBS before imaging.
Sections were imaged on a Zeiss LSM 880 Airy Scan Upright laser-scanning confocal microscope with a Plan-Apochromat 20x/0.8 M27 objective. Sections were DAPI stained, labelled with a mouse anti-TH antibody (66334-1-Ig) (ProteinTech, US) and a rabbit antibody against either KCNQ2 (ab22897), KCNQ3 (ab66640)  Single cell RNA-sequencing A single cell suspension of stellate ganglia cells was prepared via enzymatic dissociation as described under cell culture methods. Following blockade of enzymatic activity via three washes in blocking solution, the cell solution was transferred to phosphate buffered saline. The cell solutions were immediately transferred to ice and transported to the Wellcome Trust Centre for Human Genetics (WTCHG) for single cell sequencing via 10x genomics chromium (10x genomics, US) and Illumina hiseq 4000 (Illumina, US). This approach achieved 66-72K reads per cell, with a sequencing depth of 53-55% and 14-17K mean reads per cell before filtering. Initial analysis was performed by the WTHCG using the cell ranger pipeline (x10 genomics) with default parameters, before the data were exported to Seurat (v3.0) (Stuart et al., 2019) and analyzed in house. Cells were excluded in Seurat if the number of counts per cell was less than 4000 or percentage of mitochondrial genes was equal to or less than 0.3. For FindVariableFeatures we used 10000 features and the election method VST. Data was intergrated using 30 dimensions, 30 principle components were using for PCA analysis. UMAP and TSNE, FindNeighbours were ran with 19 dimensions. Findclusters was ran with a 9 resolution of 0.6. Differential expression analysis was performed via MAST(Finak et al., 2015) within Seurat.
Firing rate was taken as the maximum firing rate elicited by a range of 10 pA current injections between 10-200 pA. Membrane potential was monitored for stability during drug wash in and cells with large jumps in membrane potential were discarded.
Action potential parameters were measured from the first sequential 50 pA current step that induced an action potential. Peak amplitude (mV) was taken as the difference between the average baseline and maximum peak response of the action potential. Action potential upstroke (mV/ms) was taken as the maximum velocity from baseline to the peak amplitude.
Input resistance was calculated based upon a series of hyperpolarizing and depolarizing current injections ranging from -200 to 200 pA in amplitude in 10 pA increments (Spruston and Johnston, 1992). The average value of the final 100 ms was analyzed. As previously classified, small hyperpolarizing pulses were assumed to elicit the least active processes and any points that departed from linearity with these points or contained visible active processes in the final 200 mS of current injection were excluded.
Liquid junction potentials were calculated in JPCalcW (Barry, 1994) in Clampex (v11.0.3) (Molecular Devices, US), where ion availabilities were used instead of concentrations. Free author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2019.12.11.872580 doi: bioRxiv preprint 10 Ca 2+ , ATP, EGTA and Mg 2+ for internal solutions were estimated via MaxChelator (v8) (Bers, Patton and Nuccitelli, 2010) when relevant. For perforated-patch voltage-clamp and current clamp recordings a Liquid junction potential of 24.3 mV was calculated, without correction for the perforated patch Donnan potential (Horn and Marty, 1988). Whole cell current clamp recordings had an estimated liquid junction potential of -15.7 mV.
INa and Ik were modelled using the built in Hodgkin-Huxley kinetics. An existing model of Im was used from ModelDB (McDougal et al., 2017). The cell was morphologically represented by a ball and stick model with a somatic diameter of 22 µM. The simulated dendrites contained only passive currents, had a diameter of 1 µM and a length of 100 µM. The model cell was stimulated by a 0.1 nA current injection at the midpoint of the soma, soma0.5, for 1000 ms to elicit a train of action potentials. INa max amplitude was varied to simulate alterations in INa.
Statistics All datasets were normality tested, except for firing rate, which was taken as a discontinuous variable and treated as non-parametric data. Statistical analysis and normality tests were performed in graphpad prism (v8.2.1). The specific statistical test applied is stated in the figure legends with statistical significance accepted at p < 0.05 on two tailed tests. author/funder. All rights reserved. No reuse allowed without permission.

Stellate ganglia neurons are hyperexcitable in the prehypertensive SHR
Cultured stellate ganglia neurons of the prehypertensive SHR have a significantly higher induced firing rate than neurons cultured from normotensive age matched Wistar neurons as highlighted in Figures 1A and 1B. Of note the firing rate appears to plateau within the stimulation range used (10-200 pA) ( Figure 1B). This firing rate difference appears to be time resolved, with the majority of Wistar neurons firing action potentials within only the first 300 ms of stimulation, this phenotype is represented in Figure 1C via a raster plot of 30 Wistar and SHR neurons during a 1000 ms of 150 pA current injection.
We also observed alternate indicators of cellular hyperexcitability. The change in firing rate was accompanied by a 3.065 ± 1.128 mV mean decrease in resting membrane potential between Wistar and SHR neurons ( Figure 1D). The rheobase (minimum current injection of duration >300 ms required to reach the action potential threshold) was decreased in SHR neurons as observed via a series of 10 pA current steps of duration 1000 ms in the range 0-200 pA in amplitude ( Figure 1E). Input resistance showed a non-significant trend to increase in the SHR ( Figure 1F).

The SHR has a higher percentage of tonic firing neurons
Sympathetic neuron firing from other sympathetic and parasympathetic ganglia have been previously characterised into 3 subtypes (Cassell, Clark and McLachlan, 1986;Keast, McLachlan and Meckler, 1993;Wang and McKinnon, 1995;Weigand and Myers, 2010), as represented in Figure 1H. Here these subtypes have been assigned colour codes to allow for a visual representation of the firing rate and time course of firing in a range of conditions author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2019.12.11.872580 doi: bioRxiv preprint 12 throughout the manuscript. As per previous work (Wang and McKinnon, 1995), Phasic 1 neurons, represent neurons that fire only one action potential for a 1000 ms current injection in the range of 0-200 pA. Phasic 2 neurons fire 2-5 action potentials, all within the first 500 ms of a 1000 ms stimulation pulse for all recordings within the stimulation range 0-200 pA.
Tonic neurons fire either greater than 6 action potentials or continue to fire after 500 ms of stimulation of a 1000 ms stimulation pulse within the current injection range 0-200 pA.
When the percentage of neurons conforming to these subtypes is compared between cultured Wistar and cultured SHR neurons, Wistar neurons were found to be predominantly phasic 1 and phasic 2, whereas in the SHR, neurons were predominantly found to be of the tonic subtype ( Figure 1I). This data is supported by similar observations in Whole cell patchclamp recordings (Supplementary Figure 1).

A change in ion channel subunit expression was observed in SHR neurons
Single cell sequencing revealed a heterogenous population of cell clusters in the both Wistar and SHR dissociated stellate ganglia (Figure 2A), which mapped to known cell type markers ( Figure 2B; Supplementary Figures 3 and 4) including a large population of cells which are specific for a range of known sympathetic markers ( Figure 2C). Differential expression analysis between the Wistar and SHR sympathetic neuron populations identified in Figure 2C highlight a significant decrease in 5 ion channel subunit encoding genes that may contribute to firing rate of stellate ganglia sympathetic neurons (Table 1) (Full list clarified in supplementary 2D).

M-current is conserved and functionally reduced in SHR stellate ganglia neurons
author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2019.12.11.872580 doi: bioRxiv preprint 13 Was a decrease in KCNQ5 subunit expression the most likely explanation for the difference in phenotype? When assessed by RT-qPCR, gene expression of M-current encoding KCNQ2, KCNQ3 and KCNQ5 subunits is decreased in total RNA extracted from whole SHR ganglia ( Figure 3A). We also confirmed M-current subunit expression in samples of total RNA taken from human stellate ganglia ( Figure 3B). Using immunohistochemistry M-current encoding subunits KCNQ2, KCNQ3 and KCNQ5 were also expressed at a protein level in tyrosine hydroxylase (TH) positive cells (Supplementary Figure 5G), a classic marker for sympathetic neurons.
M-current can be recorded by deactivation curves, in this case applied from a holding potential of -25mV to -55 mV, allowing for a relaxation of current corresponding to IM. These recordings were made in perforated patch, as IM is known to rundown in whole cell recordings. To confirm that the current measured was IM we subtracted and measured current that was inhibited by IM inhibitor 10 µM XE-991 (Wang et al., 2000;Greene, Kang and Hoshi, 2018). These data were then normalised to cell capacitance. By this measure, IM was shown to be functionally present in stellate ganglia neurons and to be downregulated in SHR relative to Wistar ( Figure 3C).

M-current inhibition increased the excitability of Wistar stellate ganglia neurons
M-current pharmacology was used to assess the role of M-current in firing rate and other electrophysiological parameters in stellate ganglia neurons. M-current inhibition by 3 µM XE-991 (Wang et al., 2000;Greene, Kang and Hoshi, 2018) caused a significant increase in Wistar stellate ganglia neuron firing rate, measured as the maximum firing rate of tested neurons within a stimulation range of 0-200 pA ( Figure 3D). When this data was visualised as electrophysiological subtypes we observed an increase in tonic neurons after XE-991 author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2019.12.11.872580 doi: bioRxiv preprint 14 application and a reduction in phasic 1 and phasic 2 subtypes ( Figure 3K). This is supported by a significantly decreased rheobase (Supplementary figure 5D) and an accompanying depolarisation of resting membrane potential (Supplementary figure 5A).
Similar results were seen after application of 30 µM Linopirdine, an alternative Mcurrent inhibitor at a dose comparable in efficacy to 3 µM XE-991 (Costa and Brown, 1997).
Linopirdine increased maximum firing rate in Wistar neurons ( Figure 3E). This also caused a reduction in phasic 1 and phasic 2 subtypes, and an increase in tonic neurons ( Figure 3H). As with XE-991, Linopirdine also reduced rheobase amplitude (Supplementary figure 5E) and depolarised the resting membrane potential of these neurons (Supplementary Figure 5B).

M-current activation reduced excitability of SHR and Wistar stellate ganglia neurons
As IM was shown to be reduced, but not entirely absent via 10 µM XE-991-substracted deactivation curves ( Figure 3C), we also tested whether increasing SHR M-current via the activator retigabine (Main et al., 2000;Tatulian et al., 2001;Corbin-Leftwich et al., 2016) would be sufficient to reduce firing rate. We found that retigabine significantly reduced SHR stellate ganglia neuron maximum firing rate at all 3 doses tested ( Figure 3F). These data are visualised for maximal firing rate as subtypes in Figure Figure 5F), but at higher doses too few neurons still fired in this range to allow for a quantitative comparison of rheobase amplitude. author/funder. All rights reserved. No reuse allowed without permission.

INa, SK channels and Kv2.1 also control SHR neuron firing rate
Using single cell RNA-sequencing and a series of pharmacological inhibitors, we screened a range of calcium and voltage-gated ion channels with a known role in determining the firing rate of other neuronal populations. In Figure 4A the patterns of expression for channels implicated in firing rate (Supplementary Figure 2D) are shown against cell clusters highlighted in the Wistar and SHR stellate ganglia ( Figure 2A; Figure 2B). For these channels, either 1 or 2 pharmacological inhibitors were applied to SHR neurons to observe any effect on firing rate (Figures 4-5; Table 2). Pharmacological inhibitors which had a significant effect on the firing rate of these neurons are shown in Figures 4-5 and non-significant inhibitors are shown in Table 1.
Low dose (10 nM) Tetrodotoxin (TTX) (Tucker et al., 2012) significantly reduced firing rate in SHR neurons ( Figure 4D). In a separate population of SHR neurons, high dose TTX (300 nM) was shown to prevent firing in all tested SHR neurons (Supplementary figure 6F), confirming the absence of a large TTX insensitive INa.
The persistent Nav channel inhibitor riluzole (Urbani and Belluzzi, 2000) inhibited firing in the tested population and reduced most neurons to phasic 1 at the lowest dose tested, 3 µM ( Figure 5D).
Membrane depolarisation can limit Nav availability through increasing NaV inactivation (Ulbricht, 2005). We ensured the depolarised resting membrane potential observed in the SHR ( Figure 1D), did not limit SHR firing rate. To do so we applied a series of negative 1 second current injections in the range -10 to -100 pA followed immediately by a stimulatory 1 second 150 pA current injection. By this method, we found that there was no significant difference between these current steps or a 0 pA control in SHR neurons (Supplementary Figure 6G). author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10. 1101 Further to these data, we found that Nav1.8 inhibitors, which have previously been shown to inhibit stellate ganglia function in vivo (Yu et al., 2017), are likely to act through this mechanism. 100 nM A803467 (Jarvis et al., 2007) and 300 nM A887826 (Zhang et al., 2010) inhibited firing at tested doses (Supplementary Figure 7A) figure 7E). We also found that Nav1.8 was not present by RT-qPCR (Data not shown) nor single cell RNA-sequencing, an observation supported by previous work in the superior cervical ganglion (SCG) and the lack of TTX insensitive Nav (Supplementary Figure 6E).

Tonic neurons have less IM and more INa
By studying action potential kinetics and IM density between subtype populations we aimed to gain further insight into the mechanisms behind these subtypes. First, we found that in perforated patch clamp recordings input resistance was higher in tonic neurons than phasic 1 and phasic 2 ( Figure 6A).
Indirect measures of INa, action potential upstroke and action potential amplitude, taken from whole cell recordings of action potentials induced by a threshold 10 ms current injection were used to highlight any differences between subtypes. Whole cell recordings were used to reduce the effect of higher Rs values encountered with perforated patch on action potential kinetics. These data reveal higher INa in tonic and phasic 2 populations than in phasic 1 as measured by amplitude ( Figure 6B) or upstroke ( Figure 6C).
Further to these data, by comparing the amplitude of the first and second action potential elicited by a threshold 10 pA 1000 ms current injection, we found that in phasic 2, author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2019.12.11.872580 doi: bioRxiv preprint 18 but not tonic neurons, there was a significant decrease in the second action potential amplitude ( Figure 6D).
We confirmed that these parameters were INa dependent via the effect of low dose TTX on single action potentials as measured by action potential upstroke in perforated patch (Supplementary Figure 6A) or amplitude (Supplementary Figure 6B).
There were no significant differences in resting membrane potential between the three subtype groups when compared via perforated patch ( Figure 6E). However, there was a trend towards a more depolarised resting membrane potential in Phasic 2 and Tonic neurons. When viewed per electrophysiological subtype, we observed significantly less IM, as determined by deactivation curves, in tonic firing SHR neurons than phasic 2 SHR neurons ( Figure 6F).
We also found no significant differences between strains in either measure by whole cell patch clamp recordings of action potential upstroke (Supplementary Figure 6C) or amplitude (Supplementary Figure 6D).

Modelling confirms the roles of IM and INa
To confirm that modulation of IM and INa could theoretically lead to the three electrophysiological subtypes, we used a simple ball and stick computational model to assess The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2019.12.11.872580 doi: bioRxiv preprint

Stellate ganglia neurons of the SHR are hyperexcitable
We report three primary novel findings. First, sympathetic stellate ganglia neurons from the prehypertensive SHR are hyperexcitable, which manifests as a higher induced firing rate, depolarised resting membrane potential and reduced rheobase. Secondly, IM is downregulated in the stellate ganglia neurons of the SHR and this is the causative mechanism for membrane hyperexcitability. Thirdly, hyperexcitability can be curbed either by elevation of remaining IM or via reduction of INa, via global inhibition or selective inhibition of Nav1.1-1.3, Nav1.6 or INaP.
This phenotype of increased induced neuronal firing rate is consistent with the major description of sympathetic nerve hyperactivity in humans (Grassi, 2009b) (Manolis et al., 2014).
Together with previously reported elevated membrane calcium conductance (Li et al., 2012;Larsen et al., 2016), this will directly lead to the increased cardiac noradrenaline spill over (Esler, 2000).

Prior models of SHR sympathetic neurons
Previous work in the SHR model of hypertension has reported repetitive firing in SCG neurons, but failed to find a convincing mechanism underlying this phenomenon (Yarowsky and Weinreich, 1985;Jubelin and Kannan, 1990;Robertson and Schofield, 1999). The contemporary view is that this change may be due to changes in IA, which was reported to be larger in SCG neurons of the SHR (Robertson and Schofield, 1999). However, in the stellate ganglia IM modulation alone was enough recapitulate the SHR phenotype (Figure 3). This dominant role for IM is in alignment with reports of IM in the SCG (Wang and McKinnon, 1995). author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2019.12.11.872580 doi: bioRxiv preprint 20 Further to this, there were no observed changes in IA encoding transcript expression by single cell RNA-seq ( Figure 2D; Supplementary Figure 2D), which is further supported by assessment of IA in control SCG, suggesting that IA is the same between Phasic and Tonic neurons (Wang and McKinnon, 1995).

IM downregulation causes hyperexcitability
The phenotype observed (Figure 1) fits well with known characteristics of IM, a channel originally discovered in sympathetic neurons of the superior cervical ganglia (Brown and Adams, 1980). Alongside IM a plethora of potassium channels have been documented in other sympathetic ganglia (Dixon and McKinnon, 1996) and so we felt it was better to use an unbiased approach to search for a viable target.
As IM is a slowly activating, and non-inactivating inhibitory K + current, it provides a restriction upon firing, but only after its considerable activation period. This fits with both a general increase in firing rate ( Figure 1A; Figure 1B), and a time-dependent phenotype ( Figure   1C). IM has a powerful effect on neuronal resting membrane potential, as demonstrated in Supplementary Figure 5A-C, therefore the loss of IM in the SHR is likely to cause the depolarisation of the resting membrane potential in the SHR ( Figure 1D). Notably, reversing this depolarisation on its own does not appear to alter firing rate (Supplementary figure 6F), suggesting this variable does not contribute to the overall phenotype. IM inhibition also reduces rheobase (Supplementary Figure 5D-E), which again fits with the phenotype observed ( Figure 1E).

Prior work has investigated the systemic effect of IM modulators in vivo and found
Retigabine to reduce cardiac sympathetic activity in the SHR (Berg, 2016) (Berg, 2018). This would be expected based upon our observation of IM having a dominant role on stellate author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2019.12.11.872580 doi: bioRxiv preprint 21 ganglia neuron function in vitro ( Figure 3) and our transcriptional and functional data highlighting IM as a powerful driver of hyperactivity in the SHR. In further support of our findings, retigabine has been demonstrated to inhibit noradrenaline release from SCG neurons (Hernandez et al., 2008). Moreover, in a co-culture model of SCG neurons and neonatal cardiomyocytes, retigabine also reduced the action of SCG neurons on cardiomyocytes (Zaika, Zhang and Shapiro, 2011). Finally, human genome wide association studies have found two M-current single nucleotide polymorphisms in KCNQ3 (rs138693040-T) (Méndez-Giráldez et al., 2017) and KCNQ5 (rs12195276-T) (Evangelou et al., 2018) to be significantly associated with variation in the electrocardiographic QT interval and pulse pressure respectively, both variables which are modulated by the sympathetic nervous system.

INa and stellate ganglia neuron firing
The pattern of expression for Nav subunits is different to that noted for dorsal root ganglia or central neurons, with a stellate ganglia neuron expression pattern of Nav1.1, Nav1.2, Nav1.3, Nav1.6 and Nav1.7 ( Figure 4A). Stellate ganglia neurons therefore express peripheral channel Nav1.6, but lack Nav1.8 and Nav1.9 for which subpopulations of sensory peripheral neurons are notable (Bennett et al., 2019). As observed for other central and peripheral neuron populations, INa modulation appears to be a powerful regulator of firing rate in SHR stellate ganglia neurons ( Figure 4D; Figure 5).
Inhibition of all Nav subunits identified by single cell RNA-sequencing ( Figure 4A), apart from Nav1.7, significantly reduced SHR firing ( Figure 6). This gives some breadth to the potential mechanism of INa reduction and allows for the targeting of a range of Nav types to find a tolerable inhibitor. Of interest, thoracic epidural anaesthesia using Na channel blockers such as bupivacaine is used clinically in patients experiencing recurrent, life threatening author/funder. All rights reserved. No reuse allowed without permission.
The key role of persistent current in stellate ganglia neuron firing is demonstrated by the efficacy of INaP inhibitor riluzole ( Figure 6D), Nav1.6 channels contribute a relatively large persistent current (Herzog et al., 2003) (Smith et al., 1998;Rush, Dib-Hajj and Waxman, 2005;Chen et al., 2008), and so it is possible that riluzole is acting primarily on this channel subtype here.
These data highlight that targeting either persistent INaP, Nav1.1-1.3 or Nav1.6 are viable targets to reduce sympathetic hyperactivity. Nav1.7 inhibition does not affect firing rate by either tested inhibitor (Table 2). This suggests that Nav1.7 does not contribute to firing rate in the SHR model. Nav1.7 undergoes sustained inactivation, unlike Nav1.6 for example (Bennett et al., 2019), and is therefore less likely to support sustained firing. Nav1.7 inhibitors have been billed as a treatment for neuropathic pain, and have seen much academic and industrial attention. Thus the lack of effect on cardiac sympathetic firing can be regarded as a positive indicator that these compounds are unlikely to have sympathetic side effects. Nav1.1 (Scn1a), Nav1.2 (Scn2a) and Nav1.7 (Scn9a) are downregulated in the SHR (Table 1), but this does not appear to negatively affect SHR neuron firing rate (Figure 1), action potential amplitude (Supplementary Figure 6D) or action potential upstroke (Supplementary Figure 6C), and it may be concluded that these changes either do not cause a reduction in channel protein or that the reduction in INa is minimal. author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2019.12.11.872580 doi: bioRxiv preprint 23 Nav1.8 is not present in stellate ganglia neurons A803647, an inhibitor of the TTX-resistant Nav subunit, Nav1.8, has been previously shown to have a powerful inhibitory effect on arrhythmogenesis following stellate ganglia stimulation when applied locally in vivo (Yu et al., 2017). This was originally attributed to an effect on stellate ganglia neuron Nav1.8 function, however in our hands, we could not find molecular evidence of its expression ( Figure 4A) nor could firing occur after 300nM TTX (Supplementary figure 6F). This is in concordance with prior work from the SCG, which show no transcript expression (Akopian, Sivilotti and Wood, 1996) and a lack of TTX resistance INa (Schofield and Ikeda, 1988).
Our data, and the mechanisms described within this paper, show that the effect of this inhibitor is likely to be through an off-target reduction of INa, as shown on a cellular level for A803647 and A887826 in this manuscript ( Figure 4A; Supplementary Figure 7). This hypothesis is supported by the similar findings from the autonomic nervous system (Stone et al., 2013). If interpreted in this manner, these data support the findings in vivo. We suggest that this would be better targeted through inhibition of one of the Nav subunits shown to be present by our single cell sequencing analysis or functional work ( Figure 4A; Figure 6), rather than through off-target effects of Nav1.8 inhibitors.

Calcium-activated membrane channels in the stellate ganglia
Calcium-activated channels are likely to have a role in defining the tone of SHR neuron firing rates, as shown by the increased firing rate after Apamin inhibition of SK channels.
Apamin has reported to increase firing rate in SCG neurons, but to a lesser extent than observed here ( Figure 4D), with SCG firing only increased from an anecdotal observation of 1 to 3 Hz (Kawai and Watanabe, 1986). In the SCG, Bk channels have been reported to have a author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2019.12.11.872580 doi: bioRxiv preprint 24 key role in determining a spiking response to Bradykinin and sensitivity to NGF (Vivas, Kruse and Hille, 2014). Our single-cell RNA-sequencing data supports BK channel expression ( Figure   4A) and we observed an increase in action potential width (Data not shown) similar to that previously described in dopamine neurons (Kimm, Khaliq and Bean, 2015). We therefore suggest that BK channels are likely to have a role in the stellate ganglia, but that this may be independent from firing rate.
Overall however, the difference in subtype percentages between Wistar and SHR appears to be largely intact between whole-cell and perforated patch techniques, which would imply that this phenotype is largely independent of intracellular calcium dynamics, as they would be largely disrupted in the whole cell conformation (Shah, 2014) (Figure 1;   Supplementary Figure 1). We also found no expression differences for calcium-activated channels encoding genes in our single-cell RNA-sequencing dataset ( Figure 2D). Further to this, removal of extracellular calcium had a highly variable effect on SHR neuron firing rate, and no net difference in firing rate before and after calcium removal ( Table 2).
The calcium activated chloride channel, TMEM16A, does not appear to be involved in determining stellate ganglia neuron firing rate, nor does it appear to be present in the neuronal population (Table 2). This contrasts with the mouse SCG where the use of the broadspectrum chloride channel inhibitor, 9-AC appeared to reduce firing rate (Martinez-Pinna et al., 2018). It is unknown if this results from the difference in pharmacological inhibitor and methodology or from inter-ganglia and inter-species expression differences, but our electrophysiological and single cell sequencing appear to support each other in our study of the rat stellate ganglia. author/funder. All rights reserved. No reuse allowed without permission.

Kv2.1, KNa, T-type and HCN currents in neurons of the stellate ganglia
Prior studies have reported the importance of Kv2.1 conductance in the SCG, with a similar effect of inhibiting Kv2.1 to that observed here in the stellate ganglia (Table 2) (Malin and Nerbonne, 2002;Liu and Bean, 2014).
Slick and slack channel encoding transcripts were detected by our single-cell RNAsequencing, however, we found that substitution of Na + for Li + which should ablate the majority of the current through KNa, had little effect on firing rate (Table 2). This could suggest the effect of removing KNa was below the detectable limit in our model, or that KNa has a minor role.
Interestingly, T-type calcium, was absent in these neurons, with no effect of 2 selective inhibitors, TTA-A2 (Kraus et al., 2010) and TTA-P2 (Choe et al., 2011), (Table 2) alongside a negative result in the neuronal population via single-cell sequencing ( Figure 2C). Prior knockout studies of T-type subunits suggest that these channels regulate the cardiac sympathetic innervation via central pathways (Hansen, 2014), with our data confirming that the role of T-type channels in cardiac sympathetic activity is restricted to central input.
HCN channels appear to be expressed in stellate ganglia neurons via single cell RNAsequencing ( Figure 2C) but appear to have little control over somatic firing properties when tested via either ZD7288 or Ivabradine (Table 2). It is possible that these channels have a larger role in determining stellate ganglia neuron neurite electrophysiology, as observed in central neurons (Shah, 2014), or that these channels are relatively inactive with resting cyclic author/funder. All rights reserved. No reuse allowed without permission.

Comparison with other ganglia
Differences between the stellate ganglia and other sympathetic ganglia have now been reported by several measures including NET transporter activity (Shanks, Mane, et al., 2013), input resistance (Wang and McKinnon, 1995;Luther and Birren, 2009) and firing rate subtypes (Wang and McKinnon, 1995;. As the stellate ganglia provides the majority of cardiac sympathetic innervation, it is important that these data also describe the electrophysiological phenotype of the stellate ganglia as it is currently not characterised. A range of firing subtypes from Phasic 1 dominant to Tonic dominant have been reported in other ganglia from control models (Keast, McLachlan and Meckler, 1993). It was interesting to observe the presence of all 3 subtypes in the stellate ganglia, as in the SCG only Phasic 1 and Phasic 2 neurons were observed (Wang and McKinnon, 1995). In the coeliac (~58%) and superior mesenteric ganglia (~85%) (Wang and McKinnon, 1995), a higher percentage of tonic neurons have been observed than we observed in the stellate ganglia ( Figure 1H; ~20%). We observed that neurons of the Wistar stellate ganglia have a much larger input resistance (218.2 MΩ) in comparison to the majority of measurements from SCG (Mouse 109.7 MΩ, 75-85 MΩ; Rat 389.9 MΩ), Thoracic (Mouse 117.9 MΩ) and Coeliac (Mouse 161.9 MΩ; Guinea pig 116.9 MΩ) ganglia Anderson, Jobling and Gibbins, 2001;Lamas, Reboreda and Codesido, 2002;Martinez-Pinna et al., 2018). However, some of this difference may be attributed to interspecies differences. author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi. org/10.1101org/10. /2019 Previous studies have also observed more depolarised resting membrane potentials for the range of studied sympathetic ganglia, with our estimated value in Wistar of -68.04 mV being more hyperpolarised than most recordings from other ganglia including the SCG (Mouse SCG -49.9 mV, -54 mV; Rat SCG -58.3 mV)  (Lamas, Reboreda andCodesido, 2002)(Martinez-Pinna et al., 2018).

Electrophysiological subtypes of the sympathetic nervous system are dynamic
Stellate ganglia neuron phasic 1 neurons are likely to have low INa compared to phasic 2 and tonic neurons, which would limit sustained firing ( Figure 5; Figure 7). The difference between phasic 2 and tonic neurons is likely to be lower IM with a spectrum of current amplitudes and associated firing rates.
In other ganglia, a relationship between reduced IM and tonic firing has been reported (Wang and McKinnon, 1995;Jia et al., 2008;Luther and Birren, 2009) which we have now related to a disease context. INa has been reported to be higher in tonic firing SCG neurons (Luther and Birren, 2009), however this study provides the first non-correlative evidence that this is crucial for a tonic or phasic 2 phenotype.
Alongside IM and INa it is likely that other excitatory or inhibitory channels will shape this relationship in the stellate ganglia, including, but not limited to SK channels and Kv2.1.
Interestingly, input resistance was not reported to be different between phasic and tonic neurons in prior work (Wang and McKinnon, 1995), but this might be as phasic 1 and phasic 2 neurons were grouped together for this prior study.
The relative ease of pharmacologically converting these neural subtypes (Figures 3-5), suggests they are fluid and dynamic, and therefore do not correspond to hardwired neuron author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2019.12.11.872580 doi: bioRxiv preprint 28 subtypes. This is supported by our single cell sequencing data, where although three populations of sympathetic neurons appear ( Figure 2B), our initial probing into these data suggest that these groupings do not correlate with IM or INa expression (Data not shown).
There is no evidence at this time that these transcriptome-based groups are in any way related to electrophysiological subtype as suggested by a prior study in related ganglia, where neurochemical content did not appear to correlate with electrophysiology (Keast, McLachlan and Meckler, 1993). These transcriptome defined groups merit further study, but a detailed comparison is beyond the scope of this project.

Limitations
Whilst downregulation of IM provides an explanation for membrane hyperactivity, we have not identified a pathway by which this may occur. One possible explanation is that this results from continual presynaptic input, contributing to ganglionic LTP (Alkadhi et al., 2001;Alzoubi, Aleisa and Alkadhi, 2010). Also, it is possible that other channels may be contributing to this phenotypic difference, but as the measured variables all correlate well with our observations of IM inhibition in Wistar neurons, it seems highly likely that IM downregulation is a major cause.
The SHR model of hypertension has several issues, primarily its reliance on rodent physiology, which differs in several respects from man (Hasenfuss, 1998) and the genetic basis of the SHR pathology, which in humans only appears to contribute but not define hypertensive pathology. To address translatability of our work, we have confirmed that IM subunits are expressed in human stellate ganglia ( Figure 3B), but we have not been able to functionally confirm this. In relation to disease pathology, we use the SHR here as a model of sympathetic nerve hyperactivity, which like in humans, correlates with cardiovascular author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2019.12.11.872580 doi: bioRxiv preprint 29 disease. In this work we study sympathetic hyperactivity outside of the context of the development of SHR hypertension and chose this model due to its reliable presentation of sympathetic dysfunction.
Electrophysiological studies in this paper were performed at room temperature, due to the poor stability and viability of patch clamping these neurons at physiological temperatures in culture.
Single cell sequencing was the best approach for this study, as bulk sequencing would also incorporate contaminating cell types, for example vascular cells which are known to have ion channel expression changes in the SHR (Jepps et al., 2011) and may therefore confound results. It should be noted that we observed downregulation of all three M-current encoding subunits via RT-qPCR, but only downregulation of KCNQ5 by single cell RNA-sequencing. For KCNQ2, this may be as the detected expression level of KCNQ2 is relatively low via single cell RNA-sequencing ( Figure 2C), and therefore any chance is unlikely to be detected. For both KCNQ2 and KCNQ3 it is possible that downregulation of contaminating vascular SHR KCNQ2 and KCNQ3 is detected by our RT-qPCR of the whole stellate ganglion. Regardless, all these data support a downregulation of IM in stellate ganglia neurons of the SHR, with a functional reduction of IM supporting this.

Conclusions
We have described in detail a phenotype of sympathetic hyperactivity in stellate ganglia neurons of the SHR and have provided an electrophysiological framework for this observation, guided by single-cell RNA-sequencing of the stellate ganglia and human validation of key transcripts. Targeting key ion channels in the stellate ganglia, such as M-author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2019.12.11.872580 doi: bioRxiv preprint 30 current, may provide a reversible therapeutic opportunity to treat cardiac sympathetic hyperresponsiveness over and above interventions like surgical stellectomy. author/funder. All rights reserved. No reuse allowed without permission.