Protein Kinase C-Dependent Effects of Neurosteroids on Synaptic GABAA Receptor Inhibition Require the δ-Subunit

The dorsal motor nucleus of the vagus (DMV) contains preganglionic motor neurons important for interpreting sensory input from the periphery, integrating that information, and coding the appropriate parasympathetic (vagal) output to target organs. Despite the critical role of hormonal regulation of vagal motor output, few studies examine the role of neurosteroids in the regulation of the DMV. Of the few examinations, no studies have investigated the potential impact of allopregnanolone (Allo), a neuroactive progesterone-derivative, in the regulation of neurotransmission on the DMV. Since DMV neuronal function is tightly regulated by GABAA receptor activity and Allo is an endogenous GABAA receptor ligand, the present study used in vitro whole cell patch clamp to investigate whether Allo alters GABAergic neurotransmission to DMV neurons. Although Allo did not influence GABAergic neurotransmission during initial application (5–20 min), a TTX-insensitive prolongment of decay time and increase in frequency of GABAergic currents was established after Allo was removed from the bath for at least 30 min (LtAllo). Inhibition of protein kinase C (PKC) abolished these effects, suggesting that PKC is largely required to mediate Allo-induced inhibition of the DMV. Using mice that lack the δ-subunit of the GABAA receptor, we further confirmed that PKC-dependent activity of LtAllo required this subunit. Allo also potentiated GABAA receptor activity after a repeated application of δ-subunit agonist, suggesting that the presence of Allo encodes stronger δ-subunit-mediated inhibition over time. Using current clamp recording, we demonstrated that LtAllo-induced inhibition is sufficient to decrease action potential firing and excitability within DMV neurons. We conclude that the effects of LtAllo on GABAergic inhibition are dependent on δ-subunit and PKC activation. Taken together, DMV neurons can undergo long lasting Allo-dependent GABAA receptor plasticity.


INTRODUCTION
The efferent motor limb of the parasympathetic nervous system to subdiaphragmatic viscera originates from the brainstem nucleus known as the dorsal motor nucleus of the vagus (DMV). These DMV motor neurons send direct axonal innervations to pancreas, liver, intestine and stomach through the vagus nerve (Berthoud et al., 1991;Browning et al., 1999;Babic et al., 2012).
Despite the important role of GABA A receptors in the DMV, we still lack complete understanding of how GABA A receptors are regulated in this nucleus. In particular, little data exist on the role of non-GABA modulators of GABA A receptors and their effects on DMV neurons. One such class of GABA modulators are neurosteroids that result from the catalyzation of either cortisol or progesterone by 5α-reductase. These neurosteroids are traditionally considered secondary endogenous ligands of the GABA A receptor (Ferando and Mody, 2012) and serve as potent allosteric activators of these receptor (Bianchi and Macdonald, 2003). Neurosteroids readily cross the blood brain barrier and recently were shown to concentrate in the brainstem (Sze et al., 2018). Although no studies to date directly assess the action of neurosteroids on DMV neurotransmission and activity, inhibition of 5α-reductase activity does restrain DMV inhibitory GABAergic currents (Littlejohn et al., 2019). Additionally, in the nucleus tractus solaritus (NTS), a nucleus with significant projections to the DMV, neurosteroids influence neurotransmission enough to decrease neuronal activation to excitatory stimuli (Kim et al., 2018). Although the effects of neurosteroids vary significantly across different brain regions (Reddy, 2018), these studies together provide evidence that brainstem circuits and their motor output are regulated by neurosteroids.
Despite limited investigations into the role of neurosteroids in autonomic circuits, they are likely critical to maintaining physiologically relevant signaling. Actions of neurosteroids on GABAergic neurotransmission are directly linked to a variety of brain diseases and disorders, including depression (Maguire, 2019), epilepsy (Reddy, 2011), and anxiety (Longone et al., 2011). Elevated neurosteroid signaling is proposed to dysregulate homeostatic circuits since for example obesity and polycystic ovary syndrome (a syndrome with high risk of weight gain and hyperglycemia) present with high serum concentrations of neurosteroids (Holmberg et al., 2018). Using rodent models, administration of neurosteroids increases food intake and weight gain (Holmberg et al., 2015); while both high fat diet feeding and experimentally-induced diabetes (common animal models for the development of obesity and hyperglycemia) result in increased GABA A receptor neurotransmission to the DMV (Boychuk et al., 2015;Halmos et al., 2015;McMenamin et al., 2018;Clyburn et al., 2019). Therefore, understanding how neurosteroids alter inhibitory GABA A receptor activity in the DMV is critical to understanding the brain's role in maintaining physiological homeostasis, particularly as it relates to energy homeostatic regulation.
The present study was undertaken then to determine if neurosteroids enhanced inhibitory neurotransmission to decrease neuronal excitability in the DMV. Despite being a progesterone derivative, the neurosteroid, allopregnanolone (Allo), was chosen for the current study for its limited sex differences in in vitro signaling (Wilson and Biscardi, 1997). It was hypothesized that Allo induces an increase in inhibitory neurotransmission to DMV neurons, thereby allowing future inhibitory transmission to encode larger inhibition within the brainstem homeostatic regulatory network and decreasing DMV excitability.

MATERIALS AND METHODS
All experiments were performed on young adult male C57/Bl6J and GABA A receptor δ-subunit knockout mice (Gabrd −/− , δKO; kindly gifted from Dr. Chase Carver (Carver et al., 2014); 25.9 ± 0.37 g; 11.6 ± 0.83 weeks old). All strains were maintained on a C57/Bl6J background. C57/Bl6J mice were purchased from Jackson Research Laboratory and are now an established in-house colony at the University of Texas Health San Antonio (UTHSA). Animals were fed ad libitum and housed on a normal 14:10 light cycle. The UTHSA Animal Care and Use Committees approved all animal procedures.
Whole cell patch-clamp recordings were performed under visual control using infrared illumination and differential interference contrast optics (IR-DIC). For recordings, glass pipettes (2-5 M ; King Precision Glass, Claremont, CA) were filled with a solution containing the following (in mM): 130 Cs +gluconate (or K + -gluconate), 1 NaCl, 5 EGTA, 10 HEPES, 1 MgCl2, 1 CaCl2, 3 CsOH (or KOH), and 2-3 Mg-ATP, pH 7.2-7.4 adjusted with 5 M CsOH (or KOH). To examine inhibitory postsynaptic currents (IPSCs), voltage clamp mode was used at a holding potential of 0 mV with internal containing Cs + as the primary cation to block K + currents and minimize any influence of K + -dependent, postsynaptic GABA B receptors in recorded neurons. Action potential frequency, resting membrane potential (RMP), and input resistance (IR) were measured in current-clamp mode using a K + -gluconate internal recording solution. RMP was corrected for liquid junction potential post hoc (−7 mV). IR was measured by injecting current-steps (20 pA, 400 ms). To measure action potential response to depolarizing current pulses (400 ms duration) of increasing intensity (from 30 to 210 pA), neurons were clamped at −60 mV, to ensure a consistent starting voltage. The number of action potentials during each depolarizing current step were recorded as a count. All recordings were discarded if series resistance was > 25 M or changed by > 20% throughout the course of the experiment. Mean series resistance was 9.4 ± 0.6 M in Vehicle (Veh)treated groups and 9.2 ± 0.7 M in all Allo-treated groups. Electrophysiological signals were recorded using an Axoclamp 700B amplifier (Molecular Devices, Union City, CA), acquired at 20 kHz, low-pass filtered at 3 kHz, and stored to a computer using a Digidata 1550B converter and pClamp 10.7 software (Molecular Devices).

Drugs
Although there are several neurosteroids that are GABA A receptor ligands, Allo was used here since previous reports demonstrated limited sex differences in Allo's activity at GABA A receptors (Wilson and Biscardi, 1997). Allo ([3α,5α]-3-Hydroxypregnan-20-one, Tocris Bioscience; Minneapolis, MN) was initially dissolved in DMSO. This was added to aCSF containing 1 mM KYN for a final working concentration of 100 nM Allo containing 0.1% DMSO. This concentration of DMSO did not affect GABA A receptor-mediated IPSCs, as determined in a series of control experiments (Supplementary Figure 1).

Histology
The location of each recorded cell within the rostral-caudal aspect of the DMV was noted and recovered by staining for biocytin (0.2% added to internal recording solution; Figure 1A). After recording, slices were fixed with 4% paraformaldehyde in 0.01 M phosphate-buffered saline (PBS) overnight at 4 • C. Slices were rinsed in PBS and then immersed in avidin-conjugated to Texas red (1:400; Vector Laboratories, Burlingame, CA) in PBS containing 1% Triton X-100 and incubated for 2.5 h at room temperature to identify biocytin-filled neurons ( Figure 1A). The anatomical location of each neuron was identified using major neuroanatomical landmarks, including the shape and presence of either the central canal or 4th ventricle and the presence (or absence) of area postrema. These were compared to a known mouse brain atlas for relative location to bregma (Franklin and Paxinos, 2013).
To examine δ-subunit expression in the DMV after Allo application, brainstem slices were taken and incubated as detailed for electrophysiology. After removal from Allo or Veh bath for 30 min, slices were fixed with 4% paraformaldehyde. Serial coronal sections (30 µm) were then taken using a cryostat (Leica Biosystem) and processed for choline acetyltransferase (ChAT) and GABA A receptor δ-subunit. Sections were rinsed with 0.05 M phosphate-buffered saline (PBS; pH 7.4) and nonspecific immunoreactivity blocked with 10% normal donkey serum (Jackson Immunoresearch, West Grove, PA) in 0.3% Triton X-100 and PBS. The δ-subunit was identified using a rabbit primary antibody (1:100; Phospho-Solutions; 868A-GDN, Aurora, CO) followed by an Alexa Fluor 488-conjugated donkey-anti-rabbit IgG (1:200; Life Technologies, Carlsbad, CA). ChAT immunoreactivity was identified using a goat primary antibody (1:500; Millipore; AB144P; Burlington, MA) followed by Alexa Fluor 568-conjugated donkey-anti-goat IgG (1:200; Life Technologies). Allo-and Veh-treated tissues were run in parallel with the same antibody cocktails. Image acquisition parameters, including exposure time and intensity, were identical across groups; brightness and contrast in all images were modified together and identically for figure presentation. Negative controls were run without primary antibody. Robust labeling in the cerebellum served as a positive control due to consistently reported high δ-subunit expression in this brain region (Fritschy and Mohler, 1995;Pirker et al., 2000). All imaging was done with an Olympus BX43 microscope, and images were captured with a Retiga R6 (Teledyne Imaging, Tucson, AZ) using filters for the two fluorescent dyes.

Data Analysis
A minimum of 8-10 min following establishment of whole-cell configuration was used to allow equilibration of the intracellular and recording pipette contents. Following equilibration, spontaneous inhibitory post-synaptic currents (sIPSC) were recorded for 2 min. To determine treatment and drug effects on IPSCs, 2 min of continuous synaptic activity were analyzed offline using MiniAnalysis 6.0.3. (Synaptosoft, Decatur, CA). All events were used to assess IPSC frequency, but only unitary events (i.e., single peak) were used for IPSC amplitude and decay time measurements. For each trace, IPSCs were aligned by their rise and averaged using MiniAnalysis. The average time required for 90-37% of IPSC current to decay is reported as decay time.
To determine GABA A receptor-mediated tonic current, BIC was bath applied until a steady state was reached (7-10 min) as previously described (Littlejohn et al., 2019).
For all electrophysiological experiments, only one cell was used per slice, and no more than three cells were used per animal. Data are presented as means ± standard error of the mean (SEM). Statistical measurements were performed using GraphPad Prism 5 (GraphPad Software, La Jolla, CA) and are denoted throughout. Two-tailed unpaired Student's t-tests were used to determine statistical differences when there were only two groups. When variances were unequal, a Welch's correction was used as noted. To evaluate the distribution of neuronal responses (i.e., the number of neurons responding to ALLO), a non-parametric Mann-Whitney Test was used. A One-way ANOVA with a Tukey's multiple comparison post hoc test was used when there were more than two groups. When variances were unequal, a Kruskal-Wallis test with a Dunn's multiple comparisons post hoc test was used as noted. A One-way or Two-way repeated measures ANOVA with a Sidaks's multiple comparison post hoc test was used when there were more than two groups with repeated measures. Outliers were determined with a robust regression and outlier test (ROUT, Q = 0.1%; GraphPad). Probability values ≤ 0.05 were considered significant.

Decay Time of Inhibitory Postsynaptic Currents
The most consistent effect of Allo across multiple brain regions is an increase in the time for GABA A receptor currents to decay (Fancsik et al., 2000;Maguire and Mody, 2009;Ferando and Mody, 2012). Therefore, the following experiments determined if this was true for DMV neurons. In initial experiments during an acute (a)Allo (100 nM) application, mean sIPSC decay time at 5min (14.0 ± 0.8 ms; n = 6 from 3 mice), 10-min (14.3 ± 0.8 ms) or 20-min (16.4 ± 1.2 ms) was not significantly different from sIPSC decay time during vehicle application (time 0; 13.6 ± 0.9 ms; One-way repeated measure ANOVA; p = 0.80; Figure 1C). Dose of Allo was based on previous reports (Maguire and Mody, 2007;Abramian et al., 2014).
However, throughout the central nervous system, Allo induces longer lasting changes often associated with its removal (Maguire and Mody, 2007;Abramian et al., 2014;Carver et al., 2014). To determine then if Allo increased GABA A receptor neurotransmission over a longer duration, even after its removal, slices were incubated in either Allo (LtAllo) or Veh (aCSF with 0.1% DMSO) for 30 min before being placed in the recording chamber with aCSF perfusate. DMV neurons were recorded 30-60 min after placement in aCSF perfusate. During LtAllo, sIPSC decay times in DMV neurons from LtAllo-treated slices (35 ± 5.3 ms; n = 17 from 9 mice) were significantly increased compared to Veh-treated slices (13 ± 0.4 ms; n = 14 from 10 mice; unpaired Student's t-test; p = 0.0008; Figure 1F). To determine if this increase in decay time was mediated directly at the GABAergic synaptic contact of DMV neurons, IPSCs were examined in the presence of TTX (1 µM) to block action potential firing ( Figure 1F). In TTX, the decay time of miniature (m)IPSCs in DMV neurons from LtAllo-treated slices was again significantly longer (31 ± 2.8 ms; n = 15 from 9 mice) compared with DMV neurons from Veh-treated slices (14 ± 0.9 ms; n = 13 from 7 mice; One-way ANOVA with Tukey's post hoc; p = 0.0001; Figure 1F). Additionally, the effect of LtAllo persisted as far out as was tested (120 min after removal from Allo incubation; 31.6 ± 4.1 ms; n = 9 from 4 animals, One-way ANOVA with Tukey's post hoc, p = 0.0004; Figure 1F).
DMV neurons were also identified by anatomical location and recovered through biocytin filling and immunohistochemical analysis using avidin-Texas red ( Figure 1A; bottom right panel). Anatomical location and electrophysiological properties were compared. There was no correlation between bregma level and phasic current decay time in DMV from Veh-treated slices (n = 21, from 15 mice; Spearman correlation; p = 0.47, Figure 1I). However, phasic decay time in DMV neurons after LtAllo-treatment (n = 30, from 17 mice) significantly correlated with decreasing bregma location, suggesting that DMV neurons located more rostrally were more sensitive to LtAllo (Spearman correlation; p = 0.04, Figure 1I).

Frequency of Inhibitory Postsynaptic Currents
Similar to decay time, initial experiments examining whether aAllo application influenced the frequency of GABAergic currents demonstrated no significant difference in mean sIPSC frequency at any time point examined. During aAllo, mean sIPSC frequency at 5-min (2.0 ± 1.0 Hz; n = 6 from 3 mice), 10min (2.8 ± 1.2 Hz) and 20-min (3.5 ± 1.1 Hz) was similar to sIPSC frequency during time 0 (3.0 ± 1.7 Hz; One-way repeated measure; p = 0.60; Figure 2C). Overall, sIPSC frequencies in DMV neurons from Veh-treated slices (2.8 ± 1.0 Hz; n = 14 from 10 mice) were also not significantly different compared with neurons from LtAllo-treated slices (5.4 ± 1.2 Hz; n = 17 from 9 mice; unpaired Student's t-test with Welsh correction; p = 0.10; Figure 2F). However, when data were checked for outliers using a robust regression and outlier test (ROUT, Q = 0.1%; GraphPad), a single neuron's average sIPSC frequency was identified as an outlier in only the Veh group (denoted with an "j" in Figure 2F). With this neuron excluded, sIPSC frequencies from DMV neurons from LtAllo-treated slices were significantly higher compared with neurons from Veh-treated slices (1.9 ± 0.5 Hz; n = 13 from 10 mice; unpaired Student's t-test with Welsh correction; p = 0.018; Figure 2F). The presence of a small population of DMV neurons that exhibit relatively high frequencies is consistent with previous reports of sIPSC in DMV neurons (Boychuk et al., 2015;Boychuk and Smith, 2016). Therefore, data is presented unaltered. Similar to decay time, the LtAllo-induced increase in IPSC frequency was confirmed to occur at the direct synaptic contact of GABAergic neurons on DMV neurons using TTX. During TTX application, the mean frequency of mIPSCs in DMV neurons from LtAllotreated slices was significantly higher (2.5 ± 0.5 Hz; n = 15 from 9 mice) compared with neurons from Veh-treated slices (1.2 ± 0.3 Hz; n = 13 from 7 mice; Kruskal-Wallis with selected Dunn's post hoc; p = 0.05; Figure 2F). In slices treated with LtAllo, the number of individual DMV neurons with an average frequency higher than the average Veh value was significantly more (20/32) compared with the number of individual neurons in Veh-treated slices (9/27; Mann-Whitney U test; p = 0.04; Figure 2G), providing further confirmation that LtAllo increases IPSC frequency in DMV neurons. Unlike IPSC decay time, the LtAllo-induced increase in frequency washed out by 120 min after removal from Allo incubation (1.9 ± 0.2 Hz; n = 9 from 4 mice; Kruskal-Wallis with selected Dunn's post hoc; p = 0.12; Figure 2F). Finally, application of the selective GABA A receptor antagonist BIC (30 µM) abolished all sIPSCs between both groups (Supplementary Figure 2C), indicating that LtAllo does not recruit glycinergic IPSCs. Similar to decay time, anatomical location also influenced the overall effect of LtAllo on IPSC frequency. There was no correlation between bregma level and IPSC frequency in DMV neurons from Veh-treated slices (n = 49, from 30 mice; Spearman correlation; p = 0.174, Figure 2H). However, IPSC frequency in DMV neurons after LtAllo-treatment (n = 48, from 33 mice; Spearman correlation; p = 0.001, Figure 1I) significantly Mean ± SEM of sIPSCs frequency responses during Veh-application (3.0 ± 1.7 Hz) and aAllo-application at 5-(2.0 ± 1.0 Hz; n = 6 from 3 mice), 10-(2.8 ± 1.2 Hz), and 20-min (3.5 ± 1.1 Hz) in red. Overlaid dashed lines denote individual DMV neuronal responses (black); One-way repeated measure ANOVA. (D) Representative trace of sIPSCs from a DMV neuron during Veh (black trace) and LtAllo (blue trace). Line and arrow indicate section expanded below. (E) Representative sIPSC cumulative probability of inter-event interval (reciprocal of frequency) in a neuron from Veh-and LtAllo treatment. LtAllo left shifts sIPSC cumulative probability. (F) Mean ± SEM of IPSCs frequency in Veh-(black bars) and LtAllo-treated (blue bars) neurons. sIPSCs frequency in Veh-(2.8 ± 1.0 Hz; n = 14 from 10 mice) and LtAllo-treated (5.4 ± 1.2 Hz; n = 17 from 9 mice) neurons. The removal of the neuron labeled with "j" reveals a significantly increased sIPSC frequency after LtAllo. Right-side panel is the mean ± SEM of mIPSC frequency in Veh-treated (1.2 ± 0.3 Hz; n = 13 from 7 mice) and Allo-treated (2.5 ± 0.5 Hz; n = 15 from 9 mice) neurons, indicating that an increased frequency after LtAllo is TTX resistant. This increase did washout by 120 min after Allo incubation (1.9 ± 0.2 Hz; n = 9 from 4 mice). Overlaid points denote individual neuronal responses; j denotes the ROUT identified outlier; unpaired Student's t-test with Welsh's correction for sIPSC; Kruskal-Wallis with selected Dunn's post hoc for mIPSC. (G) Pie graphs illustrating the effect of LtAllo on sIPSC frequency responses expressed as percentage of the total group. Neurons were classified as either greater or less than the mean of Veh-treated frequencies (mean ± SEM of the IPSC frequency of the Veh-treated group). Mann Whitney U test. (H) Correlational analysis of averaged IPSC frequency across bregma level in Veh-(black with dashed line) compared to LtAllo-treated (blue with solid line) neurons recorded for this study (Veh, n = 49; Allo, n = 48). Each square represents a single neuron. # Denotes significant difference from Veh when outlier is excluded, *denotes significant difference from Veh; **denotes significant differences by Mann Whitney U-test. Significance indicates p ≤ 0.05. correlated with decreasing bregma location, with DMV neurons located more rostrally being more sensitive to the effects of LtAllo on frequency.

Tonic GABA A Receptor Currents
Since DMV neurons demonstrate significant GABAergic tonic currents (Gao and Smith, 2010a) and reports from other non-reproductive brain regions indicate that tonic GABAergic currents are increased by neurosteroids (Reddy, 2018), we also examined whether tonic GABA A receptor activity in DMV neurons is altered after Allo exposure. During the aAllo protocol, no consistent change in holding current occurred between Veh (56 ± 14 pA; n = 7 from 3 mice) and aALLO application (46 ± 11 pA; paired Student's t-test, p = 0.29; Supplementary Figure 2B). To determine differences after LtAllo, we examined GABAergic tonic currents by applying BIC (30 µM) and assessing the differences in holding current from pre-BIC to steady-state BIC application (Supplementary Figure 2C). There was no difference in BIC-sensitive tonic current from LtAllotreated DMV neurons (43 ± 7.3 pA; n = 12 from 8 mice) compared to Veh-treated neurons (32 ± 6.2 pA; n = 10 from 7 mice; unpaired Student's t-test; p = 0.27). Tonic currents were normalized to cell capacitance (tonic current density) for further analysis. There was no statistical difference in cell capacitance between either treatment group (Vehicle 30 ± 3.6 pF vs. LtAllo 35 ± 3.2 pF; unpaired Student's t-test, p = 0.31). BIC-sensitive GABAergic tonic current density from LtAllotreated DMV neurons (1.3 ± 0.3 pA/pF; n = 12 from 8 mice) was not significantly different from current density in DMV neurons from Veh-treated slices (1.1 ± 0.2 pA/pF; n = 10 from 7 mice; unpaired Student's t-test, p = 0.52; Supplementary Figure 2D). These data suggest that under these experimental conditions Allo does not mediate large changes in GABA A receptor-mediated tonic currents.

LtAllo Effects on Inhibitory Postsynaptic Current Decay Time Are Protein Kinase C-Dependent
In addition to traditional allosteric modulation, previous reports demonstrated that the effects of neurosteroids on GABAergic activity can be mediated through activation of intracellular PKC activity (Fancsik et al., 2000;Abramian et al., 2014). Therefore, we sought to determine if PKC activation was critical to the effects of LtAllo on mIPSC decay time, by either co-incubating slices with the cell permeable PKC inhibitor, GFX, during the 30 min LtAllo incubation period, or including it directly in the patch pipette. When GFX was applied during Veh incubation (regardless of within the bath or pipette), there was no significant difference between decay times compared to control conditions (One-way ANOVA; p = 0.81). Therefore, these groups were collapsed into one group. We again repeated the significant increase in mIPSC decay time after LtAllo (31.49 ± 2.76 ms; n = 15 from 9 mice) compared to neurons from Veh-treated slices (14.7 ± 0.6 ms; n = 40 from 22 mice; One-way ANOVA with Tukey's post hoc; p < 0.0001; Figures 3A,B). When compared to the average decay time from Veh-treated DMV neurons, LtAllo increased decay time in DMV neurons by 114.2 ± 18.8% ( Figure 3C). When GFX was co-incubated with LtAllo in the perfusate (bGFX), the mean mIPSC decay time was significantly elevated (21.9 ± 1.0 ms; n = 8 from 4 mice) compared to decay times from neurons in Veh-treated slices (One-way ANOVA with Tukey's post hoc; p = 0.01; Figure 3B). However, this was only a 48.9 ± 6.9% increase compared to the mean mIPSC decay times from neurons in the Veh-treated group, and this was a significant reduction in percent change compared to LtAllo alone (One-way ANOVA with Tukey's post hoc; p = 0.03; Figure 3C). When GFX was included in the patch pipette (pGFX), IPSC decay times were again significantly elevated after LtAllo (29.8 ±2.5 ms; n = 10 from 7 mice) compared to Veh-treated neurons (Oneway ANOVA with Tukey's post hoc; p < 0.0001; Figure 3B). However, unlike bGFX, this was a 102.7 ± 17.0% change, which was similar to the percent change seen during LtAllo application (One-way ANOVA with Tukey's post hoc; p = 0.85; Figure 3C). Taken together, these data indicate that ∼56% of the effect of LtAllo on decay time is PKC-dependent. They also indicate that although PKC activity contributes to the effect of LtAllo on decay time, a continued elevation in PKC activity is not required. To determine if PKC activation alone was sufficient to mediate an increase in mIPSC decay time, DMV neurons were treated with PMA. mIPSC decay time in DMV neurons from PMA-treated slices (16.0 ± 0.8 ms; n = 8 from 5 mice) was not different from mIPSC decay time in DMV neurons from Veh-treated slices (One-way ANOVA; p = 0.95; Figure 3B).

LtAllo Effects on Frequency Are Protein Kinase C-Dependent
Similar to decay time, GFX application (bath or intracellular) did not alter mIPSC frequency under Veh conditions (1.2 ± 0.3 Hz vs. 1.1 ± 0.3 Hz vs. 0.9 ± 0.3 Hz; ANOVA with Tukey's post hoc test; p = 0.78), and these groups were again consolidated. The mIPSC frequency in DMV neurons from LtAllo-treated slices (2.5 ± 0.5 Hz; n = 15 from 9 mice) were significantly higher compared to mIPSC frequency in Veh-treated slices (1.0 ± 0.2 Hz; n = 40 from 22 mice; Kruskal-Wallis ANOVA with Dunn's post hoc; p = 0.006; Figures 4A,B). This was a 151.0 ± 50.1% increase from the mean mIPSC frequency of Veh-treated DMV neurons. During bGFX, DMV neuron mIPSC frequency from LtAllo-treated slices (1.6 ± 0.4 Hz; n = 8 from 4 mice) was similar to mIPSC frequency in Veh-treated slices (Kruskal-Wallis ANOVA with Dunn's; p = 0.25; Figure 4B). To determine if PKC activation alone was sufficient to mediate the increase in mIPSC frequency after LtAllo treatment, DMV neurons were treated with PMA. When slices were incubated with PMA, DMV neuron mIPSC frequency from PMA-treated slices (7.1 ± 2.5 Hz; n = 8 from 5 mice) was significantly increased compared to mIPSC frequency in DMV neurons from Veh-treated slices (Kruskal Willis ANOVA with selected Dunn's comparisons post hoc; p = 0.0005; Figure 4B).

Protein Kinase C-Dependent Effects of LtAllo Require the δ-Subunit
The influence of neurosteroids is often attributed to interactions with the GABA A receptor's δ-subunit (MacKenzie and Maguire, 2013). To determine if LtAllo-induced PKC-mediated increases in sIPSC decay time requires the GABA A receptor's δ-subunit in the DMV, we determined the effect of LtAllo on sIPSCs in slices from δKO mice ( Figure 5A). In δKO mice, IPSC decay time was increased following LtAllo (27.0 ± 2.2 ms; n = 10 from 4 mice; One way ANOVA with Tukey's post hoc; p = 0.0007; Figures 5B,C) compared to neurons from Veh-treated slices (14.7 ± 1.2 ms; n = 9 from 6 mice). To determine if the δ-subunit was required for the PKC-dependent effects of LtAllo on decay time in DMV neurons, we tested the effects of GFX bath application on the effect of LtAllo in these mice. In DMV FIGURE 3 | Early protein kinase C (PKC) activation, but not constitutive PKC activity, is required for increased phasic GABA A receptor IPSC decay time during LtAllo. (A) Averaged trace of mIPSCs from DMV neurons after LtAllo (Allo-aCSF; blue trace), LtAllo with the PKC inhibitor Bisindolylmaleimide I co-applied in the bath (500 nM; Allo-bGFX; black trace), LtAllo with GFX restricted to the patch pipette (Allo-pGFX; green trace), and bath application of the PKC activator, phorbol 12-myristate 13-acetate in the absence of Allo application (PMA; 40 nM; gray trace). Averaged traces were normalized for amplitude. Inset graph shows mIPSC cumulative probability of decay time in a representative neuron during each treatment. (B) Mean ± SEM of mIPSCs decay time during Veh-and LtAllo-treated neurons (duplicated from Figure 1F) during control conditions, Allo-bGFX (21.9 ± 1.0 ms; n = 8 from 3 mice), Allo-pGFX (29.79 ± 2.5 ms; n = 10 from 7 mice), and PMA (16.0 ± 0.8 ms; n = 8 from 6 mice). (C) Mean ± SEM of mIPSC decay time relative change (from Veh-treatment group) in Allo-aCSF (114.2 ± 18.8%; n = 15 from 9 mice), Allo-bGFX (48.9 ± 6.9%; n = 8 from 3 mice), and Allo-pGFX (102.7 ± 17.0%; n = 10 from 7 mice). One-way ANOVA with Tukey's post hoc, * indicates significant differences from Veh; (b) indicates significant difference from Allo-aCSF. Significance indicates p ≤ 0.05.
neurons from δKO mice, the average IPSC decay time during bGFX was elevated similar to LtAllo by itself (24.7 ± 1.8 ms; n = 8 from 3 mice; p = 0.80; Figure 5C). However since only ∼56% of the action of LtAllo is dependent on PKC activation, we again examined the average percent change from Veh conditions. Unlike LtAllo in C57/Bl6 wild type animals that changed ∼114.2%, the average percent change in decay time after LtAllo in the δKO was only 56.7 ± 15.2%. Importantly, the application of GFX during LtAllo did not significantly change this percent increase (68.1 ± 11.9%; unpaired Student's t-test, p = 0.59; Figure 5D). Taken together, these data indicate while LtAllo can increase decay time in the absence of PKC and the δ-subunit, the PKC-dependent increases in IPSC decay time of LtAllo require the δ-subunit.

Allo Potentiates Expression of the δ-Subunit
Immunohistochemically labeled δ-subunits associated with the membrane to increase functional expression appear as clustered or enlarged puncta on or close to the surface of the neuron (Mangan et al., 2005;Abramian et al., 2014). Therefore, we qualitatively examined the immunohistochemical staining of the δ-subunit in DMV neurons. In general, somatic staining for the δ-subunit was present, but diffuse in DMV neurons from Veh-treated control mice (n = 4; Figure 5E), being distributed lightly throughout the soma. Labeling in the DMV from LtAllotreated mice (n = 4) was characterized by bright immunoreactive puncta in DMV neurons (Figure 5E), which was qualitatively different than what was observed in Veh-treated neurons. These findings imply that increased membrane expression of δ-subunits may be associated with the increased functional expression of IPSCs demonstrated in electrophysiological recordings from DMV neurons in LtAllo-treated neurons.
Previous reports also suggested that a repeated application of a GABA A receptor agonist decreases receptor activity as evidenced by decreasing evoked-amplitude, and that neurosteroid exposure results in increased receptor activity (i.e., increased evoked amplitude), and not decreased (Abramian et al., 2014). Therefore, we tested if LtAllo also potentiates activity of GABA A receptors in DMV neurons. We used an exposure to two brief (5 s) puffs of the δ-subunit agonist THIP (1 µM) separated by 20 min during either Veh application or LtAllo ( Figure 5F). After Veh application, the current density of a second THIP application was 97.1 ± 7.2% (n = 7 of 3 mice; Figure 5G) of the initial THIP response, representing a general lack of potentiation after activation. However, after Allo administration, the current density of the second THIP application was 213.4 ± 48.7% of the initial THIP response (n = 6 of 3 mice) and this was a significant increase compared to Veh-treatment (unpaired Student's t-test; p = 0.03). Therefore, consistent with immunohistochemical data, LtAllo significantly increased the functional expression of δ-subunit-containing GABA A receptors at the neuronal membrane, suggesting that LtAllo works to increase the expression of THIP-responsive GABA A receptors.

DISCUSSION
The present study found that DMV neurons exhibit long-term PKC-dependent increases in GABAergic neurotransmission after exposure to Allo. Specifically, this included a prolonged decay time and increased frequency after LtAllo (but not aAllo) that was mediated by actions directly at the synapse of GABAergic neurons since both parameters were TTX-insensitive. Allomediated plasticity was limited to synaptic GABA A receptor activity since tonic currents were not different. The PKCdependent actions of LtAllo required the δ-subunit of the GABA A receptor, and likely resulted from increased GABA A receptor membrane expression. LtAllo also decreased both spontaneous and evoked action potential firing, suggesting that this increase in GABAergic neurotransmission is sufficient to decrease DMV neuronal excitability. Taken together, these data are the first to demonstrate that DMV neurons have significant PKC-dependent LtAllo-induced increases in GABAergic neurotransmission. Additionally, they demonstrated that PKC-dependent increases in inhibition associated with LtAllo requires the δ-subunit of the GABA A receptor.
To our knowledge, this is the first examination of the role of Allo on inhibitory neurotransmission in the DMV. In DMV neurons, Allo failed to induce a consistent response in any parameter of GABAergic neurotransmission examined after 5-20 min of exposure. Traditionally, neurosteroids serve as allosteric modulators of the GABA A receptor resulting in modifications of GABA A kinetics (i.e., lengthening of decay time) (Haage and Johansson, 1999;Wang, 2011) via pore dilation and increased opening of the receptor's desensitization gate allowing for increased open time duration (Miller et al., 2017). Allo can also induce immediate phosphorylation of the receptor (Gyenes et al., 1994), but these effects occur rapidly (within 5 min of exposure) (Jovanovic et al., 2004). Therefore, the failure of Allo to induce immediate changes in GABAergic neurotransmission in the DMV indicates that Allo is not likely increasing GABA A receptor activity through immediate changes in receptor pore size and desensitization kinetics by simple allosteric modulation in DMV neurons.
that GABA A receptor subunits in the DMV confer strong long lasting, PKC-dependent actions of Allo on receptor kinetics since Allo prolonged decay time of GABAergic neurotransmission in DMV neurons 30-120 min after exposure and this effect was abolished after PKC inhibition. Our data further demonstrated that neither PKC activation nor inhibition was sufficient to alter GABA A receptor decay time under Veh conditions. Therefore, in the absence of Allo exposure, the influence of PKC on GABAergic receptor activity is minimal. LtAllo-induced activation of PKC also did not require elevated constitutive PKC activity since restriction of the PKC inhibitor to the patch pipette during recording did not prevent the LtAllo-induced increase in decay time. Taken together, the action of LtAllo on GABA A receptors must in part recruit a PKC intracellular signaling pathway, making the DMV a brain region with strong PKC-dependent modulation of GABA A receptors.
The PKC-dependent effects of LtAllo in the DMV also required the GABA A receptor's δ-subunit since mice lacking the δ-subunit did not demonstrate PKC-dependent increases in decay time. Although δ-subunits are traditionally considered extra-or peri-synaptic receptors (Nusser et al., 1998;Semyanov et al., 2003), the presence of the δ-subunit does confer increased IPSC decay kinetics (Haas and Macdonald, 1999;Bianchi and Macdonald, 2002), and DMV neurons have functional δ-subunit-containing GABA A receptors at GABAergic synapses (Boychuk et al., 2015). Interestingly, not only was the GABA A receptor's δ-subunit in part required for the augmentation of decay time, but LtAllo also conferred greater functional expression of the δ-subunit. Although qualitative in nature, the expression of the GABA A receptor's δ-subunit was increased in DMV neurons after LtAllo exposure. Additionally, repeated application of the δ-subunit agonist, THIP, under nACSF conditions resulted in attenuated GABAergic signaling while Allo exposure resulted in a unique potentiation of repeated THIP responses. Therefore, Allo exposure likely recodes inhibitory signaling allowing for otherwise similar inhibitory signaling to encode larger inhibition. Importantly, the lengthening of decay time during LtAllo was not associated with larger phasic GABAergic current amplitude, making it unlikely that a large increase in total GABA A receptor number occurred, but rather receptor subunits are reorganized to increase the expression of the δ-subunit.
The present results confirm that Allo induces augmented inhibition through activation of PKC signaling which requires the δ-subunit. However, the exact relationship between these three cellular components requires significantly more investigations. Since the transmembrane domain of the GABA A receptor contains multiple binding sites for neurosteroids that do not mediate changes in GABA A receptor conductance upon the immediate presence of neurosteroids (Chen et al., 2019), it is possible that the DMV prominently contains GABA A receptor subunit arrangements that confer increased GABAergic currents through less traditional allosteric binding sites. Therefore, Allo could initiate this intracellular signaling cascade by directly binding to the GABA A receptor in the DMV neuron's plasma membrane. While this binding does not mediate acute increases in receptor pore opening, it could signal changes within the intracellular compartment to activate signal transduction pathways and this dissociation from direct activation of the GABA A receptor was suggested previously (Fancsik et al., 2000). However, more recent evidence identified a neurosteroid signaling pathway through actions on metabotropic progesterone receptors that also facilitate protein kinase-dependent increases in GABAergic neurotransmission (Parakala et al., 2019). Therefore, future investigations into the role of PKC-dependent Allo-induced inhibition of DMV neurons is warranted and should attempt to resolute which receptor(s) are responsible for neurosteroid activation of intracellular PKC activity.
Although subtle changes in GABA A receptors at extrasynaptic locations cannot be completely ruled out, there was no difference in the amplitude of tonic current after Allo exposure under any condition tested. Therefore, Allo-induced modulation of GABA A receptor activity under these conditions is specific to receptors located at the synapse. This spatial specificity of Allo's actions across neuronal compartments was surprising given the relatively large conductance of tonic currents in DMV neurons (Ferreira et al., 2001;Gao and Smith, 2010a). Additionally, blocking the production of neurosteroids in female mice abolished estrouscycle dependent changes in GABA A receptor tonic, but not phasic, currents in DMV neurons (Littlejohn et al., 2019). Although sex differences do exist in Allo concentrations in the brainstem (Sze et al., 2018), Allo was specifically chosen here since previous reports demonstrated minimal sex differences compared to other neurosteroids (Wilson and Biscardi, 1997). Therefore, either sex differences do exist in the effect of Allo in the DMV or more prolonged exposure (e.g., hours to days) are need to alter receptor activity at extrasynaptic locations. Additionally, the actions of Allo can vary significantly depending on factors like ambient GABA concentration, and the presence of other GABA receptors/modulators (Khatri et al., 2019). Thus, the DMV might be a unique region of action for neurosteroids where multiple factors, notably timing, are critical for determining neurosteroiddependent activity on the GABA A receptor.
In addition to Allo effects on GABA A receptors in DMV neurons, our data further implicate LtAllo signaling in the modulation of presynaptic inhibitory signaling to the DMV. During LtAllo, there was a significant increase in IPSC frequency which was TTX-insensitive, and likely occurring directly at GABAergic synapses in the DMV. Previous reports suggest that Allo produces a significant reduction in vagal afferent signaling to the NTS (Kim et al., 2018). Reduced vagal afferent activity via deafferentation can influence signal transduction pathways in presynaptic GABAergic terminals resulting in increased statedependent GABAergic signaling to DMV neurons Browning and Travagli, 2009). Indeed, the activation of PKC through PMA recapitulated LtAllo-induced increases in IPSC frequency, suggesting a novel influence of PKC to increase inhibition in vagal circuitry. Additionally, LtAllo failed to influence IPSC frequency in mice lacking the δ-subunit making both PKC and the δ-subunit critical for the influence of LtAllo on presynaptic GABA release. Taken together, these data suggest that PKC activity is likely low at GABAergic presynaptic terminals and Allo decreases vagal afferent terminal activity resulting in the activation of PKC. PKC activation is then sufficient to increase the release of GABA quanta.
Interestingly, both the effects of LtAllo on decay time and frequency correlated with more rostrally located DMV neurons. Topographically, the DMV is traditionally defined by medial-lateral columnar organizations of different vagal branches (Fox and Powley, 1985). Organ projection target does not follow such stringent rules (Berthoud et al., 1991), and while electrophysiological parameters can be unique to projection class, these parameters are overlapping (Browning et al., 1999(Browning et al., , 2005. However, rostral to caudal topographical specificity to drug responses have been reported (Rossiter et al., 1990;Krowicki et al., 1997;Takahashi and Owyang, 1997;Zheng et al., 1999). Future studies should investigate the projection class and function of these more rostrally located neurons in relation to neurosteroid activity.
Regardless of its location of action, Allo activity, and neurosteroids in general, likely represent a state modulator in vagal circuits, especially given the timing of changes identified in the present study. Two important physiological states typically associate with large changes in neurosteroid concentration: pregnancy and stress. Both conditions require unique metabolic and cardiovascular responses. Allo has been suggested as a mechanism for baroreflex resetting associated with pregnancy (Heesch, 2011;Kim et al., 2018) and neurosteroid sensitivity is regulated during pregnancy through differential regulation of intracellular kinases activity (Koksma et al., 2003). Stress also increases levels of neurosteroids (including Allo) within the brain (Purdy et al., 1991;Sze et al., 2018), and stress alters brainstem-mediated homeostatic reflexes (Scheuer, 2010). The elevated neurosteroid signaling during these conditions could serve to recode inhibitory signaling to differentially integrate physiologically relevant stimuli. The DMV, then, is uniquely positioned for modulating integrative homeostatic responses to these conditions because of its ability to influence both metabolic and cardiovascular function. However, abnormally elevated neurosteroid activity could also dysregulate homeostatic circuits since for example obesity and polycystic ovary syndrome associate with high neurosteroid signaling (Holmberg et al., 2018). Therefore, the DMV is a unique, yet understudied, brain region of neurosteroid responses that could contribute to modulation of physiological homeostasis.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The animal study was reviewed and approved by the University of Texas Health Science Center at San Antonio Animal Care and Use Committees.

AUTHOR CONTRIBUTIONS
ELL and CRB designed, executed, analyzed all the experiments in the present manuscript, prepared, and wrote the present manuscript. Both authors contributed to the article and approved the submitted version.

FUNDING
This work was supported by AHA SDG 16SDG26590000 to CRB and a NIH grant T32 HL007446 to ELL.

ACKNOWLEDGMENTS
We would like Raehum Paik, B.S. and Selika Garza, B.S. at The Mouse Genome Engineering and Transgenic Facility at UTHSA for their help characterizing and managing our transgenic mouse line.