Regulation of rNST Responses to Afferent Input by A-type K+ Current

Abstract

Responses in the rostral (gustatory) nucleus of the solitary tract (rNST) are modified by synaptic interactions within the nucleus and the constitutive membrane properties of the neurons themselves. The potassium current I_A is one potential source of modulation. In the caudal NST, projection neurons with I_A show lower fidelity to afferent stimulation compared to cells without. We explored the role of an A-Type K+ current (I_A) in modulating the response to afferent stimulation and GABA-mediated inhibition in the rNST using whole cell patch clamp recording in transgenic mice that expressed channelrhodopsin (ChR2 H134R) in GABAergic neurons. The presence of I_A was determined in current clamp and the response to electrical stimulation of afferent fibers in the solitary tract was assessed before and after treatment with the specific Kv4 channel blocker AmmTX3. Blocking I_A significantly increased the response to afferent stimulation by 53%. Using dynamic clamp to create a synthetic I_A conductance, we demonstrated a significant 14% decrease in responsiveness to afferent stimulation in cells lacking I_A. Because I_A reduced excitability and is hyperpolarization-sensitive, we examined whether I_A contributed to the inhibition resulting from optogenetic release of GABA. Although blocking I_A decreased the percent suppression induced by GABA, this effect was attributable to the increased responsiveness resulting from AmmTX3, not to a change in the absolute magnitude of suppression. We conclude that rNST responses to afferent input are regulated independently by I_A and GABA.

Introduction

Neurons in the rostral nucleus of the solitary tract (rNST) relay taste afferent input to the forebrain via a connection with the parabrachial nucleus and to local reflex pathways through projections to the caudal nucleus of the solitary tract (cNST) and subjacent reticular formation (Beckman and Whitehead 1991; Norgren 1978; Norgren and Leonard 1971; Tokita et al. 2009; Travers 1988). Factors shaping the output signal from the rNST include convergence of primary afferents onto second-order neurons (Boxwell et al. 2018; Grabauskas and Bradley 1996; Sweazey and Smith 1987; Travers et al. 1986; Vogt and Mistretta 1990), synaptic interactions from neurons both within and outside the nucleus, and the intrinsic membrane properties of rNST neurons themselves (Chen et al. 2020; Tell and Bradley 1994) reviewed in (Bradley 2007).

Potassium channels and their associated currents are important regulators of membrane excitability (Dumenieu et al. 2017; Hille 1992; Vacher et al. 2008). One such class of currents, A-Type K+ currents (I_A) are rapidly inactivating outward currents, mediated by a number of voltage-gated K+ channels (Connor and Stevens 1971b; Noh et al. 2019; Yuan and Chen 2006). I_A influences spontaneous rate, discharge frequency and spike pattern across many neuronal populations (Imai et al. 2019; Locke and Nerbonne 1997; Simkin et al. 2015; Tarfa et al. 2017; Tell and Bradley 1994) including the rostral and caudal NST (Bailey et al. 2007; Chen et al. 2020; Dekin and Getting 1987; Strube et al. 2015; Tell and Bradley 1994). With a primarily somatodendritic location (Carrasquillo et al. 2012; Vacher et al. 2008), I_A can also influence synaptic integration including the response to excitatory input (Hoffman et al. 1997). In the rNST, I_A is mediated by KV4 channels (Chen et al. 2020) and is prominent in cells with projections to the PBN and reticular formation (Corson and Bradley 2013; Suwabe and Bradley 2009) and, to a lesser extent, in GABAergic neurons (Chen et al. 2020; Chen et al. 2016; Wang and Bradley 2010a). In the cNST, cells with I_A show less fidelity to afferent stimulation compared to cells either lacking (Bailey et al. 2007) or expressing a much reduced I_A (Strube et al. 2015). Although the presence of I_A influences membrane excitability in response to direct membrane depolarization, (Tell and Bradley 1994), its impact on responses to afferent stimulation in the rNST has not been tested.

In the present study, we explored the role of I_A in modulating the response to afferent stimulation in the rNST using whole cell patch clamp recording in transgenic mice expressing channelrhodopsin (ChR2 H134R) in GABAergic neurons. Optogenetic stimulation allowed us to differentiate GABAergic from non-GABAergic neurons (Chen et al. 2020; Chen et al. 2016). We targeted non-GABAergic neurons, a population that includes glutamatergic cells that project to the parabrachial nucleus, the major relay for forebrain pathways (Gill et al. 1999; Norgren and Leonard 1973). We determined the presence of an I_A-like response under current clamp, and evaluated the response to electrical stimulation of afferent fibers in the solitary tract (ST), before and after I_A was pharmacologically blocked with the specific KV4 channel blocker AmmTX3 (Maffie et al. 2013; Vacher et al. 2002).

To complement the pharmacologic suppression of I_A, we implemented a dynamic clamp protocol to assess the impact of adding a synthetic I_A conductance on the response to afferent stimulation. The optogenetic approach also allowed us to determine a potential interaction between GABAergic inhibition and I_A modulation in non-GABAergic neurons. In theory, a feed-forward inhibitory circuit in which primary afferents excite GABAergic interneurons that in turn contact rNST projection neurons (Chen et al. 2016; Wang and Bradley 1995), could make more Kv4.3 channels available, enhancing the resulting I_A, thus amplifying the suppressive effect of GABA. Therefore, we tested the hypothesis that there is an interaction between phasic GABAergic inhibition and I_A by measuring the efficacy of optogenetic release of GABA inhibition to suppress the response to afferent stimulation before and after blocking I_A.

Methods

Experiments were conducted in adult mice of both sexes (male = 16, female = 15, 1 unknown) that ranged in age from 34 – 116 days $(\overline{X}=64.7\pm 4.9\text{days})$. Results of male and female subjects were combined since we observed no differences between them. Mice expressed channelrhodopsin (ChR2 H134R) in GABAergic neurons by placing the expression of ChR2 under the control of the promoter for glutamic acid decarboxylase-65 (GAD65, also known as GAD2). Animals were bred by crossing a Cre-dependent ChR2-EYFP mouse line (Ai32(RCL-ChR2(H134R)/EYFP), JAX #012569) with mice expressing Cre recombinase under the control of the endogenous promoter for GAD65 (GAD2-IRES-Cre, Jax 010802). Mice received ad libitum water and food until the day of the experiment. Experimental protocols were approved by the Ohio State University Institutional Animal Care and Use Committee in accordance with guidelines from the National Institutes of Health.

Slice preparation

Following anesthetization with isoflurane, decapitation, and rapid removal and cooling of the brain, the brainstem was blocked and fixed to a ceramic block using cyanoacrylate glue. Coronal brainstem slices 250 μm thick were cut with a vibratome (model 1000, Vibratome, St. Lois, MO, USA) equipped with a sapphire blade in an ice-cold carboxygenated cutting solution containing (in mM): 110 choline, 25 NaHCO₃, 3 KCl, 7 MgSO₄, 1.5 NaH₂PO₄, 10 d-Glucose, 0.5 CaCl₂.

Slices were incubated for one hour in a carboxygenated artificial cerebrospinal fluid (ACSF) composed of (in mM): 124 NaCl, 25 NaHCO₃, 3 KCl, 1 MgSO₄, 1.5 NaH₂PO₄, 10 d-Glucose, and 1.5 CaCl₂ at 32°C, then transferred to a recording chamber and perfused with 36°C ACSF at a rate of 1–2 mL/minute. Neurons identified using DIC optics were recorded in whole-cell patch clamp mode using 4–6 MΩ glass pipettes filled with an intracellular solution (in mM: 130 K-gluconate, 10 EGTA, 10 HEPES, 1 CaCl₂, 1 MgCl₂, and 2 ATP, at pH 7.2–7.3; osmolality = 290–295 mOsm). Signals were amplified with an A-M Systems Model 2400 amplifier (Carlsborg, WA). Experimental protocols were conducted with pClamp software (v. 10.7; Molecular Devices, Sunnyvale, CA). An initial seal of greater than 1 GΩ, membrane resistance of greater than 100 MΩ, and positive action potential overshoot were inclusion criteria for seal and cell viability.

Neuron identification

We visualized the rNST under DIC optics with a Nikon E600FN microscope. Although EYFP was evident under epifluorescence, this visualization was not adequate for differentiating GABAergic (G+) from non-GABAergic (G−) neurons because clear demarcation of membranes was obscured by dense neuropil labeling. Thus, neurons were classified based on their physiological response to light, i.e. optotagging (Deubner et al. 2019; Lima et al. 2009; Pi et al. 2013), and as we have done in previous studies (Chen et al. 2020; Chen et al. 2016). A fiber optic probe (200 μM, NA = 0.39) connected to an LED (Thor Labs, model# 4100 4-channel LED Driver; 455 nm, 2.3 mW at the tip) was centered to illuminate the rNST (Fig. 1A inset). Neurons were optically stimulated with 1 s light trains (10 Hz, pulse duration = 10 ms) and classified G− if they responded with long-latency inhibitory postsynaptic potentials (IPSPs) > 9 ms (Fig. 1B1) (see (Chen et al. 2016) for details). These IPSPs likely arose from activated GABA neurons or inhibitory terminals in the nucleus. G+ neurons were identified by short-latency (< 3 ms) excitatory responses, time-locked to the light stimulation (not shown).

Fig.1.

A. Photomicrograph of patched GABAergic negative cell under DIC optics (asterisk). Inset shows configuration for stimulation and recording through a pipette (p). A twisted bipolar stimulating electrode (e) was positioned over incoming afferent fibers in the solitary tract (st). A fiber optic probe was used for optogenetic stimulation (o) and a cannula positioned for drug delivery (c). B1. Cells were categorized as non-GABAergic if they responded with inhibitory post-synaptic potentials to optogenetic stimulation. B2. Action potentials were recorded in response to 20 Hz afferent fiber stimulation. B3. Cells were identified as I_A positive (I_A+) when the onset of action potentials to a depolarizing stimulus was delayed when preceded by a hyperpolarizing pre-pulse.

Photomicrographs of the patched neuron together with the patch pipette (40×) and its location within the nucleus (4×) were taken under DIC and epifluorescence to aid in the localization of the neuron within the rNST.

Current clamp

Responses of rNST neurons to afferent stimulation were quantified as the number of action potentials (APs) evoked by electrical stimulation of the solitary tract (ST; 150 μA, 20 Hz, 1 s, Fig. 1B3). If no action potentials were elicited, we searched for another cell. 20 Hz was chosen because a previous study showed that it produced the most APs over a stimulation frequency range of 1 to 50 Hz (Chen et al. 2020). We also calculated the latency for each cell, measured from the beginning of electrical stimulation to the onset of the first spike for each stimulus train under control conditions, i.e. the absence of optogenetic stimulation and AmmTX3 (N = ~ 5 per cell). If the first response to a given train lacked a spike, we measured the onset of the first excitatory postsynaptic potential (EPSP). In all cases there was either an AP or an EPSP.

The effect of GABA release on responses to afferent stimulation was determined by simultaneous optical stimulation (2.3 mW, 10 Hz, 10 ms duration pulses) in which trials with light stimulation alternated with those without light stimulation. The mean number of spikes from multiple trials under each condition was calculated (mean number of repetitions per condition = 4.6) and responses expressed as either absolute suppression (spikes control - spikes during light) or percent suppression (spikes control - spikes during light/spikes control X 100). For these experiments, we only targeted G− neurons because the optogenetic release of GABA on responses to afferent stimulation in G+ cells was confounded with APs elicited by the optogenetic stimulation. However, 3 G+ neurons were recorded in the dynamic clamp experiments (described below) where we did not test the effects of inhibition.

We categorized cells as either expressing or not expressing I_A based on an I_A - like response in current clamp. Because we wanted to assess the impact of blocking I_A on action potential (AP) responses to afferent stimulation, we could not accurately determine the presence of I_A in voltage clamp where TTX is necessary to suppress APs. However, in a previous study conducted under voltage clamp, we determined that the highly specific Kv4 channel blocker AmmTX3 (Alomone labs, STA-305) (Maffie et al. 2013; Vacher et al. 2002) suppressed I_A, and in a separate population of cells under current clamp, eliminated the delay to spike initiation in response to depolarization preceded by hyperpolarization (Chen et al. 2020). Thus, in the present study we used a similar current clamp protocol to infer the presence of I_A and its suppression following pharmacological block. The criterion for this I_A - like response was the occurrence of a delay in the latency of spike initiation, or an increase in the first inter-spike-internal (ISI) in response to a hyperpolarizing pre-pulse (450 ms duration, −150 pA to 0 pA in 30 pA steps) followed by a depolarizing pulse (1 s duration, 150 pA: FIG. 1B2 (Bailey et al. 2007; Byrne 1980; Chen et al. 2020; Tell and Bradley 1994; Venugopal et al. 2010). For simplicity in this manuscript, we refer to neurons expressing such I_A - like responses as I_A+.

To determine the impact of I_A on responses to afferent stimulation and the modulation of these responses by optogenetic activation of the rNST GABA network, we recorded responses before and after adding AmmTX3 to the bath for a subset of neurons. AmmTX3 (0.5 μM) was delivered at a rate of 0.5 ml/min through a 33-gauge stainless steel tube positioned ~300 μm from the dorsal edge of the rNTS (Fig. 1A Inset) and connected via PE 100 tubing to a 5cc syringe mounted in a syringe pump (Instech model 2000). We recorded from only one cell per slice to avoid potential residual drug effects on subsequently recorded cells and typically recorded from only one slice per animal. The N’s provided in the text refer to the number of cells for individual experiments.

Dynamic clamp

In a subset of experiments, dynamic clamp was used to determine if adding I_A to a cell lacking I_A (N = 6) or after treatment with AmmTX3 (N = 5) could modulate the response to afferent stimulation. To implement the dynamic clamp protocol, we used values for I_A from a previous study that characterized I_A currents in rNST neurons under voltage clamp that employed both pharmacological and subtractive procedures to isolate the current (see (Chen et al. 2020)). From these recordings we obtained mean values of maximum I_A conductance (g_A), and half-max parameters for activation (θa) and inactivation (θb), as well as inactivation decay (τ_b) for G+ and G− neurons (Chen et al. 2020). These values were used to compute a synthetic I_A for real-time dynamic clamp.

The conductance-based model for I_A is given by I_A = g_A a³ b (V_M − E_K), where the gating variables a and b satisfy differential equations a’ = (a_∞ (VM) − a) /τa and b’ = (b_∞ (VM) − b) /τb, respectively. Here, $a_{\infty }\left(V_M\right)=1/(1+e^{-\frac{V_M-{\theta }_a}{{\sigma }_a}})$ and $b_{\infty }\left(V_M\right)=1/(1+e^{-\frac{V_M-{\theta }_b}{{\sigma }_b}})$. Parameter values for G+ (G−) cells are: 𝑔_𝐴 = 9 10⁻⁹(1.2 10⁻⁸) 𝑠/𝑐𝑚²; θ_𝑎 = −44(−52 ) 𝑚𝑉; θ_𝑏 = −69(−79) 𝑚𝑉; σ _𝑎 = 25.7(22.3);σ _𝑏 = −1(−3); τ_𝑏 = 140(136) 𝑚𝑆; τ_𝑎 = 2(2) 𝑚𝑆;𝐸_𝐾 = −100(−100) 𝑚𝑉.

As described at length in a previous publication, these parameter values recapitulated differences between G+ and G− neurons in the onset to AP initiation under current clamp (Chen et al. 2020).

We implemented dynamic clamp for I_A using the Linux-based Real-Time eXperimental Interface (RTXI v2.0) operating within Ubuntu 16.04.6 LTS (Xenial Xerus) on an Intel-Core i5–2500CPU @3.30Ghz × 4 processor equipped with an AMD CEDAR graphics card and National Instruments PCI-6221 Multifunction I/O card connected to the external input of the AM Systems Amplifier. Real-time computations occurred approximately every 0.25 microseconds, with peak computations taking no longer than 2.074 microseconds (RT Benchmarks Module, RTXI). Further details are provided in the Results. Software modules are available on Github (https://github.com/Trevor372/RTXI-DYNAMIC-IA).

Statistical analysis

Paired t-tests were used to evaluate differences before and after a single treatment such as optogenetic release of GABA or the introduction of a synthetic I_A under dynamic clamp. A repeated-measures ANOVA was appropriate when there were multiple treatments such as the impact of GABA before and after infusion of AmmTX3. The significance level was set at P = .05 (Systat v. 13). Exact P values are reported in the Figure Captions or the Results text for values greater than .001, otherwise P values are stated as <.001. Error bars in the figures are standard errors of the mean (SEMs).

Immunohistochemistry

At the conclusion of the experiment, tissue sections were fixed in 4% paraformaldehyde overnight then transferred to phosphate-buffered saline (PBS: 0.1 M, pH 7.4) the next morning. Slices were immunostained for the ionotropic purinergic receptor P2X2 to delineate the recording site relative to the terminal field of primary afferent fibers (Bartel 2012; Breza and Travers 2016). PBS rinses separated the following steps: Slices were treated with 1% sodium borohydride (20 min), then incubated in blocking serum (60 min, 0.3% Triton, 1% Bovine Serum Albumin, 5% Donkey Serum) prior to primary antibody incubation with a polyclonal rabbit P2X2 antibody (1:10,000, Alomone Labs, Catalog#: APR-003, Lot#: AN-1502), followed by a fluorescently tagged anti-rabbit secondary antibody (Alexa Fluor 546, Catalog #: A10040, Lot#: 1833519). Photomicrographs were taken with a DS-Ri1 Nikon camera attached to a Nikon Eclipse 600 microscope using brightfield and fluorescent illumination with appropriate fluorescent filters. These images were compared to the photomicrographs taken during recording. Locations of rNST cells were plotted on a representative section midway between the rostral pole of the nucleus and the level at which the nucleus is no longer adjacent to the 4^th ventricle. Locations were approximated by comparing the location of the recording electrode to the P2X2 field and other landmarks including the borders of the nucleus, the ST, and myelinated fibers coursing through the middle of the section.

Results

Pharmacological block of I_A increases the response to afferent stimulation

To assess the effects of ST stimulation before and after infusion of AmmTX3, we recorded from 25 G− neurons: 17 I_A+ and 8 I_A−. Based on a jitter criterion of < 200 μS as the cutoff for a monosynaptic response (Bailey et al. 2006; Peters et al. 2011), 21/25 cells were monosynaptically driven with a mean latency of 2.2 ± 0.13 ms and a mean jitter of 0.13 ± 0.01 ms. The other 4 cells (3 I_A+, 1 I_A−) had a mean latency of 3.0 ± 0.61 ms and mean jitter of 0.35 ± 0.1 ms. Across the population of 25 cells, there was no correlation between latency and jitter (P = .203), however when one outlier was removed, the correlation was significant (R = 0.53, P = .008).

In neurons with I_A, AmmTX3 produced a pronounced, significant increase in the response to afferent stimulation (compare Fig. 2A1, 2A3). The mean response in I_A+ cells to 20 Hz stimulation ranged from 0.7 to 19.0 spikes/s $(\overline{X}=6.8\pm 1.3)$, significantly larger than the response in I_A− cells $(\overline{X}=3.3\pm 0.8)$ (Fig. 2B). Across the 17 cells with I_A, AmmTX3 significantly increased the mean number of spikes by 53.3% (Fig. 2B) but did not affect the afferent response in 7 cells without I_A (3%, Fig. 2B right). Figure 2 also depicts the effects of GABAergic inhibition both before and after AmmTX3 (Fig. 2A2 and 2A4) but these effects are presented later in the manuscript. As expected from earlier work (Chen et al. 2020), AmmTX3 was highly effective in suppressing the delay to spike initiation (Fig. 3A top 2 traces), and for 10 cells, AmmTX3 nearly abolished the delays over a range of hyperpolarizing pre-pulses (Fig. 3B). For 7 cells that showed a long first ISI in response to the pre-pulse hyperpolarization protocol, AmmTX3 eliminated the longer ISI in 5 cells and shortened it in the other 2 (not shown).

Fig. 2.

A. Action potentials elicited from the same (I_A+) cell after different treatments. Action potentials elicited by afferent stimulation (A1) increased following drug treatment with AmmTX3 (A3). Following optogenetic release of GABA, responses were suppressed both before (A2) and after treatment with AmmTX3 (A4). Resting membrane potentials are shown prior to stimulation for each treatment. B. The mean number of APs elicited by afferent stimulation at 20 Hz for 1 s increased from 6.8 ± 1.3 spikes to 10.4 ± 1.5 spikes (53%) following AmmTX3 in cells with I_A (I_A+ t-test, P = .001, N = 17) but were unchanged in those without (I_A− $\overline{X}=3.3\pm 0.8$ spikes vs $\overline{X}=3.4\pm 7.3$ spikes; t-test, P = .915, N = 8). A repeated measures Power Test (Systat v. 13) yielded a power value of 0.75, suggesting sufficient power to test the effect of AmmTX3 in I_A− cells. Under control conditions (ACSF), the response to afferent stimulation was greater in I_A+ cells compared to I_A− cells (t-test, P = .028, N: I_A+ = 17, I_A− = 8). Mean and SEM indicated to the side of data points. The cell illustrated in A is marked with a star.

Fig. 3.

A. 4 traces from the same cell showing responses to 150 pA preceded by −60 pA pulse for 450 ms. A delay is evident under control conditions (ACSF) which is suppressed in the presence of AmmTX3. Under dynamic clamp, a delay is again evident, but not when dynamic clamp is turned off. B. Under dynamic clamp, systematic delays were observed in 7 cells: 3 lacking constitutive I_A and 4 in which I_A was blocked with AmmTX3 (blue line). These delays were comparable to those seen in 10 cells with I_A under control (ACSF) conditions (solid red line); ANOVA indicated no difference between delays induced by dynamic clamp and those with constitutive I_A (P = .823). In cells without I_A, turning dynamic clamp off eliminated the delay (dashed blue line, ANOVA: injected current, P < .001; dynamic clamp, P = .009; current X clamp, P < .001). AmmTX3 eliminated the delay in cells with constitutive I_A (dashed red line, ANOVA: injected current, P < .001; AmmTX3, P = .014; current X AmmTX3, P < .001). C. Schematic for implementing dynamic clamp. Membrane voltages (V_m) from cells either lacking constitutive I_A or I_A blocked with AmmTX3 were used to compute a synthetic I_A conductance using I_A parameters derived from empirical studies (Chen et al. 2020).

The increase in excitability to afferent stimulation in I_A+ cells after AmmTX3 could not be explained by alterations in the resting membrane potential (RMP) which did not change under the 2 conditions (ACSF: − 58.6 ±1.5 mV; AmmTX3: −56.2 ± 1.7 mV; P = .18, N = 17). It was notable, however, that following drug treatment, there were significant changes in action potential waveform that have been associated with changes in membrane excitability, including a 18.2% widening of the mean AP half-width (P = .008) and a 23.8% reduction in the mean after-hyperpolarization (AHP) (P = .008) (Table 1) (Carrasquillo and Nerbonne 2014; Connor and Stevens 1971a; Jerng et al. 2004; Liss et al. 2001; Tell and Bradley 1994).

Table 1.

Action Potential Waveform Before and After AmmTX3

	Control	AmmTX3	P
RMP (mV)	−58.6 (1.5)	−56.2 (1.7)	.18 N = 17
Peak (mV)	71.7 (2.8)	70.7 (3.9)	.7 N =12
AHP (mV)	−18.9 (2.0)	−14.4 (1.6)	.008 N = 12
Half-width (ms)	1.1 (.1)	1.3 (0.1)	.008 N = 12
MR (MΩ)	442.2 (43.7)	411.6 (51.7)	.366 N = 24

Values are means (SEM). RMP, resting membrane potential, AHP, after-hyperpolarization, MR, membrane resistance.

Injecting synthetic I_A with dynamic clamp reduces excitability to afferent stimulation

To further assess a causal role for I_A in modulating rNST responses to afferent stimulation, we implemented a dynamic clamp protocol to induce an I_A conductance (Fig. 3C). The parameters for this synthetic I_A were based on previous empirical studies in which we characterized the kinetic characteristics of I_A and showed that these parameters reproduced a delay to spike under dynamic clamp using a protocol of depolarization preceded by hyperpolarization (Chen et al. 2020). In the current study, the efficacy of the dynamic clamp was confirmed using a similar protocol (Fig. 3B). We recorded from 7 cells (5 G−, 2 G+) which either lacked constitutive I_A (N = 3) or in which I_A was blocked with AmmTX3 (N = 4). Dynamic clamp produced a delay in the latency to the first action potential which was not apparent in the absence of the dynamic clamp (Fig. 3A, B). These delays closely matched those in cells with constitutive I_A before and after treatment with AmmTX3 (Fig. 3B).

In addition to producing a delay to spike initiation, the synthetic I_A conductance also altered the response to afferent stimulation. The introduction of a synthetic I_A reduced the number of spikes to afferent stimulation and was reversed when dynamic clamp was turned off (Fig. 4A). For 11 cells, synthetic I_A significantly reduced the response to 20 Hz afferent stimulation by 12.8%. Indeed, in all but one cell, dynamic clamp reduced the number of spikes in response to afferent stimulation (Fig. 4B). Although the magnitude of the effect was small it was highly consistent across cells.

Fig. 4.

A. The response of a cell to afferent stimulation under control conditions (dynamic clamp off) was suppressed when a synthetic I_A was introduced under dynamic clamp. The response returned to normal when dynamic clamp was turned back off. Traces are continuous in time. B. Effects of a synthetic I_A conductance in 11 cells during afferent stimulation at 20 Hz. Six cells lacked constitutive I_A (including 2 G+ cells) and 5 had I_A blocked pharmacologically with AmmTX3 (including 1 G+ cell). Dynamic clamp parameters were adjusted for G+ and G− cells (see Methods). The effect of dynamic clamp produced a small but consistent suppression to afferent stimulation in 10/11 cells. (dynamic clamp off, $\overline{X}=7.0\pm 1.7$ spikes/s; dynamic clamp on, $\overline{X}=6.1\pm 1.7$ spikes/s; t-test: P = .004). Mean and SEM indicated to the side of data points. The cell illustrated in A is indicated by a star in B.

Impact of AmmTx3 on GABAergic suppression

We next asked whether GABA-induced suppression of ST-evoked responses was accentuated by the outward K+ current afforded by the activation of I_A channels. If true, neurons with I_A should show greater GABA suppression and blocking this current would result in a loss of inhibition during optogenetic GABA release. However, optogenetic release of GABA suppressed the response to afferent stimulation in both I_A+ and I_A− cells to a similar degree (Fig. 5). Moreover, although blocking I_A channels with AmmTX3 increased the response to afferent stimulation in I_A+ cells (Fig. 2A3), GABA was still effective in suppressing this response (Fig. 2A4, 6A). To more closely scrutinize the interaction between GABA and I_A, we evaluated GABA suppression in the two states by quantifying both percent suppression and the absolute reduction in the number of spikes in individual cells. Based on the percent suppression measure, there was an apparent decrease in the efficacy of optogenetic inhibition after blocking I_A (P = .025). However, this effect was dependent on two cells in which the response to afferent stimulation was entirely eliminated by GABA prior to AmmTX3, making the extent of the actual suppression unknown. Without these cells, the effect was in the same direction but not significant (Fig. 6B1). Indeed, the apparent decrease in GABA efficacy following treatment with AmmTX3 was likely due to the increased excitability of cells to afferent stimulation rather than a decrease in GABA suppression. When the magnitude of optogenetic inhibition was calculated as the difference in spike number (Fig. 6B2), AmmTX3 had no effect, i.e. blocking I_A did not change the efficacy of GABA.

Fig. 5.

A. The mean number of APs elicited by afferent stimulation at 20 Hz for 1 s decreased following optogenetic release of GABA for both cells with and without I_A (I_A+: $\overline{X}=6.8\pm 1.3$ spikes vs 3.4 ± 1.0 spikes; t-test, P < .001, N = 17; I_A−: $\overline{X}=3.3\pm 1.5$ spikes vs 1.5 ± 0.4 spikes, t-test, P = .034, N = 8). Mean and SEM indicated to the side of data points. Cell illustrated in figure 2A is indicated by a star. B. On average, the response to afferent stimulation was reduced 59% in I_A+ cells, whereas responses in cells without I_A were suppressed to a nominally smaller degree, 45% (P = .116).

Fig. 6.

A. Following optogenetic release of GABA (light) there was a significant reduction in the afferent-evoked response compared to no light (control) in I_A+ cells both before and after treatment with AmmTX3. (ANOVA: drug, P <.001; GABA, P < .001, drug × GABA, P = .372, N = 17). B1. Despite the lack of a significant interaction between GABA and AmmTX3 using spike counts, the mean percent suppression due to GABA was smaller following AmmTX3 (paired t-test, P = .025, N = 17). However, this only approached significance when 2 cells with a floor effect (prior to drug treatment) were removed (P = .069, N = 15). B2. Moreover, when the magnitude of inhibition was calculated as the difference in the absolute number of spikes before and after optogenetic stimulation (Δ spikes), AmmTX3 had no effect.

Location of cells responding to afferent stimulation

A majority (10/17) cells with I_A responding to afferent stimulation were in the central subdivision (Fig. 7). Three of the 8 in the medial subdivision expressed I_A, 2/5 in the ventral subdivision, and 2/2 in the lateral subdivision). Many of the cells responding to afferent stimulation tended to be near the central subdivision where most primary afferents terminate.

Fig. 7.

Photomicrograph of the rostral solitary nucleus stained for P2X2 which effectively demarcates the central subdivision (C). Incoming fibers of the solitary tract (st) were also labeled with P2X2. Many cells with I_A were located in the central subdivision (black symbols). Despite the failure to identify cells in the central subdivision without I_A in this sample (white symbols), such cells were recorded in our previous study (Chen et al. 2020). An asterisk designates damage to the tissue where a stimulating electrode was positioned adjacent to the solitary tract. Abbreviations: C, central subdivision; IV, 4^th ventricle; L, lateral subdivision; M, medial subdivision; V, ventral subdivision.

Discussion

Pharmacologically blocking Kv4 channels enhanced the response to afferent stimulation in a population of non-GABAergic rNST neurons. Although there is evidence for GABAergic feed-forward inhibition in the rNST, this inhibition does not appear to potentiate I_A to amplify the effects of inhibition, i.e. blocking Kv4 channels did not modulate GABAergic inhibition. These results suggest that GABAergic inhibitory synapses and the transient, rapidly inactivating outward K⁺ current (I_A) function independently to regulate the purely excitatory input of gustatory afferent fibers into the rNST.

Regulation of excitatory input by IA

I_A is well represented in the rNST, and pharmacological block with either 4-AP or the specific KV4 channel blocker AmmTX3 suppressed I_A currents (Chen et al. 2016; Tell and Bradley 1994). Indeed, I_A currents have been identified in both rNST projection neurons and GABAergic interneurons, although a subset of GABAergic neurons in the ventral subdivision appear not to express it (Corson and Bradley 2013; Suwabe and Bradley 2009; Wang and Bradley 2010b). In the present study, 17/25 G− cells expressed I_A (68%). This is similar to our previous report that 57% of G− cells express I_A (Chen et al. 2020) and a report from Bradley’s group that 72% of cells with projections to the PBN express IA (Suwabe and Bradley 2009). Because many of the non-GABAergic neurons with I_A that we recorded were in the central subdivision, the location of a prominent glutamatergic projection to the PBN (Gill et al. 1999; Halsell and Travers 1997; Whitehead 1990), many of these cells are likely PBN projection cells. Despite the robust presence of I_A in the rNST, a functional role in sensory processing has been unclear. In the cNST, I_A is more pronounced in neurons with projections to the hypothalamus compared to those with local brainstem reflex projections to the ventrolateral reticular formation (Bailey et al. 2007; Strube et al. 2015). Because local projection cells more faithfully followed afferent nerve stimulation compared to those with ascending projections, it was proposed that I_A provided cells that project to the hypothalamus with a mechanism for greater state-dependent modulation of visceral signals, compared to neurons with a purely reflex function (Bailey et al. 2007). However, in contrast to the results reported by Bailey et al., rNST neurons with I_A had larger (more faithful) responses to afferent stimulation than those lacking I_A. Indeed, cells expressing I_A were found not only in the central subdivision (with prominent ascending projections), but also in the ventral subdivision, where local reticular projection neurons are plentiful (Chen et al. 2020; Corson and Bradley 2013; Halsell et al. 1996; Suwabe and Bradley 2009). Thus, cells contributing to both pathways can potentially be modulated by this current. Neurons in both these subdivisions receive monosynaptic afferent input from fibers in the VIIth and IXth cranial nerves (Corson and Bradley 2013; Corson and Erisir 2013; Suwabe and Bradley 2009), and despite the large variation in the magnitude of the response to afferent stimulation (Figs. 2 & 3), latency and jitter measurements suggest that most of the postsynaptic responses we recorded were monosynaptic. The large variation in response magnitude may have resulted from natural variation in the connectivity or cellular properties of the neurons, as well as technical limitations due to the coronal slice preparation that likely reduced convergent input to 2^nd order cells, and the use of a single suprathreshold level of stimulation that does not take into account differences in neuron thresholds (Grabauskas and Bradley 1996; Wang and Bradley 1995).

That I_A is involved in regulating afferent input was shown by Hoffman and colleagues who demonstrated a gradient of I_A in hippocampal pyramidal neuron dendrites that increased from proximal to distal (Hoffman et al. 1997). Blocking I_A with 4-AP increased the amplitude of an evoked EPSP by over 50%, suggesting that IA channels acted as a “shock absorber” to regulate excitatory inputs (Hoffman et al. 1997). Although we were unable to measure EPSP amplitudes following treatment with AmmTX3 due to the presence of APs, we note that the increase in excitability reflected in AP frequency was, likewise, just over 50%. Similarly, Cai and colleagues showed that focal application of 4-AP at the branch point of terminal dendrites of hippocampal pyramidal cells facilitated AP firing (Cai et al. 2004). Although we observed a clear increase in firing frequency at 20 Hz following AmmTX3, additional studies are required to determine if blocking I_A would differentially alter firing frequencies at other rates of stimulation. It has been suggested, for example, that I_A might act as a low frequency filter to reduce small transient inputs to the NST (Corson and Bradley 2013). In addition to suppressing dendritic EPSPs, I_A channels influence excitability by altering AP waveform (Carrasquillo and Nerbonne 2014; Connor and Stevens 1971a; Imai et al. 2019; Jerng et al. 2004; Liss et al. 2001; Tell and Bradley 1994). Similar to a previous report using 4-AP to block I_A channels in rNST neurons (Tell and Bradley 1994), the specific I_A channel blocker AmmTX3 produced an increase in the AP half-width and a decrease in the AHP with no change in the resting membrane potential. Despite these changes in spike waveform, however, the results of our previous rNST study did not find a change in the AP threshold to injected current, nor an increase in the number of APs elicited by injected current following treatment with AmmTX3 (Chen et al. 2020). Thus, the increase in the responsiveness to afferent stimulation observed in the current study after blocking KV4 channels likely reflects a role for I_A in dendritic excitability that ultimately influences AP initiation.

An I_A window current, i.e., one that is operational at the resting membrane potential (Ficker and Heinemann 1992; Frolov et al. 2017; Kloppenburg et al. 1999), is evident in both the cNST and rNST by the overlap of activation and inactivation curves for isolated I_A currents (Bailey et al. 2002; Chen et al. 2020; Strube et al. 2015). Evidence that this mechanism can directly modulate excitability was shown by a shortened latency to AP initiation in response to a depolarizing stimulus in the absence of a hyperpolarizing pre-pulse (Chen et al. 2020). In the present study, the increase in responsiveness to afferent stimulation following treatment with AmmTX3 further suggests that an I_A current was active at rest and increasing an I_A conductance with dynamic clamp decreased the response to ST stimulation. Although the magnitude of the change under dynamic clamp is considerably smaller than the magnitude of the increase observed with the pharmacological block, dynamic clamp studies have demonstrated that I_A can influence AP waveform generated at the soma, (Ma and Koester 1996; Sakurai et al. 2006), but may be less suitable to mimic such currents in distal dendrites (Taylor et al. 2009).

Interaction between GABA and I_A

Bradley and colleagues were the first to propose a feedforward circuit in the rNST based on the presence of IPSPs in 2^nd order rNST neurons following afferent stimulation, as well as enhanced EPSPs when GABAergic inhibition was blocked (Grabauskas and Bradley 1996; Tell and Bradley 1994; Wang and Bradley 1995). In addition, there is evidence that rNST GABAergic neurons receive monosynaptic afferent input (Boxwell et al. 2013; Wang and Bradley 2010a). Thus, with a window current close to the RMP, a phasic inhibitory input might be expected to further de-inactivate I_A channels such that incoming excitatory afferent input that activates the outward K+ channel would add to the suppressive effect of the phasic inhibition itself (Dekin and Getting 1984; Strube et al. 2015). Despite the evidence for a feed-forward inhibitory circuit and a widespread substrate for both GABA inhibition and I_A currents in rNST, we saw little evidence for an interaction between these two regulatory mechanisms. When we factored out the general increase in excitability when I_A was blocked and calculated only the absolute number of spikes associated with inhibition, there was no evidence for decreased inhibition in the absence of functioning KV4 channels.

The lack of a dynamic interaction between I_A and GABAergic inhibition in the rNST may reflect the location of these respective channels and receptors. I_A channels in the hippocampus and cortex are located in close proximity to GABAergic synapses (Burkhalter et al. 2006; Jinno et al. 2005) and an interaction between these two mechanisms in the cortex was shown by blocking I_A in cortical pyramidal cells with 4-AP (Chang and Higley 2018). Likewise, an interaction between GABAergic inhibition and I_A was observed in midbrain dopaminergic neurons (Tarfa et al. 2017), which have inhibitory inputs to both dendritic and somal compartments (Edwards et al. 2017). We did not we see a decrease in the efficacy of GABAergic inhibition to suppress responses to afferent stimulation when I_A was blocked. However, in the rNST, KV4.3 channels predominate in the neuropil of the rNST (Chen et al. 2020) and may be concentrated on distal dendrites as seen in hippocampal pyramidal neurons (Hoffman et al. 1997; Trimmer and Rhodes 2004). Indeed distal dendrites are also the location of excitatory afferent inputs to the rNST that might activate KV4 channels, but are distant from GABAergic synapses that are dominant on the soma and more proximal dendrites (Whitehead 1993), making interactions less likely.

Although our evidence to date suggests that I_A and GABAergic pathways work independently to regulate rNST responses to afferent stimulation, this conclusion is subject to several methodological considerations. GABAergic synapses in the rNST derive from multiple sources including both intrinsic local and external sources. Our optogenetic activation of GABAergic neurons/terminals did not discriminate among these sources. As such, full-on release of GABA might have obscured more subtle interactions had only one source of GABA been activated. Additional studies focused on specific GABAergic pathways may clarify if there are any conditions under which GABA mediated hyperpolarization interacts with I_A. Nevertheless, despite the lack of a dynamic interaction between GABAergic inhibition and I_A, it seems likely that an I_A window current works in concert with GABAergic inhibition to exert major effects on afferent responses in the rNST.

Highlights.

1. I_A K+ currents suppress the response of rNST neurons to afferent stimulation.

2. GABAergic inhibition also suppresses afferent-induced responses.

3. GABAergic inhibition and I_A appear to operate independently.