Inspiratory rhythm generation is stabilized by I_h

Abstract

Cellular and network properties must be capable of generating rhythmic activity that is both flexible and stable. This is particularly important for breathing, a rhythmic behavior that dynamically adapts to environmental, behavioral, and metabolic changes from the first to the last breath. The pre-Bötzinger complex (preBötC), located within the ventral medulla, is responsible for producing rhythmic inspiration. Its cellular properties must be tunable, flexible as well as stabilizing. Here, we explore the role of the hyperpolarization-activated, nonselective cation current (I_h) for stabilizing PreBötC activity during opioid exposure and reduced excitatory synaptic transmission. Introducing I_h into an in silico preBötC network predicts that loss of this depolarizing current should significantly slow the inspiratory rhythm. By contrast, in vitro and in vivo experiments revealed that the loss of I_h minimally affected breathing frequency, but destabilized rhythmogenesis through the generation of incompletely synchronized bursts (burstlets). Associated with the loss of I_h was an increased susceptibility of breathing to opioid-induced respiratory depression or weakened excitatory synaptic interactions, a paradoxical depolarization at the cellular level, and the suppression of tonic spiking. Tonic spiking activity is generated by nonrhythmic excitatory and inhibitory preBötC neurons, of which a large percentage express I_h. Together, our results suggest that I_h is important for maintaining tonic spiking, stabilizing inspiratory rhythmogenesis, and protecting breathing against perturbations or changes in network state.

NEW & NOTEWORTHY The I_h current plays multiple roles within the preBötC. This current is important for promoting intrinsic tonic spiking activity in excitatory and inhibitory neurons and for preserving rhythmic function during conditions that dampen network excitability, such as in the context of opioid-induced respiratory depression. We therefore propose that the I_h current expands the dynamic range of rhythmogenesis, buffers the preBötC against network perturbations, and stabilizes rhythmogenesis by preventing the generation of unsynchronized bursts.

INTRODUCTION

The generation of rhythmicity is a fundamental property of the nervous system and is associated with many brain functions ranging from motor behaviors to cognition (1–3, 4, 5). A myriad of integrated cellular and network properties contributes to rhythmogenesis. The hyperpolarization-activated inward cation current (I_h) is one of the ion channels that plays a key role in the generation of rhythmic activity across multiple neural networks including the thalamus, striatum, and hippocampus (6–15). This voltage-dependent mixed cationic current is activated upon phasic hyperpolarization of neurons (16, 17). The activation begins near resting membrane potential of most neurons (−55 mV) and produces a slow depolarization (18). These properties allow I_h to play an important role in terminating hyperpolarization, initiating depolarization, and activating other critical voltage-dependent depolarizing currents, which is critical for mediating phase transitions. However, the role of this current is more diverse as it can also control synaptic processing in the dendrites (19–21) and as a “subthreshold” current it can add plasticity to neuronal networks by altering spike dynamics and synchronization (14, 22, 23).

Plasticity and the ability to reconfigure and adapt to changes in metabolic, environmental, and behavioral demands are critical for all neuronal networks (24–26, 27). Yet, this flexibility must be counterbalanced by the need to maintain stability (28). This is exemplified by the neuronal networks that control breathing, a vital rhythmic motor activity that must be dynamically regulated to achieve a wide range of flexibility while remaining robust (29). Here, we find that I_h plays a critical role in maintaining stability and flexibility of respiratory rhythmogenesis.

The respiratory rhythm is generated in the ventral respiratory column within the medulla. A group of neurons collectively known as the preBötzinger complex (preBötC) assemble in a network that is both necessary and sufficient to maintain inspiratory rhythm generation (4, 30–34). The preBötC contains a heterogeneous group of neurons that are differentially active throughout the respiratory cycle (29, 35, 36). This includes neurons active in phase with inspiration and expiration, as well as neurons that are silent or tonically active across all phases of respiratory activity (37). Mechanisms of rhythm generation have historically been debated (4, 38–40), but increasing evidence suggests that rhythmic burst activity emerges from interactions between synaptically coupled cells (41) with the contribution of burst-promoting conductances within a subset of neurons (27). Numerous voltage-dependent and ligand-gated channels contribute significantly to the generation and maintenance of rhythmic activity within the network (26, 42, 43). I_h is expressed in ∼50% of rhythmic cells within the synaptically intact network, and 86% of I_h-expressing rhythmic neurons seem to be capable of intrinsic bursting (44). To date, I_h was specifically studied in rhythmically active neurons that were undefined (44). The introduction of optogenetic and transgenic approaches allows for a more defined characterization of neurons within the preBötC. Herein, we sought to determine the distribution of I_h among genetically defined preBötC subpopulations. In addition, we tested whether removing I_h in vitro and in vivo renders the network more susceptible to inhibitory influences. We find that I_h plays a critical role in imbuing the respiratory network with stability against the inhibitory influence of neuromodulators such as opioids.

METHODS

Animal Strains

All experiments were performed on male and female adult and neonatal (p4–p12 for in vitro, p90–p120 for in vivo) mice bred at Seattle Children’s Research Institute in accordance with National Institutes of Health guidelines under a protocol approved by the Seattle Children’s Research Institute Animal Care and Use Committee. Mice were group-housed and given access to food and water ad libitum. Light/dark cycles were maintained at 12 h each and temperature was controlled at 22 ± 1°C.

Control experiments were performed using C57BL/6 mice. Studies including visual identification of DBX1⁺ neurons occurred using DBX1^{Cre-ERT2/Cre-ERT2}; Rosa26^Ai14/Ai14 mice on the C57BL/6 background, as described previously (45). This strain was obtained by first outcrossing Dbx1^CreERT2 mice on a CD1 background (46) [gift of Christopher Del Negro (The College of William and Mary)] to a C57BL/6 background. Subsequent mice were bred to B6.Cg-Gt(ROSA)26Sor^{tm14(CAG–tdTomato)Hze}/J (Ai14) mice (47) from Jackson laboratories (Stock No. 007914) for visualization of DBX1⁺ neurons. For optogenetic stimulation of cells expressing VGAT (vesicular GABA transporter), homozygous breeder lines for Vgat-ires-Cre were obtained from Jackson Laboratories (Stock No. 016962) and crossed with homozygous mice containing a floxed STOP channelrhodopsin2 fused to a EYFP (Ai32) reporter sequence donated by Dr. Hongkui Zeng (Allen Brain Institute, WA), as described previously (48).

In Vitro Medullary Slice Preparation and Recording

Horizontal medullary slices containing the ventral respiratory column were prepared from neonatal mice between P4–P12, as described previously (29, 49). Whole brainstems were dissected following rapid decapitation, and the dorsal surface glued onto an agar block cut at ∼15° angle. Brainstems were sectioned stepwise at 200 μm in the transverse plane until the facial nerves (VII) were visualized. The block was then reoriented to position the ventral surface upward with the rostral surface adjacent to the vibratome blade. The blade was aligned flush on the rostral end of the tissue with the ventral surface. After blade alignment, a single 850-μm section was obtained to generate the horizontal slice. The horizontal slice contains the ventral respiratory column extending from the spinal cord to the prominence of the facial nerve. Nuclei within the ventral portions of the medulla are preserved including the ventrolateral medulla and the postinspiratory complex (PiCO). Dissection and tissue sectioning occurred in ice-cold artificial cerebrospinal fluid (aCSF) with an osmolarity between 305–312 mosmol/kgH₂O, in mM: 118 NaCl, 3.0 KCl, 25 NaHCO₃, 1 MgCl₂, 1.5 CaCl₂, and 30 d-glucose equilibrated with carbogen (95% O₂, 5% CO₂). pH remained 7.4 ± 0.05 when equilibrated with carbogen.

Slices were placed in a recording chamber with circulating aCSF (15 mL/min, 30°C) for study. aCSF [K⁺] was elevated to 8 mM for all in vitro studies. Extracellular population recordings were obtained at the preBötC, as described previously (29, 49). The preBötC within horizontal slices is located ½–¾ of the distance between the midline and lateral edge of the slice at the level of the rostral end of central canal opening on an ∼850 μm brainstem slice. Extracellular population activity was recorded using glass suction pipettes (tip resistance <1 MΩ) filled with aCSF positioned on the surface of the slice. Signals were recorded at 10 kHz through an AC (alternating current) differential amplifier (A-M Systems), amplified by 10 K, bandpass-filtered between 300 Hz and 5 kHz, rectified, integrated, digitized (Digidata 1550 A, Axon Instruments), acquired in pClamp software (Molecular Devices), and analyzed post hoc with Clampfit software. For drug application, drugs were bath applied following a 30-min control period into the perfusion solution. ZD7288 (Tocris, Cat. No. 1000) was solubilized into deionized H₂O and applied at 100 µM for all studies. DAMGO ([d-Ala², N-MePhe⁴, Gly-ol]-enkephalin) (Tocris, Cat. No. 1171) was solubilized into deionized H₂O and applied at concentrations between 50 and 200 nM depending on study condition with 10 min between applications. CNQX (6-Cyano-7-nitroquinoxaline-2,3-dione) (Tocris, Cat. No. 0190) was solubilized into deionized H₂O and applied at concentrations between 0.001 and 20 µM depending on study condition.

The visual patch-clamp approach was used to record activity from single neurons. Recording electrodes were pulled from borosilicate glass (4–8 MΩ tip resistance) using a P-97 Flaming/Brown micropipette puller (Sutter Instrument Co., Novato, CA) and filled with intracellular patch electrode solution containing (in mM): 140 potassium gluconate, 1 CaCl₂, 10 EGTA, 2 MgCl₂, 4 Na₂ATP, and 10 HEPES (pH 7.2, osmolarity: 309–314 mosmol/kgH₂O). Neuronal activity was recorded in both current-clamp and voltage-clamp approaches with whole cell configuration using video-enhanced Dodt-IR optics and fluorescence for tdTomato (Ai14) on a Zeiss Axio Examiner.A1 microscope with a ×40 water immersion objective. Patch-clamp recordings were acquired using an Axopatch 1 D amplifier and pClamp software (Molecular Devices, Sunnyvale, CA), digitized at 10 KHz, and filtered at 2 KHz.

Identification of cells expressing DBX1 during visual-patch recording was confirmed by fluorescent labeling of the recorded cell from DBX1^{Cre-ERT2/Cre-ERT2}; Rosa26^Ai14/Ai14mice. VGAT+ cells were characterized using optogenetic stimulation from horizontal slices obtained from VGAT^cre/Ai32 mice. VGAT expression was assumed if cells showed activation of an inward current within voltage-clamp configuration following stimulation with blue light (473 nm laser illumination, 0.75 MW).

Neuropixel Recordings

In six slices, high-density Neuropixel probes (IMEC) (50) were used to simultaneously record many single neurons. The Neuropixel probes have been described in detail elsewhere (50). Briefly, they consist of 384 recordings sites packed along a 70 μm × 4 mm probe with a pitch of 20 μm, allowing for simultaneous, extracellular recording of many single neurons. Horizontal slices were prepared as described in In Vitro Medullary Slice Preparation and Recording. An extracellular population recording was first obtained to determine the approximate location of the preBötzinger complex as evidenced by rhythmic population bursting in 8 mM K⁺ aCSF. A single Neuropixel probe was inserted contralaterally to the population electrode, at approximately a 45° angle to the horizontal surface of the brain slice, through the preBötzinger complex. Approximately 100 of the 384 available channels were in contact with the brain tissue during recordings. Recordings from the Neuropixel probe and the integrated contralateral population recording were acquired with SpikeGLX (https://billkarsh.github.io/SpikeGLX v20200520). Neuropixel probes were set to AP gain = 500, LFP gain = 250, recorded from bank 0, and an external reference was soldered to a silver wire and submerged in the slice bath. Neuropixel traces were filtered first with a local common average reference filter, and spikes were sorted offline with Kilosort2 (51) to obtain spike times of simultaneously recorded, putative single neurons.

In Vivo Recordings and Microdialysis

Adult mice were anesthetized with isoflurane (3%) before urethane administration (1.5 mg/kg ip). Mice were positioned supine on a custom-heated surgical table to maintain body temperature ∼37°C throughout the protocol. The trachea was exposed from a ventral approach and cannulated with a curved (180°) tracheal tube (24 G) caudal to the larynx. Mice spontaneously breathed 100% O₂ throughout the remainder of the experimental protocol. The trachea and esophagus were removed rostral to the tracheal tube and underlying tissue was removed to expose the ventral aspect of the occipital bone. The portion of the occipital bone and underlying dura was removed over the ventral medullary surface. Following exposure, the ventral medullary surface was superfused with aCSF (∼37°C) equilibrated with carbogen (95% O₂, 5% CO₂). The hypoglossal nerve was isolated unilaterally, cut, and activity was recorded through a suction electrode connected to a pulled glass pipette filled with aCSF. Electrical activity from the XII nerve was amplified (10,000×), band-pass filtered (low pass 300 Hz, high pass 5 kHz), rectified, integrated, and digitized (Digidata 1550 A, Axon Instruments). Microdialysis probes (CMA 7, Harvard Bioscience) were inserted unilaterally into the preBötzinger complex based upon anatomic location referenced previously (48) (1.35-mm lateral from the basilar artery, 1.29-mm caudal to the caudal cerebellar artery, and 0.30-mm rostral to the intersection of the vertebral arteries). aCSF was dialyzed (3 µL/min) before ZD7288 (3 µL/min) solubilized into aCSF (100 µM). EMG electrodes were placed into the lateral aspect of the diaphragm to assess diaphragmatic activity. Anesthetic depth was monitored throughout via lack of heart rate and respiratory responses to toe pinch. Morphine was delivered through intraperitoneal (ip) administration at 150 mg/kg from a stock solution of 10 mg/mL (Patterson Vet.)

Computational Modeling

To investigate the role of and predict the effects of I_h block in respiratory rhythm and pattern generation, we added I_h into a contemporary computational model of the preBötC inspiratory network (52) that robustly reproduces a wide array of existing experimental data (53, 54) and has been recently experimentally validated (55). The voltage-dependent properties of I_h are adapted from Ref. 56. Individual neurons are simulated with a single compartment that incorporates Hodgkin-Huxley style conductances. The membrane potential of each neuron is governed by the following differential equation:

$$C_m\frac{d{V}_m}{dt}+I_{Na}+I_K+I_{NaP}+I_{Ca}+I_{CAN}+I_h+I_{Leak}+I_{Syn}+I_{GIRK}=0,$$

where C_m = 36 pF is the membrane capacitance and each I_i represents a current, with i denoting the current’s type. The currents include the action potential-generating Na⁺ and delayed-rectifying K⁺ currents (I_Na and I_K), persistent Na⁺ current (I_NaP), voltage-gated Ca²⁺ current (I_Ca), Ca²⁺-activated nonselective cation (CAN) current (I_CAN), the hyperpolarization-activated inward cation current (I_h), K⁺-dominated leak current (I_Leak), excitatory synaptic current (I_Syn), and µ-opioid receptor-activated G protein-coupled inwardly rectifying K⁺ leak current (I_GIRK). The channel currents are defined as follows:

$$I_{Na}={\overline{g}}_{Na}·m_{Na}^3·h_{Na}·(V_m-E_{Na})$$

$$I_K={\overline{g}}_K·m_K^4·(V_m-E_K)$$

$$I_{NaP}={\overline{g}}_{NaP}·m_{NaP}·h_{NaP}·(V_m-E_{Na})$$

$$I_{Ca}={\overline{g}}_{Ca}·m_{Ca}·h_{Ca}·(V_m-E_{Ca})$$

$$I_{CAN}={\overline{g}}_{CAN}·m_{CAN}·(V_m-E_{CAN})$$

$$I_h=g_h·m_h·\left(V_m-E_h\right)$$

$$I_{Leak}={\overline{g}}_{Leak}·(V_m-E_{Leak})$$

$$I_{Syn}=g_{Syn}·\left(V_m-E_{Syn}\right)$$

$$I_{GIRK}=g_{GIRK}·\left(V_m-E_K\right)$$

where g_i is the maximum conductance, E_i is the reversal potential, m_i and h_i are voltage-dependent gating variables for channel activation and inactivation, respectively, and i∈{Na, K, NaP, Ca, CAN, H, Leak, Syn, GIRK}. The parameters g_i and E_i are given in Table 1.

Table 1.

Model parameter values

Channel	Parameters
INa	gNa = 150.0 nS, ENa = 55.0 mV,
	$V_{m_{1/2}}$ = −43.8 mV, km = 6.0 mV,
	$V_{\tau m_{1/2}}$ = −43.8 mV, $k_{{\tau }_m}$= 14.0, ${\tau }_{m_{max}}$= 0.25 ms,
	$V_{h_{1/2}}$ = −67.5 mV, kh = −10.8 mV,
	$V_{\tau h_{1/2}}$ = −67.5 mV, $k_{{\tau }_h}$= 12.8 mV, ${\tau }_{h_{max}}$= 8.46 ms
IK	gK = 160.0 nS, EK = −94.0 mV,
	Aα = 0.01, Bα = 44.0 mV, κα = 5.0 mV
	Aβ = 0.17, Bβ = 49.0 mV, κβ = 40.0 mV
INaP	gNaP ∈ [0.0,5.0] nS,
	$V_{m_{1/2}}$= −47.1 mV, km = 3.1 mV,
	$V_{\tau m_{1/2}}$ = −47.1 mV, $k_{{\tau }_m}$ = 6.2, ${\tau }_{m_{max}}$ = 1.0 ms,
	$V_{h_{1/2}}$ = −60.0 mV, kh = −9.0 mV,
	$V_{\tau h_{1/2}}$ = −60.0 mV, $k_{{\tau }_h}$= 9.0 mV, ${\tau }_{h_{max}}$= 5,000 ms
ICa	gCa = 0.1 nS, ECa = R·T/F·ln([Ca]out/[Ca]in),
	R = 8.314 J/(mol·K), T = 308.0 K,
	F = 96.485 kC/mol, [Ca]out = 4.0 mM
	$V_{m_{1/2}}$= −27.5 mV, km = 5.7 mV, τm = 0.5 ms,
	$V_{m_{1/2}}$= −52.4 mV, kh = −5.2 mV, τh = 18.0 ms
ICAN	gCAN ∈ [0.5,1.5] nS, ECAN = 0.0 mV,
	Ca1/2 = 0.00074 mM, n = 0.97
Ih	gh = 2.0 nS, Eh = −40 mV,
	$V_{m_{1/2}}$ = −78 mV, km = 7 mV, τm = 500 ms,
ILeak	gLeak = 2.75 nS, ELeak = −68.0 mV,
Cain	αCa = 2.5·10−5 mM/fC, PCa = 0.01, Camin = 1.0·10−10 mM, τCa = 50.0 mS
ISyn	GTonic = 0 – 0.5 nS, ESyn = −10.0 mV, τSyn = 5.0 ms
IGIRK	gGIRK = 0 – 0.1 nS

The channel kinetics, intracellular Ca²⁺ dynamics, and the corresponding parameter values were derived from previous models and experimental studies (see Refs. 56, 57, and the references therein).

For I_Na, I_K, I_NaP, and I_Ca, the dynamics of voltage-dependent gating variables m_i, and h_i are defined by the following differential equation:

$${\tau }_{\eta }\left(V\right)·\frac{d\eta }{dt}={\eta }_{\infty }\left(V\right)-\eta ;\eta \in \{m_i,h_i\}$$

where steady-state activation/inactivation η_∞ and time constant τ_η are given by:

$${\eta }_{\infty }\left(V\right)={\left(1+e^{-(V-V_{{\eta }_{1/2}})/k_{\eta }}\right)}^{-1}$$

$${\tau }_{\eta }(V)={\tau }_{{\eta }_{max}}/\text{cosh }((V-V_{\tau {\eta }_{1/2}})/k_{{\tau }_{\eta }}).$$

For the voltage-gated potassium channel, steady-state activation m_K∞(V) and time constant τ_mK(V) are given by:

$$m_{K\infty }\left(V\right)=\frac{{\alpha }_{\infty }\left(V\right)}{{\alpha }_{\infty }\left(V\right)+{\beta }_{\infty }\left(V\right)}$$

$${\tau }_{mK}\left(V\right)=1/[{\alpha }_{\infty }\left(V\right)+{\beta }_{\infty }\left(V\right)]$$

where

$${\alpha }_{\infty }\left(V\right)=A_{\alpha }·(V+B_{\alpha })/\{1-\exp \left[-\frac{V+B_{\alpha }}{{\kappa }_{\alpha }}\right]\}$$

$${\beta }_{\infty }\left(V\right)=A_{\beta }·\mathrm{e}\mathrm{x}\mathrm{p}[-(V+B_{\beta })/{\kappa }_{\beta }].$$

The parameters $V_{{\eta }_{1/2}}$, $V_{\tau {\eta }_{1/2}}$, κ_η, ${\kappa }_{\tau \eta }{\tau }_{{\eta }_{max}}$, A_α, A_β, B_α, B_β, κ_α, and κ_β are given in Table 1. I_CAN activation is dependent on the intracellular calcium concentration [Ca]_in and is given by:

$$m_{CAN}=1/(1+{(C{a}_{1/2}/{\left[Ca\right]}_{in})}^n).$$

The parameters Ca_1/2 and n, given in Table 1, represent the half-activation calcium concentration and the Hill coefficient, respectively.

Calcium enters the neurons through voltage-gated calcium channels (I_Ca) and/or synaptic channels, where a percentage (P_Ca) of the synaptic current (I_Syn) is assumed to consist of Ca²⁺ ions. A calcium pump removes excess calcium with a time constant τ_Ca and sets the minimum calcium concentration Ca_min. The dynamics of [Ca]_in is given by the following differential equation:

$$\frac{d{\left[Ca\right]}_{in}}{dt}=-{\alpha }_{Ca}(I_{Ca}+P_{Ca}·I_{syn})-({\left[Ca\right]}_{in}-C{a}_{min})/{\tau }_{Ca}.$$

The parameter α_Ca is a conversion factor relating current and rate of change in [Ca]_in, see Table 1 for parameter values.

Simulations of the preBötC network consisted of 100 neurons coupled through excitatory synapses. The synaptic conductance of the i^th neuron ($g_{Syn}^i$) in the population is described by the following equation:

$$g_{Syn}^i=g_{Tonic}+{\displaystyle \sum }_{j,n}{w}_{ji}·C_{ji}·H(t-t_{j,n})·e^{-(t-t_{j,n})/{\tau }_{syn}}$$

where g_Tonic represents the tonically active component of the synaptic conductance, w_ji is the weight of the synaptic connection from cell j to cell i, C is a connectivity matrix (C_ji = 1 if neuron j makes a synapse on neuron i, and C_ji = 0 otherwise), H(.) is the Heaviside step function, t is time, τ_Syn is the exponential decay constant, and t_j,n is the time at which an action potential n is generated in neuron j and reaches neuron i.

Heterogeneity is introduced into the network by randomly assigning g_NaP and g_CAN based on a uniform distribution over the ranges [0.0,5.0]nS and [0.5,1.5]nS, respectively. In addition, the weight of each synaptic connection was uniformly distributed over the range w_ji ϵ [0,W_max] where W_max = 0.032 nS. The elements of the network connectivity matrix, C_ji, are randomly assigned values of 0 or 1 such that the probability of any connection between neuron j and neuron i being 1 is equal to the network connection probability P_Syn = 0.15.

Data Analysis and Definitions

The time of an action potential was defined as when the membrane potential of a neuron crosses −35 mV in a positive direction. The network burst frequency was determined by identifying peaks and calculating the inverse of the interpeak interval in the integrated population voltage.

Numerical Integration Methods

All simulations were performed locally on an 8-core computer running the Ubuntu 20.04 operating system. Simulation software was custom written in C++ and compiled with g++ version 9.3.0. Numerical integration was performed using the exponential Euler method with a fixed step-size (dt) of 0.025 ms. In all simulations, the first 10 s of simulation time was discarded to allow for the decay of any initial condition-dependent transients.

Statistical Analysis

Differences in burst frequency, burstlet frequency, composite rhythm frequency, slope of tonic activity following drug application, membrane potentials, and burst characteristics were assessed with paired two-tailed t tests and linear regressions. Changes in integrated tonic activity and total tonic spike activity were assessed with one-way repeated-measures ANOVA with Sidak’s multiple comparisons. Differences across conditions in burst frequency and breathing following increasing concentrations of DAMGO, CNQX, and morphine were assessed with two-way repeated-measures ANOVA with Sidak’s multiple comparisons. All statistical analyses were conducted through GraphPad Prism.

RESULTS

In Silico Modelling of Blocking I_h in the PreBötC Network

We first explored the role of I_h in respiratory circuits from a theoretical perspective. To specifically test the effects of I_h blockade, we incorporated I_h into an existing, well-established computational model of the preBötC (58) that robustly reproduces a wide array of experimental data (52, 55, 58). The voltage-dependent properties of I_h were adapted from Ref. 56 and are illustrated in Fig. 1A. At resting membrane potentials, I_h is only weakly activated (Fig. 1A) and generates a small depolarizing current (Fig. 1B). Unsurprisingly, if I_h is present, hyperpolarization activates I_h resulting in a depolarizing voltage “sag” during the negative current injection (Fig. 1C). Including I_h into this model did not impact the patterns of rhythmic activity produced by the model either at the level of an isolated neuron (Fig. 1, D1 and D2) or at the level of the integrated network (Fig. 1, E1 and E2). However, I_h decreased the level of excitability (controlled by the parameter g_Tonic) required to generate the same burst frequency (Fig. 1, D2 and E2). This shift is due to the depolarizing current generated by I_h. Therefore, our model predicts that a block of I_h should result in hyperpolarization at the cellular level and a decrease in respiratory frequency at the network level.

Figure 1.

Computational modeling of I_h into the PreBötC network. The impact of I_h on preBötC rhythmicity and the predicted effects of I_h block in a computational model. A: voltage-dependence of steady-state I_h activation and current (g_h = 2 nS). B: example traces showing the magnitude of I_h in an isolated intrinsically bursting neuron g_Tonic = 0.38 nS, g_NaP = 4.0 nS. C: voltage responses to negative current injections (−20 pA steps) in a simulated neuron with (g_h = 2.0 nS) and without (g_h = 0 nS) I_h. Notice the predicted 1.5 mV hyperpolarization and loss of the depolarizing voltage “sag,” with I_h removed (I_h−). Voltage traces illustrating the dependence of an isolated neuron (D1) and integrated network activity patterns (E1) on I_h and excitability, controlled here by the parameter g_Tonic. D2 and E2: the relationship between g_Tonic and burst frequency in an isolated neuron and network. Notice that loss of I_h is predicted to decrease burst frequency and increase the level of excitability required to generate bursting. Simulated effect of opioid activation of GIRK (G protein-coupled inwardly-rectifying potassium) channels (F) and CNQX block of excitatory synapses (G) in the model preBötC network with I_h intact (control) and I_h blocked. *P < 0.05. I_h, hyperpolarization-activated inward cation current; PreBötC, preBötzinger complex.

Next, we characterized how loss of I_h impacts the sensitivity of the model network to opioids or CNQX. Simulation of opioid application was achieved by introducing an opioid-activated suppression of presynaptic strength and a hyperpolarizing current carried by an inwardly rectifying potassium channel (I_GIRK) (see methods for full description) (35, 55). Increasing concentrations of opioids were simulated by concurrently decreasing synaptic strength and increasing the potassium channel conductance (g_GIRK). Our simulations predict that the preBötC network will be more sensitive to opioids after I_h block (Fig. 1F). This is due to the hyperpolarization caused by the loss of I_h. Finally, the block of excitatory synapses by CNQX was simulated by blocking synaptic currents (I_Syn) in the model. Our simulations predict that I_h block will only slightly increase the sensitivity of the preBötC network to CNQX (Fig. 1G).

I_h Block Increases Burstlet Frequency and Has Time-Dependent Effects within the In Vitro PreBötC Network

To test the predictions of our model network, we assessed the effects of removing I_h from the far more complex in vitro PreBötC network using pharmacological blockade of I_h within rhythmically active horizontal medullary slices. Under typical control conditions, the preBötC reliably produces regular population-wide burst activity (59) during which most inspiratory-modulated preBötC neurons exhibit a synchronized bout of action potentials. However, occasionally the network generates population bursts with smaller amplitudes that have been attributed to incomplete synchronization of the network and lower frequency action potential firing across respiratory neurons (4). These bursts, known as burstlets (59), are easily distinguished from normal bursts by their significantly smaller amplitudes. Figure 2A demonstrates the phenotypic differences between bursts and burstlets, as well as the separation in amplitudes observed when plotting the normalized values across animals (Fig. 2A).

Figure 2.

ZD7288 increases the occurrence of incomplete synchronized “burstlets” from the PreBötC network. Representative trace from integrated PreBötC activity displaying the increase in burstlets (red arrows) following I_h current block with 100 µM ZD7288 (A). Right displays the normalized amplitude of PreBötC activity across 17 animals following 100 µM ZD7288. Notice the clear distinction between full synchronized bursts (black) and burstlets (red). There was minimal effect of I_h block on the frequency of fully synchronized bursts from the PreBötC (P = 0.35, n = 17, paired t test) (B, left). However, there was a small, but significant frequency-dependent effect on burst frequency primarily contributed to by the few slowest and fastest baseline rhythms (P < 0.01, R²= 0.65, n = 17, simple linear regression) (B, middle, right). Despite the minimal change in burst frequency, there was a significant (P < 0.01, n = 17, paired t test) increase in the burstlet fraction and burstlet frequency (P < 0.01, n = 17, paired t test) following 100 µM ZD7288 (C, left, middle). The increase in burstlets is also visualized from the combined histograms of burst amplitude before and after treatment with 100 µM ZD7288 (C, right) (n = 17, shaded area represents SD). Due to the minimal change in burst frequency and increase in burstlets, there was a significant increase in the composite rhythm (P < 0.01, n = 17, paired t test) (D, left, middle), which did not display a baseline frequency effect (P = 0.14, R²= 0.14, n = 17, simple linear regression) (D, right). Interestingly, I_h block induced a biphasic response in the baseline tonic spiking recorded from the integrated PreBötC activity (E, left). Following 100 µM ZD7288, there was a significant increase (P < 0.01, n = 17, one-way repeated-measures ANOVA) in baseline activity 5 min after drug application. However, this was followed by a significant reduction thereafter following 10 min after drug application (P < 0.01, n = 17, one-way repeated-measures ANOVA) (E, middle, right). *P < 0.05. PreBötC, preBötzinger complex.

Application of the I_h antagonist ZD7288 in horizontal slices showed minimal change in the frequency of fully synchronized population bursts, but significantly increased the number of burstlets recorded from the PreBötC network (Fig. 2, B and C). This was also demonstrated by an increase in the fraction of burstlets relative to normal bursts, as well as the total “composite” frequency (burstlets plus bursts) of the inspiratory rhythm (Fig. 2C). Figure 2C (right) displays the average histogram of composite frequency recorded over a 1-min period. ZD7288 induces a bimodal distribution within the histogram of total activity, further demonstrating the increase in burstlet activity after I_h block. Due to the increase in burstlet frequency, the combined activity of the bursts and burstlets (composite rhythm) shows a significant increase following ZD7288, which occurs independent of the baseline population frequency (Fig. 2, C and D). These results conflict with the predictions of our in silico model, suggesting that the (in)direct consequences of I_h blockade on preBötC neurons are likely more complex than hyperpolarization.

In addition to the effects on rhythmic activity, bath application of ZD7288 had a notable biphasic effect on the level of integrated tonic activity recorded from the preBötC, quantified as a change in the amplitude of the baseline activity between rhythmic bursts on the integrated preBötC recording. The baseline activity represents the tonic spiking throughout the network during the interburst interval. Following I_h blockade, tonic activity was initially increased at 5 min and subsequently suppressed by 10 min (Fig. 2E). However, aside from changes in tonic activity, effects on the inspiratory rhythm did not show a similar biphasic response to I_h blockade (Supplemental Fig. S1; see https://doi.org/10.6084/m9.figshare.19763527.v1).

Preconditioning with ZD7288 Increases Sensitivity of the preBötC to Opioids and CNQX

To explore whether inhibition of I_h sensitizes the respiratory network to manipulations that suppress respiratory rhythmogenesis, we exposed rhythmic horizontal brainstem slices to increasing doses of the μ-opioid receptor agonist DAMGO ([d-Ala², N-MePhe⁴, Gly-ol]-enkephalin). Bath application of 50, 100, and 200 nM DAMGO under baseline conditions resulted in a 28 ± 8%, 51 ± 8%, and 64 ± 7% reduction in burst frequency, respectively (Fig. 3A). Following pretreatment with ZD7288, baseline burst frequency was unchanged. However, subsequent application of the same concentrations of DAMGO resulted in a 43 ± 13%, 86 ± 8%, and 97 ± 3% reduction in burst frequency, respectively (Fig. 3A). Similar results were observed following inhibition of glutamatergic signaling with the AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) receptor antagonist CNQX (6-Cyano-7-nitroquinoxaline-2,3-dione). Under control conditions, bath application of increasing concentrations of CNQX reduced burst frequency by 8 ± 16%, 3 ± 18%, 10 ± 17%, 14 ± 18%, 27 ± 18%, 54 ± 18%, 98 ± 2%, and 100 ± 0% at 0.01, 0.05, 0.1, 0.5, 1.0, and 10 µM CNQX, respectively (Fig. 3B). Pretreatment with 100 µM ZD7288 increased the sensitivity of the inspiratory rhythm to CNQX, such that burst frequency was reduced by 16 ± 5%, 60 ± 13%, 84 ± 18% in 0.01, 0.05, and 0.1 µM CNQX, respectively (P < 0.01). Thereafter, all brainstem slices pretreated with 100 µM ZD7288 were silenced following 0.5 µM CNQX (Fig. 3B). This suggests that I_h stabilizes the respiratory network against the suppressive effects of opioids or reduction in synaptic excitation. These results agree with our model predictions, although the magnitude of hypersensitivity induced by loss of I_h was much higher in the in vitro network; again, indicating that I_h blockade likely leads to more complex effects in preBötC neurons than hyperpolarization.

Figure 3.

ZD7288 increases sensitivity of the PreBötC to opioids and CNQX. Bath application of DAMGO onto rhythmic horizontal brainstem slices results in a dose-dependent reduction in burst frequency (A, top). However, pretreatment with 100 µM ZD7288 results in a significantly (P = 0.02, n = 28 control, n = 12 ZD7288 treated, two-way repeated-measures ANOVA) greater reduction in burst frequency following DAMGO administration (A, top). In addition to the greater reduction in burst frequency, pretreatment with ZD7288 results in a significantly (P = 0.01, n = 28 control, n = 12 ZD7288 treated, paired t test) greater reduction in the tonic activity measured as the Δamplitude of interburst spiking/Δ[DAMGO] from the integrated PreBötC population activity following DAMGO application (A, bottom). Similar results were found following blockade of excitatory synaptic transmission with CNQX (B). Pretreatment of rhythmic slices with 100 µM ZD7288 resulted in a significantly (P < 0.01, n = 5, two-way repeated-measures ANOVA) greater reduction in burst frequency following bath application of CNQX, compared with control (B, top). In addition, pretreatment with ZD7288 resulted in a significantly (P = 0.02, n = 5, paired t test) greater reduction in the tonic activity measured as the Δamplitude of interburst spiking/Δ[DAMGO] from the integrated PreBötC recording, following CNQX application (B, bottom). Following ventral brainstem visualization, microdialysis probes were inserted into the PreBötC unilaterally into anesthetized mice and respiratory activity was assessed with diaphragmic EMG and XII nerve rootlet activity (C). Dialysis of artificial cerebrospinal fluid (aCSF) into the PreBötC had minimal effect on respiratory frequency (D) (P = 0.99, Sidak’s multiple comparison from two-way repeated-measures ANOVA). Following aCSF dialysis, animals were either dialyzed with 100 µM ZD7288 (ZD7288 treated) or continued with aCSF dialysis (control). Dialysis of ZD7288 had minimal effects on recorded respiratory activity following 20 min (D) (P = 0.86, Sidak’s multiple comparison from two-way repeated-measures ANOVA). ZD7288 or aCSF dialysis continued while 150 mg/kg morphine was administered intraperitoneally. Administration of morphine resulted in a significant reduction in respiratory frequency of mice receiving control aCSF dialysis (D) (P = 0.04, Sidak’s multiple comparison from two-way repeated-measures ANOVA). However, morphine administration to animals receiving dialysis of ZD7288 resulted in a fatal apnea in all animals (D) (P = 0.03, Sidak’s multiple comparison from two-way repeated-measures ANOVA). Shaded areas in D show SE. Representative traces showing diaphragmatic, and XII activity are shown in E. *P < 0.05. PreBötC, preBötzinger complex.

The increased sensitivity to the effects of opioids and CNQX, following exposure to ZD7288 was associated with a greater reduction of integrated tonic activity during opioid and CNQX exposure. The slope of the change in baseline amplitude relative to the change in [DAMGO] was 207 ± 17% greater following pretreatment with ZD7288, compared with control (Fig. 3A). Similarly, the slope of the change in baseline amplitude relative to the change in [CNQX] was 373 ± 128% greater following pretreatment with ZD7288, compared with control (Fig. 3B). These results suggest that I_h plays a critical role in the generation of tonic activity, which may be important for the network to maintain rhythmic activity in the presence of depressive or destabilizing influences.

The in vivo respiratory network represents a much more complex system compared with the in vitro slice preparation. For example, interdependence among neuromodulators can compensate for the lack of a single neuromodulator by concurrent shifts in the concentrations of others (60, 61). This creates a more robust system that is less susceptible to perturbation. To test whether the in vivo respiratory network becomes more susceptible to OIRD (opioid-induced respiratory depression) following I_h block, we microdialyzed either ZD7288 or artificial cerebrospinal fluid (aCSF) into the preBötC during systemic administration of 150 mg/kg morphine (Fig. 3C). Microdialysis of aCSF alone had no effect on the frequency of respiratory output as measured in the diaphragm EMG. Similarly, microdialysis of ZD7288 alone had no effect on the rhythmic EMG activity. However, injection of 150 mg/kg morphine during aCSF microdialysis reduced respiratory frequency by 42 ± 22% (Fig. 3D), whereas the same morphine injection during ZD7288 microdialysis lead to terminal apnea in all mice (Fig. 2, D and E). These results indicate that the loss of I_h renders respiratory rhythmogenesis more susceptible to opioid-induced respiratory depression even in vivo.

Characterization of I_h across Neurons of the preBötzinger Complex

Neurons in the preBötC are heterogeneous with distinct functional roles in the control of inspiratory rhythmogenesis (35, 62, 4). Thus, we sought to determine the cell-type-specific distribution of I_h within the preBötC by using an optogenetic approach to distinguish between excitatory (Dbx1) and inhibitory (Vgat) neurons. Dbx1 is a transcription factor expressed in excitatory neurons that form the core inspiratory oscillator of the preBötC (63, 64), whereas Vgat is expressed in GABAergic and glycinergic neurons (48, 65, 66).

Whole cell patch-clamp recordings were simultaneously performed with contralateral extracellular population recordings to determine whether the neurons exhibited rhythmic, tonic, or silent spiking activity. The presence of I_h was indicated by depolarizing sag potentials in response to negative current injections (Fig. 4A). In addition, the activation potential of I_h was determined to be approximately −50 mV by recording in voltage-clamp configuration during negative voltage steps from a holding potential of −30 mV (Fig. 4B). We found that 20% (13/69) of Dbx1+ neurons were rhythmically active in phase with the integrated inspiratory rhythm, whereas 55% (38/69) were tonically active and 25% (17/69) were silent throughout all phases of the respiratory cycle (Fig. 4C). Among the Dbx1+ neurons, tonically active neurons were the most likely to have I_h (68%; 26/38), whereas I_h was present in 41% (7/17) of silent Dbx1+ neurons and 38% (5/13) of rhythmic Dbx1+ neurons. Among inhibitory Vgat+ neurons, 13% (3/24) displayed rhythmic activity in phase with the inspiratory rhythm, 62% (15/24) had a tonic spiking pattern, and 25% (6/24) remained silent. Similar to the Dbx1+ neurons, I_h was most prevalent among tonically active Vgat+ (67%, 10/15), with less expression in silent (50%, 3/6) and rhythmic (33%, 1/3) Vgat+ neurons (Fig. 4C). This is consistent with previous reports of the prevalence of I_h in nonidentified rhythmic neurons of the preBötC (44) (Fig. 4C), but slightly higher than the 20% reported by Picardo et al. (64). Importantly, our results suggest that I_h is evenly distributed among excitatory and inhibitory preBötC neurons, with a surprisingly high prevalence among tonic neurons.

Figure 4.

Characterization of I_h across different types of preBötC neurons. Intracellular patch-clamp recordings were obtained with simultaneous population activity from the contralateral PreBötC to determine the pattern of single-cell activity (A, left, middle). The presence of I_h was confirmed by the emergence of a sag potential following hyperpolarizing current injections (A, right). Activation of I_h was estimated to be ∼ 50–60 mV based upon the presence of an inward current following hyperpolarizing voltage steps from a holding potential of −30 mV (B). The distribution of cells expressing I_h among DBX1⁺ excitatory cells and VGAT⁺ inhibitory cells within the PreBötC revealed that the greatest expression was within tonically active cells throughout the network (C). PreBötC, preBötzinger complex.

ZD7288 Preferentially Suppresses Tonic Spiking within the preBötC

Our findings that loss of I_h decreases integrated tonic activity at the network level (see Figs. 2 and 3) and that I_h is most prevalent in tonically active preBötC neurons, suggest that tonic spiking activity may be preferentially suppressed by the blockade of I_h. To test this more directly, we recorded from tonic neurons in whole cell configuration to determine expression of I_h and the corresponding changes in spiking activity elicited following application of ZD7288. In tonic neurons with I_h (I_h⁺), sag potentials were eliminated as expected (Fig. 5), and tonic spiking activity was suppressed, suggesting I_h plays an important role in its generation. However, to our surprise, the suppression of tonic spiking activity was associated with an unexpected depolarization in membrane potential (+12 ± 2 mV depolarization) (Fig. 5). This depolarization could explain the biphasic effects of ZD7288 observed at the population level (see Fig. 2E), since spiking activity increases initially as these tonic neurons begin to depolarize but then becomes silent as they enter depolarization block. Importantly, this effect was specific to I_h+ neurons as ZD7288 did not change the spiking activity or V_m of I_h− tonic neurons (Fig. 5). These results indicate that reduced tonic spiking activity within the preBötC largely occurs due to depolarization block of the tonically active I_h⁺ cells within the network. Our results suggest that the loss of I_h has a paradoxical depolarizing effect on intrinsic activity. To the best of our knowledge, such a depolarization has not been described before, however, it is consistent with the increased excitability reported in bursting respiratory neurons (44). These unexpected findings also point to novel I_h-dependent mechanisms not yet captured within our in silico preBötC model, contributing to its inability to reproduce some of the effects of I_h block we observed experimentally.

Figure 5.

ZD7288-induced suppression of tonic spiking activity is paradoxically associated with depolarization. Following bath application of ZD7288, tonic spiking within I_h⁺-expressing neurons became silenced (top). I_h block was confirmed following 100 µM ZD7288 by the loss of sag potential from hyperpolarizing current steps (top). Interestingly, application of ZD7288 resulted in a 11.4 ± 2 mV depolarization of membrane potential (P < 0.01, n = 5, paired t test). In contrast, tonic spiking from cells lacking I_h were unaffected (bottom) (P = 0.22, n = 4, paired t test). *P < 0.05.

To further explore how blocking I_h alters spiking activity within the network, we used high-density Neuropixel probes to simultaneously record from many neurons. This approach allowed us to assess the tonic spike activity of 50–67 single units across three slices (n = 181 neurons total) before and after treatment with ZD7288 (Fig. 6). Consistent with our integrated population recordings (Fig. 2E), spiking from tonically active neurons initially increased +38 ± 12% within minutes following ZD7288 (“early ZD”) (Fig. 6B). However, following ∼10 min tonic spiking began to decline by 88 ± 15% and was nearly silenced thereafter (“late ZD”). The change in tonic spiking is exemplified in the representative spike raster plots shown in Fig. 6B. To determine whether this effect is dependent on synaptic interactions with the rhythmically active network, we blocked respiratory rhythmic activity with CNQX to block AMPA-dependent excitatory interactions in the network. Three additional horizontal brainstem slices with 42–60 single units (n = 154 neurons total) were exposed to 10 µM CNQX followed by 100 µM ZD7288. CNQX slightly but significantly increased tonic spiking across the recorded neurons by 27 ± 16% (Fig. 5C). The effect of ZD7288 in the presence of CNQX was similar to the intact network. In the presence of CNQX, ZD7288 initially increased spiking by 40 ± 16% in tonically active neurons within minutes (early ZD), followed by a reduction of 88 ± 24% and nearly complete silencing of tonic activity thereafter (late ZD) (Fig. 6C). These results demonstrate that I_h plays a critical role in the intrinsic regulation of tonic spiking activity within the preBötC.

Figure 6.

ZD7288 suppresses tonic spiking activity within the preBötC. Neuropixel multiunit recordings were used to simultaneously assess multiple single-unit activity simultaneously throughout the preBötC. Concurrent contralateral recording of integrated population activity from the preBötC was used to determine the pattern of spiking across individual units. Representative traces from a rhythmic and tonic neuron compared with the integrated population activity is shown in A, left, middle. Peri-stimulus time histograms (PSTH) aligning spike activity with the onset of population bursts shows the difference in activity between rhythmic (left) and tonic (right) spike activity from representative cells (A, right). Following bath application of 100 µM ZD7288, tonic spiking throughout the network showed a biphasic effect (B). Initially, within ∼5 min (early ZD), the average frequency of tonic spiking increased within the PreBötC (P < 0.01, n = 165, one-way repeated-measures ANOVA with Sidak’s multiple comparisons), followed by a reduction of spike activity thereafter (late ZD) (P < 0.01, n = 165, one-way repeated-measures ANOVA with Sidak’s multiple comparisons) (B, left). Representative spike raster plots are shown to display activity across the tonic population throughout the PreBötC (B, right). Similar results were found in the presence of CNQX to block excitatory synaptic transmission (C). Bath application of 10 µM CNQX resulted in an insignificant increase (P = 0.72, n = 154, one-way repeated-measures ANOVA with Sidak’s multiple comparisons) in overall tonic spiking throughout the PreBötC. In the presence of CNQX, 100 µM ZD7288 initially (∼5 min) resulted in a further increase in the average tonic spike activity within the PreBötC (P = 0.04, n = 154, one-way repeated-measures ANOVA with Sidak’s multiple comparisons), followed by a significant reduction in the average tonic spiking thereafter (P < 0.01, n = 154, one-way repeated-measures ANOVA with Sidak’s multiple comparisons) (C, left). Representative raster plots demonstrating the change in tonic spiking during CNQX and ZD7288 administration are displayed (C, right). *P < 0.05. PreBötC, preBötzinger complex.

DISCUSSION

The preBötC network represents the minimal circuity necessary to produce rhythmic inspiratory activity (4, 67, 68). Within this network, rhythm arises through a combination of excitatory synaptic interconnections and intrinsic membrane properties that dictate the neuronal firing patterns of individual neurons (39, 27, 69, 70). I_h has been shown to play a major role in the regulation of rhythmic brain networks such as the thalamus and hippocampus (71, 72) and activation of the current upon hyperpolarization makes it a putative candidate for a key role in rhythm generation (13). Some PreBötC rhythmic neurons have been shown to express I_h, and the effect of blocking I_h has been studied by Refs. 44 and 56 in transverse slice preparations. Our study reveals that blocking I_h in the horizontal slice preparation results in minimal change to the fully synchronized burst activity, but increases the incidence of burstlet activity within the preBötC. Burstlets produced from the preBötC are not transduced to the motor neuron pool of the XII nucleus (73, 74). However, since horizontal slices do not contain the XII motor nucleus, we could not assess the effect on the XII motor output in the present study.

Previous estimates suggest that ∼38%–50% of rhythmic preBötC neurons express I_h (44), with lower expression (∼20%) among identified Dbx1 neurons (64). However, all previous studies focused on the role of rhythmic neurons thus little is known about the role of I_h in tonic spiking neurons. Here, we report that I_h expression is most prevalent among tonically active preBötC neurons (∼65+%). In many computational models of the PreBötC, a tonic excitatory drive parameter is often included, to stabilize rhythmicity that emerges through reciprocal inhibition between inspiratory, postinspiratory, and expiratory neurons (75). Other models, like the one we propose here, assume that tonic excitation is important for rhythmogenesis by elevating the level of g_tonic required to produce rhythmic activity at the population level (52). In this model, tonic spiking does not contribute directly to intrinsic bursting per se, but to the excitation level required to activate the synaptically coupled network. However, if tonic drive indeed plays a role in rhythmogenesis, it likely does not simply provide excitation to the respiratory network, since I_h is present in both excitatory and inhibitory tonically active neurons. Thus, we hypothesize that any potential tonic influence will likely depend on the balance between tonic excitatory and inhibitory drives. This can have a direct effect on the phenotypic pattern of postsynaptic cells by shifting activity between silent, bursting, or tonic spiking (52), as well as activating or inhibiting other native currents present within the cells.

Our data showed that blocking I_h within the preBötC resulted in silencing of the majority of tonic spiking within the network. This effect was associated with increased network sensitivity to 1) the hyperpolarizing effects of opioids and 2) the blockade of excitatory synaptic drive with CNQX. We, therefore propose that I_h plays a critical role in generating tonic activity within the preBötC and stabilizing the respiratory network. However, while I_h is most abundant in tonic neurons, they share the same genetic markers as the rhythmically active neurons. Thus, the currently available methods do not allow us to further test the network effects of the tonic neurons by e.g., specifically manipulating them using optogenetics.

However, our data allow us to conclude that rhythm generation cannot be maintained at lower levels of excitability following I_h blockade within the preBötC. At baseline (8 mM K⁺), the level of excitement is sufficient to maintain a rhythm with or without the presence of I_h. This is why there is little effect when blocking I_h under control conditions. However, when excitability of the network is reduced by the addition of opioids or reduction in excitatory synaptic transmission (CNQX), inspiratory activity is no longer present following blockade of I_h, whereas similar doses of DAMGO and CNQX are not sufficient to cease inspiratory activity in the presence of I_h. This suggests that I_h creates a more robust network, evident by the ability to maintain rhythm across a wider range of excitation. Whether this is due to the tonic spiking across the network and/or I_h-dependent activation of other persistent or voltage-dependent excitatory currents, such as I_NaP, among bursting neurons remains unresolved. Montandon et al. (76) previously demonstrated that blocking I_h or I_NaP alone showed minimal effects in the in vivo freely behaving rat, while concurrent block resulted in significant respiratory suppression. Interestingly, this effect was only noted in the anesthetized or sleep states, and not during the awake state that further highlights the state-dependent actions of I_h. Our study similarly highlights this state-dependent role of I_h in rhythm generation. I_h becomes increasingly necessary for rhythm generation during a depressing stimulus, such as seen in this study with opioids and excitatory synaptic block. Future studies will need to determine to what extent this effect depends on interactions between I_h and I_NaP, or other voltage-dependent activation of ion channels. Indeed, reduced activation of persistent excitatory currents, such as I_NaP, following I_h block may plausibly explain the inability of the network to maintain rhythmic activity following opioid administration. Accordingly, Burgraff et al. (29) demonstrated that reduced I_NaP activation increases the sensitivity to opioid-induced respiratory depression in the in vitro brainstem slice.

The blockade of I_h also results in a significant increase in the membrane resistance of cells expressing I_h (20, 71, 77). This effect has been well documented within cortical circuits through compaction of the dendritic space and was suggested to account for the changes in burst frequency observed by Thoby-Brisson et al. (44). Increasing the membrane resistance of preBötC neurons increases the change in voltage for any given change in ionic current, and blocking I_h facilitates intrinsic bursting. This could explain why blocking I_h increases the number of burstlets, as it may increase the amount of ectopic bursting among intrinsically bursting neurons (44). It is unknown whether changes in the membrane resistance of neurons throughout the preBötC may account for the observed increase in sensitivity to hyperpolarizing stimuli following I_h block. However, studies specifically targeting changes in membrane resistance throughout the network are needed to directly assess whether this may also be a contributing factor. The relationship between burstlet production and tonic spiking throughout the preBötC, however, is less clear. Although blocking I_h depolarizes I_h⁺ tonic cells throughout the network and silences their spike activity, this effect is seen in both excitatory and inhibitory cells. Thus, the current study can answer the question whether removing tonic spiking activity has any effect on burstlet production. Similarly, further experiments specifically targeting individual excitatory and inhibitory pools of tonic neuron throughout the preBötC are needed. However, to the best of our knowledge, the molecular markers to specifically and differentially manipulate these neurons have so far not been identified.

Our experimental findings described above help explain some of the obvious discrepancies between the effects of I_h block in the preBötC in vitro versus in our in silico model. First, our experiments reveal that blocking I_h results in a paradoxical depolarization in vitro, which we suspect is caused by interactions between I_h and other currents. Yet, in our model, based on well-known I_h dynamics, removing I_h leads to hyperpolarization. Second, we found that I_h was roughly equally distributed among excitatory and inhibitory preBötC neurons, but our in silico model only contains excitatory neurons. Third, I_h was found to be most prevalent among tonically active neurons, but our model does not include a similar tonically active, I_h-expressing population. And finally, I_h block increased the prevalence of burstlets, yet our model does not include distinct rhythm and pattern generating populations as recently incorporated into newer models (55). Considering these differences, and others [e.g., potential dendritic effects of I_h block, other currents not included in the model (78), etc.] that are not captured by our model, it may not be surprising that functional effects of I_h block in the preBötC were more complex than those predicted by our simplified model network. However, these discrepancies are important to highlight because they emphasize and support our experimental findings that the role of I_h in the preBötC is far more complex than simply the regulation of V_m among a rhythmic excitatory population. Future computational models that attempt to incorporate these findings will be necessary to further unravel the specific roles of I_h in different types of preBötC neurons and its interactions with the various excitatory neurotransmitters known to regulate rhythmogenesis (78–84).

In summary, we conclude that I_h plays multiple roles within the preBötC. Some of these roles are complex and not consistent with the traditionally assumed role of the I_h current in rhythmogenesis. We find that the I_h current is important for promoting tonic spiking activity in excitatory and inhibitory neurons and for preserving rhythmic function during conditions that dampen network excitability, such as in the context of opioid-induced respiratory depression. We therefore propose that I_h expands the dynamic range of rhythmogenesis, buffers the preBötC against network perturbations, and stabilizes rhythmogenesis by preventing the generation of unsynchronized bursts.

SUPPLEMENTAL DATA

Supplemental Fig. S1: https://doi.org/10.6084/m9.figshare.19763527.v1.

GRANTS

Funding for this work was provided by the National Heart, Lung, and Blood Institute grants: HL-126523 (JMR), HL-090554 (JMR), HL-145004 (NAB), HL-154558 (NJB).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors. Jan-Marino Ramirez is an editor of Journal of Neurophysiology and was not involved and did not have access to information regarding the peer-review process or final disposition of this article. An alternate editor oversaw the peer-review and decision-making process for this article.

AUTHOR CONTRIBUTIONS

N.J.B., R.S.P., and J-M.R. conceived and designed research; N.J.B., R.S.P., and L.J.S. performed experiments; N.J.B., R.S.P., and N.E.B. analyzed data; N.J.B., R.S.P., N.A.B., and J-M.R. interpreted results of experiments; N.J.B. prepared figures; N.J.B. drafted manuscript; N.J.B., N.A.B., and J-M.R. edited and revised manuscript; N.J.B., R.S.P., L.J.S., N.E.B., N.A.B., and J-M.R. approved final version of manuscript.