Abstract
The homeostatic regulation of pulmonary ventilation, and ultimately arterial PCO2, depends on interactions between respiratory chemoreflexes and arousal state. The ventilatory response to CO2 is triggered by neurons in the retrotrapezoid nucleus (RTN) that function as sensors of central pH, which can be identified in adulthood by the expression of Phox2b and neuromedin B. Here, we examine the dynamic response of genetically defined RTN neurons to hypercapnia and arousal state in freely behaving adult male and female mice using the calcium indicator jGCaMP7 and fiber photometry. We found that hypercapnia vigorously activates RTN neurons with a low CO2 recruitment threshold and with response kinetics that match respiratory activity whereas hypoxia had little effect. RTN activity increased transiently during wakefulness and respiratory-related arousals and rose persistently during rapid eye movement sleep, and their CO2 response persisted under anesthesia. Complementary studies using inhibitory optogenetics show that RTN activity supports eupneic breathing under anesthesia as well as during states of high arousal, but their activity is redundant for voluntary breathing patterns. Collectively, this study demonstrates that CO2-activated RTN neurons are exquisitely sensitive to the arousal state, which determines their contribution to alveolar ventilation in relation to arterial PCO2.
Significance Statement
Respiratory chemoreceptors stimulate neural respiratory motor output to regulate arterial PCO2 and PO2, thereby maintaining optimal gas exchange. Central chemoreceptor neurons expressing Phox2b and neuromedin B in the retrotrapezoid nucleus (RTN) are required for the hypercapnia ventilatory response. However, the dynamic activity of RTN neurons in conditions of normal and elevated carbon dioxide has not been described in unanesthetized conditions. Here, we use a genetically encoded calcium indicator to demonstrate that RTN neurons are exquisitely sensitivity to hypercapnia and variations in arousal and sleep state in freely behaving mice. This work has implications for understanding the central control of breathing across arousal states, particularly during sleep and anesthesia.
Introduction
The homeostatic regulation of ventilation to maintain blood gas homeostasis is controlled by central and peripheral respiratory chemoreceptors (Dempsey et al., 2012; Guyenet and Bayliss, 2022; López-Barneo, 2022). Neurons located in the retrotrapezoid nucleus (RTN) are putative central respiratory chemoreceptors that encode arterial pH/PCO2 (Guyenet and Bayliss, 2022) and regulate pulmonary ventilation through connections within the pontomedullary respiratory network (Abbott et al., 2011; Ruffault et al., 2015; Souza et al., 2023; Souza and Abbott, 2024). In adult rodents, RTN neurons are genetically defined by the expression of the transcription factor Phox2b (Stornetta et al., 2006; Ruffault et al., 2015; Cardani et al., 2024) and the neuropeptide neuromedin B (Nmb; Shi et al., 2017; Souza et al., 2023). In addition to encoding arterial PCO2, RTN neurons also integrate synaptic inputs from respiratory and nonrespiratory regions of the brainstem, hypothalamus, and spinal cord (Guyenet et al., 2005; Rosin et al., 2006; Fortuna et al., 2009; Kanbar et al., 2016; Li et al., 2020). Thus, RTN neurons may be a hub for polymodal signaling that stimulates ventilation. However, the activity of RTN neurons has been primarily characterized using in vitro and anesthetized preparations, which are conditions that disrupt sensorimotor function and obscure the contribution of arousal state to RTN function.
A defining characteristic of chemoreceptor neurons in the RTN is an increase in firing rate during hypercapnia in vivo and acidification in vitro (Mulkey et al., 2004; Guyenet et al., 2005; Lazarenko et al., 2009). The nature of this response varies greatly with both experimental condition and developmental stage. Single-unit electrophysiological recordings in anesthetized, mechanically ventilated rats show that RTN neurons display tonic firing at a rate of 0–2 Hz in normocapnic conditions that increases linearly with hypercapnia at a rate of ∼2 Hz per 1% increase in end-tidal CO2 up to a maximum firing rate of ∼10 Hz (Guyenet et al., 2005). This suggests RTN neurons are sensitive to small perturbations in arterial PCO2 around the set point for blood gas homeostasis. Consistent with this, functional studies indicate RTN neurons contribute to the recruitment of ventilation in both normocapnic and hypercapnic conditions (Souza et al., 2023; Souza and Abbott, 2024). However, direct evidence that RTN neurons are activated by hypercapnia in freely behaving conditions currently is limited to c-Fos expression (Sato et al., 1992; Teppema et al., 1997; Shi et al., 2017; Dereli et al., 2024), while single-cell calcium imaging studies suggest that very few neurons in the parafacial region respond to hypercapnia in a manner that would be expected for central chemoreceptors (Bhandare et al., 2022). As such, the response of RTN neurons to hypercapnia in conditions free from anesthetics remains unclear. Our first hypothesis is that RTN neurons exhibit a graded response to hypercapnia in freely behaving mice, which we test using genetically targeted fiber photometry to selectively measure RTN neuron activity across distinct states of arousal.
Breathing, like most sensorimotor functions, is heavily influenced by arousal state and is particularly sensitive to anesthesia (Brown et al., 2011; Nattie, 2011). Waking is associated with increased respiratory rate and voluntary breathing patterns whereas anesthesia suppresses arousal and causes respiratory depression (Brown et al., 2011; Nattie, 2011). Variations in sleep–wake state are known to modulate the contribution of RTN neurons to breathing (Burke et al., 2015; Souza et al., 2023), but whether this reflects changes in their activity is not known. Moreover, RTN neurons receive inputs from brain regions that promote arousal and stress responses (Rosin et al., 2006; Li et al., 2020) suggesting that variations in arousal state might drive changes in RTN activity. RTN neurons are also activated by isoflurane in vitro (Lazarenko et al., 2010) and are unaffected by morphine when arterial PCO2 is controlled by mechanical ventilation (Fortuna et al., 2009), suggesting that RTN activity is resistant to the depressant effects of some anesthetics. Based on this, we also hypothesized that RTN activity is modulated by changes in arousal accompanying natural sleep–wake cycles and by anesthesia, which we test by combining fiber photometry recordings of RTN activity with EEG/EMG recordings.
Materials and Methods
Animals
All experiments were conducted in accordance with the National Institutes of Health's Guide for the Care and Use of Laboratory Animals and approved by the University of Virginia Animal Care and Use Committee (protocol #4312). We used transgenic NmbCre/+ described in Souza et al. (2023) backcrossed with C57BL/6J mice acquired from The Jackson Laboratory (strain #000664, RRID: IMSR_JAX:000664). Hemizygous adult (8–24 weeks) male and female mice Cre-positive littermates (NmbCre/+) and Cre-negative (Nmb+/+) were used as experimental and control groups, respectively.
Central microinjections, viral vector, and fiber placement
Mice of both sexes (age: 8–12 weeks) were anesthetized with a mixture of ketamine (100 mg/kg) and dexmedetomidine (0.2 mg/kg) given intraperitoneally. The depth of anesthesia was assessed by an absence of the hindpaw withdrawal reflex. Additional anesthetic was administered as necessary (20% of the original dose, i.p.). The following procedures were performed under aseptic conditions. Following skin shaving and disinfection, the skin overlying the left mandible was cut to reveal a segment of the mandibular branch of the facial nerve. The mice were then placed on a stereotaxic apparatus adapted for CNS microinjections (ear bar adaptor, model EB-5N, Narishige Scientific Instrument Lab; bite bar, model 926 mouse adaptor set at −3 mm below the interaural line for a flat skull, David Kopf Instruments). Body temperature was maintained at 37°C with a servocontrolled heating pad and a blanket. A small craniotomy, approximately 1.5 mm in diameter, was drilled unilaterally into the occipital bone.
To selectively transduce Nmb-expressing neurons in the RTN, we used an adeno-associated Cre-dependent vector encoding the calcium indicator, jGCaMP7s (pGP-AAV1-syn-FLEX-jGCaMP7s-WPRE, Addgene plasmid 104491; titer of injected virus solution, 4.4 × 1012 GC/ml). jGCaMP7s present onset kinetics with half-rise time in response to 10 action potential trains of 70 ± 2 ms, offset kinetics with a half-decay time of 1,690 ± 55 ms, and a 373 ± 11.5% increase in ΔF/F0 (Dana et al., 2019). The AAV was loaded into a glass pipette with a 1.2 mm internal diameter pulled on a vertical pipette puller, and the tip was cut to an external diameter of 25 µm. The facial nerve was stimulated (0.1 ms, 100–300 mA, 1 Hz) to evoke antidromically evoked field potentials in the facial motor nucleus (Brown and Guyenet, 1984). These field potentials were used to map the facial motor nucleus and help identify the location of the RTN neurons, which reside 100–200 µm below this nucleus. Once localized, two microinjections (60 nL each) were performed: one placed in the most caudal aspect of RTN and the other 300 µm rostral to this point. The microinjections were followed by the implant of an optical fiber (400 µm, NA = 0.66, Doric Lenses) with the tip of the fiber placed 200–300 µm above the injection site.
For the optogenetic inhibition experiments, a vector encoding eArch3.0 (AAV2-EF1α-DIO-eArch3.0-eYFP, UNC Vector Core, titer of injected virus solution, 4.0 × 1012 GC/ml) was microinjected in the left RTN in three sites spanning the rostrocaudal length of the facial nucleus (1,400 µm lateral to midline, 100 µm below the bottom of the facial nucleus) and a vector encoding taCasp3 (AAV5-FLEX-taCasp3-TEVp; titer, 4.2 × 1012 GC/ml, UNC Vector Core) was injected in the contralateral RTN as in Souza et al. (2023). The microinjections were followed by the implant of an optical fiber (200 µm, 0.39 NA) with the tip of the fiber placed 400–500 µm above the injection site. Unilatera ablation of RTN neurons increases the effectiveness of optogenetic inhibition (Souza et al., 2023).
Following viral injections, a subset of mice was implanted with an electroencephalogram (EEG) and neck electromyogram (EMG) head stage to monitor sleep patterns consisting of six-pin connectors soldered to three intracranial screws positioned in the left frontal and left and right parietal regions for EEG recording and two stainless-steel wires (A-M Systems) implanted in the neck muscles for EMG recording.
In all mice, incisions were closed with sutures and surgical cyanoacrylate adhesive. All mice received postoperative boluses of atipamezole (α2-adrenergic antagonist, 2 mg/kg, s.c.) and ketoprofen (4 mg/kg, s.c.) and were then placed in a clean warmed home cage (37°C) until consciousness was regained before being returned to the vivarium. Ketoprofen was administered immediately after surgery and at 24, 48, and 72 h following the procedure. Mice were housed in the vivarium for 4 weeks before experimentation to allow sufficient expression of the calcium indicator transgene.
Experimental setup and signal acquisition
Experimental recording was conducted more than 1 month after the surgical procedures. Mice were briefly anesthetized with isoflurane (2.0% in normoxia) to connect the optical fiber and EEG/EMG connecting leads. Mice were then placed in an unrestrained whole-body plethysmography chamber (Buxco-style with a 400 ml internal volume, Data Sciences International) and allowed to recover for 1 h before experiments began. Recording sessions were conducted between 900 and 1,600 h and lasted <6 h. The chamber was continuously flushed with dry room temperature air delivered by three computer-driven mass flow regulators controlling the flow of O2, N2, and CO2 (1 l/min).
Fiber photometry was performed as described previously (Souza et al., 2022). We employed a 460 or 470 nm LED with a modulation frequency of 211 Hz to induce excitation in GCaMP, while a 405 nm LED with a modulation frequency of 531 Hz was utilized for the isosbestic channel. Both light sources were combined into a single-core multiple-mode fiber (Doric Lenses, 400 µm, 0.66 NA) using a fluorescence minicube from Doric Lenses [ilFMC4-G3_IE(400-410)_E(460-490)_F(500-550)_S]. Lock-in amplifiers (SR810 DSP, Stanford Research Systems) were used to modulate and demodulate each LED. The power output of each modulated LED was set between 5 and 25 μW when measured with a power meter (Thorlabs). Fluorescence emission signals were collected by a photoreceiver (Newport, model 2151) or the light sensor integrated into the fluorescence minicube from Doric Lenses. Demodulated fluorescence signals were acquired at 1,000 Hz and smoothed with a time constant of 1 s prior to calculating F/F0 with the following formula: F/F0 = (F460/F460 baseline/F405/F405 baseline) × 100. Baseline values consisted of the 30 s period preceding the event being analyzed.
EEG and EMG signals were amplified and bandpass filtered (EEG: bandpass, 0.1–100 Hz; gain, ×2,000; sampling rate, 1 kHz; EMG: filter setting, 300–1,000 Hz; gain, ×300; sampling rate, 1 kHz; CWE, BMA-400) and then acquired, processed, and analyzed using Spike 2 (v9.07, Cambridge Electronic Design).
The pattern of breathing was measured by whole-body plethysmography using a differential pressure transducer with the signal amplified and bandpass filtered (bandpass, 0.1–100 Hz; gain, ×2,000) and acquired at 1 kHz. Owing to the conditions of the experiment, we were unable to accurately measure tidal volume.
Experimental protocols and data analysis
Experimental protocol 1: RTN response to hypercapnia and hypoxia
Mice were exposed to recurrent 5 min bouts of hypercapnia (3, 6, and 9% CO2 in 60% O2), hypoxia (10% O2), and hypoxic hypercapnia (5% CO2 in 10% O2) with a minimum recovery of 10 min between trials. Between two and four trials for each gas condition were averaged to generate both time course and grouped data for ΔF/F0 and fR.
Experimental protocol 2: state-dependent RTN neuron activity
EEG and EMG data were collected in parallel with F/F0 for 3–4 h without any perturbations of the mice’s behavior. We manually scored periods of wake, nonrapid eye movement (NREM) sleep, and rapid eye movement (REM) sleep using raw recordings and hypnograms of EEG and integrated EMG. REM sleep was defined by a high theta/delta ratio, low EMG power, and low delta power. NREM sleep was defined as periods of high delta power, low theta/delta ratio, and low EMG power. Wake was considered states with low delta power and high EMG and high gamma power. Microarousals were defined as periods of EEG desynchronization lasting <10 s with or without EMG activation occurring during periods of NREM sleep.
The effect of anesthesia on RTN activity was tested using isoflurane and ketamine/dexmedetomidine. Prior to induction of anesthesia, mice were exposed to 9% CO2 for 2 min in normoxic conditions to generate preanesthesia values for F/F0. Isoflurane was administered in normoxic gas (21% O2 balance N2) at a concentration of 2% to induce anesthesia. Once a stable plane of anesthesia was achieved (∼7–10 min after the beginning of isoflurane induction), the concentration of inspired CO2 was increased to 9% (21% O2 balance N2) for 2 min followed by a return to baseline conditions. After at least 2 h of recovery, mice were re-exposed to CO2 (9% for 2 min), and after this period, ketamine and dexmedetomidine (Ket/Dex, 50 mg/kg and 0.1 mg/kg, i.p.) were administered. After reaching a stable plane of anesthesia judged by EEG/EMG activity, inspired CO2 was increased to 9% (21% O2 balance N2) for 2 min. After recovery from hypercapnia, mice were administered atipamezole (α2-adrenergic antagonist, 2 mg/kg, s.c.) to accelerate recovery of consciousness.
Experimental protocol 3: state-dependent effect of RTN inhibition on breathing
For optogenetics experiments, anesthesia was not used when connecting mice to the optic fiber and EEG/EMG head stage. A green diode laser (peak emission wavelength, 532 nm; power output, 10 mW at the tip of the fiber) was used to excite archaerhodopsin for optogenetic inhibition of RTN neurons. For optogenetic experiments, mice were connected to a connecting optical fiber and EEG/EMG leads without anesthesia. Once mice were tethered for recording, they were placed immediately into the unrestrained whole-body plethysmography chamber and recordings of breathing were initiated. Laser stimulation was delivered for 5 s every 60 or 120 s for 90 min. Raster plots for fR were triggered from the onset of laser. The z-score for fR was calculated for 2.5 s epochs over a 25 s window centered around the laser-on period. Periods of sham stimulation were triggered from a timepoint 30 s after the laser onset. To classify data according to arousal state, EEG and EMG were used to determine the state, with breathing patterns used to distinguish periods of active wakefulness (fR > 200 breaths/min with high variability) or eupneic breathing (fR < 200 breaths/min with low variability). The protocol used to test the effect of RTN inhibition under anesthesia matched the one used for the fiber photometry experiment (Experimental protocol 2). In this stage of the protocol, laser stimulation was manually triggered. fR values in all experiments reflect data from a 5 s baseline and during the 5 s of laser stimulation.
Histology
Following experiments, mice used for histology were deeply anesthetized by intraperitoneal administration of a combination of ketamine (150 mg/kg) and dexmedetomidine (0.3 mg/kg). Transcardial perfusion was performed with a 4% paraformaldehyde solution in 100 mM phosphate buffer (pH 7.4) for tissues used with FISH or a pH-buffered 10% formalin solution (pH 7.0) for tissues using IHC. Following perfusion, the brains were extracted and underwent a postfixation for 12–24 h at 4°C in the same fixative solution. Tissues were sectioned to a thickness of 30 μm on a vibrating microtome (VT-1000S, Leica Biosystems), and sections were stored in cryoprotectant [30% ethylene glycol (v/v), 20% glycerol (v/v), 50% 100 mM phosphate buffer, pH 7.4 (v/v)] at −20°C.
Immunohistochemistry was carried out on free-floating sections at room temperature, unless otherwise specified. All washes used Tris-buffered saline (TBS, 100 mM Tris, 150 mM saline) In brief, a one-in-three series of tissue was washed and then blocked in a solution containing 10% horse serum (v/v) and 0.1% Triton X-100 (v/v) in TBS. Subsequently, these sections were incubated with primary antibodies for 60 min at room temperature, followed by an overnight incubation at 4°C. After washing, the sections were incubated in secondary antibodies for 60 min at room temperature before being mounted on slides using ProLong Gold antifade mounting medium (P36931, Thermo Fisher Scientific). Specific antibodies used are detailed in Table 1.
For fluorescent in situ hybridization (FISH) using RNAscope (V1 kit, Advanced Cell Diagnostics), sections were washed briefly in phosphate-buffered saline, placed on charged slides, dried overnight, and then processed according to the manufacturer's instructions. Immunohistochemistry was conducted after the RNAscope procedure using an abbreviated immunohistochemistry protocol labeling with an antibody against GFP and a secondary antibody.
Microscopy and imaging procedures were performed using Neurolucida software (MBF Bioscience) in conjunction with an Axio Imager.M2 microscope (Carl Zeiss Microscopy). Grayscale digital images were captured with a Hamamatsu C11440 ORCA-Flash 4.0LT digital camera. Further image adjustments, including pseudocoloring, optimization for presentation, and uniform adjustments to brightness and contrast, were carried out using Fiji software (Schindelin et al., 2012). Cell counting and fiber placement were conducted on a one-in-three series aligned to Paxinos and Franklin's fourth edition.
Statistical analysis
Statistical analysis was conducted in GraphPad Prism (v10). The distribution of the data was assessed using either the D’Agostino–Pearson, Shapiro–Wilk, or Kolmogorov–Smirnov test, and the outcome of these tests was used to determine if parametric or nonparametric tests were used to determine significant differences between groups. The cross-correlation between fR and GCaMP fluorescent intensity (ΔF/F0) was carried out using R (v4.2.1) across a range of ±15 s at 1 s intervals. Statistical tests are specified for each test in the results and figure legends. Significance was attributed to differences with p < 0.05. Data are presented as mean ± SD unless stated otherwise.
Results
Genetically targeted GCaMP expression in RTN neurons
To measure RTN activity, a Cre-dependent construct driving GCaMP expression was microinjected unilaterally in the RTN of Nmb Cre-positive and Cre-negative mice and a 400 µm fiber-optic cannula was inserted within 200 µm of the site of injection (Fig. 1A). This procedure resulted in expression of GCaMP in 82–135 neurons (median, 108.0, n = 5) as measured by a one-in-three series with 94 ± 1% of these cells expressing Phox2b (Fig. 1B,C). In situ hybridization for Nmb combined with immunofluorescence staining for GCaMP revealed that 90 ± 1% of neurons expressing GCaMP also expressed Nmb, with 89 ± 6% of Nmb+ neurons in the parafacial region expressing GCaMP (Fig. 1D). This shows that GCaMP was expressed selectively in RTN central chemoreceptor neurons (Guyenet et al., 2016). No GCaMP staining was observed in Cre-negative control mice injected with the Cre-dependent GCaMP construct. High-quality recordings required the tip of the fiber optic to be proximal to the caudal cluster of RTN chemoreceptor neurons (where most of the Nmb+ neurons reside), within 400 µm of the ventral surface of the medulla (Fig. 1E,F).
Genetically targeted recording of Nmb-expressing RTN neurons. A, Experimental design for genetically targeted recordings of Nmb-expressing RTN neurons using fiber photometry. B, Image showing the position of the fiber optic relative to GCaMP-expressing Phox2b neurons in the RTN. Scale bar, 500 μm. C, Enlarged image from B showing colocalization between GCaMP and Phox2b expression in the RTN. Scale bars: 200 μm for larger image, 100 μm for smaller insets. D, Image showing that neurons expressing GCaMP also express transcript for Nmb. Scale bar, 50 μm for larger image and smaller insets. E, Drawings from a representative GCaMP case showing the distribution of GCaMP-expressing neurons and the position of the fiber optic. The placement of the blue line marking the center of the tip of the fiber optic is indicative of the approach used to map the placement of the fiber optic in F. F, Fiber-optic placement in GCaMP cases.
RTN neurons increase their activity in response to CO2 in freely behaving mice
Fiber photometry recordings of RTN fluorescence were conducted in an unrestrained whole-body plethysmography chamber allowing concurrent measurement of respiratory activity and control of inspired gases (Fig. 2A). In response to hypercapnia, there was a reproducible and sustained increase in the fluorescence emitted from RTN neurons in GCaMP-expressing cases correlating with an elevation in breathing frequency (Fig. 2B,C,F). Following hypercapnia exposure, RTN fluorescence returned to baseline levels with a kinetic that paralleled respiratory frequency (Fig. 2B,D). In GCaMP cases, the increase in RTN emitted fluorescence during steady-state normoxic hypercapnia was significantly greater in GCaMP cases for 3, 6, and 9% CO2 than in controls in which there was no expression of GCaMP. Moreover, the magnitude of the response to hypercapnia in GCaMP cases was significantly different for each level of CO2 (Fig. 2E). Also, the change in RTN fluorescence during hypercapnia was linearly related to inspired CO2 (mean slope of linear regression between CO2 and ΔF/F0 in GCaMP cases, 1.40 ± 0.70% per % of CO2; goodness of fit, 0.83–0.97; n = 7). Together, these data indicate that, in freely behaving mice, RTN neurons exhibit a sustained and graded response to hypercapnia with response kinetics that closely match the pattern of the hypercapnic ventilatory response (HCVR).
RTN neurons are sensitive to different concentrations of inspired CO2 in conscious mice. A, Experimental setup for the simultaneous recording of RTN fluorescence and breathing frequency in freely behaving mice. The concentration of CO2 inside the chamber was measured and represented as a percentage of the composition of gases in the inspired air. B, Representative experimental recording of RTN corrected fluorescence (ΔF/F0) and respiratory activity during recurrent episodes of hypercapnia (CO2 = 9% for 5 min). RTN activity displays a fast onset kinetics and remains elevated during CO2 stimulation, mirroring the respiratory activity. C, Changes in corrected fluorescence in GCaMP (blue, n = 7) and control (gray, n = 7) cases during different levels of inspired CO2. The top panel shows the mean ΔF/F0 (± SEM) from the GCaMP and control groups. The middle panels show a heat-map for corrected fluorescence in GCaMP cases as a z-score from each individual case. The bottom panels show the mean respiratory frequency (fR) ± SEM over time during exposure to hypercapnia for GCaMP cases. D, Overlay of ΔF/F0 from GCaMP cases during hypercapnia, showing the graded response to different levels of CO2. E, Average values of ΔF/F0 from GCaMP and control cases during the plateau phase observed at different levels of CO2 (repeated-measures two-way ANOVA, interaction between levels of CO2 vs GCaMP, F(2, 24) = 22.1, p < 0.0001; effect of CO2, F(1.5, 17.6) = 26.0, p < 0.0001; effect of GCaMP, F(1, 12) = 24.6, p = 0.0003). *p < 0.05; **p < 0.01; ***p < 0.001; GCaMP vs control mice by Bonferroni's multiple comparisons. †p < 0.05; ††p < 0.01, when the average fluorescence is compared with the previous level of CO2 by Bonferroni's multiple comparisons. F, Average values of fR from GCaMP and control cases during the plateau over different levels of CO2. Repeated-measures two-way ANOVA, interaction between levels of CO2 versus GCaMP, F(2, 24) = 1.57, p = 0.2293; effect of CO2, F(1.6, 19.5) = 186.3, p < 0.0001; effect of GCaMP, F(1, 12) = 0.013, p = 0.9104. †p < 0.05; ††p < 0.01, when fR compared with the previous level of CO2 by Bonferroni's multiple comparisons.
RTN neuron activity parallels fluctuations in respiratory activity during eupneic breathing
During quiescent breathing in room air conditions (normoxic normocapnia), we observed transient increases in RTN fluorescence that coincided with brief hyperpneas and sighs (Fig. 3A,E). Cross-correlation revealed a robust positive correlation between RTN fluorescence and fR, which was significantly greater in the GCaMP group compared with controls (Fig. 3B–D). The most prominent peaks in RTN fluorescence during resting breathing were coupled with a sigh or augmented breath (Fig. 3A,E–H). However, the onset of the increase in RTN fluorescence coincided with the onset of an increase in respiratory frequency rather than the sigh in most cases (Fig. 3A,E). There was some variation in the pattern of recovery following spontaneous sighs, both between trials and between cases. The most common was a monotonic return of RTN fluorescence to baseline, but we occasionally observed abrupt reductions in RTN activity coinciding with post-sigh apnea in individual trials in three of six cases.
RTN activity correlates with breathing frequency at rest. A, Representative experimental recording of RTN fluorescence (ΔF/F0) and respiratory frequency (fR) during periods of eupneic breathing in conscious mice. The state of arousal is identified based on the respiratory pattern (see Materials and Methods). The bottom panels provide enlarged examples of common patterns of RTN fluorescence. B, Experimental trace showing a correlation between RTN fluorescence and fR during behaviorally active periods. C, Grouped data for the cross-correlation between GCaMP and fR and control and fR during active wakefulness. D, Pearson's r at a lag of t = 0 from cross-correlation analysis in GCaMP and control (t test, t = 2.894, df = 10, p = 0.016). *p < 0.05. E, Examples of RTN fluorescence during spontaneous sighs. Examples i and ii are from the same mouse, whereas example iii is a distinct case. F, Heat-map for corrected fluorescence in GCaMP cases as a z-score centered on the onset of spontaneous sighs. The rows represent the values from several individual examples grouped by case number. G, Grouped data for ΔF/F0 from GCaMP (n = 6) and control (n = 6) cases during spontaneous sighs. Zero on the abscissa denotes the onset of the sigh event. H, Changes in ΔF/F0 from GCaMP and control cases during spontaneous sigh events (t test, t = 3.867, df = 10, p = 0.0031).
State-dependent activity of RTN neurons
To characterize changes in RTN activity with arousal state, we recorded from RTN neurons simultaneously along with EEG and EMG. We observed a prominent increase in RTN fluorescence during periods of REM sleep (Figs. 3A, 4A,B,D). On average, RTN activity during REM sleep showed no correlation with breathing frequency during this sleep state (r2 = 0.098, at lag = 0). RTN activity increased during transitions from NREM sleep to wakefulness and was more dynamic and variable during active wake compared with NREM sleep (Fig. 4C,F). RTN fluorescence also gradually decreased during transitions from waking to NREM sleep (Fig. 4H) and abruptly decreased with transitions between REM sleep and wakefulness (Fig. 4J). Finally, many spontaneous peaks in RTN fluorescence associated with changes in breathing patterns and sighs appear to reflect microarousals during NREM sleep (Figs. 4A,B, 5), a prevalent periodic feature of sleep microstructure involving both cortical, motor, and autonomic activation (Ramirez, 2014). In summary, these data demonstrate that RTN activity is dynamic with changes in arousal state and coupled with breathing patterns during NREM sleep and wake. However, during REM sleep, RTN activity and breathing pattern appear to be uncoupled.
RTN activity is dependent on the arousal state. A, Example of RTN activity recording during periods of nonrapid eye movement (NREM), rapid eye movement (REM) sleep, and active wakefulness in a freely behaving mouse. Arousal states were classified based on EEG and EMG recordings. B, Sleep–wake recording showing the increase in RTN activity during the transition from NREM to REM sleep. C, Sleep–wake recording showing the increase in RTN activity during the transition from NREM sleep to active wakefulness. Note the increase in EMG activity. D, Average RTN activity during the transition from NREM to REM sleep in GCaMP (n = 6) and control (n = 5) cases. E, Changes in corrected fluorescence (ΔF/F0) in GCaMP and control cases during the transition from NREM to REM sleep (t test, p = 0.0019, t = 4.321, df = 9). F, Average RTN activity during the transition from NREM sleep to wake in GCaMP and control cases. G, Changes in corrected fluorescence (ΔF/F0) in GCaMP and control cases during the transition from NREM sleep to wake (t test, p = 0.0141, t = 3.035, df = 9). H, Average RTN activity during the transition from wake to NREM sleep in GCaMP and control cases. I, Changes in corrected fluorescence (ΔF/F0) in GCaMP and control cases during the transition from wake to NREM sleep (t test, p = 0.0052, t = 3.670, df = 9). J, Average RTN activity during the transition from REM sleep to wake in GCaMP and control cases. K, Changes in corrected fluorescence (ΔF/F0) in GCaMP and control cases during the transition from REM sleep to wake (t test, p = 0.0152, t = 2.990, df = 9).
RTN activity during respiratory microarousal. A, Example of RTN activity during a period of NREM sleep with multiple microarousals associated with respiratory events. The arrowhead indicates the microarousal from NREM sleep marked by EEG desynchronization, an increase in EMG, and a respiratory event. B, Examples of microarousal events. Note that RTN activity is coupled with microarousal regardless of whether the event is associated with a sigh or a transient tachypnea.
Effect of anesthesia on RTN neuron activity
Anesthesia profoundly influences brain state (Franks, 2008; Brown et al., 2011) and the hypercapnic ventilatory response (HCVR; Massey and Richerson, 2017), which may be related to alterations in chemoreceptor function (Nattie, 2011). Here, we tested the effect of anesthesia on basal RTN activity and the response of these neurons to hypercapnia using isoflurane and ketamine/dexmedetomidine (Fig. 6). Infusion of 2% isoflurane into the chamber in normoxic normocapnia led to a reduction in EEG activity interspersed with bursts of EEG activity (Antunes et al., 2003), a loss of motor activity, and slight increase in breathing frequency. During induction of isoflurane anesthesia, we observed a comparable decrease in baseline fluorescence in GCaMP (n = 6) and control cases (n = 5; −4.6 ± 2.2% vs −4.3 ± 2.6%, Student's t test, t = 0.2136, df = 9, p = 0.8356). Isoflurane anesthesia abolished spontaneous variations in RTN activity (Fig. 6A). In contrast, the increase in RTN fluorescence in response to CO2 persisted under isoflurane anesthesia and was of an amplitude similar to that found when the animal was conscious (Fig. 6A–C). The increase of RTN fluorescence in GCaMP cases under isoflurane anesthesia was significantly greater than in control cases (Fig. 6D) and similar in kinetics to the conscious state (Fig. 6E).
Effect of anesthesia on the response of RTN neurons to hypercapnia. A, Neck electromyogram (EMG), electroencephalogram (EEG), RTN activity, respiratory flow, and concentration inspired CO2 in conscious and isoflurane- (2%) and ketamine/dexmedetomidine (Ket/Dex, 50/0.1 mg/kg)-anesthetized mice. Note that the different anesthetics produced distinct changes in EEG pattern and completely silenced EMG. During hypercapnia (CO2 = 9% for 2 min), RTN activity increases in both anesthetized states, but the increase is less pronounced during Ket/Dex anesthesia. Note that the scale of respiratory flow amplitude is not matched between examples. Under isoflurane anesthesia, hypercapnia causes a reduction in respiratory frequency and an increase in amplitude, whereas under Ket/Dex, both respiratory frequency and amplitude are unresponsive to hypercapnia. B, Average change in fluorescence emitted from GCaMP (n = 6) and control (n = 5) mice in response to hypercapnia (CO2 = 9%) in conscious and anesthetized states. C, Mean GCaMP signal during hypercapnia in three different states. Note that the amplitude of the increase in RTN activity during hypercapnia is minimally affected by isoflurane and reduced by Ket/Dex anesthesia. D, Changes in corrected fluorescence in GCaMP (n = 6) and control (n = 5) in response to hypercapnia (two-way ANOVA, state vs GCaMP; interaction, F(2, 18) = 5.17, p = 0.017. State, F(1.6, 14.1) = 3.23, p = 0.079. GCaMP, F(1, 9) = 69.9, p < 0.0001). *p < 0.05; **p < 0.01; ***p < 0.001; GCaMP versus control mice by Bonferroni's multiple comparisons. E, Changes in the 1/slope of a linear regression from time point zero to maximum value of GCaMP emitted fluorescence during CO2 exposure in each state (n = 6, one-way ANOVA, F(1.4, 7.0) = 12.4, p = 0.0070). *p < 0.05, conscious versus isoflurane versus Ket/Dex by Tukey's multiple comparisons.
A subanesthetic dose of ketamine/dexmedetomidine (Ket/Dex, 50 mg/kg and 0.1 mg/kg, i.p.) led to stable slow-wave EEG activity (Xi et al., 2018), an absence of voluntary motor activity, and a reduction in breathing frequency. Induction of Ket/Dex anesthesia led to a small but significant increase in corrected RTN fluorescence in GCaMP (n = 6) compared with control cases (n = 5; (GCaMP vs control; 5.2 ± 2.9% vs 1.3 ± 2.7%, unpaired t test, t = 2.43, df = 9, p = 0.0379). Similar to isoflurane anesthesia, Ket/Dex anesthesia abolished spontaneous variations in RTN activity (Fig. 6A), whereas hypercapnia produced a significant increase in RTN fluorescence in GCaMP cases compared with controls (Fig. 6B,D). However, the response of RTN neurons to CO2 was attenuated under Ket/Dex anesthesia compared with the conscious state and under isoflurane anesthesia (Fig. 6D,E). In summary, the response of RTN neurons to hypercapnia was preserved under anesthesia, which demonstrates that the response of these cells to CO2 is not contingent on arousal-related inputs. Conversely, spontaneous changes in RTN activity in normoxic conditions are eliminated under anesthesia, indicating that this activity is dependent on arousal-related inputs.
State-dependent effects of RTN inhibition on breathing
The functional contribution of RTN neurons to ventilation varies between NREM sleep, REM sleep, and quiet wakefulness (Burke et al., 2015; Souza et al., 2023). Here, the fiber photometry data show that RTN neurons are highly active during periods of active wakefulness (Fig. 4C). To investigate the relationship between breathing patterns during active wakefulness and RTN activity, we used unilateral optogenetic inhibition of the RTN combined with ablation of the contralateral RTN. We intermittently inhibited RTN neurons over a 90 min period after mice were introduced to the plethysmography chamber (Fig. 7B). During this period, mice exhibit a mild stress response associated with variable and high-amplitude eupneic breathing patterns that give way to quiet wake and sleep breathing patterns as mice accommodate to the chamber, as reflected by average fR (Fig. 7C). The effect of RTN inhibition on fR was more pronounced during quite wakefulness and largely ineffective during active wakefulness state (Fig. 7D). But on average, fR was significantly reduced by RTN inhibition, but not sham inhibition periods (Fig. 7F), regardless of differences in mean fR (Fig. 7C,G). When trials of RTN inhibition were grouped based on the baseline breathing patterns and arousal state, we observed no significant reduction in absolute fR during active wakefulness and a significant reduction in fR during periods of quite wake and NREM sleep (Fig. 7H). In summary, RTN neurons contribute modestly and appear to be redundant to respiratory motor drive in high arousal states.
State-dependent effects of RTN inhibition on breathing. A, Experimental approach for genetically targeted inhibition of Nmb-expressing RTN neurons by optogenetics. B, Experimental design for testing the effects of RTN inhibition of breathing during an elevated arousal state. C, Mean fR during each 30 min epoch of the 90 min testing period. There was a significant effect of time on fR by repeated-measures ANOVA (F(1.3, 3.9) = 12.97, p = 0.0220). *p < 0.05 by Tukey's multiple comparisons test. D, Experimental traces of individual trials of RTN inhibition during the first 30 min epoch of the 90 min testing period. Note that RTN inhibition did not slow breathing when applied during periods of active wakefulness (top trace and bottom trace). E, Cycle-triggered raster plot of breaths during RTN inhibition (indicated by green shading) applied every minute during a 90 min testing period for one case. F, The z-score of mean fR (2.5 s bins) centered on the period of RTN inhibition and sham inhibition during each 30 min epoch of the 90 min testing period for four separate GCaMP cases. Epochs inside the green box indicate the laser-on period. G, Grouped results for data in F. The z-score of mean fR was significantly reduced during RTN inhibition compared with sham inhibition for each 30 min epoch of the 90 min testing period (repeated-measures two-way ANOVA, interaction between time and presence of RTN inhibition, F(2, 6) = 5.5, p = 0.04; effect of time, F(2, 6) = 11, p = 0.01; effect of RTN inhibition, F(1, 3) = 56, p < 0.01). **p < 0.01 by Bonferroni's multiple comparisons. H, Mean absolute fR before and during RTN inhibition classified by the arousal state. fR was not significantly reduced by RTN inhibition during active wakefulness, whereas RTN inhibition reduced fR in other conditions. Repeated-measures two-way ANOVA, interaction between state and RTN inhibition, F(2, 15) = 5.602, p = 0.0152; effect of state, F(2, 15) = 64.15, p < 0.0001; effect of RTN inhibition, F(1, 15) = 35.15, p < 0.0001. **p < 0.01 by Bonferroni's multiple comparisons. I, Experimental traces of the effect of RTN inhibition on respiratory activity in conscious and anesthetized mice in normocapnic (top panel) and hypercapnic (bottom panel) conditions. J, Mean absolute fR before and during RTN inhibition classified in conscious and anesthetized mice in normocapnic conditions. Repeated-measures two-way ANOVA, the interaction between anesthetic and RTN inhibition, F(2, 6) = 11.82, p = 0.008; effect of state, F(2, 6) = 19.68, p = 0.002; effect of RTN inhibition, F(1, 3) = 83.80, p = 0.002. *p < 0.05, **p < 0.01 by Bonferroni's multiple comparisons. K, Mean absolute fR before and during RTN inhibition classified in conscious and anesthetized mice in hypercapnic conditions. Repeated-measures two-way ANOVA, interaction between anesthetic and RTN inhibition, F(2, 6) = 1.661, p = 0.27; effect of state, F(2, 6) = 33.11, p = 0.0006; effect of RTN inhibition, F(1, 3) = 30.01, p = 0.012. *p < 0.05, **p < 0.01 by Bonferroni's multiple comparisons. L, Change in fR during RTN inhibition normalized to baseline fR in conscious (Con) and anesthetized [isoflurane (Iso) or ketamine/dexmedetomidine (KD)] mice in normocapnic and hypercapnic conditions. Two-way ANOVA, interaction between anesthetic and CO2 condition, F(2, 18) = 10, p = 0.001; effect of state, F(2, 18) = 8.6, p = 0.002; effect of CO2 condition, F(1, 18) = 22, p < 0.001). *p < 0.05, **p < 0.01 by Tukey's multiple-comparisons test.
Our fiber photometry recordings show that RTN neurons retain a response to CO2 under anesthesia, but it is not clear from these data whether RTN neurons drive breathing under these conditions. To test this possibility, we inhibited RTN neurons under isoflurane and Ket/Dex anesthesia. In these experiments, RTN neurons were inhibited under conditions matching our recordings of RTN activity, with EEG/EMG activity used to monitor brain state. Under anesthesia, RTN inhibition resulted in a significant reduction in fR in normoxic normocapnic conditions (Fig. 7I,J). However, the magnitude of hypoventilation during RTN inhibition normalized to baseline fR was enhanced by Ket/Dex when compared with the conscious state (Fig. 7L). In hypercapnic conditions, RTN inhibition effectively reduced fR in both conscious and anesthetized conditions (Fig. 7I,K), but the magnitude of hypoventilation during RTN inhibition normalized to baseline fR was enhanced by both anesthetics (Fig. 7L). In summary, RTN neurons promote respiratory activity in both conscious and anesthetized conditions, but their contribution to fR is dependent on the class of anesthetic and the presence of hypercapnia.
Effect of hypoxia on RTN neuron activity
Previous studies indicate that RTN neurons are inhibited during hypoxia in freely behaving rodents (Basting et al., 2015; Souza et al., 2023). Here, we tested the effects of hypoxia (O2 = 10%) on RTN activity measured by fiber photometry (Fig. 8A). Hypoxia led to a reduction in RTN fluorescence in four of six cases, with the remaining two cases showing either an increase or no change in fluorescence (Fig. 8B). In control mice that lack expression of GCaMP, we observed a consistent reduction in corrected fluorescence (Fig. 8B). On average, there was no significant difference in the ΔF/F0 in response to hypoxia between GCaMP and control cases (Fig. 8B,H). Previous studies indicate that the inhibition of RTN activity during hypoxia is related to changes in arterial PCO2. As such, we reasoned that RTN activity during hypoxia might have a relationship to their response to hypercapnia, so we generated scatter plots for the change in fluorescence in response to 9% CO2 versus 10% O2 for GCaMP and control cases. This revealed no clear evidence for a relationship between the hypercapnia and hypoxia response across cases (Fig. 8C). Interestingly, we noted that during hypoxia the increase in RTN fluorescence associated with sighs was attenuated relative to normoxic conditions (Fig. 8D,E). To exclude the possibility that GCaMP activity in RTN neurons is occluded by hypoxia, we exposed mice to hypercapnic hypoxia, which resulted in a marked activation of RTN activity in GCaMP cases and little effect in controls (Fig. 8G–I). In summary, these data show that RTN neurons are not markedly inhibited by hypoxia, but the coupling between respiratory motor activity and RTN activity is dampened during hypoxia.
Effect of hypoxia on RTN neuron activity measured by fiber photometry. A, Representative experimental recording of RTN fluorescence during 10% hypoxia. Note the reduction in fluorescence during hypoxia. B, Grouped time course data for ΔF/F0 during hypoxia from GCaMP and control cases. The top trace represents the mean ± SEM for GCaMP (n = 6) and control (n = 7) cases. The bottom traces represent ΔF/F0 for each case. C, Scatter plot for the ΔF/F0 during hypercapnia and hypoxia for GCaMP and control cases. Correlation between ΔF/F0 during CO2 versus ΔF/F0 during hypoxia for GCaMP (r2 = 0.045, p = 0.685) and control (r2 = 0.36, p = 0.152) cases. D, Representative example of RTN fluorescence during a sigh in normoxia (top panel) and hypoxia (bottom panel). E, Grouped time course data for ΔF/F0 during sigh events in normoxia versus hypoxia in GCaMP cases (mean ± SEM). F, Grouped data of peak fluorescence from RTN neurons during a sigh in normoxia and hypoxia condition (Wilcoxon matched-pairs signed-rank test, p = 0.0313). *p < 0.05. G, Grouped time course data for ΔF/F0 during hypoxia (in blue) and during hypercapnic hypoxia (in purple) from GCaMP cases (n = 6–5). H, Grouped time course data for ΔF/F0 during hypoxia (in blue) and during hypercapnic hypoxia (in gray) from control cases (n = 6–5). I, Grouped data for ΔF/F0 during hypoxia and hypercapnic hypoxia in GCaMP (n = 6–5) and control cases (n = 6 and 5; two-way ANOVA, interaction between condition and GCaMP, F(1, 19) = 6.5, p = 0.0198; effect of condition, F(1, 19) = 17.7, p = 0.0005; effect of GCaMP, F(1, 19) = 7.7, p = 0.0123). **p < 0.01 for GCaMP versus control by Bonferroni's multiple comparisons. †††p < 0.001 for hypoxia versus hypoxia hypercapnia.
Effect of hypercapnia and anesthesia on tissue autofluorescence
Hypercapnia and isoflurane cause cerebral vasodilation leading to increased cerebral blood flow (Ainslie and Duffin, 2009; Sullender et al., 2022), which is known to influence fiber photometry (Simpson et al., 2024). Compounding this issue, RTN neurons have a specialized anatomical and functional relationship with blood vessels in the parafacial region (Lazarenko et al., 2009; Cleary et al., 2020; Fig. 9A). In our recordings, we observed reductions in fluorescence in the 405 nm “isosbestic” channel in response to hypercapnia (Fig. 9B,C) and isoflurane anesthesia (Fig. 9F) in both GCaMP and control cases, which likely reflect changes in cerebral blood flow. Hypercapnia exposure led to a concentration-dependent reduction in isosbestic fluorescence (Fig. 9B). In contrast, hypoxia resulted in a small variable change in isosbestic fluorescence, while hypercapnic hypoxia led to a reduction in isosbestic fluorescence that was comparable in magnitude to 6% hypercapnia (Fig. 9B). The magnitude of the decrease in isosbestic fluorescence during hypercapnia and hypoxia was comparable in GCaMP and control cases (Fig. 9B). Furthermore, the scatter plot of the change in fluorescence from the 405 and 470 nm channels during hypercapnia shows that changes in fluorescence are similar between channels in controls (Fig. 9D), as expected. Conversely, in GCaMP cases, the change in fluorescence from the 470 nm channel was consistently right-shifted relative to that in the 405 nm channel (Fig. 9D). Similar results were obtained during hypoxia (Fig. 9E). We also consistently observed a marked reduction in isosbestic fluorescence in response to isoflurane anesthesia in control and GCaMP cases (GCaMP vs controls; −8.1 ± 2.9 vs −5.6 ± 3.7% from baseline, unpaired t test, t = 1.225, df = 9, p = 0.2516; Fig. 9F), whereas Ket/Dex resulted in increases or decreases in isosbestic fluorescence with no significant difference in the average response between GCaMP and control cases (GCaMP vs control; −1.6 ± 4.5 vs 0.67 ± 3.5% from baseline, unpaired t test, t = 0.919, df = 9, p = 0.3818). Of note, the reduction in isosbestic fluorescence during hypercapnia was blunted during Ket/Dex anesthesia and completely lost under isoflurane anesthesia (Fig. 9F). Finally, we also observed a small reduction in isosbestic fluorescence during periods of REM sleep that was coincident with a reduction of similar magnitude in the 470 nm channel in control cases (Fig. 9G), presumably reflecting an increase in cerebral blood flow during REM sleep (Bergel et al., 2018).
Effect of hypercapnia and anesthesia on tissue autofluorescence. A, Images showing the proximity of blood vessels to GCaMP-expressing neurons and the fiber-optic tip in a representative case. Scale bars: 500 μm for the left image, 250 μm for the right image. B, Group averages for the change in 405 nm channel fluorescence (isosbestic) during hypercapnia and hypoxia in GCaMP (n = 6) and control (n = 5) cases. Two-way ANOVA, interaction between gas condition and presence of GCaMP, F(4, 44) = 0.87, p = 0.49; effect of gas condition, F(4, 44) = 28, p < 0.001; effect of presence of GCaMP, F(1, 44) = 2.1, p = 0.16. There were no significant differences between GCaMP versus control mice by Bonferroni's multiple comparisons. C, Representative experimental recording of corrected fluorescence and the raw emitted fluorescence in the 470 and 405 nm channels during hypercapnia in GCaMP (top panel) and control (bottom panel) cases. D, Scatter plot of the ΔF/F0 for fluorescence in the 470 and 405 nm during hypercapnia in GCaMP (n = 6) and control (n = 5) cases. Each case is represented by three data points reflecting data from 3, 6, and 9% CO2. The red line indicates the line of identity where changes in fluorescence in the 470 and 405 nm are equal. E, Scatter plot of the ΔF/F0 for fluorescence in the 470 and 405 nm during hypoxia in GCaMP (n = 6) and control (n = 5) cases. Each case is represented by two data points. The red line indicates the line of identity where changes in fluorescence in the 470 and 405 nm are equal. F, Experimental recording of corrected fluorescence and the raw emitted fluorescence in the 470 and 405 nm channels during isoflurane in GCaMP (top panel) and control (bottom panel) cases. G, Experimental recording of corrected fluorescence and the raw emitted fluorescence in the 470 and 405 nm channels during REM sleep in GCaMP (top panel) and control (bottom panel) cases.
Discussion
This study demonstrates that Nmb-expressing RTN neurons are activated by CO2 in a concentration-dependent manner, a phenomenon that parallels the increase in pulmonary ventilation in freely behaving adult mice. These CO2-activated RTN neurons have dynamic activity in normocapnic conditions that correlate with breathing patterns and arousal state, with increased activity during REM sleep and active wakefulness relative to quiet resting and NREM sleep. Finally, we confirmed that the activation of RTN neurons by CO2 persists under anesthesia and showed that their contribution to breathing is dependent on the interaction between brain state and inspired CO2.
Technical considerations
Our study demonstrates the utility of fiber photometry for genetically targeted recordings of RTN neuronal activity with high temporal-resolution in freely behaving rodents. Our approach guarantees a high degree of specificity for RTN chemoreceptor neurons as defined by Nmb expression. There are fewer than 700 RTN neurons (<350 per side) in the adult mouse ventrolateral medulla based on Nmb RNA expression (Shi et al., 2017). Therefore, our recordings reflect the population-level activity of fewer than 350 neurons, which is only a small fraction of all neurons present within the parafacial region. Nevertheless, we observed a robust and reproducible increase in fluorescence during hypercapnia, indicating that a majority of RTN neurons are activated in unison during this stimulus. A limitation of our approach is that it cannot resolve heterogeneity among individual RTN neurons and the kinetics of GCaMP do not allow us to quantify breath-to-breath patterns of RTN activity, which vary between cells based on single-cell recordings under anesthesia (Guyenet et al., 2005). As such, our recordings provide a population-level readout of RTN central chemoreceptor neuron activity that overlooks putative heterogeneity in this population. Moreover, due to the kinetic properties of GCaMP, we cannot precisely establish the temporal relationship between RTN activity and respiratory activation.
Another important consideration when using fiber photometry is the effect of cerebral blood flow and blood oxygenation on emitted fluorescence (Ma et al., 2016; W. T. Zhang et al., 2022; Simpson et al., 2024). Both the excitation and emission paths in fiber photometry are subject to absorption by hemoglobin in blood vessels. Moreover, hemoglobin absorption is wavelength dependent and varies as a function of its oxygenation status (W. T. Zhang et al., 2022). Therefore, changes in cerebral blood flow and oxygen saturation have the potential to interfere with the emitted fluorescence in fiber photometry recordings. This issue is particularly important given that hypercapnia and hypoxia affect both parameters (Ainslie and Duffin, 2009). We employed two controls to account for the confounding effects of cerebral blood flow and oxygenation. First, we utilize an isosbestic control channel (405 nm) to measure nonspecific variations in fluorescence concurrently with GCaMP fluorescence. Secondly, we conducted GCaMP-negative controls to validate the matching of the isosbestic and GCaMP excitation wavelengths across conditions. These controls ensure that the conclusions drawn from this study are based on the activity of RTN neurons (Simpson et al., 2024).
RTN neurons are activated by hypercapnia in conscious adult mice
We demonstrate that RTN neurons are potently activated by hypercapnia in freely behaving adult mice. This study provides data showing that the response of RTN neurons to CO2 in vivo has a low recruitment threshold (∼3% CO2) and kinetics that match the HCVR, suggesting that RTN neurons encode arterial PCO2 in vivo. The effect of hypercapnia on RTN activity at a population-level resembles the response of a small number of unidentified CO2-activated neurons found by single-cell calcium imaging using a miniscope in the mouse parafacial region (Bhandare et al., 2022). Moreover, a graded activation of RTN neurons during hypercapnia is supported by the functional effects of RTN inhibition of breathing (Souza et al., 2023). The effects of hypercapnia on RTN neurons in vivo are likely to reflect the convergent effects of intrinsic proton-sensing mechanisms (Wang et al., 2013; Kumar et al., 2015; Gonye et al., 2024) along with paracrine and synaptic signaling (Nattie, 2011; Gourine and Dale, 2022; Guyenet and Bayliss, 2022). Together, this evidence supports the proposal that RTN neurons function as central CO2 chemoreceptors that underpin the HCVR in mammals (Guyenet and Bayliss, 2022; Souza and Abbott, 2024).
RTN activity is coupled with arousal state
A novel finding in this study is that the spontaneous activity of RTN neurons is significantly influenced by the arousal state. During stable NREM sleep, RTN activity remained consistent, with brief bursts of activity associated with tachypnea, sighs, and often coinciding with microarousals. RTN activity tended to increase during waking periods and became highly dynamic during episodes of active wakefulness breathing patterns. The coupling of RTN activity with arousal may be derived from synaptic inputs from arousal-promoting centers in the brainstem and hypothalamus or feedback from the pontomedullary respiratory network (Guyenet and Bayliss, 2022). Importantly, we found that inhibiting RTN neurons was largely ineffective at reducing fR during active wakefulness, whereas the same intervention caused marked hypoventilation during eupneic breathing while awake as well as during NREM sleep. Thus, while RTN neuron activity increases with elevated arousal, these neurons are not essential for the pattern of breathing that accompanies elevated arousal, such as sighs and control of breathing during active wakefulness.
In contrast to wakefulness, REM sleep is associated with a reduction in respiratory motor output and a blunting of the HCVR (Horner et al., 2002; Peever and Fuller, 2016). Interestingly, our data indicate that RTN activity increases during REM sleep. Previous functional data using inhibitory optogenetics showed that the contribution of these neurons to fR are attenuated during REM sleep in rodents (Burke et al., 2015; Souza et al., 2023). The effect of optogenetic stimulation of RTN neurons on minute ventilation is also attenuated during REM sleep (Abbott et al., 2013; Burke et al., 2015). Our results, together with these previous studies, suggest that RTN neurons are functionally decoupled from the respiratory central pattern generator during REM sleep. Several mechanisms may explain the increase in RTN activity during REM sleep. For example, RTN neurons may be activated by synaptic inputs from pontomedullary REM-active neurons (Peever and Fuller, 2016) or respiratory-related neurons in the ventral respiratory column (J. M. Orem et al., 2005). RTN neurons could also be influenced by changes in neuromodulatory tone during REM sleep (Peever and Fuller, 2016) or potentially by changes in arterial PCO2 that are known to accompany REM sleep (Douglas et al., 1982; J. Orem et al., 2000). In summary, while RTN activity in normocapnic conditions is dynamically coupled with changes in arousal states, the primary function of these neurons is to generate a tonic respiratory drive that supports eupneic breathing.
RTN control of breathing under anesthesia
Anesthetics blunt several reflexes (Franks, 2008; Brown et al., 2011), including the HCVR (Nattie, 2011). The blunting of the HCVR is considered to be multifactorial, with changes in respiratory chemoreceptor function being a contributing factor (Comroe, 1967; Nattie, 2011). Our recordings suggest that the basal activity of RTN neurons in intact mice is not markedly inhibited by isoflurane and increases with Ket/Dex anesthesia. An important consideration for these results is that RTN activity under these conditions reflects the direct actions of the anesthetic on RTN neurons, general depressant effects on other CNS cells, and changes in arterial blood gas due to hypoventilation. Notably, hypercapnia exposure resulted in an increase in RTN activity under both anesthetic regimes, demonstrating that the activation of RTN neurons by hypercapnia does not rely on inputs silenced by anesthesia, including other potential central chemoreceptors such as serotonin, noradrenaline, and orexin (Franks, 2008; Brown et al., 2011; Massey et al., 2015). However, the kinetics of RTN activity during hypercapnia were affected by Ket/Dex anesthesia. The altered response of RTN neurons to hypercapnia under Ket/Dex anesthesia could reflect differences in baseline PCO2 in normocapnic conditions, a disruption of cerebral blood flow affecting CO2 and pH concentrations within the RTN, or the silencing of critical inputs that facilitate hypercapnia's effect on RTN neurons. Interestingly, the effect of RTN inhibition on respiratory frequency was present under anesthesia and was enhanced in the presence of hypercapnia, indicating that RTN activity is an important driver of ventilation under anesthesia, as previously suggested (Lazarenko et al., 2010; Bourgeois et al., 2019).
Are RTN neurons inhibited by hypoxia?
Electrophysiological single-unit recordings in anesthetized mechanically ventilated animals show RTN neurons are activated by carotid body stimulation (Takakura et al., 2006). On the other hand, evidence from loss-of-function optogenetics indicates that RTN neurons are silenced during hypoxia via alkalosis (Basting et al., 2015; Souza et al., 2023). In contrast, our photometry recordings reveal that RTN activity does not consistently increase or decrease during hypoxia in freely behaving mice. Two factors influence the interpretation of this result. First, the discharge frequency of RTN neurons in normocapnic conditions (the baseline condition) is low (Mulkey et al., 2004; Guyenet et al., 2005), making it difficult to detect a reduction in GCaMP fluorescence with the variant of GCaMP used in these experiments (Y. Zhang et al., 2023). Second, it is possible that only a subset of RTN neurons is inhibited by hypoxia, while another subset could potentially be activated. We found that sighs during hypoxia were not coupled with an increase in RTN activity, which may indicate that sighs during hypoxia are driven by mechanisms that differ from those driving sighs associated with microarousals, with only the latter modulating RTN activity. Overall, while hypoxia appears to influence RTN function in vivo, our recordings suggest that the effect of hypoxia on RTN neurons in conscious animals is variable.
Conclusion
In conclusion, this study demonstrates that RTN chemoreceptor neurons integrate intrinsic CO2/pH sensitivity with potential CO2-dependent inputs, as well as arousal and respiratory-related inputs, in freely behaving adult mice. Additionally, our data show that the contribution of RTN neurons to respiratory frequency depends on the interaction between brain state and arterial PCO2. RTN neurons are critical for ventilation during NREM sleep, under anesthesia, and during eupneic breathing in the wake state. However, RTN neurons are not required to drive breathing during active wakefulness and appear to be functionally decoupled from breathing during REM sleep. These findings have significant implications for understanding the central control of breathing across various physiological states and pathophysiological disorders, particularly during sleep and anesthesia.
Antibodies and ISH probes
Footnotes
This work was supported by the National Institutes of Health grant HL148004 to S.B.G.A.
The authors declare no competing financial interests.
- Correspondence should be addressed to Stephen B.G. Abbott at sba6t{at}virginia.edu.
