Is plasticity within the retrotrapezoid nucleus responsible for the recovery of the $P_{\mathrm{C}{\mathrm{O}}_2}$ set‐point after carotid body denervation in rats?

Abstract

Key points

• Arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ is kept constant via breathing adjustments elicited, at least partly, by central chemoreceptors (CCRs) and the carotid bodies (CBs).

• The CBs may be active in a normal oxygen environment because their removal reduces breathing. Thereafter, breathing slowly returns to normal. In the present study, we investigated whether an increase in the activity of CCRs accounts for this return.

• One week after CB excision, the hypoxic ventilatory reflex was greatly reduced as expected, whereas ventilation and blood gases at rest under normoxia were normal.

• Optogenetic inhibition of Phox2b‐expressing neurons including the retrotrapezoid nucleus, a cluster of CCRs, reduced breathing proportionally to arterial pH. The hypopnoea was greater after CB excision but only in a normal or hypoxic environment. The difference could be simply explained by the loss of fast feedback from the CBs.

• We conclude that, in rats, CB denervation may not produce CCR plasticity. We also question whether the transient hypoventilation elicited by CB denervation means that these afferents are active under normoxia.

Abstract

Carotid body denervation (CBD) causes hypoventilation and increases the arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ set‐point; these effects eventually subside. The hypoventilation is attributed to reduced CB afferent activity and the $P_{\mathrm{C}{\mathrm{O}}_2}$ set‐point recovery to CNS plasticity. In the present study, we investigated whether the retrotrapezoid nucleus (RTN), a group of non‐catecholaminergic Phox2b‐expressing central respiratory chemoreceptors (CCRs), is the site of such plasticity. We evaluated the contribution of the RTN to breathing frequency (F _R), tidal volume (V _T) and minute volume (V _E) by inhibiting this nucleus optogenetically for 10 s (archaerhodopsinT3.0) in unanaesthetized rats breathing various levels of O₂ and/or CO₂. The measurements were made in seven rats before and 6–7 days after CBD and were repeated in seven sham‐operated rats. Seven days post‐CBD, blood gases and ventilation in 21% O₂ were normal, whereas the hypoxic ventilatory reflex was still depressed (95.3%) and hypoxia no longer evoked sighs. Sham surgery had no effect. In normoxia or hypoxia, RTN inhibition produced a more sustained hypopnoea post‐CBD than before; in hyperoxia, the responses were identical. Post‐CBD, RTN inhibition reduced F _R and V _E in proportion to arterial pH or $P_{\mathrm{C}{\mathrm{O}}_2}$ (ΔV _E: 3.3 ± 1.5% resting V _E/0.01 pHa). In these rats, 20.7 ± 8.9% of RTN neurons expressed archaerhodopsinT3.0. Hypercapnia (3–6% FiCO₂) increased F _R and V _T in CBD rats (n = 4). In conclusion, RTN regulates F _R and V _E in a pH‐dependent manner after CBD, consistent with its postulated CCR function. RTN inhibition produces a more sustained hypopnoea after CBD than before, although this change may simply result from the loss of the fast feedback action of the CBs.

Key points

• Arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ is kept constant via breathing adjustments elicited, at least partly, by central chemoreceptors (CCRs) and the carotid bodies (CBs).

• The CBs may be active in a normal oxygen environment because their removal reduces breathing. Thereafter, breathing slowly returns to normal. In the present study, we investigated whether an increase in the activity of CCRs accounts for this return.

• One week after CB excision, the hypoxic ventilatory reflex was greatly reduced as expected, whereas ventilation and blood gases at rest under normoxia were normal.

• Optogenetic inhibition of Phox2b‐expressing neurons including the retrotrapezoid nucleus, a cluster of CCRs, reduced breathing proportionally to arterial pH. The hypopnoea was greater after CB excision but only in a normal or hypoxic environment. The difference could be simply explained by the loss of fast feedback from the CBs.

• We conclude that, in rats, CB denervation may not produce CCR plasticity. We also question whether the transient hypoventilation elicited by CB denervation means that these afferents are active under normoxia.

Abbreviations

CB

carotid body

CBD

carotid body denervation

CCR

central chemoreceptor

EMG

electromyogram

F_R

breathing frequency

HVR

hypoxic ventilatory reflex

LVV

lentiviral vector

REM

rapid eye movement

RTN

retrotrapezoid nucleus

V_E

minute ventilation

V_T

tidal volume

Introduction

The carotid bodies (CBs) encode arterial $P_{{\mathrm{O}}_2}$ in a pH‐dependent manner and their activation during hypoxia stimulates breathing and the cardiovascular system (Nurse, 2014; Prabhakar & Semenza, 2015). The CBs may also contribute to resting ventilation under normoxic conditions (Dejours, 1962; Blain et al. 2010; Smith et al. 2010; Hodges & Forster, 2012). The main supportive evidence for CB contributing to resting ventilation under normoxia is that carotid body denervation (CBD) produces a long‐lasting (days to weeks) hypoventilation and elevates the steady‐state level of arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ ($P_{\mathrm{aC}{\mathrm{O}}_2}$) in all mammals, including humans (Forster & Smith, 2010; Hodges & Forster, 2012; Mouradian et al. 2012). Also, in unanaesthetized dogs, resting ventilation is reduced and arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ increases when an isolated CB is perfused with hyperoxic blood (Blain et al. 2009, 2010).

The hypoventilation and hypercapnia consecutive to CBD decline with time and the hypoxic ventilatory reflex (HVR) recovers to a variable extent. The degree and time‐course of the recovery are species‐dependent (Martin‐Body et al. 1986; Forster, 2003; Timmers et al. 2003; Hodges & Forster, 2012; Mouradian et al. 2012; Angelova et al. 2015). The recovery of the HVR has two potential causes. Accessory peripheral chemoreceptors (ectopic glomus cells and aortic bodies, among others) (Dejours, 1962; McDonald & Blewett, 1981) may gradually become more sensitive to hypoxia, develop new CNS connections and eventually compensate for the absence of CB input. In rats, the brainstem respiratory network may develop the ability to respond directly to hypoxia after CBD (Angelova et al. 2015), although this does not appear to be the case in mice; in this species, the HVR was permanently abolished by deleting a gene that encodes a protein required for O₂ sensing by the CBs but absent from the brain (Olf58 knock out) (Chang et al. 2015).

The surgical procedure used to remove the carotid bodies and their innervation may influence both the degree and duration of the subsequent hypoventilation and hypercapnia (Olson et al. 1988; Mouradian et al. 2012). Therefore, collateral damage to structures other than the carotid bodies may also contribute to the breathing changes. Indeed, the return to normal ventilation and $P_{\mathrm{C}{\mathrm{O}}_2}$ appears to be poorly correlated with the recovery of a ventilatory reflex to cyanide or hypoxia (Forster, 2003; Hodges et al. 2005; Mouradian et al. 2012). This discrepancy implies that some form of central plasticity of the respiratory network develops that is unrelated to oxygen sensing. This interpretation is supported by evidence of neurochemical changes in the brainstem, as well as changes in the respiratory stimulation elicited by activating raphe neurons (Hodges et al. 2005; Mouradian et al. 2012; Miller et al. 2013).

In the present study, we investigated whether a compensatory increase in the activity of central respiratory chemoreceptors contributes to the recovery of ventilation and the return to control of the arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ set‐point after CBD. We focused on a well characterized group of chemoreceptors: the retrotrapezoid nucleus (RTN). RTN neurons encode CNS $P_{\mathrm{C}{\mathrm{O}}_2}$ and operate as a nodal point for the regulation of breathing by CO₂ (Gourine et al. 2010; Basting et al. 2015; Guyenet & Bayliss, 2015; Kumar et al. 2015). Inhibiting these neurons briefly (10 s) produces hypopnoea of a magnitude that is a measure of their contribution to ventilation at the precise instant when the inhibitory stimulus is applied (Basting et al. 2015). In the present study, we use the same optogenetic approach to test two hypotheses (i) after CBD, the contribution of RTN neurons to ventilation remains proportional to arterial pH or $P_{\mathrm{C}{\mathrm{O}}_2}$ and (ii) 1 week after CBD, breathing returns to normal because RTN neurons are more active and this increased activity compensates for the loss of CB input.

Methods

Animals and ethical approval

All procedures conformed to the NIH Guide for the Care and Use of Laboratory Animals and were also approved by the University of Virginia Animal Care and Use Committee. The investigators understand the ethical principles under which The Journal of Physiology operates and our work complies with the animal ethics checklist provided by the journal. Experiments were performed on male Sprague–Dawley rats (n = 24; weighing 400–550 g; Taconic, Hudson, NY, USA). Animals were housed under a standard 12:12 h light/dark cycle with ad libitum access to food and water.

Lentiviral construct and vector preparation

We used a lentiviral vector (LVV) that expresses the photoactivatable proton pump ArchT3.0 (Chow et al. 2010; Han et al. 2011; Mattis et al. 2012) under the Phox2b‐responsive promoter PRSx8 (pLenti‐PRSX8 ArchT3.0‐eYFP). PRSx8‐ArchT‐eYFP was generated by replacing the CamKIIɑ promoter present in an existing pLenti CamKIIɑ ArchT 3.0 eYFP construct (courtesy of K. Deisseroth via Addgene, Cambridge, MA, USA) by the PRSx8 promoter (kindly provided by M. Raizada, University of Florida, Gainesville, FL, USA). The LVV was produced by the University of North Carolina virus core at 3.0 × 10⁸ viral particles per ml.

Injections of vector

The LVV was microinjected bilaterally under the facial motor nucleus in 14 rats as described previously (Basting et al. 2015). Rats were anaesthetized with a mixture of ketamine (75 mg kg⁻¹), xylazine (5 mg kg⁻¹) and acepromazine (1 mg kg⁻¹) given i.p. The depth of anaesthesia was assessed by an absence of the corneal and hind‐paw withdrawal reflex. Additional anaesthetic was administered as necessary (25% of the original dose, i.p. or i.m. during surgery). Body temperature was maintained close to 37°C with a servocontrolled heating pad and a blanket. All surgical procedures were performed under aseptic conditions. The hair over the skull, neck and cheek was removed and the skin was disinfected. A skin incision was made over the left mandible to expose the facial nerve. This nerve was later stimulated to elicit anti‐dromic field potentials in the facial motor nucleus (Brown & Guyenet, 1985). The rat was then placed prone on a stereotaxic apparatus (bite bar set at –3.0 mm for flat skull; David Kopf Instruments, Tujunga, CA, USA). A 1.5 mm diameter hole was drilled into the occipital plate on both sides caudal to the parieto‐occipital suture. The LVV solution was loaded into a 1.2 mm internal diameter glass pipette broken to a 25 μm tip (external diameter). Anti‐dromic field potentials were recorded through the pipette to identify the location of the RTN, which resides immediately below the facial motor nucleus. All injections were made 50–100 μm below this nucleus. The vector was pressure‐injected bilaterally into five sites separated by 200 μm (total volume 500–700 nl side^–1). The caudal‐most site was located below the caudal end of the facial motor nucleus (mean stereotaxic co‐ordinate: 2.6 mm caudal to the parieto‐occipital suture, 2.0 mm lateral to the midline, 8.3 mm below the cerebellar surface; individual co‐ordinates varied by up to 400 μm in every direction in accordance with the location of the facial motor nucleus).

Next, we implanted electrodes for recording the EEG and neck electromyogram (EMG) and we inserted two optical fibre/ferrule assemblies for light stimulation of ArchT‐expressing neurons (one on each side, for further details, see Basting et al. 2015). The bottoms of the optical fibres (200 μm diameter) were placed 300 μm dorsal to the middle of the LVV injection sites. The head‐stage (EEG and EMG connections) and optic fibre‐ferrules were secured to the skull using a two‐part epoxy adhesive (Loctite Corp., Düsseldorf, Germany). Incisions were then closed in two layers (muscle and skin) with absorbable sutures and vet bond adhesive. Rats received post‐operative ampicillin (125 mg kg⁻¹, i.p.) and ketoprofen (3–5 mg kg⁻¹, s.c.) and were monitored daily. Rats recovered for 4 weeks for functional expression of the opsin.

CB excision and CBD

Eleven rats were subjected to CBD and seven animals had sham surgery. The first group included seven of the rats that had received LVV injections 4–6 weeks previously. CBD and sham surgery were performed under aseptic conditions with the body temperature maintained close to 37°C. Anaesthesia was induced with 5% isoflurane anaesthesia in pure oxygen and maintained with 2% isoflurane in oxygen during surgery. Depth of anaesthesia was assessed by an absence of the corneal and hind‐paw withdrawal reflex. The CBs were isolated following a mid‐line incision in the ventral surface of the neck and physically removed with forceps. In addition, the carotid sinus nerves were cut and phenol was applied to the carotid bifurcation. This surgical procedure also destroyed the carotid baroreceptors. The incisions were then closed in two layers (muscle and skin) with absorbable sutures and vet bond adhesive. Rats received post‐operative ampicillin (125 mg kg⁻¹, i.p.) and ketoprofen (3‐5 mg kg⁻¹, s.c.) and were monitored daily. Sham surgery consisted of exposing the carotid bifurcation bilaterally without damaging the carotid bodies or the surrounding nerves. All rats were given 6–7 days to recover from surgery before their respiratory chemoreflex and ventilatory response to bilateral RTN inhibition was retested.

Physiological experiments in freely‐behaving rats with unrestrained whole body plethysmography

The animals were tested in a Buxco plethysmography chamber (Data Sciences International, St Paul, MN, USA) modified to allow tethered EEG/EMG recordings and optical stimulation. Before the actual experiments were run, the rats were repeatedly habituated to these surroundings, which were visually‐isolated and with low‐ambient noise. Low breathing rates under normoxia (65–80 breaths min^–1) and a regular sleep pattern were taken as evidence of low stress on the experimental days. On those days, rats were lightly anaesthetized with isoflurane (induction with 5%, maintenance with 2% in pure oxygen for <1 min) to permit cleaning of hardware and connection to the ferrule and EEG/EMG recording assembly. A 200 μm‐thick multimode optical fibre terminated with a ferrule was mated to each of the two implanted ferrules with a zirconia sleeve. Optical matching gel (Fibre Instrument Sales, Oriskany, NY, USA) was applied at the ferrule junction to reduce light loss. A minimum of 1 h was allowed for recovery from anaesthesia and the emergence of stable sleep/wake patterns. Recordings were made between 10.00 h and 18.00 h, with a minimum of 3 days of rest between tests. The ventilatory response to RTN inhibition was assessed using barometric, unrestrained whole‐body plethysmography (EMKA Technologies, Paris, France). The plethysmography chamber was continuously flushed with 1.5 l min⁻¹ of O₂ balanced with N₂, and CO₂ (3%) was added occasionally. The gas mixture was regulated by computer‐driven mass flow controllers for O₂, N₂ and CO₂ (Alicat Scientific Inc., Tucson, AZ, USA). Temperature and humidity within the plethysmography chamber were kept constant (25°C and 42% relative humidity).

Photoinhibition of ArchT3.0‐expressing neurons was achieved with a green laser (532 nm; Shanghai Laser and Optics Century Co. Ltd, Shanghai, China) controlled by a Grass model S88 stimulator (Astro‐Med, West Warwick, RI, USA). The green light was applied in 10 s episodes of continuous illumination. The transmission efficiency of each implantable optical fibre was tested prior to implantation with a light meter (Thorlabs, Newton, NJ, USA). Based on this test, the power of the laser was adjusted so that the final output at the tip of the implanted fibre was estimated between 6–7 mW in vivo.

Indwelling femoral artery catheter implantation and blood gas measurements

Blood gases were measured in 16 rats. This cohort included six fully instrumented rats subjected to CBD 7 days previously, another six fully instrumented rats that had been sham‐operated (also 7 days previously) and four uninstrumented rats (without vector injection, optical fibre or head‐set for EEG measurements). An indwelling femoral artery catheter was implanted when the rats were anaesthetized with isoflurane (5% induction and 2% for the maintenance of anaesthesia in pure oxygen). The hair on the rat's upper back, neck and right inner leg was shaved and the skin disinfected. One incision on the rat's inner leg was made to expose the femoral artery and another was made between the shoulder blades to exteriorize the catheter. A catheter (PE50) filled with 100 U ml⁻¹ heparinized saline was implanted into the right femoral artery. The catheter was threaded s.c. from the femoral artery along the rat's back to emerge between the rats shoulder blades. Incisions were then closed in two layers (muscle and skin) with absorbable sutures and vet bond adhesive, bupivacaine was infiltrated around the skin incisions, and the rats received the standard post‐operative doses of ketoprofen and ampicillin (see above). The animal was then placed in the plethysmography chamber and the catheter was connected through a harness to a swivel system that allowed the animal to move freely within the chamber and the experimentor to withdraw blood without disturbing the animal. After 3 h of recovery within the chamber, the animal were exposed to the various hypoxic or hypercapnic gas mixtures of interest for 30 min to insure that a steady‐state had been achieved. Arterial blood (0.2 ml) was then withdrawn and arterial blood gases and other parameters (pH, $P_{\mathrm{C}{\mathrm{O}}_2}$, $P_{{\mathrm{O}}_2}$ and HCO₃ ⁻) were measured with a VetScan Blood Gas Analyser (Abaxis, Union City, CA, USA). To maintain patency of the line, the catheter was flushed with 100 U ml^–1 heparinized saline after each blood withdrawal. After all of the blood sampling had been performed (5–6 h after catheter implantation), the animals were killed with an overdose of anaesthetic (5% isoflurane delivered to the chamber followed by an intra‐arterial injection of 70–90 mg kg⁻¹ of Nembutal (Lundbeck, Copenhagen, Denmark). The rats were then perfused with aldehyde and their brains harvested for histology (see below).

Data acquisition and analysis

Physiological signals were processed using Spike, version 7.03 (Cambridge Electronic Design, Cambridge, UK). EEG and EMG signals were amplified and bandpass filtered (EEG: 0.1–100 Hz, ×1000. EMG: 300–3000 Hz, ×1000) and acquired at a sampling frequency of 1 kHz. The signal generated by the differential pressure transducer connected to the plethysmography chamber was amplified and bandpass filtered (0.1–15 Hz, ×100) and acquired at a sampling frequency of 1 kHz. Satisfactory EEG/EMG recordings were obtained in eight of 14 rats. In these rats, periods of wake or natural sleep were classified on the basis of EEG, EMG activity, patterns of breathing activity and animal posture. During non‐rapid eye movement (REM) sleep, EEG spectra were dominated by delta activity (0.5–4 Hz) with little or no EMG activity and a stable breathing pattern. Quiet wake was characterized by a reduction in total power with EMG tone and labile breathing. REM sleep was characterized by theta activity (6–7 Hz), loss of EMG and labile breathing. In the other six rats for which EEG/EMG recordings were not available, RTN inhibition was performed when the breathing rate was regular and the animals were either immobile or exhibiting the curled up posture characteristic of sleep. Light was not applied during REM sleep, a state that was instantly recognizable, with or without EEG recording, by the combination of a sleeping posture and breathing irregularities. A minimum of six photoactivation trials was conducted in non‐REM and/or quiet wake at each level of inspired O₂ (FiO₂ 65%, 21%, 18%, 15% or 12%, balance N₂, or FiO₂ 21% + FiCO₂ 3% balance N₂) and the responses were averaged. Trials were conducted during non‐REM sleep or quiet waking only. No trial was conducted during REM sleep, a state in which breathing is labile and the contribution of RTN to breathing is drastically reduced (Burke et al. 2015). Trials during which the animals moved, as indicated by EMG activity, were also ignored. For each trial, breathing parameters were measured during the 10 s preceding light delivery (baseline parameters) and when the light was applied (10 s). Breathing frequency (F _R, breaths min^–1) and tidal volume (V _T, ml per 100 g body weight) were calculated using Spike, version 7.3 (CED). V _T was derived from the integral of the flow signal during inspiration following calibration to waveforms generated by injecting 5 ml of dry air into the chamber. Minute ventilation (MV = F _R × V _T,) was expressed as ml (100 g body weight)^–1 min^–1. Event triggered breathing responses (Fig. 6) were an average of three to six photoactivation trials at each level of inspired O₂ for each of the seven rats both before and after CBD. Sigh data were collected in CBD‐denervated or sham‐denervated rats for a total of 10 min under each of the gas conditions described above when the animal was quietly awake or in non‐REM sleep.

Figure 6. Kinetics of the hypopnoea elicited by bilateral inhibition of ArchT‐expressing neurons .

Graphs represent event‐triggered averages of the frequency and amplitude components of the breathing response observed at three levels of FiO₂ before and after CBD. Four to six responses were collected at each FiO₂ level in every rat (n = 7) to generate a single average per rat. Seven such responses were finally averaged. The continuous lines represent these grand averages and the dotted lines indicate the 95% confidence intervals.

Histology

The animals that had received injections of vector were deeply anaesthetized with 5% isoflurane followed by a lethal dose of ketamine/xylazine/acepromazine (75:5:1 mg kg⁻¹ i.m.) or, in the case of the animals instrumented with intra‐arterial catheters, with 5% isoflurane followed by a lethal dose of Nembutal (70–80 mg intra‐arterially). These animals were then perfused transcardially with 4% paraformaldehyde, and the brains removed and processed as described previously (Abbott et al. 2012). Immunohistochemistry with antibodies against tyrosine hydroxylase (sheep anti‐TH, dilution 1: 2000; Millipore, Billerica, MA, USA), eYFP (to detect ArchT3.0; chicken anti‐GFP, dilution 1:1000; AVES Labs, Tigard, OR, USA) and Phox2b, a marker of RTN chemoreceptors (rabbit anti‐Phox2b, dilution 1:8000; a gift from J. F. Brunet, Ecole Normale Superieure, Paris, France) were performed as described previously (Abbott et al. 2009). Cell mapping and counting and photography were carried out using the Neurolucida system (MicroBrightfield, Inc., Colchester, VT, USA) with an Axioskop microscope driven stage and an AxioCam MRc camera (Carl Zeiss, Oberkochen, Germany). Cell counts were taken from a one‐in‐six series of sections and only profiles containing a nucleus were counted.

Statistical analysis

Differences within and between groups were determined using a two‐tailed paired Student's t test, a one‐way repeated measures ANOVA with Dunnett's multiple comparisons or two‐way repeated measures ANOVA with Sidak's multiple comparisons in Prism, version 6 (GraphPad Software Inc., San Diego, CA, USA). All group data, between and within groups, passed the Shapiro–Wilkes test on error residuals for normality as determined using the R statistical software package (R Foundation for Statistical Computing, Vienna, Austria). The results are generally expressed as the mean ± SD. The mean without the SD is represented in plots in which individual values are shown. Statistical significance is indicated as appropriate (*P < 0.05; **P < 0.01; ***P < 0.005).

Results

Arterial blood gases in control and CBD rats exposed to various O₂ concentrations

Blood gases ($P_{\mathrm{C}{\mathrm{O}}_2}$,$P_{{\mathrm{O}}_2}$), pH and [HCO₃ ⁻] were measured when the rats were exposed to 65%, 21%, 15%, 12% FiO₂ and 12% FiO₂ plus 3% FiCO₂ within the plethysmography chamber. This particular FiCO₂ level was selected because it almost compensates for the respiratory alkalosis elicited by 12% FiO₂. The measurements were made in six of the seven CBD rats 7 days after neck surgery and in six rats with intact carotid bodies. The rats were habituated to their surroundings and blood was remotely withdrawn when the animals were quiescent, as monitored by visual inspection and the plethysmography signal. Under hyperoxia and normoxia, rats were either asleep or quietly awake. Under hypoxia (12% FiO₂), rats were typically quietly awake.

In the six control rats, graded hypoxia produced graded respiratory alkalosis (blood alkalization, hypocapnia and reduced [HCO₃ ⁻]) whereas hyperoxia caused a minor respiratory acidosis (Fig. 1 A, B and D and Table 1). These effects were evaluated by one‐way repeated‐measures ANOVA across four levels of FiO₂ and were highly significant (for pH: P = 0.001; Dunnett's multiple comparisons for pH: 21% vs. 65% FiO₂, P = 0.0104; 21% vs. 15% FiO₂, P < 0.0001; 21% vs. 12% FiO₂, P < 0.0001; for $P_{\mathrm{C}{\mathrm{O}}_2}$: P < 0.0001; Dunnett's multiple comparisons for $P_{\mathrm{C}{\mathrm{O}}_2}$: 21% vs. 65% FiO₂, P = 0.0551; 21% vs. 15% FiO₂, P = 0.0184; 21% vs. 12% FiO₂, P = 0.0004; for HCO₃: P < 0.0001; Dunnett's multiple comparisons for HCO₃: 21% vs. 65% FiO₂, P = 0.4159; 21% vs. 15% FiO₂, P = 0.0120; 21% vs. 12% FiO₂, P < 0.0001). In six CBD rats, one‐way repeated measures ANOVA also showed a significant overall effect of FiO₂ on arterial blood pH, $P_{\mathrm{C}{\mathrm{O}}_2}$ and [HCO₃ ⁻] but Dunnett's multiple comparison test revealed that significant group differences from room air conditions were only present at 12% FiO₂ (pH: F = 18.29, P = 0.0004, 21% vs. 12% FiO₂, P = 0.008; $P_{\mathrm{C}{\mathrm{O}}_2}$: F = 32.51, P < 0.0001, 21% vs. 12% FiO₂, P = 0.0009; HCO₃ ⁻: F = 12.99, P = 0.0068; 21% vs. 12% FiO₂, P = 0.0326) (Fig. 1 A, B and D).

Figure 1. Arterial blood parameters at four levels of FiO₂ in intact vs. CB denervated (CBD) rats .

A, arterial pH vs. FiO₂. B, arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ vs. FiO₂. C, arterial $P_{{\mathrm{O}}_2}$ vs. FiO₂. D, arterial [HCO₃] vs. FiO₂. Control: n = 6; CBD: n = 7). *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001 (two‐way ANOVA).

Table 1.

Blood parameters in intact rats and in CB‐denervated rats

FiO2/FiCO2	pH	$P_{\mathrm{C}{\mathrm{O}}_2}$	$P_{{\mathrm{O}}_2}$	HCO3
Intact rats (n = 6)
65%/0%	7.4 ± 0.03*	49.2 ± 5.7	270.8 ± 20***	30.4 ± 2.7
21%/0%	7.44 ± 0.02	41.8 ± 4.7	78.3 ± 7.4	29 ± 1.8
15%/0%	7.51 ± 0.02***	32.6 ± 2.2*	51.5 ± 6.3***	26.1 ± 1.9*
12%/0%	7.59 ± 0.03***	25 ± 1.4***	43 ± 5.9***	23.9 ± 2.2***
12%/3%	7.48 ± 0.04**	34.5 ± 2.03	54.6 ± 7.4***	26 ± 3.2
CBD rats (n = 6)
65%/0%	7.45 ± 0.01	44 ± 4.1	270.5 ± 20.8***	30.6 ± 2.3
21%/0%	7.44 ± 0.02	42.7 ± 5.1	79.3 ± 10.4	28.9 ± 2.9
18%/0%	7.43 ± 0.01	43.78 ± 4.7	60 ± 6.1**	28.9 ± 2.9
15%/0%	7.46 ± 0.02	39.8 ± 3.5	45.7 ± 5.3***	27.8 ± 2.3
12%/0%	7.49 ± 0.03	35.5 ± 3.1***	34 ± 4.4***	27.5 ± 3*
12%/3%	7.43 ± 0.01	42 ± 4.6	48.8 ± 5.8***	27.6 ± 2.7

All values are compared with those found in rats exposed to 21% FiO₂. *P < 0.05; *P < 0.01; ***P < 0.005; one‐way repeated measures ANOVA followed by Dunnett's multiple comparisons test.

The effects of CBD on arterial blood parameters (pH, $P_{\mathrm{C}{\mathrm{O}}_2}$, $P_{{\mathrm{O}}_2}$ and [HCO₃ ⁻]) were evaluated with two‐way repeated‐measures ANOVA for two factors (FiO₂, four levels; rat group i.e. CBD vs. intact; n = 6 per group) (Fig. 1). There were significant differences of arterial pH between control and CBD animals except under normoxia (effect of FiO₂ on pH, P < 0.0001; effect of CBD on pH, P < 0.0322; interaction, P < 0.0001; Sidak's multiple comparisons for pH before and after CBD: 65% FiO₂, P = 0.0048; 21% FiO₂, P > 0.9; 15% FiO₂, P = 0.008; 12% FiO₂, P < 0.0001). Significant differences in arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ were also identified (effect of FiO₂ on $P_{\mathrm{C}{\mathrm{O}}_2}$, P < 0.0001; effect of CBD on $P_{\mathrm{C}{\mathrm{O}}_2}$, P = 0.0987; interaction, P < 0.0001; Sidak's multiple comparisons for arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ in intact vs. CBD rats: 65% FiO₂, P = 0.119; 21% FiO₂, P > 0.9; 15% FiO₂, P = 0.0128; 12% FiO₂, P = 0.0002) (Fig. 1 B). Significant differences in arterial HCO₃ ⁻ were also found after CBD (effect of FiO₂ on HCO₃ ⁻, P < 0.0001; effect of CBD on HCO₃ ⁻, P = 0.3204; interaction, P < 0.0019). The statistical analysis revealed no effect of CBD on arterial $P_{{\mathrm{O}}_2}$ (effect of CBD on $P_{{\mathrm{O}}_2}$, P = 0.3803; interaction, P = 0.6704) (Fig. 1 C). However, CBD animals had consistently lower $P_{{\mathrm{O}}_2}$ levels under 12% hypoxia (effect of FiO₂ on $P_{{\mathrm{O}}_2}$, P < 0.0001) (Table 1).

In summary, exposure to hypoxia in intact quietly resting rats produced the expected respiratory alkalosis. After CBD, the alkalosis was much reduced but not eliminated. At 12% FiO₂, hypoxia still produced a slight but significant rise in arterial pH and a decrease of both arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ and [HCO₃ ⁻] relative to room air conditions. Arterial $P_{{\mathrm{O}}_2}$ was slightly lower at 15% and 12% FiO₂ in CBD vs. intact rats (Fig. 1 C), although the difference did not reach statistical significance.

Effect of FiO₂ on breathing in CBD or sham‐operated rats

The effects of CBD on the ventilatory response to hypoxia or hyperoxia (Fig. 2 A) were evaluated with two‐way repeated measures ANOVA for two factors (FiO₂, four levels; rat group, pre‐ vs. post sham or CBD surgery; seven rats per group). There was a significant effect of FiO₂ and CBD on F _R (FiO₂ effect, P < 0.0001; CBD effect, P = 0.0009; interaction, P = 0.0105; Sidak's multiple comparisons for F _R before vs. after CBD: 65% FiO₂, P = 0.006; 21% FiO₂, P = 0.004; 15% FiO₂, P < 0.0001; 12% FiO₂, P < 0.0001). There was a significant overall effect of FiO₂ on V _T (P < 0.0001; the trend was towards smaller V _T under hypoxia) but no overall effect of CBD on V _T (P = 0.6256) and no interaction (P = 0.3445). Finally, there was a significant effect of FiO₂ and CBD on V _E (effect of FiO₂, P = 0.0002; effect of CBD, P = 0.0322; interaction, P = 0.0002; Sidak's multiple comparisons for V _E before vs. after CBD: 65% FiO₂, P = 0.3062; 21% FiO₂, P > 0.9; 15% FiO₂, P = 0.0028; 12% FiO₂, P < 0.0001) (Fig. 2 A). Sham denervation had no effect on the ventilatory response to hypoxia (Fig. 2 B) (two‐way repeated measures ANOVA: FiO₂ effect, P < 0.0001; sCBD effect, P = 0.3932; interaction, P = 0.8742). Before denervation, F _R was significantly higher at 12% FiO₂ than at 21% FiO₂ (P = 0.0002) and V _E was significantly higher at 12% FiO₂ compared to normoxia as well (P = 0.0013). However, FiO₂ had no effect on V _T in these intact rats. After denervation, hypoxia had very little effect on any of the breathing parameters, with only a slight increase in F _R in 12% FiO₂ (one‐way repeated measures ANOVA: after denervation 21% FiO₂ vs. 12%, P = 0.0013). Of note, CBD significantly reduced F _R under normoxia but had no effect on V _T or V _E.

Figure 2. Effect of FiO₂ on breathing before and after CBD or sham surgery .

Aa–Ac, from top to bottom, effect of FiO₂ on breathing frequency, tidal volume and minute volume before and 7 days after CBD (n = 7). Ba–Bc, effect of FiO₂ on breathing before and 7 days after sham surgery (n = 7). *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001 (two‐way ANOVA).

In brief, the hypoxic ventilatory response consisted primarily of an increase in F _R. CBD but not sham surgery attenuated the hypoxic ventilatory response. The F _R increase in hypoxia, although greatly reduced, remained statistically significant after CBD in 12% FiO₂. When combined with the slight respiratory alkalosis revealed by blood gas analysis, this result suggests that a small ventilatory stimulation persisted at 12% FiO₂ in CBD rats.

Effects of CBD on hypoxia‐induced sighs

Sigh frequency was measured before and after CBD under six different gas conditions (65% FiO₂, 21% FiO₂, 21% FiO₂ + 3% FiCO₂, 15% FiO₂, 12% FiO₂, 12% FiO₂ + 3% FiCO₂). Before CBD, sigh incidence increased significantly above control (FiO₂ 21%) at 15% and 12% FiO₂. The addition of 3% FiCO₂ had no effect on the sigh frequency under normoxia but eliminated the increase sigh incidence caused by 12% FiO₂ (Fig. 3). CBD attenuated the increase in sigh frequency produced by hypoxia (Fig. 3). Statistical analysis (two‐way repeated measures ANOVA with factors FiO₂ and CBD) showed a highly significant effect of FiO₂ on sigh incidence, P < 0.0001, a highly significant effect of CBD on sighs, P = 0.0086, and a highly significant interaction, P < 0.0001. Sidak's multiple comparisons before vs. after CBD showed no effect of CBD on sigh incidence at 65% FiO₂ (P = 0.8115), 21% FiO₂ (P = 0.2018) or in 21% FiO₂ + 3% CO₂ (P > 0.9). Sigh incidence was markedly reduced by CBD at 15% FiO₂ (P = 0.0122) and 12% FiO₂ (P < 0.0001) but unaffected when the rats were exposed to 12% FiO₂ + 3% FiCO₂ (P = 0.9085) (Fig. 3). In CB intact rats, sigh frequency was significantly elevated above the normoxic baseline only under 12% FiO₂ (P = 0.014). The addition of 3% FiCO₂ returned sigh frequency to baseline. The difference between 15% FiO₂ and 12% FiO₂ + 3% FiCO₂ was small but statistically significant (P = 0.03).

Figure 3. Sighing at four levels of FiO₂ in intact vs. CB denervated (CBD) rats .

Hypoxia increased sigh frequency before but not after CBD (n = 7). Sighs were suppressed by adding 3% FiCO₂. Effect of CBD: *P ≤ 0.05; ***P ≤ 0.001 (two‐way ANOVA).

Effect of CBD on the breathing response to ArchT3.0 photoactivation

To assess the instant contribution of ArchT‐transduced neurons to breathing in conscious rats, we measured the degree to which F _R, V _T and V _E were reduced during 10 s of light‐induced bilateral inhibition of these neurons. These measurements were made before and 6–7 days after CBD in seven rats. As illustrated by a representative example (Fig. 4), ArchT activation inhibited both F _R and V _T in intact rats. F _R and V_T inhibition was greater in hyperoxia than in normoxia and smaller in hypoxia than in normoxia. These results are identical to those reported in a previous study (Basting et al. 2015). Following CBD, breathing inhibition was about the same as before CBD under hyperoxia but considerably larger under normoxia or hypoxia. To control for the possibility that the observed differences might have resulted from the passage of time (e.g. because of a higher level of expression of ArchT), we performed the identical experiment in seven rats that were operated but in which the CBs and their innervation were left intact. Breathing responses before and after sham surgery were indistinguishable (representative case not illustrated).

Figure 4. Breathing reduction elicited by bilateral ArchT activation at four FiO₂ levels before and after CBD .

From top to bottom in each panel: breathing frequency, tidal volume and the raw plethysmography trace (airflow velocity). All results are from the same rat before and after CBD. Resting F _R is indicated by a dashed red line.

The group data are illustrated in Fig. 5. In CBD rats, ArchT activation produced significantly more breathing inhibition after surgery (effect of FiO₂ on delta F_R, P < 0.0001; effect of CBD on delta F _R, P = 0.0169; interaction, P = 0.6274; Sidak's multiple comparisons for delta F _R in sham vs. CBD rats: 65% FiO₂, P = 0.2698; 21% FiO₂, P = 0.0119; 15% FiO₂, P = 0.0228; 12% FiO₂, P = 0.0085; effect of FiO₂ on delta V _T, P = 0.0002; effect of CBD on delta V _T, P = 0.0421; interaction, P = 0.0604; Sidak's multiple comparisons for delta V _T in sham vs. CBD rats: 65% FiO₂, P = 0.0067; 21% FiO₂, P = 0.0008; 15% FiO₂, P < 0.0001; 12% FiO₂, P < 0.0447; effect of FiO₂ on delta V _E, P < 0.0001; effect of CBD on delta V _E, P = 0.0056; interaction, P = 0.0754; Sidak's multiple comparisons for delta V _E in sham vs. CBD rats: 65% FiO₂, P = 0.9855; 21% FiO₂, P = 0.0113; 15% FiO₂, P = 0.0049; 12% FiO₂, P = 0.0042) (Fig. 5 A). Of note, light‐induced inhibition of F _R and V _E was increased after CBD in normoxia and in hypoxia but but not under hyperoxia (Fig. 5 A).

Figure 5. ArchT‐induced hypopnoea at four levels of FiO₂ before and after CBD .

Aa–Ac, from top to bottom, effect of FiO₂ on the reduction in breathing frequency, tidal volume and minute volume elicited by inhibiting ArchT‐expressing neurons before and 7 days after CBD (n = 7). Ba–Bc, identical experiment conducted in an additional seven rats before and 7 days after sham surgery. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001 (two‐way ANOVA). Note that sham surgery had no effect, whereas breathing inhibition was enhanced after CBD except under hyperoxia.

By contrast, sham‐operated rats responded identically to ArchT activation 1 week after the operation (effect of FiO₂ on delta F _R, P < 0.0001; effect of CBD on delta F _R, P = 0.3117; interaction, P = 0.8023; effect of FiO₂ on delta V _T, P < 0.0001; effect of CBD on delta V _T, P = 0.6660; interaction, P = 0.2065; effect of FiO₂ on delta V _E, P < 0.0001; effect of CBD on delta V _E, P = 0.6387; interaction, P = 0.1864) (Fig. 5 B).

In summary, photoinhibition of ArchT‐expressing neurons reduced breathing to the same extent before and 6–7 days after sham neck surgery. As reported previously, the inhibition was greatest under hyperoxia and progressively smaller with increasing degrees of hypoxia (Basting et al. 2015). CBD resulted in an increase of the light‐induced breathing inhibition at all FiO₂ levels except in hyperoxia. Breathing frequency was greatly reduced by photostimulation after CBD, indicating that the photoinhibited RTN neurons regulate F _R even in the absence of CB input.

Kinetics of the breathing response to ArchT activation in intact and CBD rats

As illustrated in Fig. 4, CBD changed the time course of the breathing responses to photoinhibition of ArchT‐transduced neurons and this change was much more obvious under normoxia and hypoxia than under hyperoxia. To examine the kinetics of the breathing response with greater precision, event triggered responses (F _R and V _T) were averaged over 30 s, a time frame that included the 10 s preceding and 10 s following the photostimulation period. The responses were first averaged within animals (three to five repeats) and the resulting averages were grand‐averaged across seven rats (Fig. 6). Under hyperoxia, the decrease of F _R and V _T produced by the photoinhibition was sustained throughout the entire illumination phase (10 s) and CBD made no difference to the kinetics (Fig. 6). In pre‐ and post‐CBD animals under hyperoxia, at the end of the inhibitory period, both F _R and V _T returned slowly towards the baseline with kinetics that could be modelled by an exponential function (preCBD t _1/2 = 22.4 ± 7.8; post CBD t _1/2 = 22.8 ± 2.2 s; n = 4 rats, three trials per rat).

In normoxia, the kinetics of the response to ArchT stimulation were notably different. Before CBD, after an initially large hypopnoea, respiration started to recover towards the baseline within 3 s of the laser onset and returned close to the pre‐illumination baseline by the end of the light stimulus. Breathing frequency rose slightly above control level when the light was switched off and returned to control level a few seconds later. After CBD, breathing inhibition during normoxia resembled the pattern observed under hyperoxia. Inhibition was sustained throughout the 10 s light delivery period and recovered approximately exponentially at the same rate as before CBD after the light was switched off (t _1/2 : 19.6 ± 9 s; n = 4 rats, three trials per rat). At 12% FiO_2, breathing inhibition by ArchT activation was negligible before CBD. After CBD, breathing inhibition was sustained during the entire laser on episode and recovered instantly after the laser was switched off.

Relationship between ArchT‐induced breathing inhibition and blood gases in intact and CBD rats

We aimed to determine whether CBD changes the relationship between arterial blood gases and the contribution of ArchT‐transduced neurons to F _R, V _T and V _E. To do so, we analysed the relationship between the light‐induced hypopnoea (ΔF _R, ΔV _T and ΔV _E relative to baseline values recorded immediately prior to illumination) and the level of arterial pH, $P_{{\mathrm{O}}_2}$, $P_{\mathrm{C}{\mathrm{O}}_2}$ and [HCO₃ ⁻] (Fig. 7). Before surgery, rats were exposed to 65%, 21%, 15%, 12% FiO₂ and 12% FiO₂ + 3% FiCO₂; after CBD, they were exposed to the same gas mixtures, as well as to 18% FiO₂.

Figure 7. Relationship between ArchT‐induced hypopnoea and blood gases .

A–D, from top to bottom: reduction of breathing frequency (delta F _R), tidal volume (delta V _T) and minute volume (delta V _E) elicited by photostimulating ArchT plotted as a function of arterial pH (A), arterial $P_{{\mathrm{O}}_2}$ (B), arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ (C) and arterial [HCO₃ ⁻] (D). Results (from six rats) were obtained under five conditions: 65%, 21%, 18%, 15% and 12% FiO₂ balance N₂ and 12% FiO₂ plus 3% FiCO₂ (the latter condition is indicated by #). A, C and D, straight lines indicate that a significant linear correlation between the change in breathing parameter and the blood parameter was identified. B, curved lines indicate a difference in the non‐linear exponential regression correlation between before and after CBD.

ΔF _R during ArchT activation was negatively correlated with arterial pH and positively correlated with arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ and [HCO₃ ⁻] both before and after CBD. These relationships appeared approximately linear. The slopes were consistently steeper after CBD but the slope change was not statistically significant (difference in slopes of ΔF _R vs. pH before and after CBD: F = 1.217, P = 0.2736; difference in slopes of ΔF _R vs. $P_{\mathrm{C}{\mathrm{O}}_2}$ before and after CBD: F = 2.9746, P = 0.08881; difference in slopes of ΔF _R vs. [HCO₃ ⁻] before and after CBD: F = 0.8541, P = 0.3584). ΔV _T and ΔV _E during ArchT activation were linearly related to arterial pH, $P_{\mathrm{C}{\mathrm{O}}_2}$ and [HCO₃ ⁻] in the same way as ΔF _R. The slope of these relationships were also not detectably different after CBD (difference in slopes of ΔV_E vs. pH before and after CBD: F = 0.6697, P = 0.2645; difference in slopes of ΔV _E vs. $P_{\mathrm{C}{\mathrm{O}}_2}$ before and after CBD: F = 1.559, P = 0.2158; difference in slopes of ΔV _E vs. [HCO₃ ⁻] before vs. after CBD: F = 0.1343, P = 0.7151. difference in slopes of ΔV _T vs. pH before and after CBD: F = 0. 0.2709, P = 0.6043; difference in slopes of ΔV _T vs. $P_{\mathrm{C}{\mathrm{O}}_2}$ before and after CBD: F = 1.899, P = 0.1724; difference in slopes of ΔV _T vs. [HCO₃ ⁻] before vs. after CBD: F = 0.8586, P = 0.3572). The relationship between ΔV _E and arterial $P_{{\mathrm{O}}_2}$ was left‐shifted after CBD (Fig. 7 B). A non‐linear regression analysis of these curves revealed a statistically significant effect of CBD (difference in curves of ΔF _R before vs. after CBD in relation to $P_{{\mathrm{O}}_2}$: F = 5.09, P = 0.003; difference in curves of ΔV _T before vs. after CBD in relation to $P_{{\mathrm{O}}_2}$: F = 7.804, P = 0.0001; difference in curves of ΔV_E before vs. after CBD in relation to $P_{{\mathrm{O}}_2}$: F = 4.8, P = 0.0042).

Histology

As reported previously (Basting et al. 2015), ArchT‐expressing neurons (EYFP‐immunoractive) were located at or close to the ventral medullary surface and included TH‐immunoreactive neurons and neurons devoid of TH immunoreactivity. Both cell types express the transcription factor Phox2b (Basting et al. 2015). The EYFP‐immunoreactive TH‐negative neurons were defined as RTN neurons (i.e. as presumptive central respiratory chemoreceptors) (Guyenet & Bayliss, 2015). Caudally, the TH‐positive neurons are C1 cells; rostrally, these cells belong to the A5 noradrenergic cluster.

The rostrocaudal distribution of the two populations of Arch‐tranduced neurons identified in the seven rats subjected to CBD and in six of the sham‐operated rats is shown in Fig. 8 B. The estimated total number of ArchT⁺ RTN neurons per rat (neurons counted in a one‐in‐six series of sections multiplied by 6) was 433.7 ± 188.6 (range: 252–750) for the seven rats subjected to CBD and 751 ± 221 (range: 336–924) in the six sham‐operated rats. The corresponding number of ArchT⁺ catecholaminergic neurons was 399.4 ± 124.8 (range: 222–540) in CBD rats and 552 ± 294.9 (range: 264–870) in sham‐operated rats. Thus, 44.7 ± 3.7% of the ArchT⁺ neurons were catecholaminergic and the remainder (55.3 ± 3.4%) were RTN neurons. The total number of RTN neurons is ∼2100 per rat (Takakura et al. 2008); therefore we also estimate that 20.7 ± 8.9% of RTN neurons expressed ArchT in the seven rats subjected to CBD and 35.8 ± 10.5% of RTN neurons expressed ArchT in the six sham‐operated rats.

Figure 8. Location and distribution of ArchT‐expressing neurons .

A, coronal section through the medulla (level ∼11.4 mm caudal to bregma) illustrating the selective expression of ArchT‐EYFP in Phox2b‐immunoreactivity nuclei close to the ventral medullary surface (indicatde by a dashed line). B, distribution of EYFP‐immunoreactive neurons after bilateral injections of PRSx8‐ArchT‐EYFP lentiviral vector below the caudal half of the facial motor nucleus (one‐in‐six series of 30 μm‐thick transverse sections). The plane of section is calibrated after a standard atlas where level 11.6 mm is defined as the caudal end of the facial motor nucleus (Paxinos & Watson, 1998). TH‐immunoreactive neurons consist mostly of C1 neurons caudally and A5 neurons rostrally. (sections 180 μm apart). C, relationship between the number of all RTN neurons transduced (counted in a 1:6 coronal series) and the percentage decrease of V _E elicited by Arch photostimulation. D, relationship between the number of all catecholaminergic (C1) neurons transduced and the percentage decrease of V _E elicited by Arch photostimulation (n = 13, combined experimental and sham animals).

Relationship between number of ArchT‐expressing RTN neurons and light‐induced hypopnoea

After CBD, the V _E reduction elicited by ArchT activation was inversely related to the number of ArchT‐expressing RTN neurons (r ² = 0.348; P = 0.0338) (Fig. 8 C). The linear correlation between the hypopnoea and the number of ArchT‐expressing C1 neurons was not statistically significant (r ² = 0.258; P = 0.0766) (Fig. 8 D).

Next, we attempted to estimate the total contribution of RTN neurons to the central chemoreflex. We assumed that the slope of the relationship between ArchT‐induced hypopnoea (delta V_E) and arterial pH (or $P_{\mathrm{C}{\mathrm{O}}_2}$) (Fig. 7 A and B) represented the instant contribution of the ArchT‐transduced RTN neurons to the chemoreflex and we expressed these values as a percentage of the mean resting V _E under normoxia (30.1 ml 100 g^–1 min^–1). The slope of the relationship was 3.3 ± 1.3% resting V _E/0.01 arterial pH or 2.4 ± 0.8% resting V _E/mmHg arterial $P_{\mathrm{C}{\mathrm{O}}_2}$. To estimate the total contribution of RTN neurons (ArchT‐expressing or not) to the reflex, we corrected the above values by the mean percentage of RTN neurons that expressed ArchT. The resulting estimate of the contribution of RTN to the chemoreflex was 15.9 ± 6.7% resting V _E/0.01 arterial pH or 12 ± 3.9% resting V _E/mmHg arterial $P_{\mathrm{C}{\mathrm{O}}_2}$. This estimate was then compared with the slope of the hypercapnic chemoreflex in four CBD rats. These rats were exposed to room air supplemented with 0%, 3% and 6% CO₂ in the plethysmography chamber and their blood gases and ventilation measured after 35 min exposure to each condition (Table 2). The gain of the hypercapnic ventilatory reflex measured under these conditions was 17.3 ± 2.4% resting V _E/0.01 arterial pH (12 ± 1.72% resting V _E/mmHg arterial $P_{\mathrm{C}{\mathrm{O}}_2}$).

Table 2.

Hypercapnic ventilatory reflex 7 days after CBD

FiCO2	pH	$P_{\mathrm{C}{\mathrm{O}}_2}$	$P_{{\mathrm{O}}_2}$	HCO3	F R	V T	V E
0%	7.4 ± 0.02	47.4 ± 2.4	77.8 ± 10.4	29.5 ± 0.8	78.5 ± 15.9	0.38 ± 0.04	29.8 ± 5.3
3%	7.38 ± 0.03**	50.5 ± 2.9*	100.8 ± 5.3*	30.1 ± 1.2	102 ± 17.2	0.49 ± 0.09	50.3 ± 9.8**
6%	7.33 ± 0.02**	58.9 ± 2.6***	124.3 ± 16.4**	30.8 ± 0.8	136 ± 11.2**	0.53 ± 0.03**	72.1 ± 6.1***

CBD rats (n = 4) were exposed to room air (21% FiO₂) supplemented with various levels of FiCO₂ (0%, 3% or 6%). Breathing and blood parameters shown as the mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.005; one‐way repeated measures ANOVA followed by Dunnett's multiple comparisons test.

Discussion

We report and discuss new observations along with confirmatory findings regarding the generation of sighs in hypoxia. Seven days after CBD, blood gases and ventilation in resting rats breathing room air did not differ from pre‐CBD. The hypopnoea elicited by inhibiting RTN neurons and nearby C1 cells remains linearly related to arterial pH and $P_{\mathrm{C}{\mathrm{O}}_2}$ after CBD. RTN/C1 inhibition produces a greater degree of hypopnoea after than before CBD, except under hyperoxic conditions.

We interpret the results as follows. The hypopnoea produced by combined optogenetic inhibition of C1 and RTN neurons is probably caused by inhibition of the latter. One week after CBD, the arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ set‐point of rats breathing room air returns to normal. RTN neurons still encode arterial pH/$P_{\mathrm{C}{\mathrm{O}}_2}$ after CBD, consistent with the notion that these neurons are a nodal point for the regulation of breathing by CO₂. RTN neuron inhibition causes a more sustained hypopnoea after CBD, although this phenomenon can be explained by a loss of the rapid countervailing influence of the carotid bodies without invoking plasticity.

Recovery of normal ventilation and blood gases 1 week after CBD despite persisting reduction of the HVR

In intact rats breathing room air, blood gases were as previously reported for quietly resting unstressed rats (Olson et al. 1988). Weight normalized values for basal ventilation on room air were similar to prior published values for the same rat strain (Strohl et al. 1997; Mouradian et al. 2012). Also as described previously, the HVR of intact rats consisted of a selective increase in breathing frequency (Hodges et al. 2002). The magnitude of the breathing stimulation elicited at 12% FiO₂ in intact rats (average 48.8 ± 19.1% increase in V _E) was consistent with values reported in a previous study investigating the same strain (Hodges et al. 2002). Six to 7 days after bilateral CBD, the HVR was greatly attenuated but not eliminated, also as reported previously (Mouradian et al. 2012). Specifically, there was still a significant overall effect of FiO₂ on F _R in CBD rats (one‐way ANOVA) and F _R at 12% FiO₂ was modestly but significantly higher than in room air. The slight plasma alkalization and reduced arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ observed in CBD rats exposed to 12% FiO₂ confirmed the persistence of a small ventilatory stimulation even though hypoxia did not produce a statistically significant increase in V _E after CBD.

In rats, arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ increases immediately after CBD but eventually returns towards control. This recovery took 10 weeks in one study (Olson et al. 1988) and between 4 and 10 days in another (Mouradian et al. 2012). In the present study, blood gases in resting rats breathing room air were at control levels 7 days after CBD, whereas the HVR was still greatly depressed. Differences in the surgical procedure used to denervate the carotid bodies may account for the variable recovery kinetics. In the present study, we purposefully chose to perform our experiments 1 week after surgery because, at later times, the HVR recovers completely in rats (Curran et al. 2000; Timmers et al. 2003; Teppema & Dahan, 2010; Hodges & Forster, 2012; Angelova et al. 2015).

Experimental limitations

The interpretation of the present optogenetic data relies on three assumptions. The first, which has been verified previously (Basting et al. 2015), is that the ArchT‐expressing RTN neurons are silenced by green light and instantly recover their original discharge rate when the light is switched off. The second assumption is that ArchT is only expressed by neurons; this premise has also been verified previously by histological means (Basting et al. 2015). The third assumption is that, although both RTN and C1 neurons expressed ArchT, the light‐induced inhibition of breathing results predominantly from silencing RTN neurons. This assumption is highly plausible but not fully demonstrated. Stimulation of the C1 cells produces some breathing stimulation (Burke et al. 2014); therefore, inhibiting these cells could conceivably have the converse effect. However, combined inhibition of ArchT‐expressing RTN and C1 cells produces very little blood pressure reduction in quietly resting rats (<5 mmHg; T. Basting, I. Wenker and P. Guyenet, unpublished results), which suggests that the C1 cells have a very low resting discharge under normoxia. However, C1 cells are robustly activated by hypoxia (Reis et al. 1989; Sun & Reis, 1994; Sun et al. 1996; Reis et al. 1997; Paton et al. 2009; Teppema & Dahan, 2010). If the breathing response caused by inhibiting a mixed population of C1 and RTN neurons had been predominantly caused by C1 cell inhibition, one would have expected the strongest inhibition of breathing to have occurred under hypoxic conditions, whereas the exact opposite was observed (Basting et al. 2015; present study). Thus, in the present experiments, RTN inhibition was probably primarily responsible for the observed breathing reduction. This interpretation is supported by the tight correlation between the degree of hypopnoea elicited by inhibiting ArchT+ neurons with the number of RTN rather than C1 cells.

RTN inhibition reduces breathing frequency after CBD

Consistent with previous studies, we observed that hypoxia stimulates predominantly breathing frequency in rats and that this stimulation requires intact carotid bodies (Roux et al. 2000; Coles et al. 2002; Mouradian et al. 2012). Several studies report that central respiratory chemoreceptors regulate tidal volume selectively (Nattie et al. 2001; Wakai et al. 2015). In unaesthetized rodents at least, this is clearly not the case. RTN stimulation or hyperoxic hypercapnia (to activate preferentially central chemoreceptors) produces large increases in breathing frequency in quietly awake rats and mice habituated to their environment or even asleep (Abbott et al. 2011; Holloway et al. 2015; Kumar et al. 2015). Indeed, before birth, selective stimulation of RTN neurons increases breathing frequency in preparations that are devoid of CB input (Onimaru & Homma, 2003; Ruffault et al. 2015). Furthermore, even in the absence of CB input, the breathing frequency is still robustly increased by hypercapnia in rats, optogenetic inhibition of RTN neurons continues to reduce breathing frequency and, more importantly, this frequency reduction is still proportional to arterial pH or arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ (present study). The latter characteristic strongly suggests that the frequency regulating RTN neurons are activated by CO₂. Whether these particular RTN neurons are directly chemosensitive (Guyenet & Bayliss, 2015; Kumar et al. 2015) or activated by brain $P_{\mathrm{C}{\mathrm{O}}_2}$ via synaptic inputs and surrounding astrocytes (Gourine et al. 2010; Nattie, 2012), or both, is not addressed by the results of the present study.

Breathing does not require RTN neurons to be active

In intact rats exposed to hypocapnic hypoxia, the breathing reduction (both F _R and V _T) elicited by optogenetic inhibition of RTN neurons is inversely correlated with the FiO₂ and greatly reduced at 12% FiO₂ (Basting et al. 2015). These results were replicated in the present study. Ventral surface cooling in awake goats produces somewhat less hypopnoea (10%) when the animals breathe a hypoxic vs. hypercapnic mixture, suggesting that the hypoventilation could have partly resulted from RTN neuron inhibition (Pan et al. 1995). In rats, the breathing inhibition elicited by optogenetic inhibition of RTN is restored if the plasma is reacidified, suggesting that hypoxia probably inhibits RTN neurons via respiratory alkalosis (i.e. by removing the direct or indirect stimulatory effect of CO₂ on these neurons) (Gesell et al. 1940; Duffin, 2005; Ainslie & Duffin, 2009; Basting et al. 2015). Finally, hypoxic stimulation of the carotid bodies can compensate for a reduction in activity of the medullary respiratory centres in awake goats (Pan et al. 1995). Collectively, these results suggest that, under hypoxia, breathing is normally sustained by inputs from both CBs and central chemoreceptors. Furthermore, when peripheral chemoreceptors are sufficiently stimulated, breathing does not require RTN neurons to be active, and, perhaps by extension, no central chemoreceptor needs to be active. This interpretation is also congruent with the results of Fiamma et al. (2013) obtained in an arterially perfused rat preparation; in this model, fictive breathing could still be maintained by CB stimulation when the brain was subjected to a degree of hypocapnia so severe that, theoretically, central chemoreceptors should have been inactivated. Finally, the selective genetic loss of RTN is not lethal in mice; the survival of an animal probably relies on a ventilatory stimulus originating in the carotid bodies (Ramanantsoa et al. 2011).

Central chemoreflex plasticity after CBD: does the phenomenon exist in rats and do RTN neurons contribute?

Carotid body plasticity is a well documented phenomenon (Kumar & Prabhakar, 2012; McBryde et al. 2013; Nurse, 2014). By contrast, central chemoreceptor plasticity is more elusive because, at present, this notion is based on interpretations of changes in the threshold and the gain of the hypercapnic ventilatory reflex rather than on direct measurement of the properties of central chemoreceptors (Hodges & Forster, 2012). The principal evidence for central chemoreceptor plasticity is that, some time after CBD (∼1 week in rats), resting ventilation and blood gases return to normal (in normoxia) despite a severe persisting reduction of the peripheral chemoreflex (Mouradian et al. 2012). The question raised in the present study is whether this return to the status quo is a neuroplasticity phenomenon and whether the RTN is responsible. Specifically, are the carotid bodies tonically active in normal quiescent rats breathing room air and does RTN hyperactivity compensate for the loss of this tonic input after CBD?

Initially, this interpretation appears to be well supported by the results; indeed, 7 days after CBD, blood gases and ventilation in 21% O₂ were at control pre‐CBD levels and the hypopnoea elicited by inhibiting RTN neurons for 10 s was significantly larger than before CBD, suggesting that breathing was maintained at control levels by an increased excitatory drive from RTN. However, this interpretation is less convincing when the response kinetics are scrutinized. Specifically, the immediate hypopnoea (i.e. the first few seconds) elicited by RTN inhibition in rats breathing room air was very similar before and after CBD (Figs 4 and 6). The difference was that, in intact rats, the inhibition was less sustained. The simplest explanation for the return of breathing towards control when the laser was still on is that the initial hypopnoea activated the peripheral chemoreceptors within this short time frame and this activation gradually reduced the hypopnoea elicited by optogenetic inhibition of RTN. Consistent with this interpretation, breathing did not recover during ArchT photoactivation in intact rats exposed to 65% FiO₂ or in CBD rats under any oxygen condition. The slow time constant of the response of central chemoreceptors to changes in arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ probably explains why ventilation stayed constant throughout the period of RTN inhibition (Fatemian et al. 2003; Smith et al. 2006).

One explanation of the return of breathing and arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ to control levels 1 week after CBD could be that, in resting rodents, unlike in larger mammals, the CBs do not contribute significantly to resting ventilation in normoxia; therefore, the absence of the carotid bodies would make no difference, also at rest, once the animals are fully healed. Further support for this hypothesis is that hyperoxia, which is assumed to silence the carotid bodies, produces very little hypopnoea in intact rodents (Olson et al. 1988; Basting et al. 2015; present study). However, if this hypothesis is true, some explanation is required for the transient period of hypopnoea and hypercapnia that follows CBD (Olson et al. 1988; Mouradian et al. 2012). One possibility is that surgical deafferentation is not equivalent to merely silencing the CBs. CBD increases the baroreflex gain despite collateral damage to carotid baroreceptors (Del Rio et al. 2013; McBryde et al. 2013), may damage the sympathetic innervation of the CNS vasculature, and probably causes temporary inflammation and synaptic rearrangement within the nucleus of the solitary tract with unpredictable consequences on metabolism and the respiratory system. A non‐surgical method capable of quickly, reversibly and selectively inhibiting the carotid bodies of intact animals will be required to measure accurately the contribution of the CBs to breathing under normoxia in fully intact mammals. We speculate that the CBs of rats might have a slightly higher discharge threshold to arterial $P_{{\mathrm{O}}_2}$ and/or pH compared to that of larger species (goats, dogs, humans). Such a difference could explain the variablilty of the contribution of the CBs vs. central chemoreceptors to resting ventilation under normoxia. Given the apparent absence of plasticity of CO₂‐responsive central chemoreceptors (rats; present study), a slight change in CB sensory transduction could also in theory explain why CBD produces such a long‐lasting hypoventilation and rise in $P_{\mathrm{C}{\mathrm{O}}_2}$ setpoint in large species and not in rats (Forster, 2003; Miller et al. 2013; Angelova et al. 2015). A major species difference in the central circuitry underlying the chemoreflexes need not be invoked to explain these observations.

In sum, the loss of the peripheral chemoreflex could be sufficient to explain the increased hypopnoea elicited by inhibiting RTN neurons 7 days after CBD in rats breathing room air or a hypoxic mixture. The CB afferents of normal quietly resting rats breathing room air may be silent and the temporary breathing changes that follow CBD may have other causes than the mere silencing of these sensory afferents. We acknowledge that, in goats, in which CBD produces an extremely long‐lasting hypopnoea, neurochemical changes that could, at least theoretically, underlie these persistent changes in breathing, have been observed in the brainstem (Pan et al. 1998; Miller et al. 2013). We did not find clear evidence of RTN plasticity in rats 1 week after CBD but the contribution of the carotid bodies to resting ventilation could be highly species‐dependent and it is also possible that, following CBD, adaptive changes in respiratory circuitry are species‐specific or occur somewhere else than in RTN.

Long‐lasting breathing inhibition in response to RTN inhibition

As discussed above, the kinetics of the breathing response produced by RTN inhibition varied depending on the prevailing FiO₂ and on whether the carotid bodies were intact or denervated. The breathing responses observed under these conditions also differed during the recovery phase. In hyperoxia (intact rats), or indeed under all oxygen conditions in CBD rats, breathing returned relatively slowly to control after the light was switched off (t _1/2: ∼20 s). This feature is counterintuitive; when the light was switched off, a ventilation overshoot would have been expected because of the preceding hypopnoea and resulting CO₂ accumulation. The slow recovery is probably not caused by a protracted inhibition of RTN neurons because, at least in anaesthetized rats, the discharge rate of these neurons returns instantly to control at the end of light induced inhibition (Basting et al. 2015). In addition, a similarly protracted ventilatory response (t _1/2: 12.4 s) but of opposite sign is observed in anaesthetized ventilated rats when RTN neurons are activated with channelrhodopsin (Abbott et al. 2009). RTN neurons instantly recover their initial discharge rate after stimulation with channelrhodopsin; therefore, the long‐lasting ventilatory stimulation is not caused by a persistent increase in the discharge rate of these neurons (Abbott et al. 2009). In sum, the persistence of the ventilatory effects observed after a short period of excitation or inhibition of RTN neurons probably reflects the cellular and integrative properties of the respiratory rhythm and pattern generating network downstream from RTN neurons. Alternately, this phenomenon could result from the release of a slow transmitter by RTN neurons (Stornetta et al. 2009).

Linear relationship between RTN activity and pHa/$P_{\mathrm{C}{\mathrm{O}}_2}$ persists after CBD

Below, RTN ‘activity’ refers to the contribution of ArchT‐expressing RTN neurons to ventilation as measured by the breathing reduction elicited when these neurons are photoinhibited. As illustrated in Fig. 7, the activity of RTN neurons remained a linear function of arterial pH, $P_{\mathrm{C}{\mathrm{O}}_2}$ and [HCO3⁻] after CBD. By a linear relationship, we mean that the data are statistically compatible with such a linear relationship. Although the chemodenervation meant that the range of pH and $P_{\mathrm{C}{\mathrm{O}}_2}$ elicited by hypoxia or 3% FiCO₂ was much reduced (0.06 pH; 8.5 mmHg $P_{\mathrm{C}{\mathrm{O}}_2}$), the correlation coefficients between light‐induced hypopnoea (delta F _R and delta V _E) and blood gases (pH, $P_{\mathrm{C}{\mathrm{O}}_2}$) or HCO3⁻ were still high. The slope of these relationships was always slightly greater after CBD, although the change did not reach statistical significance. The trend towards a steeper slope is presumably related to the underestimation of the ‘activity’ of RTN neurons in intact rats because of the countervailing effect of the peripheral chemoreflex as discussed above. After CBD, the slope of this relationship was on average 3.3 ± 1.3% of resting V _E per 0.01 pH (average resting V _E = 31 ml/min/100 g). Based on the assumption that the breathing inhibition caused by photoinhibiting RTN is proportional to the fraction of the neurons that express ArchT and only ∼20% of total RTN neurons in the present experiment expressed ArchT, we speculate that the entire nucleus could potentially elicit a change in ventilation equal to 15.9% resting V _E/0.01 arterial pH or 12% resting V _E/mmHg arterial $P_{\mathrm{C}{\mathrm{O}}_2}$. These prorated values are almost identical to the average gain of the hypercapnic ventilatory reflex measured in four CBD rats of the same strain (17.3% resting V _E/0.01 pHa or 12% resting V _E/mmHg arterial $P_{\mathrm{C}{\mathrm{O}}_2}$). However, the statistical power was inadequate to rule out a type II error (false negative) and we could not use these determinations to assess what pecentage of the chemoreflex is mediated by RTN.

In intact rats, a small but consistent decrease in F_R and V_T in response to photoinhibition of ArchT⁺ neurons persisted even under the most severe respiratory alkalosis (pHa ∼7.6, $P_{\mathrm{C}{\mathrm{O}}_2}$ ∼25 mmHg) (Fig. 7). Thus, RTN neurons still retained some activity under such conditions. Under anaesthetized normoxic rats or in slices, RTN neurons would be silenced by this degree of alkalosis (Guyenet et al. 2005). However, in the present study, the alkalosis was produced by hypoxia. The residual activity of these neurons probably resulted from the robust excitatory input that RTN neurons receive from the carotid bodies (Takakura et al. 2006). After CBD, the slopes of the relationship between F _R (or V _T) and arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ (or pH) was steeper (Fig. 7) and their intersect with the zero line occurred at more acidic levels (lower pH, higher $P_{\mathrm{C}{\mathrm{O}}_2}$). This observation suggests that, in the absence of CBs, a lesser degree of alkalosis is probably required to silence RTN neurons; this is precisely what would be expected from the loss of the excitatory input from the carotid bodies to RTN neurons. Yet, although plausible, the above interpretations are tentative. At least two alternative explanations of the residual activity of RTN neurons during hypoxia‐induced alkalosis can be invoked. The first is that a fraction of RTN neurons may be only moderately pH‐sensitive. Alternately, some fraction of the hypopnoea observed under hypoxic conditions was caused by C1 neuron inhibition. The first possibility is supported by evidence from RTN neurons recording in slices (Lazarenko et al. 2009); the second is supported by the finding that C1 cells (mice and rats) are activated by hypoxia and that their selective activation produces some hyperpnoea (Sun & Reis, 1993; Abbott et al. 2013; Burke et al. 2014).

Hyperbolic relationship between RTN activity and $P_{{\mathrm{O}}_2}$

Not surprisingly, the relationship between delta F _R or delta V _E and arterial $P_{{\mathrm{O}}_2}$ was hyperbolic and significantly left‐shifted after CBD. This shift is consistent with the observation that, for any given level of $P_{{\mathrm{O}}_2}$ below room air, $P_{\mathrm{C}{\mathrm{O}}_2}$ is lower in intact than in CBD rats; therefore, the resting activity of RTN neurons is also lower and photoinhibition of these neurons produces less hypopnoea.

CBD and hypoxia‐induced sighs

Sighs (augmented breaths) may result from a brief reconfiguration of the pre‐Bötzinger complex circuitry (Lieske et al. 2000; Ramirez, 2014). Sigh incidence is markedly increased by hypoxia. This effect requires intact carotid bodies and is attenuated by supplementing the hypoxic breathing mixture with CO₂ or by inducing metabolic acidosis (Bartlett, 1971; Bell et al. 2009; Bell & Haouzi, 2010). After CBD, hypoxia no longer increased sigh frequency in our experimental animals. We take this result as further evidence that we successfully eliminated the carotid bodies. We also confirmed previous results (Bell et al. 2009; Bell & Haouzi, 2010) showing that, in intact rats exposed to hypoxia, sighing is eliminated by adding CO₂ to the breathing mixture. The addition of a mere 3% FiCO₂ returned sigh frequency to the baseline level found in normoxia. Sigh frequency in 12% O₂ supplemented with 3% CO₂ was significantly below the level observed under 15% O₂. However, blood gas analysis showed that the addition of 3% CO₂ to 12% FiO₂ caused arterial $P_{{\mathrm{O}}_2}$ to rise slightly above the level observed in animals exposed to a 15% FiO₂. Therefore, under our experimental conditions, the addition of CO₂ could have eliminated sighs simply by reducing CB stimulation, although alternative interpretations, as considered below, are also possible.

Sigh‐like events are elicited by hypoxia in brain slices (Lieske et al. 2000; Ramirez, 2014) but, in unaesthetized rats subjected to CBD, hypoxia no longer triggers sighs. The latter in vivo findings argue against a role of CNS hypoxia in triggering sighs, although the interpretation of these findings is predicated on the assumption that, in unanaesthetized animals exposed to a given low FiO₂, brainstem parenchymal $P_{{\mathrm{O}}_2}$ is necessarily lower after CBD than in intact animals. This could be wrong. In 12% hypoxia, arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ is substantially higher in CBD rats than in intact rats (11 mmHg in the present case). Because hypercapnia causes cerebral vasodilatation, brainstem $P_{{\mathrm{O}}_2}$ could conceivably be higher after CBD despite the slightly reduced arterial $P_{{\mathrm{O}}_2}$. This possibility would also explain why the addition of a small amount of CO₂ (3% FiCO₂) suppresses the sighs (Bell & Haouzi, 2010; present study). In sum, the suppression of sighs by CBD does not eliminate the possibility that brainstem hypoxia could contribute to sigh generation in intact unaesthetized rodents. In an intact animal, the sigh generating circuitry could operate as a coincidence detector requiring both a high level of CB activity plus some degree of brainstem hypoxia. Finally, as shown in the present study, hypoxia does not silence RTN neurons in CBD rats. Complete RTN inhibition could conceivably be required for CB stimulation to produce sighs. This possibility would also explain why hypoxia‐induced sighs are suppressed by hypercapnia or metabolic acidosis.

Summary and conclusions

One week after CBD, the HVR was greatly attenuated, sighs were no longer elicited by hypoxia, and blood gases and resting V_E were normal, whereas F _R was marginally reduced and V _T was marginally increased. RTN inhibition produced greater breathing inhibition after than before CBD, although this difference was not observed under hyperoxic conditions and could be explained by the absence of the rapid countervailing influence of the peripheral chemoreflex. Accordingly, we did not find compelling evidence that CBD modifies RTN function and, by extrapolation, central respiratory chemoreception.

The hypopnoea elicited by RTN, especially the reduction of breathing frequency and minute ventilation, remains linearly and highly correlated with arterial $P_{\mathrm{C}{\mathrm{O}}_2}$ and pH after CBD. These observations further strengthen the notion that RTN neurons are central respiratory chemoreceptors or, at the very least, operate as a nodal point for the regulation of breathing by central chemoreceptors. Finally, in conscious rats, RTN and, by extension, central chemoreceptors regulate both the frequency and the amplitude of breathing.

Additional information

Competing interests

The authors declare that they have no competing interests.

Author contributions

All authors contributed to design of the work; acquisition, analysis or interpretation of data for the work; and drafting of the work, or revising it critically for important intellectual content. All authors approved the final version of the manuscript and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed.

Funding

This work was supported by grants from the National Institutes of Health (HL28785 and HL74011 to PGG). TMB was supported by the American Heart Association 15PRE24870004. CA was supported by the Japan Society for the Promotion of Science Postdoctoral Fellowships for Overseas Researchers.