Selective optogenetic stimulation of the retrotrapezoid nucleus in sleeping rats activates breathing without changing blood pressure or causing arousal or sighs

Abstract

Combined optogenetic activation of the retrotrapezoid nucleus (RTN; a CO₂/proton-activated brainstem nucleus) with nearby catecholaminergic neurons (C1 and A5), or selective C1 neuron stimulation, increases blood pressure (BP) and breathing, causes arousal from non-rapid eye movement (non-REM) sleep, and triggers sighs. Here we wished to determine which of these physiological responses are elicited when RTN neurons are selectively activated. The left rostral RTN and nearby A5 neurons were transduced with channelrhodopsin-2 (ChR2⁺) using a lentiviral vector. Very few C1 cells were transduced. BP, breathing, EEG, and neck EMG were monitored. During non-REM sleep, photostimulation of ChR2⁺ neurons (20s, 2-20 Hz) instantly increased V̇e without changing BP (13 rats). V̇e and BP were unaffected by light in nine control (ChR2⁻) rats. Photostimulation produced no sighs and caused arousal (EEG desynchronization) more frequently in ChR2⁺ than ChR2⁻ rats (62 ± 5% of trials vs. 25 ± 2%; P < 0.0001). Six ChR2⁺ rats then received spinal injections of a saporin-based toxin that spared RTN neurons but destroyed surrounding catecholaminergic neurons. Photostimulation of the ChR2⁺ neurons produced the same ventilatory stimulation before and after lesion, but arousal was no longer elicited. Overall (all ChR2⁺ rats combined), ΔV̇e correlated with the number of ChR2⁺ RTN neurons whereas arousal probability correlated with the number of ChR2⁺ catecholaminergic neurons. In conclusion, RTN neurons activate breathing powerfully and, unlike the C1 cells, have minimal effects on BP and have a weak arousal capability at best. A5 neuron stimulation produces little effect on breathing and BP but does appear to facilitate arousal.

hypoxia and hypercapnia, alone or in combination, activate breathing and, if severe enough, produce arousal (17, 60, 62, 63). The carotid bodies are largely responsible for the hypoxic ventilatory stimulation and for hypoxia-induced arousal because both responses are eliminated by ablating these organelles (17). How and where central nervous system (CNS) hypercarbia elicits the hypercapnic ventilatory reflex (HCVR) are still debated, and the CNS pathways responsible for arousal to hypoxia or hypercapnia are even more conjectural. The prevailing view is that both the HCVR and CO₂-induced arousal result from the combined action of CO₂ on myriads of CNS pH-sensitive neurons, including serotonergic, noradrenergic, and orexinergic neurons (13, 21, 29, 41, 49, 59, 67).

The retrotrapezoid nucleus (RTN) is a key structure for the HCVR. RTN neurons are glutamatergic and seem to operate both as a central respiratory chemoreceptor and chemoreflex integrator, i.e., these neurons detect acid, directly and via surrounding astrocytes, and receive convergent inputs from other neurons that regulate the HCVR, e.g., CNS orexinergic and aminergic neurons and the carotid bodies (4, 5, 9, 36, 65). In the present study we examine whether the RTN could also play a role in CO₂-induced arousal and blood pressure (BP) control.

At present, the most selective way to manipulate RTN neurons in adult rodents in vivo is to use PRSx8-promoter-containing lentiviral vectors that express their actuator (opsins, allatostatin receptor, etc.) selectively in Phox2b-positive neurons (1, 4, 5, 44, 53). The main disadvantage of these vectors is that they transduce both RTN and surrounding catecholaminergic neurons (C1 and A5 neurons). Simultaneous activation of these two classes of neurons produces massive breathing stimulation and other effects including BP increase, arousal, and sighs (1, 47). All these effects, including breathing stimulation, can be evoked, albeit to different degrees, by activating the C1 neurons selectively (22). The best documented function of the C1 neurons is their contribution to the baroreflex (36, 38), but many of them also receive powerful excitatory inputs from the carotid bodies (77). Increased sympathetic vasomotor tone, arousal, and sighs are characteristic effects of hypoxia (10, 63), raising the possibility that the C1 neurons mediate these responses with some selectivity (22). However, the C1 neurons belong to the reticular core of the medulla oblongata and activation of many other types of neurons within this region could conceivably also cause arousal and raise BP, the RTN included.

The C1 and A5 catecholaminergic neurons that are coextensive with RTN are bulbospinal and can be selectively destroyed by spinal injections of the ribosome-inactivating toxin anti-dopamine β-hydroxylase-saporin (anti-DβH-saporin) (5, 24, 64, 71). We took advantage of this toxin to identify which physiological effects are elicited by selective activation of RTN neurons. We transduced both RTN and nearby catecholaminergic neurons with channelrhodopsin-2 (ChR2) using a previously described PRSx8-containing lentiviral vector (LVV) (5, 9, 47). We first examined the effects produced by photostimulation of ChR2-expressing neurons and then repeated the experiments after lesioning spinally projecting catecholaminergic neurons (5). The responses that persisted after such lesions were therefore attributed to the selective activation of ChR2-transduced RTN neurons. The specific questions addressed in this study are whether these residual responses include hyperpnea, BP increase, and arousal from non-rapid eye movement (non-REM) sleep or sighs.

MATERIALS AND METHODS

Animals.

Experiments were performed on male Sprague-Dawley rats (n = 26; 400–550 g; Taconic). All procedures conformed to the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the University of Virginia Animal Care and Use Committee. Animals were housed under standard 12-h light-dark cycle with ad libitum access to food and water.

Viral constructs and virus preparation.

We used a previously described LVV encoding the photoactivatable cation channel ChR2 (H134R) fused to mCherry under the control of the Phox2-responsive promoter PRSx8 (PRSx8-ChR2-mCherry) (5, 44). The LVV was produced by the University of North Carolina virus core and used at a concentration of 3.0 × 10⁸ viral particles/ml.

Injections of vectors and animal instrumentation.

The rats were anesthetized with a mixture of ketamine (75 mg/kg), xylazine (5 mg/kg), and acepromazine (1 mg/kg) given intraperitoneally. Depth of anesthesia was assessed by an absence of the corneal and hindpaw withdrawal reflex. Additional anesthetic was administered when necessary (25% of the original dose, ip or im during surgery). Body temperature was kept close to 37°C with a servo-controlled heating pad and a blanket. All surgical procedures were performed under aseptic conditions. The mandibular branch of the facial nerve was exposed on the left side. The rat was then placed prone on a stereotaxic apparatus (bite bar set at −3.0 mm for flat skull; David Kopf Instruments). A ∼1.5 mm-wide hole was drilled into the left parietal bone 0.2–2 mm rostral to the lambdoid suture. The LVV was loaded into a 1.2-mm internal diameter glass pulled pipette broken to a 25-μm tip (external diameter) that was introduced into the brain at a 12° angle pointing towards the back of the animal. This angled approach was used to access the RTN without damaging the transverse sinus and to minimize the possibility of transducing C1 neurons by retrograde leak of viral solutions up the pipette track. The facial motor nucleus was identified by recording antidromic field potentials evoked by facial nerve stimulation (0.1 ms, 0.2–0.8 mA) (18). Two injections of LVV, 150 nl/each, were made, the first below the rostral tip of this nucleus and the second 200 μm caudal to the first site.

We then implanted an optical fiber for light stimulation plus electrodes to record the electroencephalogram (EEG) and neck electromyogram (EMG). For EEG recordings, stainless steel jeweler screws (Plastics One) were implanted extra-durally (0.5–1 mm anterior, 1 mm lateral to bregma, and 3 mm posterior, 2–2.5 mm lateral to bregma above the contralateral hemisphere). Two additional screws were implanted for structural support of the head stage and for grounding. Teflon-coated braided stainless steel wire (A-M systems) was stripped at the tip and wrapped around the implanted screws. Two additional wires were stripped at the tips and implanted in the superficial muscles of the neck for EMG recordings of postural activity. All wires were crimped to amphenol pins (A-M systems) and inserted into a plastic headstage (Plastics One). The optical fibers (200 μm, numerical aperture-0.39; Thorlabs) were fitted with ferrules as described previously (72). An ∼1-mm-wide hole was drilled into the occipital bone 1.2–2.2 mm caudal to the lambdoid suture, and a recording pipette was advanced at a 14° angle pointing forward to remap the facial motor nucleus. The optic fiber-ferrule assembly was then implanted in this orientation (14° pointing forward) with the fiber tips 0.5–0.8 mm dorsal to the vector injection sites. This angled approach enabled the fiber/ferrule assemblies and the EEG/EMG tether to project back and upwards from the skull. This configuration prevented damage to our hardware and optic fibers when the subjects tucked their heads under their bodies for sleep.

The head-stage and optic fiber-ferrule assemblies were secured to the skull using a two-part epoxy (Loctite). Incisions were then closed in two layers (muscle and skin) with absorbable sutures and vet bond adhesive. Rats received postoperative ampicillin (125 mg/kg ip) and ketoprofen (3–5 mg/kg sc) and were monitored daily. Rats recovered for a minimum of 4 wk for functional expression of the ChR2-mCherry. Rats were tested for breathing responses to light and then implanted with radiotelemetry probes (PA-C10; Data Sciences International) to record BP from the descending aorta via the right femoral artery. These rats recovered for yet another week before physiological experiments began.

After completing the physiological tests, a subset of animals (n = 6) received intraspinal injections of anti-DβH-saporin (Advanced Targeting Systems) to destroy the ChR2-expressing C1/A5 neurons that project to the spinal cord (71). With the use of fine-tipped glass pipettes (20 μm external diameter), the toxin was pressure-injected bilaterally into the region of the lateral horn of the second thoracic segment (1.0–1.2 mm lateral of midline, 1 mm below lateral sulcus; 2 injection sites per side; 100 nl of 0.22 μg/μl toxin solution per site). Animals were maintained for a further 2–3 wk before they were re-examined. The surgical procedures and toxin injections produced no observable respiratory or motor deficits, and these rats gained weight normally.

Physiological experiments in freely behaving rats.

The rats were tested in a plethysmography chamber (Buxco) modified to allow tethered EEG/EMG recordings and optical stimulation (1, 22). Before the actual experiments were run the rats were repeatedly habituated to these surroundings, which were visually isolated and with low ambient noise. On the day of the experiment, rats were lightly anesthetized with isoflurane (induction with 5%, maintenance with 2.5% 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 fiber terminated with a ferrule was mated to the implanted ferrule with a zirconia sleeve. Optical matching gel (Fiber Instrument Sales) was applied at the ferrule junction to reduce light loss. A minimum of 1 h was allowed for recovery from anesthesia and the emergence of stable sleep/wake patterns. Recordings were made between 10 AM and 6 PM, over multiple days, with a minimum of 3 days' rest between tests. The ventilatory response to RTN and catecholaminergic neuronal stimulation was assessed using barometric, unrestrained whole body plethysmography (EMKA Technologies). The plethysmography chamber was continuously flushed with 1.5 l/min of 21% O₂ balanced with N₂ regulated by computer-driven mass flow controllers for O₂, N₂, and CO₂ (Alicat). Ambient temperature and humidity within the plethysmography chamber were kept constant (±0.5°C, ±10% relative humidity) and varied from 22 to 24°C and 40 to 60% relative humidity between experiments.

Photoactivation of ChR2-expressing RTN/catecholaminergic neurons was achieved with pulses of blue light (5 ms, 2–20 Hz, ∼9 mW) delivered continually for 20 s from a blue laser (473 nm; CrystaLaser) controlled by TTL-pulses from a Grass model S88 stimulator (AstroMed) (1, 22). The transmission efficiency of each implantable optical fiber was tested before implantation with a light meter (Thorlabs). The light output at the tip of the implanted fibers was 9 mW before implantation into the brain.

Data acquisition and analysis.

Physiological signals were acquired and processed using Spike v7.03 software (Cambridge Electronic Design). EEG and EMG were amplified and band pass filtered (EEG: 0.1–100 Hz, ×1,000; EMG: 100-3,000 Hz, ×1,000) and acquired at a sampling frequency of 1 KHz. The signal generated by the differential pressure transducer connected to the plethysmography chamber was amplified, band pass filtered (×200, 0.1–20 Hz), and acquired at a sampling frequency of 1 KHz. The BP signal from the radio telemetry probe was acquired at a sampling frequency of 0.2 KHz. Mean arterial pressure and heart rate (HR) were extracted from pulsatile BP recordings from the descending aorta based on values calibrated before implantation of the telemetry probe. Periods of wake or natural sleep were classified on the basis of EEG, EMG activity, and the patterns of cardiovascular and breathing activity. During non-REM sleep, EEG spectra were dominated by delta activity (0.5–4 Hz), with little or no EMG activity, and a stable breathing pattern, BP, and HR. REM sleep was characterized by stable theta oscillations (6–8 Hz), neck muscle atonia, an irregular breathing pattern, and bradycardia with BP fluctuations.

A minimum of eight photoactivation trials were conducted in each rat in each sleep state and at each stimulus frequency (2, 10, 20 Hz). Average breathing values were extracted from all trials not contaminated by body movements, as indicated by EMG activity. Respiratory and cardiovascular parameters (BP, HR) were averaged during the 20-s photostimulus. Baseline values were measured during the 20 s preceding light delivery. Respiratory frequency (f_R, breaths/min) and tidal volume [vt, area under the curve during the inspiratory period calibrated to waveforms generated by injecting 5 ml of dry air from a syringe during the experiment, expressed in ml per 100 g body weight (BW)] were calculated using Spike software. These values were used to calculate minute ventilation (V̇e = f_R × Vt, expressed as ml/100 g BW/min). The light-induced changes in breathing or cardiovascular parameters are denoted as Δf_R, ΔVt, etc.

All data sets were tested for normality using the Shapiro-Wilk test and then differences within and between groups were determined using either unpaired Student's t-test, one- or two-way ANOVA with Bonferroni's or Tukey's multiple comparisons using PRISM software (v.6, GraphPad Software). Two-tailed Pearson's correlation was used to test for a relationship between the number of ChR2⁺ RTN or ChR2⁺ catecholaminergic neurons and the light-induced change in minute ventilation (ΔV̇e) or probability for arousal. All values are expressed as means ± SE and significance is indicated (*P < 0.05; **P < 0.01; ***P < 0.005; ****P < 0.0001).

Histology.

Animals were deeply anesthetized with sodium pentobarbital and perfused transcardially with 4% paraformaldehyde, and brains were removed and processed as described previously (22). Immunohistochemistry with antibodies against tyrosine hydroxylase (sheep anti-TH, 1:2,000; Millipore), mCherry (rabbit anti-dsRed, 1:500; Clontech No. 632496; Clontech Laboratories), and Phox2b (rabbit anti- Phox2b, 1:8,000; a gift from J. F. Brunet, Ecole Normale Superieure, Paris, France) were performed as previously described (5, 9). Cell mapping and counting and photography were done using the Neurolucida system (MicroBrightfield, Colchester, VT) with a Zeiss Axioskop microscope with computer driven stage and Zeiss MRc camera. Cell counts were taken from a one in six series of sections and only profiles containing a nucleus were counted.

RESULTS

Spontaneous arousal and sighs in acclimatized rats.

In unstressed rats habituated to the plethysmography recording chamber, non-REM sleep was characterized by large amplitude delta power EEG (0.5–4.5 Hz), a low and regular breathing rate (f_R = 72 ± 4 beats/min; n = 13) and tidal volume (Vt = 0.43 ± 0.02 ml·100 g⁻¹·min⁻¹; n = 13), and a low HR (HR = 311 ± 7 beats/min; n = 9). Spontaneous arousals from non-REM sleep were typically brief and followed a stereotyped pattern (Fig. 1A). Arousal began with cortical EEG desynchronization, which was followed by a rise in HR and, in most cases, by a sigh (augmented breath). Sighs occurred in 54 ± 2% percent of all spontaneous arousals and were followed by a postsigh apnea lasting 3.5 ± 0.2 s (range: 1.5–5.7 s; n = 6; >20 arousals/rat). Latency from arousal to sleep onset was 30 ± 2 s (range: 5–140 s; n = 6; >20 arousals/rat).

Fig. 1.

Cardiorespiratory correlates of spontaneous arousal in Sprague-Dawley rats. A: spontaneous arousal from non-rapid eye movement (REM) sleep. Non-REM sleep was characterized by large amplitude, slow-wave EEG (delta frequencies: 0.5–4.5 Hz), a regular breathing rate (respiratory frequency; f_R, breaths/min) and tidal volume (Vt; ml/100 g), steady blood pressure (BP; mmHg, via telemetry), and heart rate (HR; beats/min). Arousal followed a stereotypical sequence beginning with EEG desynchronization (loss of delta power; arrow #1), followed by tachycardia (arrow #2) and, typically, a sigh (arrow #3) followed by an apnea. B: spontaneous arousal from REM sleep. REM sleep was characterized by a highly synchronized EEG in the theta band (6–8 Hz), more rapid and irregular breathing rate, bradycardia, with increased BP and HR variability. Arousal from REM sleep followed the same sequence: EEG desynchronizarion, tachycardia, and sigh. Regular breathing and steady BP resumed shortly thereafter during quiet wake or with the recommencement of non-REM sleep. Air flow represents the raw whole body plethysmography signal (Flow). The rat had been habituated to the environment through repeated sojourns in the plethysmography chamber; au, arbitrary units.

REM sleep was characterized by a strong concentration of EEG power in the theta band (6–8 Hz), more rapid and irregular breathing (REM f_R: 97 ± 4 breaths/min; n = 13; paired Student's t-test comparing f_R during non-REM and REM sleep, P < 0.0001), bradycardia (REM HR = 300 ± 5 beats/min; n = 9; paired Student's t-test comparing HR during non-REM and REM sleep, P = 0.03), and increased BP and HR variability (Fig. 1B). Arousal from REM sleep followed the same pattern as arousal from non-REM sleep, namely EEG desynchronization (loss of theta power), tachycardia, and, typically, a sigh or augmented breath (Fig. 1B). Sighs occurred in 67 ± 10% of all spontaneous arousals from REM sleep and were followed by an apnea lasting 4 ± 0.3 s (range: 2–6 s; n = 6; >5 arousals/rat). Latency from arousal to sleep onset was 47 ± 5 s (range: 15–137 s; n = 6; >5 arousals/rat).

Optogenetic stimulation of rostral RTN and nearby catecholaminergic neurons.

During non-REM sleep, optogenetic stimulation of ChR2-transduced neurons produced an instantaneous and robust increase of breathing frequency and tidal volume (Fig. 2A). This stimulus also produced an arousal that was not accompanied by sighs. The EEG desynchronization was followed by tachycardia and small, generally downward BP fluctuations, a response akin to the spontaneous arousal (see Fig. 1A) with the exception that sighs were not elicited (Fig. 2A). In REM sleep, photostimulation of ChR2-transduced neurons did not produce an increase in f_R or produce arousal or sighs (Fig. 2B).

Fig. 2.

Optogenetic stimulation of channelrhodopsin-2 (ChR2)-transduced neurons in intact rats. A: photostimulation (20 Hz, 20 s; gray bar) during non-REM sleep. The stimulus produced an immediate increase in breathing frequency (f_R) and tidal volume, followed (∼3 s later) by EEG desynchronization (box in EEG trace is expanded at bottom) and, 4 s later, by tachycardia and a slight drop in BP. No sigh occurred with the arousal or the stimulus. B: photostimulation during REM sleep (20 Hz; 20 s; same rat in A) had no effect besides a slight increase in Vt. The box in the EEG trace is expanded at bottom to illustrate that the theta rhythm was unaffected by the stimulus.

The respiratory stimulation produced by photostimulation of ChR2-transduced neurons was frequency dependent and significant at and above 10 Hz (Fig. 3, A–C). In contrast, mean BP and HR were unchanged by photostimulation of combined RTN and catecholaminergic neurons (mostly A5 neurons; Fig. 2A). At rest, BP was 111 ± 3 mmHg and HR was 311 ± 7 beats/min; during 20 s of 20-Hz light stimulation, BP was 112 ± 5 mmHg and HR was 316 ± 7 beats/min (n = 9; not significant by paired Students t-test).

Fig. 3.

Optogenetic stimulation of ChR2-transduced neurons in non-REM sleep stimulates breathing and produces arousal. A–C: breathing parameters during non-REM sleep before (baseline) and during photostimulation at 2, 10, or 20 Hz for 20 s [n = 6; 2-way repeated-measures (RM) ANOVA with Bonferroni's multiple comparisons]. Group data from the same 6 rats that later received intraspinal anti-dopamine β-hydroxylase-saporin (anti-DβH-saporin). V̇e, minute ventilation. D: cumulative arousal probability elicited by 20-s photostimulation (grey bar) at 2, 10, and 20 Hz in ChR2⁺ animals (n = 13) or 20-Hz photostimulation in control animals (n = 9). Control subjects (ChR2⁻) were fully instrumented rats in which no neuron was transduced and the photostimulus was delivered. E: cumulative arousal probability in experimental ChR2⁺ (n = 13) and ChR2⁻ control rats (n = 9). The 20-s photostimulus was delivered during periods of established non-REM sleep at time 0 and the cumulative arousal probability was measured at 40 s (see D); significance by two-way ANOVA with Bonferroni's multiple comparisons. F: cumulative probability of a sigh during the same 40-s window in experimental ChR2⁺ (n = 13) and control ChR2⁻ rats (n = 9); not significant (ns) by two-way ANOVA. In A–F: **P < 0.01, ***P < 0.001, ****P < 0.0001.

Stimulation of ChR2-transduced neurons during non-REM sleep produced arousal (Figs. 2A and 3, D–F). Arousal probability was frequency dependent and peaked towards the end of the stimulus period (Fig. 3, D and E; two-way ANOVA; effect of ChR2 on arousal: F_1,56 = 53.45, P < 0.0001; effect of photostimulus frequency on arousal: F_2,56 = 10.43, P = 0.0001). With 20-Hz photostimulation, arousal occurred in 62 ± 5% of trials (n = 13). The control stimulations represent the probability of arousal and sigh when light was applied to the RTN in rats in which no neuron was transduced with ChR2 (ChR2⁻; Fig. 3, D–F; n = 9). The incidence of sighs in the ChR2-expressing cohort (n = 13) was not statistically different from control (n = 9; Fig. 3F; two-way ANOVA; effect of ChR2 on sigh incidence: F_1,56 = 0.84, P = 0.4).

Photostimulation during REM sleep increased tidal volume but did little else (see also Ref. 23). The breathing rate, BP, or HR were unchanged and neither arousal nor sighs were elicited (Fig. 2B; Vt at baseline: 0.35 ± 0.2 ml/100 g; 20 s at 20-Hz light stimulation: 0.44 ± 0.03 ml/100 g; n = 13; paired Student's t-test, P = 0.005; f_R at baseline: 97 ± 4 beats/min; 20 s at 20-Hz light stimulation: 101 ± 4 beats/min; n = 13; paired Student's t-test, P = 0.2).

Optogenetic stimulation of rostral RTN after lesion of catecholaminergic neurons.

Six of the most robust respiratory responders (i.e., largest ΔV̇e elicited by photostimulation) received spinal injections of the catecholaminergic neuron-selective toxin anti-DβH-saporin and were retested 2 wk later (Fig. 4). In these six rats, 20-Hz photostimulation before bulbospinal catecholaminergic lesion had produced a significant increase in arousal (69 ± 5%) and sighs (36 ± 5%) compared with control ChR2⁻ rats (n = 9; Fig. 5, A–C; one-way ANOVA: arousal: F_2,18 = 50.40, P < 0.0001; Tukey's multiple comparisons: control vs. intact, P < 0.0001; sigh: F_2,18 = 11.83, P = 0.0005; Tukey's multiple comparisons: control vs. intact, P = 0.001); however, note that sigh incidence across all ChR2-expressing rats (26 ± 4%; n = 13) was not statistically different from control (16 ± 3%; n = 9) (see above). The lesion of bulbospinal catecholaminergic neurons produced a dramatic reduction in the probability of arousal and sighs with 20-Hz photostimulations (Figs. 4 and 5, A–C). Compared with ChR2⁻ controls, the lesioned subjects showed a modest increase in arousal probability with 20-Hz photostimulations (Fig. 5B; P = 0.03) but had no increase in sighs (Fig. 5C; P = 0.9). Yet, this stimulus produced similar increases in breathing frequency, tidal volume, and minute ventilation before vs. after lesion (Fig. 5, D–I).

Fig. 4.

Optogenetic stimulation of ChR2-transduced neurons during non-REM sleep after lesion of bulbospinal catecholaminergic neurons. Photostimulation (20 Hz, 20 s; gray bar) in non-REM sleep 2 wk after administration of the saporin conjugate into the thoracic spinal cord (same animal shown in Fig. 2A). This stimulus produced a similar large increase in breathing but had no effect on EEG, HR, or BP.

Fig. 5.

Optogenetic stimulation of ChR2-transduced neurons during non-REM sleep before and after lesion of bulbospinal catecholaminergic neurons: arousal, sighs, and respiratory stimulation. A: arousal probability elicited by a 20-Hz/20-s photostimulus (gray bar) measured 40 s after the onset of the stimulus in 6 rats before (intact ChR2⁺) and 2 wk after administration of the catecholaminergic neuron-specific toxin (lesion ChR2⁺) into the thoracic spinal cord. Control rats (ChR2⁻; n = 9) were fully instrumented rats in which no neuron was transduced. B: arousal probability in control rats or in the experimental rats before (intact) and after lesion; one-way ANOVA with Tukey's multiple comparisons test. C: probability of sigh during the photostimulation period in control rats or experimental rats before (intact) and after lesion; one-way ANOVA with Tukey's multiple comparisons test. D–F: breathing parameters during non-REM sleep at rest and during photostimulation (20 s/20 Hz) in 6 rats determined before (intact) and after lesion of bulbospinal catecholaminergic neurons; two-way RM ANOVA with Bonferrroni's multiple comparisons test. G–I: change (Δ) in breathing parameters evoked by photostimulation in 6 rats before (intact) and after catecholaminergic neuron lesion. Breathing stimulation was identical before and after the lesion (paired Students t-test). In A–I: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Histology.

In control rats, the rostrocaudal distribution of catecholaminergic neurons within the ventrolateral medulla is bimodal (Fig. 6A). The rostral cluster (Fig. 6; ∼8.5–11.0 mm caudal to bregma) contains the A5 noradrenergic neurons; the caudal cluster (posterior to bregma 11.4 mm) consists almost exclusively of C1 neurons (see Ref. 71). Cell counts were taken from a 1 in 6 series of 30-μm-thick sections from the left side (i.e., the transduced side); the numbers reported below are neurons counted in the 1 in 6 series multiplied by 6 to give the approximate absolute number of relevant neurons.

Fig. 6.

Anatomical distribution of ChR2-transduced neurons before and after lesion of bulbospinal catecholaminergic neurons. A: rostrocaudal distribution of tyrosine hydroxylase (TH)-immunoreactive neurons in 8 intact and 6 lesioned rats. Counts were made on the left side of the brain in a 1-in-3 series of 30-μm-thick transverse sections. The section closest to the caudal end of the facial motor nucleus was assigned the level 11.6 mm caudal to bregma and served as reference point to align the brain sections from all the rats. The rostral cluster (bregma 11-8.5) consists of A5 noradrenergic neurons and is massively destroyed by the toxin. The caudal cluster (C1 neurons) consists of adrenergic/glutamatergic neurons that are more modestly damaged. Its rostral portion (bregma 11.5–13.5 mm) contains the bulk of the spinally projecting C1 cells. B: rostrocaudal distribution of ChR2-transduced neurons (mCherry-immunoreactive) in 8 intact rats. Transduced catecholaminergic neurons (TH⁺) and noncatecholaminergic neurons [putative retrotrapezoid nucleus (RTN) chemoreceptors] are separately represented. C: rostrocaudal distribution of ChR2-transduced neurons (mCherry-immunoreactive) in 6 lesioned rats. Transduced catecholaminergic neurons (TH⁺) and noncatecholaminergic neurons (putative RTN chemoreceptors) are represented separately. All transduced neurons were on the left side.

In the six rats treated with anti-DβH-saporin, the number of A5 neurons (total catecholaminergic neurons identified between 8.5 and 11 mm caudal to bregma on the left side) was reduced by 76% (583 ± 68 vs. 140 ± 18; unpaired Student's t-test, P = 0.002; Fig. 6A). Similarly, the number of C1 neurons (total catecholaminergic neurons identified between bregma 11.0 and 12.5 mm) was reduced by 52% in toxin-treated rats (762 ± 86 vs. 368 ± 62; unpaired Student's t-test, P = 0.004; Fig. 6A). The selective deletion of the catecholaminergic neurons conformed to expectations (71). RTN neurons do not project to the spinal cord and are insensitive to spinal injections of this toxin (5). By contrast a large proportion of A5 neurons and rostral C1 neurons innervate the thoracic spinal cord and the bulk of these catecholaminergic neurons are destroyed by intraspinal injections of anti-DβH-saporin (71).

As previously described, the LV vector used here selectively transduces Phox2b-expressing neurons when microinjected as described previously (for details see Refs. 5, 9). The ChR2-transduced neurons were located under the facial motor nucleus and between this nucleus and the exiting root of the facial motor nerve (i.e., the noradrenergic A5 region; not illustrated). In nonlesioned rats, 51% of ChR2-transduced neurons were noncatecholaminergic (190 ± 25 noncatecholaminergic vs. 179 ± 32 catecholaminergic; n = 8; Fig. 6B). Most transduced catecholaminergic neurons were within the A5 noradrenergic region (rostral to bregma 11 mm) but a few were within the rostral C1 region (11.0–12.0 mm caudal to bregma). In the six toxin-treated rats, the number of ChR2-expressing catecholaminergic neurons was lower than in the control rats (103 ± 14; n = 6; unpaired Student's t-test, P = 0.04). The number of ChR2-expressing noncatecholaminergic neurons was not significantly different between intact and lesioned rats (intact: 190 ± 25; lesioned: 227 ± 26; unpaired Students t-test, P = 0.3).

Breathing stimulation correlates with the number of ChR2⁺ RTN neurons whereas arousal correlates with the number of ChR2⁺ catecholaminergic neurons.

The hyperpnea elicited by activating ChR2 strongly correlated with the number of transduced RTN neurons (n = 10; Pearson r = 0.83, P = 0.003) but not with the number of transduced catecholaminergic neurons (n = 10; Pearson r = −0.3, P = 0.5; Fig. 7A). In contrast, the probability of arousal during photostimulation correlated with the number of transduced catecholaminergic neurons (n = 10; Pearson r = 0.69, P = 0.03) but not with the number of transduced RTN neurons (n = 10; Pearson r = −0.48, P = 0.2; Fig. 7B). There was no relationship between the total number of ChR2-expressing neurons and breathing stimulation (n = 8; Pearson r = 0.37, P = 0.4) or arousal probability (n = 10; Pearson r = 0.41, P = 0.3; Fig. 7, A and B, right).

Fig. 7.

Correlations between the number and phenotype of ChR2-transduced neurons and the physiological response elicited by photostimulation. A: x–y plots of the increase in minute ventilation (ΔV̇e) vs. the number of ChR2⁺ RTN neurons (left), ChR2⁺ catecholaminergic neurons (middle), or total ChR2⁺ neurons (right). Photostimulation was 20 s/20 Hz in non-REM sleep. The degree of hyperpnea was correlated only with the number of ChR2⁺ RTN neurons. Pearson's r and significance (P) are indicated. B: plots of arousal probability elicited by 20-s/20-Hz photostimulation vs. the number of ChR2⁺ RTN neurons, ChR2⁺ catecholaminergic neurons, or all ChR2⁺ neurons. Arousal probability was only correlated with the number of ChR2⁺ catecholaminergic neurons.

DISCUSSION

This study describes two novel findings. First, combined activation of ChR2-transduced RTN and A5 neurons, with minimal C1 cell involvement, stimulates breathing in rats, without raising BP and causes a moderate probability of arousal from non-REM sleep in rats. Second, selectively stimulating RTN neurons increases breathing but elicits neither arousal nor sighs nor BP increase or tachycardia. Based on this evidence, we conclude that RTN controls breathing selectively and, when moderately activated, this nucleus has very little influence on resting BP or state of vigilance. Furthermore, the data suggest that activation of the noradrenergic A5 neurons produces arousal from non-REM sleep. Thus it would appear that increased sympathetic tone, arousal, and sighs are elicited by activating C1, possibly A5, but not RTN neurons.

RTN regulates breathing selectively.

In the present study, we identified the function of RTN neurons by optogenetically activating these neurons after selectively destroying the majority of surrounding catecholaminergic neurons. Any effect of RTN that might be mediated or facilitated by these catecholaminergic neurons, including arousal, could have been underestimated or overlooked by this approach. Keeping this limitation in mind, the present study strongly suggests that the hyperpnea resulting from combined activation of RTN and surrounding catecholaminergic neurons (mostly A5 neurons in the present case) derives almost exclusively from the activation of RTN. This interpretation is based on two observations: 1) the degree of hyperpnea was highly correlated with the number of transduced RTN neurons, and 2) breathing stimulation was unaffected by lesioning the catecholaminergic neurons. These results extend to conscious animals similar observations made in anesthetized rats (5).

Selective C1 cell stimulation does produce hyperpnea, but this response is suppressed by hypocapnic hypoxia, which suggests that it is mediated by neurons that are silenced by respiratory alkalosis (2, 22). RTN neurons are likely candidates because they are activated by stimulating the C1 cells (22) and are silenced by respiratory alkalosis (9). The presence of phenylethanolamine N-methyl transferase (PNMT)-immunoreactive terminals in the RTN supports the existence of an input from adrenergic neurons, possibly C1 cells (69). In this study, we minimized the number of transduced C1 cells by targeting ChR2 to the rostral RTN.

Finally, the present data further illustrate the power of the excitatory drive from RTN to the respiratory pattern generator (4). Selective activation of at most 227 ± 26 RTN neurons (11.5% of the total population which is ∼2,000 in rats; Ref. 78) increased V̇e by an average of 83% (+ 20 ml·min⁻¹·100 g⁻¹; see Fig. 5, F and I), the equivalent of exposing rats to ∼4% Fi_CO2 (23, 50).

RTN has little effect on BP.

Combined activation of RTN and C1 cells or selective activation of the C1 cells produces large rises in BP (1, 22). Here, combined stimulation of RTN and catecholaminergic neurons (51/49 ratio in unlesioned animals) or selective RTN stimulation (69/31 ratio after catecholaminergic lesion) produced no effect on mean BP. Because very few C1 cells were transduced in the present experiments, these data strongly suggest that the C1 neurons are selectively responsible for the rise in BP elicited by combined activation of C1 and RTN neurons and that the latter have very little effect on BP, even in animals with intact catecholaminergic neurons.

These observations do not mean that RTN is completely devoid of any effect on the cardiovascular system. Although stimulation of the ChR2-transduced neurons produced virtually no change in mean BP, minor BP fluctuations were observed. These fluctuations could have been caused by an arousal from sleep, as depicted in Fig. 1 or by the mechanical effects of increased ventilation on cardiac return, or perhaps by regional blood flow redistribution and differential activation of various sympathetic efferents (visceral >> skeletal muscle) caused by activation of the “A5 region” (46, 52, 73).

How does CNS hypercarbia activate sympathetic vasomotor outflow?

Hypercapnia likely increases BP partly by activating presympathetic C1/non-C1 rostral ventrolateral medulla (RVLM) neurons via the carotid bodies (75). However, in anesthetized animals at least, hypercapnia also increases sympathetic tone in carotid bodies-denervated animals, an effect that is independent of the vagus nerve and therefore presumably caused by CNS hypercarbia. According to Lioy and Trzebski (51), brainstem hypercarbia increases BP via a pathway independent of the respiratory pattern generator because the pressor effect of CO₂ is observed considerably below the CO₂ respiratory recruitment threshold. Also in anesthetized ventilated rats, isoxic hypercapnia (10% Fi_CO2) doubles sympathetic tone and markedly increases the discharge of the bulbopsinal C1/non-C1 RVLM neurons (57). Because this effect of hypercarbia is resistant to ionotropic glutamate receptor blockade within the C1 region (57), RTN neurons, which are presumably glutamatergic (58), are an unlikely source of excitatory drive to presympathetic RVLM neurons during hypercapnia. Supporting this view, in the present study RTN stimulation produced no increase in BP, even with intact catecholaminergic neurons. Alternate potential mechanisms include an astrocyte-dependent control of the C1 cells (54, 80) or inputs from CO₂-responsive neurons other than RTN (e.g., serotonergic, catecholaminergic, orexinergic neurons) (59).

Combined RTN/A5 activation has no effect on arousal during REM sleep.

Combined stimulation of caudal RTN/C1 cells or selective activation of the C1 cells does not produce arousal from REM sleep (1, 22). We add here that combined stimulation of RTN (rostral part) plus A5 neurons does not produce arousal from REM sleep either. All these observations could be explained by a generalized reduction in the ability of any stimulus (intero- or exteroceptive) to produce arousal during REM sleep or to regulate breathing (23, 28, 74).

Arousal from non-REM sleep occurs in response to activation of catecholaminergic neurons but not RTN neurons.

In sleeping rats, brief selective activation of the C1 neurons produces a rise in BP, a mild hyperpnea, and a very high probability of arousal (85%) with sighs (72%) (22). As shown here, combined activation of rostral RTN and surrounding A5 noradrenergic neurons also produces arousal (62%) but with a lower probability than selective stimulation of C1 cells. Moreover, the probability of sighs (26%) with arousal was not statistically different from control ChR2⁻ animals, at least in the main cohort. Following lesion of spinally projecting catecholaminergic neurons, for which the noradrenergic A5 neurons are especially sensitive (71), the arousal/sigh response was virtually eliminated.

The present results do not exclude the possibility that selective RTN stimulation could produce arousal in animals with intact bulbospinal catecholaminergic neurons. We think this is unlikely based on neuroanatomical evidence that RTN neuronal projections specifically target the respiratory central pattern generator and eschew brainstem serotonergic and catecholaminergic neuronal clusters (16, 19, 69). A more plausible scenario is that the arousal from non-REM sleep is triggered by selective stimulation of the noradrenergic A5 neurons, whose collateral projections are extensive and target higher structures regulating vigilance and arousal states [i.e., paraventicular thalamic nucleus, central amygdala, bed nucleus of the stria terminalis, lateral hypothalamus (perfornical region), dorsomedial and paraventicular nuclei of the hypothalamus, and ventrolateral periaqueductal gray, among other targets (25, 68)]. In support of this, we show here that the arousal sensitivity was correlated to ChR2-expressing catecholaminergic neurons but not to the RTN neurons. Moreover, we observed arousals without sighs, which would indicate that activation of C1 cells were not responsible for the arousal.

The fact that selective RTN stimulation did not trigger sighs is also noteworthy. In our rats, sighs were most frequently associated with arousal from sleep, sighs were observed in 52% of all spontaneous arousals from non-REM sleep, and selective RTN stimulation produced no marked increase in arousal. Furthermore, there is evidence that sighs induced by hypoxia can be suppressed by increasing Fi_CO2 or by arterial acidification with acetazolamide, a carbonic anhydrase inhibitor (10, 11). Both treatments are potent stimulants of RTN activity (9); hence, increased RTN activity may actually suppress sighs, as CO₂ does.

A5 neurons and breathing.

A5 neurons, or the “A5 region,” may exert a tonic inhibitory effect on breathing in neonatal rats (40) and may facilitate phasic expiratory inhibition in adult rats exposed to hypoxia (posthypoxic frequency decline) (30, 32, 70). The present study did not reveal any effect of brief A5 neuronal stimulation on breathing. We may have activated too few A5 cells to produce detectable effects. Furthermore, we estimated the effect of A5 on breathing by comparing the ventilatory responses elicited before and after lesion of these neurons, which is not conducive to detect subtle changes.

Hypoxemia vs. hypercarbia; arousal and sigh.

Spontaneous arousal from sleep in adult rats involves a stereotyped sequence of events starting with EEG desynchronization followed by a tachycardia and, in most cases, by a sigh (Fig. 1; see also Ref. 22). Arousals in adult humans exhibit a similar pattern, whereas in infants this sequence is reversed (sigh/augmented breath, tachycardia, cortical arousal, and movement/agitation) (26, 56). Sighs prevent atelectasis and sigh genesis is linked to wake-promoting neuronal circuits and vigilance (35, 66). Hypoxia is a strong arousing stimulus, in part because of the frequent sighing that it induces (11, 17, 60, 61, 63). Sigh incidence is inversely correlated with the Fi_O2 and these events are triggered by carotid body input (8, 22, 27). We have previously reported that selective stimulation of C1 cells produces reliable arousal from non-REM sleep with sighs (22). In the RVLM, both the C1 and non-C1 bulbospinal sympathoexcitatory neurons are strongly activated by carotid body stimulation (55, 77) and mildly activated by CNS hypercarbia (see A5 neurons and breathing). Hence, these cells likely contribute to hypoxia or asphyxia-induced arousal (Fig. 8). However, evidence that silencing these neurons reduces asphyxia-induced arousal will be required to shore up this hypothesis. Indeed many other CNS neurons are activated by carotid bodies stimulation afferents, including the A5 neurons, and some of them (locus coeruleus, lateral parabrachial region) can definitely contribute to arousal (12, 37, 48). The present study also provides evidence that A5 neurons facilitate arousal from non-REM sleep. This observation is consistent with our hypothesis that central noradrenergic neurons contribute to C1 cell-mediated arousal (3, 38, 42) (Fig. 8).

Fig. 8.

Summary and working model. A: current finding: selective RTN stimulation increases breathing without producing arousal from non-REM sleep, sighs, or increasing BP, whereas stimulation of the noradrenergic A5 neurons produces arousal with little apparent effects on breathing or mean BP. B: working model for the contribution of A5, C1, and RTN neurons to arousal, breathing, and autonomic activation by hypoxia and hypercapnia: hypoxia activates C1 and A5 cells by a pathway stimulated by the O₂-sensing carotid body (CB) afferents and relayed via caudal commissural commissural nucleus of the solitary tract (NTS) neurons. Systemic hypoxia may also activate C1 cells via intrinsic and local glial mechanisms (54, 76, 80). C1 cell stimulation reproduces most of the effects of hypoxia: arousal from non-REM sleep, sighs, and increases in breathing and BP (22). C1 cells innervate and excite major wake-promoting systems, the orexinergic and noradrenergic neurons, including the locus coeruleus (A6), A5, A2, and A1 (3, 15, 42). Sighs may be triggered upon arousal by norepinephrine release at the pre-Bötzinger complex via β-adrenergic signaling (79); sighs are probably linked to hypoxia via the same wake-promoting/vigilance systems. RTN regulates breathing selectively; CO₂-induced arousal from non-REM sleep is not produced by RTN stimulation but is likely caused by 5HT and other wake-promoting systems that do not trigger sighs (not represented here; see Ref. 21). A5 neurons also innervate sympathetic preganglionic neurons (SPNs; not represented), which may cause some blood flow redistribution bu t little change in mean BP. CPG, respiratory central pattern generator.

The prevailing view is that CO₂-induced arousal results from the combined and direct effect of acid on many types of CNS pH-sensitive neurons, including serotonergic, noradrenergic, orexinergic, and RTN neurons (13, 21, 29, 41, 49, 59, 67). Lesions of serotonergic or the locus coeruleus noradrenergic neurons attenuate the CO₂-induced arousal and the HCVR but have minimal effects on the hypoxia-induced arousal or the hypoxic ventilatory reflex (13, 14, 21, 41, 45). To date, the interpretation of this evidence is that subsets of these aminergic neurons are directly responsive to protons (i.e., central chemoreceptors) and unaffected by hypoxia (20). However, the very low CO₂ sensitivity of aminergic neurons so far identified in vivo and the fact that bulk administration of serotonin reverses the deficit caused by complete destruction of serotonergic neurons seem more consistent with the view that serotonergic neurons modulate rather than mediate the HCVR (31, 33, 41, 45, 58). On the other hand, arousal, which requires relatively high Pa_CO2 levels (e.g., 54 mmHg in dogs), could conceivably result from the direct effect of protons on CNS neurons with low pH sensitivity, viz. serotonergic, orexinergic, and locus coeruleus noradrenergic neurons (59, 62).

We were somewhat surprised by the fact that selective RTN stimulation caused no arousal. Indeed, the average V̇e increase elicited by RTN stimulation (83%; +20 ml·min⁻¹·100 g⁻¹) was equivalent to the breathing stimulation elicited by ∼4% Fi_CO2 at steady state (23, 43, 50). This result suggests that the degree of hyperventilation may not be critical to CO₂-induced arousal. In paralyzed mechanically ventilated awake humans, a minor rise in Pa_CO2 (∼10 mmHg) produces an extraordinary level of discomfort (7, 34) and a comparable rise in Pco₂ (15 mmHg) produces arousal from non-REM sleep in high cervical spinal cord transected ventilated humans (C3, or higher) (6). A mismatch between central chemoreceptor stimulation and lung ventilation is therefore presumably much more important for dyspnea than the absolute level of chemoreceptor stimulation (39). Such mismatch could conceivably also serve as arousal signal during sleep in spite of the initial absence of awareness. In other words, although the present experiments show that RTN stimulation does not produce arousal when ventilation is allowed to increase normally, the evidence does not rule out the possibility that RTN neurons might contribute to asphyxia-induced arousal when lung ventilation is limited or compromised.

Conclusions.

The main conclusions of the study are illustrated in Fig. 8. RTN controls breathing selectively and, at least when activated to a moderate extent, has negligible effects on BP, HR, and sleep state. These observations support the view that the core function of RTN is Pa_CO2 stabilization via changes in breathing (“chemoreflex integrator” concept) (9, 36). By contrast, C1 neurons seem to be a primeval stress-response-inducing system whose acute activation raises BP and produces arousal partly by recruiting the brainstem noradrenergic neurons including locus coeruleus and A5. Finally, A5 neuron stimulation has little effect on breathing and mean BP but does produce arousal from non-REM sleep. Thus activation of C1 and A5 neurons by hypoxia may contribute to the sleep disruption and deleterious cardiovascular consequences of sleep apnea.

GRANTS

This work was supported by National Heart, Lung, and Blood Institute Grants HL28785 and HL74011 to P.G.G.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: P.G.B., R.K., R.L.S., and P.G.G. conception and design of research; P.G.B., R.K., and K.E.V. performed experiments; P.G.B., R.K., R.L.S., and P.G.G. analyzed data; P.G.B., R.K., R.L.S., and P.G.G. interpreted results of experiments; P.G.B. and P.G.G. prepared figures; P.G.B. and P.G.G. drafted manuscript; P.G.B. and P.G.G. edited and revised manuscript; P.G.B., R.K., R.L.S., and P.G.G. approved final version of manuscript.