Hyperexcitability and plasticity induced by sustained hypoxia on rectus abdominis motoneurons

Abstract

Key points

• Acute hypoxia induces active expiration in rectus abdominis (RA) muscles in conscious freely moving rats, although its overall contribution is smaller than in internal oblique (IO) muscles.

• Tonically active and silent RA motoneurons were identified in in vitro preparations of rat spinal cords.

• Sustained hypoxia (SH) increased the synaptic strength and induced morphological changes in tonically active RA motoneurons.

• Expiratory RA motoneurons were recorded in the in situ preparation and SH enhanced both the excitability and the synaptic transmission in those firing during the stage 2 expiration.

• The present study contributes to a better understanding of the mechanisms involved in SH recruitment of RA motoneurons to induce active expiration in rats.

Abstract

Rectus abdominis (RA) motoneurons translate the complex respiratory brainstem inputs into effective muscle contractions. Despite their fundamental role in respiration, their functional and morphological properties are not fully understood. In the present study, we investigated for the first time the contribution of RA muscle to active expiration and characterized RA motoneurons regarding their electrical, molecular and morphological profiles in control rats and in rats submitted to sustained hypoxia (SH), which induces chronic recruitment of abdominal muscles. Electromyographic experiments in conscious freely moving control rats and SH rats showed that RA contributes to active expiration induced by acute hypoxia, although its contribution is smaller than in internal oblique muscles. in vitro whole‐cell patch clamp recordings from RA motoneurons revealed two populations of cells: tonically active and silent. SH induced hyperexcitability in the tonically active cells by changing their action potential properties, and EPSCs. Three‐dimensional morphological reconstructions of these cells showed that SH increased the dendritic complexity, stimulated the appearance of dendrite spines, and increased the somatic area and volume. Physiologically identified RA motoneurons, firing in two distinct phases of expiration, were recorded in the brainstem–spinal cord in situ preparation of rats. SH increased the firing frequency and EPSCs of neurons firing during stage 2 expiration. Taken together, our results show that RA motoneurons reconfigure their biophysical properties, morphology and synaptic strength to produce an appropriate expiratory drive in response to SH in rats.

Key points

• Acute hypoxia induces active expiration in rectus abdominis (RA) muscles in conscious freely moving rats, although its overall contribution is smaller than in internal oblique (IO) muscles.

• Tonically active and silent RA motoneurons were identified in in vitro preparations of rat spinal cords.

• Sustained hypoxia (SH) increased the synaptic strength and induced morphological changes in tonically active RA motoneurons.

• Expiratory RA motoneurons were recorded in the in situ preparation and SH enhanced both the excitability and the synaptic transmission in those firing during the stage 2 expiration.

• The present study contributes to a better understanding of the mechanisms involved in SH recruitment of RA motoneurons to induce active expiration in rats.

Introduction

Breathing is a precisely regulated physiological function that requires orchestrated interactions between brainstem neurons, and respiratory muscles. Motoneurons intermediate this process, working as an integrative centre responsible for translating complex brainstem inputs into effective respiratory muscles contractions (Kirkwood, 1995). It appears to be a particularly complex task because the majority of synaptic contacts on these cells are made on their dendritic tree (Cullheim et al. 1987). Therefore, motoneurons should integrate, amplify and propagate all dendritic electrical oscillations to the cell soma to ensure an appropriate action potential generation. In this context, their membrane properties play a critical role in processing brainstem inputs, determining the respiratory motoneurons output and resultant respiratory motor activity (Iscoe, 1998; de Almeida et al. 2010; Wright & Calabrese, 2011; Road et al. 2013).

Well‐designed studies have described both extrinsic (synaptic inputs) and intrinsic (membrane) properties of spinal cord and brainstem inspiratory motoneurons, from perinatal development to adulthood (Jodkowski et al. 1987, 1988; Greer & Funk, 2005; Adachi et al. 2010; Moraes & Machado, 2015). However, expiratory spinal motoneurons have received much less attention, probably because expiration is considered as a passive mechanical process at rest. Evidence indicates that expiratory muscles are only recruited during physiological reflexes, such as a coughing and sneezing (Grelot & Milano, 1991; Ono et al. 2010), or during high chemical drive, as observed in physical exercise and after hypoxic episodes (Willett et al. 2001; Moraes et al. 2013; Moraes et al. 2014).

Previous studies from our laboratory demonstrated that prolonged exposure (hours or days) to a hypoxic environment, an increase in the expiratory phase of the respiratory cycle occurs, which is characterized by chronic recruitment of the abdominal nerve and muscles, generating active expiration (Zoccal et al. 2008; Moraes et al. 2014). However, the contribution of the rectus abdominis (RA) muscles to active expiration, as well as the electrical, molecular and morphological properties of the RA motoneurons, have been poorly explored in rats, even at rest. Therefore, the key questions addressed in the present study were: (i) do RA muscles contribute to active expiration in response to acute and sustained hypoxia (SH) and (ii) how do RA motoneurons integrate and process the synaptic inputs during rest and after SH?

In the present study, we used electromyographic recordings of respiratory muscles in conscious freely moving rats to investigate the contribution of RA muscle to active expiration. in vitro and in situ whole‐cell patch clamp recordings, immunofluorescence, single‐cell quantitative RT‐PCR (qRT‐PCR) and 3D motoneurons reconstructions were also used to characterize the electrophysiological, molecular and morphological properties of RA projecting motoneurons from control rats and rats submitted to SH.

Methods

Ethical approval

Experiments were performed in male Wistar rats (weighing 80–90 g) provided by the Animal Care Facility of the University of São Paulo, campus of Ribeirão Preto, Brazil. All protocols used in the present study were approved by the Ethical Committee for Animal Experimentation of the same institution (CEUA protocols #093/2009, #082/2015). The animals were maintained under standard environmental conditions (12:12 h dark/light cycle at 23 ± 2°C) with unrestricted access to food and water. The authors understand the principles and standards for reporting animal experiments under which this journal operates, and this work complies with those guidelines (Grundy, 2015).

Electromyography

An electromyogram (EMG) of respiratory muscles was obtained aiming to evaluate the inspiratory and expiratory indices in conscious freely moving rats (Moraes et al. 2013). For this, rats were anaesthetized with ketamine (80 mg kg⁻¹) and xylazine (10 mg kg⁻¹; Aldrich, Milwaukee, WI, USA) and bipolar Teflon‐coated stainless steel electrodes were implanted in the RA, internal oblique (IO) and diaphragm (DIA) muscles. The wires were tunneled under the skin and attached to an electrical socket positioned on the back of the neck of a rat. Five days later, animals were placed into a plethysmograph chamber (5 litres) and the electrical socket of the EMG was connected to an amplifier (1700 amplifier; A‐M Systems, Sequim, WA, USA). The chamber was flushed with room air at a flow rate of 1 L min^–1 during the baseline condition. In a first set of experiments, the respiratory parameters were recorded in conscious freely moving rats under normoxia for 30 min. The gas was then switched to a hypoxic mixture ($F_{\mathrm{I}{\mathrm{O}}_2}$ = 0.1, N₂ balance; Pegas 4000; Columbus Instruments, OH, Columbus, USA) for 10 min. Pulmonary ventilation (V _E), a product of tidal volume (V _T) and respiratory frequency (fR), was obtained by using whole body plethysmography. A volume calibration was performed by injecting a known volume of air (1 mL) inside the chamber. V _T was calculated using the formula described in Bartlett & Tenney (1970). At the end of the experiment, the animals were submitted to the SH protocol (see below) and, 24 h later, the same parameters were recorded. The EMGs (0.3–5 kHz of bandpass) and breathing‐related pressure oscillations (MLT141 spirometer; ADInstruments, Bella Vista, NSW, Australia) were acquired using a data acquisition system (5 KHz; PowerLab; ADInstruments) controlled by a computer running LabChart software (ADInstruments).

EMGs were recorded and expressed in absolute units (μV) and analyses were performed off‐line from rectified and integrated (∫) signals (time constant: 50 ms). DIA_EMG burst activity was assessed as fR. Based on the absolute values of DIA_EMG, RA_EMG and IO_EMG (μV), we determined the percentage of changes aiming to compare, in each animal, their activities (magnitude) during the acute hypoxic protocol. Active expiration was defined by the presence of RA_EMG or IO_EMG expiratory activity, namely a rhythmic burst of activity (between inspiratory EMG activities recorded from DIA) above tonic levels (Magalhães et al. 2018). This allowed an assessment of the incidence of active expiration and to compare the contribution of RA and IO to active expiration in the same animal. All data were normalized to their largest peak amplitude of EMG expiratory activities (i.e. with ‘1’ being considered as EMG active expiration at its maximal value observed in each animal).

Retrograde labelling of RA motoneurons

RA motoneurons were identified using retrograde fluorescent microspheres (green beads‐Lumafluor Inc, New City, NY, USA). Briefly, Wistar rats (45–50 g) were anaesthetized with ketamine (80 mg kg⁻¹) and xylazine (10 mg kg⁻¹; Aldrich) and small incisions were made on the skin to expose the abdominal muscles. The tracer was bilaterally injected in three different points of the RA muscles (2 μL point^–1) using a Hamilton syringe (10 μL). At the end of surgery, fluxina meglumine (analgesic and antipyretic; 1 mg kg^–1; Banamine, Schering‐Plough, Kenilworth, NJ, USA) was administered i.m.

Sustained hypoxia

Three days later, the animals were divided into two experimental groups: control (normoxia) and SH. The rats were placed inside Plexiglas chambers equipped with gas injectors and sensors for O₂, CO₂, humidity and temperature. In the SH group, pure N₂ was injected into the chamber to reduce the fraction of inspired O₂ to 0.1 ($F_{\mathrm{I}{\mathrm{O}}_2}$) and this level was maintained for 24 h. The injections of N₂ into the chamber were regulated by a solenoid valve system with opening–closing being controlled via a computer running Oxycycler software (Biospherix, Lacona, NY, USA). All gases were injected at the upper level of the chamber to avoid direct gas jets on the rats, which prevent any unnecessary stress. Control group was kept in room air ($F_{\mathrm{I}{\mathrm{O}}_2}$ = 0.21).

Spinal cord slice preparation

Spinal cord slices were prepared as previously described by Carp et al. (2008). In short, after 24 h of normoxia or SH protocols, the animals were anaesthetized with tribromoethanol [2.5%, 250 mg kg⁻¹, 1 mL (100 g)⁻¹ i.p.; Aldrich] and laminectomy was performed. In this step, the rats received 100% of oxygen by nasal mask to avoid any ischaemic episode. After exposing the spinal cord, the animals were placed in a supine position and transcardially perfused with a modified artificial cerebrospinal fluid (aCSF) at 2– 4°C containing (in mm): 212.5 sucrose, 3.5 KCl, 26 NaHCO₃, 1.3 MgSO₄, 1.2 KH₂PO₄, 10 glucose, 1.2 CaCl₂, 2 MgCl₂ and 0.4 ascorbic acid. This solution was previously saturated with carbogen (95% O₂/5% CO₂) (pH 7.4) and osmolality ranged between 310 and 315 mosmol kg^–1 H₂O (Mark 3 Osmometer; Fiske, Norwood, MA, USA). In sequence, a suture thread was placed around the superior thoracic segment of the spinal cord, which was lightly raised in the process of cutting the roots in a rostral–caudal direction. The spinal cord segment (from T12 to L2) was transferred to a Petri dish glass to remove the dura using fine forceps and roots were cut as close as possible to facilitate the slice production. The spinal cord was glued with cyanoacrylate to a 4% agarose block and then placed onto to the stage of a vibratome (VT 1200; Leica, Wetzlar, Germany) set to obtain slices at a thickness of 280 μm. Slices were incubated for at least 60 min at 34°C in aCSF with the composition (in mm): 125 NaCl, 3.5 KCl, 26 NaHCO₃, 1.3 MgSO₄, 1.2 KH₂PO₄, 2.4 CaCl₂, 10 glucose and 0.4 sodium ascorbate (pH 7.4 and osmolality = 295 mosmol kg^–1 H₂O) constantly gassed with carbogen. After incubation, a single slice was transferred to a recording chamber placed on the stage of an upright microscope (BX‐51; Olympus, Tokyo, Japan) for the electrophysiological recordings.

Brainstem–spinal cord in situ preparation

In this set of experiments, we used the brainstem–spinal cord in situ preparation of rats as described previously (Paton, 1996; Moraes et al. 2013). Briefly, rats were deeply anaesthetized with halothane (Astra Zeneca do Brazil Ltda, Cotia, SP, Brazil), transected caudally to the diaphragm, exsanguinated and submerged in an ice‐cold Ringer solution containing (in mm): 125 NaCl; 24 NaHCO₃, 3 KCl, 2.5 CaCl₂, 1.25 MgSO₄, 1.25 KH₂PO₄ and 10 glucose. The skin was removed and decerebration was performed at the pre‐collicular level. Therefore, with this procedure, the animals no longer perceived painful information. We also made a thoracolumbar laminectomy in the spinal cord to record RA motoneurons on the ventral horn of the spinal cord. All preparations were transferred to a recording chamber, and the descending aorta was cannulated and perfused with Ringer solution containing an oncotic agent (molecular mass 20,000, 1.25% polyethylene glycol; Sigma, St Louis, MO, USA). The solution was aerated with 95% O₂ and 5% CO₂ and continuously supplied with the aid of a peristaltic pump (520S; Watson‐Marlow, Falmouth, UK). A neuromuscular blocker was added to the perfusion solution to avoid any muscular movement (vecuronium bromide, 3–4 μg mL⁻¹; Cristália Produtos Química e Farmaceitica Ltda, Sao Paulo, Brazil).

Electrophysiology

in vitro and in situ whole‐cell recordings were obtained from RA motoneurons using an Axopatch 200B amplifier (Molecular Devices, Foster City, CA, USA). Data were filtered at 2 kHz and digitized at 10 kHz with a Digidata 1440A controlled by PClamp, version 10 (Molecular Devices). Borosilicate glass capillaries were used to pull patch pipettes with a tip resistance in the range 4–5 MΩ when filled with pipette solution (mm): 140 potassium gluconate, 10 KCl, 0.3 CaCl₂, 1.0 MgCl₂, 10 HERPES, 1.0 EGTA, 2.0 Na‐ATP and 0.25 Na‐GTP (pH 7.4 and osmolality = 285–295 mosmol kg^–1 H₂O). Ultrapure water was treated overnight with diethylpyrocarbonate (Sigma) and used to prepare pipette solutions for posterior single cell qRT‐PCR analysis (see below). The junction potential between the pipette and bath solution was calculated (–15 mV) and the value subtracted from baseline during offline analysis. Membrane resistance was determined by measuring the transmembrane voltage change (ΔV) evoked by currents pulses (ΔΙ) applied across the cell membrane. The cell input resistance was (R _n) calculated from the slope of the Δ(ΔV/ΔΙ) relationship. Depolarized current injection (0.10 nA for 2 s) was used to analyse the cellular excitability. The intrinsic electrical properties of motoneurons were measured in the presence of synaptic transmission blockers (GABA_A = 30 μm picrotoxin; glutamate = 10 μm DNQX, 30 μm AP‐5 and glycine = 1 μm strychnine). Frequency histogram distributions were used to discriminate the RA motoneuron subpopulations based on the best fit of Gaussian functions to the data points. In the brainstem–spinal cord in situ preparation, 50 μm bicuculine was used instead of picrotoxin.

RA motoneurons were labelled bilaterally and the motor phenotype was confirmed by adding biocytin (Life Technologies, Grand Island, NY, USA) into the patch pipette solution (final concentration =  0.2%). Slices were fixed in 4% paraformaldehyde and stored in a cryoprotectant solution at –20°C for further processing at the end of each electrophysiological recording. The cholinergic motor phenotype was confirmed by the double‐labelling immunoreactivity for biocytin and choline acetyltransferase (ChAT) (see details below).

In brainstem–spinal cord in situ preparations, the inspiratory phrenic motor nerve was isolated and recorded using a bipolar glass electrode. RA motoneurons were physiologically identified by antidromic stimulation of RA abdominal nerve (see below), and electrophysiological recordings were performed using blind a whole‐cell patch clamp to study the effect of SH on the electrical activity, as well as with respect to the spontaneous EPSCs of RA motoneurons. For this, the electrodes were mounted on a micromanipulator (PatchStar; Scientifica, Uckfield, UK) and placed on the ventral surface of the spinal cord with the aid of a microscope (Seiler, St Louis, MO, USA). EPSCs were isolated by local application of bicuculline and strychnine. Motoneurons on the ventral surface of the thoracic region of the spinal cord were mapped by searching for antidromic field potentials after RA nerve stimulation (5–10 V, 1 Hz, 0.2 ms pulse duration) as previously described by de Britto & Moraes (2017). All recorded signals were amplified, filtered (from 300 Hz to 5 kHz) and acquired (20 kHz) using an analogue‐to‐digital converter (CED 1401; Cambridge Electronic Design, Cambridge, UK) controlled by Spike 2 software (Cambridge Electronic Design).

Single‐cell qRT‐PCR

After the in vitro electrophysiological recordings, the cytoplasm of each RA motoneuron diluted into the pipette solution was collected for mRNA reverse transcription (High Capacity cDNA Reverse Transcription Kit; Invitrogen, Waltham, MA, USA). Preparation of single cell cDNAs was performed as described previously (da Silva et al. 2014). cDNA from a single cell was pre‐amplified using the TaqMan PreAmp Master Mix Kit (Thermo Fisher, Waltham, MA, USA) and the following hydrolysis probes: voltage‐gated sodium channels (Nav 1.2 Rn00680558_m1 Snc2a1; Nav 1.3 Rn01485332_m1 Scn3a) and voltage‐gated potassium channels (Kv 1.3 Rn02532059_s1 Kcna4; Kv 4.2 Rn00581941_m1 Kcnd2; Kv 4.3 Rn04339183_m1 Kcnd3).

All single‐cell qRT‐PCR was performed in simplex and in triplicate (StepOne Plus System; Applied Biosystems, Foster City, CA, USA) using the probes described before and a TaqMan Universal PCR Master Mix kit (Thermo Fisher) in accordance with the manufacturer's instructions. β‐actin (NM_031144.2) was used as a housekeeping gene to normalize the results. The relative quantitation was determined by the threshold cycle (Ct) for each ion channel in each cell and normalized to the β‐actin Ct (ΔCt = Ct unknown – Ct housekeeping gene). Data are presented as ion channel mRNA expression relative to the control gene (ΔCt).

Motoneurons motor phenotype and three‐dimensional reconstruction

Immunofluorescence was performed in free‐floating sections. After three PBS washes, the slices were incubated overnight with primary antibody goat anti‐ChAT (dilution 1:1000; Millipore, Billerica, USA) for further confirmation of the motor phenotype of the recorded cells. Alexa‐488‐conjugated streptavidin (dilution 1:1.000; Invitrogen) and Alexa Fluor‐647 donkey anti‐goat (dilution 1:250; Molecular Probes, Carlsbad, CA, USA) were chosen as secondary antibodies. Sections were washed and mounted with Fluoromount (Sigma) on glass slides. A confocal scanning laser microscope (TCS SP5; Leica) was used to produce optical sections (0.04 μm) of fluorescence labelled cells. These sections were superimposed to confirm the motor phenotype of RA motoneurons by colocalization of biocytin labelled cells with ChAT. The images were also used to reconstruct the motoneuron morphology in three dimensions using the Neurolucida 360 software (MicroBrightField, Williston, VT, USA). Detailed morphometric parameters, such as area and volume of soma, dendrite length, number of nodes and dendritic complexity, were analysed using Neuroexplorer (https://www.neuroexplorer.com) and compared between control and SH. The dendritic complexity index was measured as described by Pillai et al. (2012) as:

$$\mathrm{Complexity}=\frac{(\mathrm{Sum}\mathrm{of}\mathrm{the}\mathrm{terminal}\mathrm{orders}+\mathrm{Number}\mathrm{of}\mathrm{terminals})\times (\mathrm{Total}\mathrm{dendritic}\mathrm{length})}{\mathrm{Number}\mathrm{of}\mathrm{primary}\mathrm{dendrites}}$$ (1)

Ion channel blockers

The ion channels blockers used were: 4‐aminopyridine (4‐AP; 1 mm), tetraethylammonium (TEA; 1 mm), NiCl₂ (50 μm), iberiotoxin (IBTX; 50 nm), quinidine (100 μm) (Sigma) and ZD7288 (30 μm) (Tocris Bioscience, Ellisville, MO, USA).

Statistical analysis

Statistical comparisons between data were performed using Prism, version 5 (GraphPad Software Inc., San Diego, CA, USA). All results are reported as the mean ± SEM. In each case, the normality of data distribution was verified, and group comparisons were made using parametric or non‐parametric tests, as appropriate: Mann–Whitney test or Student's paired or unpaired t test. In the case of EMG, where multiple groups were compared, we applied the repeated‐measures ANOVA followed by a Bonferroni post hoc test. P < 0.05 was considered statistically significant.

Results

Contribution of the RA muscles to active expiration in conscious freely moving rats

Initially, we investigated the contribution of RA to active expiration in response to acute hypoxia in conscious freely moving rats (n = 6) and whether 24 h of SH increases its contribution. Figure 1 A, shows representative integrated (∫) and raw records from RA_EMG, IO_EMG and DIA_EMG, as well as barometric respiratory movements from one rat exposed to room air and acute hypoxic condition ($F_{\mathrm{I}{\mathrm{O}}_2}$ = 0.1). Active expiration was not observed during the baseline conditions (i.e. no rhythmic expiratory RA_EMG or IO_EMG activities were observed) (Fig. 1 A and Aa). Acute hypoxia increased fR (94.5 ± 2.1 vs. 148.2 ± 3 cpm; P < 0.0001), V _T (4.1 ± 0.1 vs. 6.6 ± 0.1 mL kg^–1; P < 0.0001), V _E (390.3 ± 13.3 vs. 988.6 ± 27.6 mL kg^–1 min^–1; P < 0.0001) and DIA magnitude (ΔDIA_EMG amplitude: 23.2 ± 0.7%) and evoked active expiration in both RA_EMG and IO_EMG (Fig. 1 Ab). However, active expiration was more incident in IO than in RA, as noted by the higher values of expiratory IO_EMG magnitude (385.9 ± 6.3 vs. 92.9 ± 2.7%, P < 0.0001) and incidence (0.6 ± 0.02 vs. 0.42 ± 0.02; P = 0.006) in response to acute hypoxia (Fig. 1 C–F).

Figure 1. Effects of acute and sustained hypoxia on inspiratory and expiratory activities of conscious freely moving rats.

Integrated (∫) and raw electromyographic responses of RA (RA_EMG), internal oblique (IO_EMG) and diaphragm (DIA_EMG) to acute hypoxia ($F_{\mathrm{I}{\mathrm{O}}_2}$ = 0.1) from one representative rat, before (control, A) and after (B) 24 h of SH. Red rectangles in (A) and (B) represent the site of the magnification of respiratory cycles before (Aa or Ba) and during (Ab or Bb) acute hypoxia (grey squares). Note the delayed presence of the rhythmic expiratory activity (active expiration) in the RA_EMG compared to IO_EMG in response to acute hypoxia. SH induced active expiration in both RA_EMG and IO_EMG during baseline conditions, and also increased their responses to a new episode of acute hypoxia. Grouped data showing the magnitude and incidence of RA_EMG (C and D, respectively) and IO_EMG (E and F, respectively) during acute hypoxia in control and SH rats (n = 6). Red arrow indicates an absence of active expiration; green arrow indicates the presence of active expiration. Unpaired t test. [Color figure can be viewed at wileyonlinelibrary.com]

Twenty‐four hours of SH also evoked active expiration, which was more incident in the IO_EMG than in the RA_EMG (0.47 ± 0.02 vs. 0.31 ± 0.02; P = 0.001), in conscious freely moving rats during baseline condition (Fig. 1 B and Ba). Rats exposed to SH also exhibited higher values of fR (115.7 ± 2.5 vs. 94.5 ± 2.1 cpm; P < 0.0001), V _T (5.1 ± 0.1 vs. 4.1 ± 0.1 mL kg^–1; P < 0.0001) and V _E (597.8 ± 17.2 vs. 390.3 ± 13.3 mL kg^–1 min^–1; P < 0.0001) than those observed in baseline conditions in control rats (Fig. 1 A, Aa, B and Ba). During a new episode of acute hypoxia, the SH animals also exhibited increases in fR (115.7 ± 2.5 vs. 175.3 ± 3 cpm; P < 0.0001), V _T (5.1 ± 0.1 vs. 7.4 ± 0.1 mL kg^–1; P < 0.0001), V _E (597.8 ± 17.21 vs. 1298 ± 25.9 m: kg^–1 min^–1; P < 0.0001) and DIA magnitude (Δ DIA_EMG amplitude: 22.6 ± 0.6%), as well as in the expiratory activity of RA_EMG and IO_EMG (Fig. 1 Bb). Furthermore, the magnitude of increases in some of these parameters (fR: 148.2 ± 3 vs. 175.3 ± 3 cpm, P < 0.0001; V _T: 6.6 ± 0.1 vs. 7.4 ± 0.1 mL kg^–1, P < 0.0001; V _E: 988.6 ± 27.6 vs. 1298 ± 25.9 mL kg^–1 min^–1, P < 0.0001; RA_EMG magnitude: 92.9 ± 2.7 vs. 119.6 ± 3.2%, P < 0.0008; RA_EMG incidence: 0.42 ± 0.02 vs. 0.65 ± 0.03, P < 0.0007; IO_EMG magnitude: 385.9 ± 6.3 vs. 456.4 ± 9%, P < 0.001; IO_EMG incidence: 0.6 ± 0.02 vs. 0.75 ± 0.02, P < 0.0001) (Fig. 1) in response to acute hypoxia were higher than those observed before the exposure to SH. Although the responses of RA were smaller than IO, these experiments demonstrate that RA effectively contributes to active expiration during acute hypoxia in conscious freely moving control rats and SH rats.

Electrical properties of RA motoneurons

To investigate whether the electrical properties of RA motoneurons were affected by SH, in vitro whole‐cell patch clamp recordings were performed in 135 (Control = 100; Hypoxia = 35) RA motoneurons from juvenile Wistar rats. These cells were identified as RA motoneurons by the presence of fluorescent microspheres in their cytosol 3 days after microinjections (Fig. 2 A–D) and the cholinergic phenotype was confirmed by double labelling immunoreactivity for biocytin and ChAT (Fig. 2 E–H).

Figure 2. RA motoneurons.

A, photomicrography showing retrograde labelled motoneurons that send projections to abdominal muscles. Bilateral distribution of stained motoneurons is observed after 3 days of microinjections. B, motoneurons visualized with a 3× digital zoom. C, labelled motoneurons visualized by epifluorescence microscope. D, the same motoneuron visualized in bright field. Dotted lines delimit the patch pipette. Patched motoneuron loaded with biocytin and revealed by streptavidin is shown in (E) and (F) (digital zoom). G, motoneuron immunoreactive for ChAT. H, double labelled confirmation of phenotype by colocalization with biocytin labelled recorded motoneuron and ChAT. [Color figure can be viewed at wileyonlinelibrary.com]

Electrophysiological recordings of pre‐identified RA motoneurons indicate that these cells can be grouped in two distinct populations (Fig. 3): (i) tonically active firing (n = 16) and (ii) silent motoneurons (n = 21). In the first group, motoneurons showed a tonic firing frequency (7.29 ± 1 Hz) and resting membrane potential of –69 ± 0.9 mV. On the other hand, silent motoneurons did not exhibit spontaneous activity and their resting membrane potential was more hyperpolarized compared to tonically active ones (–73.7 ± 0.8 mV; P = 0.0008) (Fig. 3 A–B). Frequency histograms of the membrane potential were better described assuming a bimodal distribution, demonstrating the presence of two populations of RA motoneurons with V _m = –68 ± 0.4 mV and V _m = –73 ± 0.4 mV, respectively (Fig. 3 C and D).

Figure 3. Subpopulations of RA‐projecting motoneurons.

Aa and Ba, representative traces of the firing frequency and resting membrane potential of a tonically active (n = 16) and silent motoneurons (n = 21). The individual values of the resting membrane potential were obtained from amplitude histograms. Ab and Bb, summarize the mean values for each subpopulation. C and D, frequency histogram of the membrane potential showing a bimodal distribution of motoneurons recorded. Ea and Eb, input resistance calculated for each group of motoneuron with their respective representative tracings. Fa and Fb, cellular excitability assessed by the injection of a brief depolarizing current pulse (+ 0.10 nA). Statistical significance is indicated by the asterisks (data are represented as the mean ± SEM, ^*,** P < 0.05; ^***,**** P < 0.0001. A parametric or non‐parametric test was applied based on the Gaussian distribution of each condition. [Color figure can be viewed at wileyonlinelibrary.com]

Significant differences between these populations were also detected in the R _n (tonically active = 0.34 ± 0.04 GΩ ; silent = 0.20 ± 0.04 GΩ;   P = 0.01) (Fig. 3 Ea, Eb), as well as in the instantaneous firing frequency in response to a brief depolarizing current pulse (+0.10 nA). Figure 3 Fa and Fb shows that tonically active motoneurons are more excitable than silent motoneurons (21.2 ± 2.6 Hz vs. 2.8 ± 1.5 Hz; P < 0.001).

Ion channels and the generation of action potentials in tonically active RA motoneurons

In an attempt to understand the involvement of ion channels in the spontaneous activity of tonically active RA motoneurons, we evaluated the contribution of sodium, potassium, calcium, subthreshold and background potassium channels in the generation of their action potentials. The action potentials properties were analysed after selective pharmacological blockade of: (i) delayed rectifier and voltage‐gated potassium channels using 4‐AP (n = 6) and TEA (n = 10), respectively; (ii) voltage‐gated sodium channels using TTX (n = 8); (iii) high and low voltage‐gated calcium channels using high concentration of NiCl₂ (n = 5); (iv) large conductance voltage‐gated calcium‐activated potassium channels (BK_Ca) using IBTX (n = 7); (v) subthreshold cation current mediated by hyperpolarization‐activated cyclic nucleotide‐gated channels (HCN) using ZD7288 (n = 9); and (vi) leak potassium channels using quinidine (n = 6). All of the results obtained in this set of experiments are summarized in Table 1 and representative tracers for each experimental condition are shown in Fig. 4.

Table 1.

Pharmacological blockade of ion channel and the evaluation of several electrophysiological parameters

Ion channel blockers
Parameters analysed	Control vs. 4‐AP (n = 6)	Control vs. TEA (n = 10)	Control vs. IBTX (n = 7)	Control vs. NiCl2 (n = 5)	Control vs. ZD7288 (n = 9)	Control vs. quinidine (n = 6)
Instantaneous frequency (Hz)	43 ± 10/20 ± 7.6*	57 ± 7.6/ 37 ± 4.4*	40 ± 10/41 ± 10	46 ± 5.7/56 ± 7	37.7 ± 1.6/5 ± 3*	32 ± 4.8/18.6 ± 3.3**
V m (mV)	−70 ± 2/−70 ± 2	−69 ± 0.9/−68 ± 0.9	−68 ± 1.6 /−68 ± 1.3	−64 ± 1.1/−68 ± 2	−64 ± 1.5/−70 ± 1.7***	−67 ± 2/−64 ± 2**
ΔAHP (mV)	22.3 ± 2.7/10 ± 1.7**	20 ± 1.3/6.1 ± 1.3***	19.3 ± 2.6/16 ± 2.2	22 ± 0.7/25 ± 1.7	17 ± 0.7/20 ± 0.5**	24.8 ± 2/15 ± 2**
R i (GΩ)	0.63 ± 0.09/ ‐	0.58 ± 0.1/0.79 ± 0.1**	0.59 ± 0.06/0.6 ± 0.03	0.53 ± 0.1/0.65 ± 0.1*	0.58 ± 0.06/1.1 ± 0.19*	0.51 ± 0.07/0.87 ± 0.14*
Half‐width (ms)	0.6 ± 0.1/1 ± 0.1*	0.57 ± 0.03/0.81 ± 0.06**	0.57 ± 0.03/0.57 ± 0.04	0.65 ± 0.04/0.65 ± 0.05	0.69 ± 0.07/0.8 ± 0.14	0.58 ± 0.09/0.55 ± 0.09
Rise time (ms)	0.42 ± 0.1/0.64 ± 0.2	0.38 ± 0.03/0.56 ± 0.1	1.06 ± 0.7/1.47 ± 1.1	0.56 ± 0.2/0.53 ± 0.1	0.62 ± 0.2/0.62 ± 0.2	1.4 ± 0.6/2.0 ± 0.5
Decay time (ms)	3.25 ± 0.9/7.47 ± 2	4.59 ± 1.3/9.9 ± 6.3	5.8 ± 1.9/7.05 ± 2.3	3.31 ± 1/3.33 ± 1	2.1 ± 0.9/1.9 ± 0.5	0.97 ± 0.1/2.2 ± 0.5*

Data are expressed as the mean ± SEM in the control condition and after ion channel blocker. Instantaneous firing frequency, resting membrane potential (V _m), after hyperpolarization potential (ΔAHP), input resistance (R _i) and kinetic parameters of action potential were analysed (half‐width, rise time, decay time) and are specified on the right. ^* P < 0.05; ^** P < 0.01; ^*** P < 0.001; paired t test.

Figure 4. Ion channels involved in the tonically active RA motoneurons action potential.

Representative tracings of action potentials, afterhyperpolarization potential (AHP) peak and input resistance of a tonically active motoneuron before and after the pharmacological blockade of potassium delayed rectifiers channels (A–C), voltage‐dependent potassium channels (D–F), calcium‐activated voltage‐gated potassium channels (G–I), low voltage‐gated calcium channels (J–L), hyperpolarization‐activated cyclic nucleotides‐gated channel (M–O) and leak channels (P–R). The mean values are presented in Table 1. [Color figure can be viewed at wileyonlinelibrary.com]

4‐AP induced a marked decrease in the excitability of tonically active motoneurons, by increasing their half‐maximal spike amplitude (half‐width) and decreasing the afterhyperpolarization peak amplitude of the action potential [afterhyperpolarization potential (AHP)]. No changes were observed in the resting membrane potential of these cells. It was not possible to measure R _n in the presence of 4‐AP because of the presence of action potentials during hyperpolarized current steps. Similarly, the inhibition of voltage‐gated potassium channels by TEA resulted in a decrease in the excitability of the motoneurons and the AHP, as well as an increase in the half‐width and the R _n (Fig. 4 A–F).

On the other hand, pharmacological blockade of BK_Ca by IBTX did not change the action potential waveform (Fig. 4 G–I). High concentrations of NiCl₂, used to inhibit high and low voltage‐gated calcium channels, increased R _n, without significant changes in any other parameter of the action potential (Fig. 4 J–L). TTX, a blocker of voltage‐gated sodium channels, fully abolished action potentials in tonically active motoneurons.

Subthreshold conductances, such as those induced by HCN channels and background potassium channels, have been described as responsible for setting the resting membrane potential in several cell types (Dobler et al. 2007; Kase & Imoto, 2012). In our experiments, ZD7288 produced a hyperpolarization of resting membrane potential, with a consequent decrease in the excitability of tonically active motoneurons. A significant increase in the AHP was also observed (Fig. 4 M–O and Table 1). On the other hand, quinidine produced a depolarization of the resting membrane potential, as well as a decrease in the excitability and in the AHP (Fig. 4 P–R and Table 1). Both ZD7288 and quinidine induced an increase in R _n. Rise time was not affected by these blockers, and quinidine, increased the decay time of the action potential (Fig. 4 and Table 1).

Intrinsic electrophysiological properties of RA motoneurons

Because the electrical excitability of most cells types is influenced by synaptic inputs, we also evaluated the intrinsic electrical activity of both tonically active and silent motoneurons in the absence of fast synaptic transmission (see Methods). We observed that ionotropic receptors antagonists induced a hyperpolarization of the resting membrane potential in both tonically active, with consequent reduction in the firing frequency (7.3 ± 1 Hz vs. 0.9 ± 0.1 Hz; P < 0.0001), and silent cells (–71.8 ± 0.9 mV and –74 ± 0.8 mV, respectively) (Fig. 5 Ba and Bb). Although tonically active motoneurons significantly decreased their firing frequency after the synaptic blockade, spontaneous intrinsic activity still persisted in the majority of cells (12 out 16) (Fig. 5 Aa and Ab).

Figure 5. Intrinsic properties of RA motoneuron subpopulations.

Aa and Ab, representative traces and mean date of the firing frequency of tonically active (circle n = 16) and silent (square, n = 21) motoneurons. Note that, even in the presence of synaptic blockers, tonically active motoneurons still trigger action potentials. Ba and Bb, resting membrane potential values, calculated by amplitude histograms, for each subpopulation after synaptic blockers. Ca and Cb, representative plot and mean values of the input resistance for each subpopulation. Da and Db, cellular excitability of tonically active and silent motoneurons after synaptic blockers. Data are represented as the mean ± SEM, n = 12, ^* P < 0.05; ^** P < 0.01; ^**** P < 0.001; Mann–Whitney test. [Color figure can be viewed at wileyonlinelibrary.com]

Membrane properties, such as Rn and cellular excitability, were also analysed in each group of motoneurons after the synaptic blockade (Fig. 5 C and D). In tonically active motoneurons, R _n (0.38 ± 0.04 GΩ vs. 0.22 ± 0.04 GΩ; P = 0.007) and the instantaneous frequency (23 ± 4.3 Hz vs. 7 ± 4.1 Hz; P = 0.0008) were reduced. These results lend support to the conclusion that RA motoneurons can be grouped in two distinct populations based on their electrophysiological properties.

Impact of SH on the electrical properties of RA motoneurons

Recent studies from our laboratory have shown that SH induces chronic recruitment of abdominal muscles (i.e. active expiration) (Moraes et al. 2013; Moraes et al. 2014). Thus, to investigate whether chronic recruitment of RA muscles is related to changes in the electrophysiological properties of RA motoneurons, we analysed the intrinsic properties and local network circuitry of these cells (n = 16) in rats submitted to SH. As shown in Fig. 6 A and B, the resting membrane potential of tonically active motoneurons is depolarized (–69 ± 0.9 mV vs. –65 ± 0.8 mV; P = 0.01), with a consequent increase in their firing frequency (7.29 ± 1 Hz vs. 17.5 ± 3 Hz; P = 0.006). This population also showed a significant decrease in its R _n (0.34 ± 0.04 GΩ vs. 0.14 ± 0.01 GΩ; P = 0.0005) and became more excitable after SH (21 ± 2.6 Hz vs. 31 ± 3.4 Hz; P = 0.03) compared to motoneurons from the control group (Fig. 6 C and D). However, the addition of synaptic blockers in the bath recording solution significantly reduced the firing frequency (0.9 ± 0.1 Hz vs. 1.0 ± 0.56 Hz; P = 0.83) and R _n (0.30 ± 0.04 GΩ vs. 0.19 ± 0.2 GΩ; P = 0.052)  of tonically active motoneurons to levels similar to that of control rats (Fig. 6, filled dots). Nevertheless, their resting membrane potential remained at more depolarized values (–71.82 ± 0.9 mV vs. –68.7 ± 0.7 mV; P = 0.01), whereas their excitability decreased to levels below of those observed in the control rats (23.5 ± 4.3 Hz vs. 7.58 ± 3.3 Hz; P = 0.03).

Figure 6. Sustained hypoxia selectively modulates tonically active RA motoneurons.

Aa and Ba, representative traces of the firing frequency and resting membrane potential of a tonically active RA motoneuron from rats submitted to sustained hypoxia before and after the synaptic blockade, as indicated at the top of each tracing. A summary is provided for (Ab) frequency and (Bb) resting membrane potential values for each experimental condition. The values were obtained from the construction of amplitude histograms. Note the apparent reduction in the number of action potentials after synaptic blockade (full symbols). Ca and Da, representative traces of the input resistance and excitability of a tonically active RA motoneuron before (open symbols) and after synaptic blockers (full symbols). The mean values are plotted in (Cb) and (Db), respectively. Data are represented as the mean ± SEM and a parametric or non‐parametric test was applied based on the Gaussian distribution of each situation. ^* P < 0.05; ^** P < 0.005; ^*** P < 0.01. [Color figure can be viewed at wileyonlinelibrary.com]

We did not observe any significant changes in the electrophysiological properties of silent motoneurons after SH (data not shown). Accordingly, all data presented below were acquired from tonically active RA motoneurons.

SH changes the kinetic parameters of the action potential of tonically active RA motoneurons

For a better understanding of the mechanisms underlying the changes produced by SH in tonically active motoneurons, we measured some kinetic parameters of their action potentials, such as half‐widths and both rise and decay time constants. This last type of analysis provided to us a better understanding of the speed of the depolarization and repolarization phases of the action potential, respectively. As shown in Fig. 7 A–D, SH increased the half‐width (0.50 ± 0.02 ms vs. 0.68 ± 0.06 ms; P = 0.02), as well as the rise time (0.30 ± 0.02 ms vs. 0.47 ± 0.03 ms; P = 0.002), although it slowed the decay time of tonically active motoneurons (0.32 ± 0.01 ms vs. 0.7 ± 0.1 ms; P < 0.0001).

Figure 7. SH changes the kinetic parameters of the action potential of tonically active RA motoneurons.

A, representative trace of the averaged action potential from 30 seconds of recording of tonically active expiratory motoneuron in normoxia (black) and after 24 h of sustained hypoxia (grey). B–D, mean ± SEM obtained for rise time, decay time and half‐width under control and hypoxic conditions. E, phase plane diagram (dV/dt vs. V) of a tonically active motoneuron in normoxia (black) and after 24 h of sustained hypoxia (grey). F, zoom of the dashed area for a better visualization of the points displacement caused by hypoxia. The mean ± SEM is shown in (G). Note the hyperpolarization of the threshold value in the hypoxia group (F and G). Maximum dV/dt variation reflecting the maximum sodium conductance is plotted in (H). Correlation analyses of R _i vs. V _m and threshold vs. V _m in control (I and J) and after SH (K and L). Data are represented as the mean ± SEM. ^* P < 0.05; ^** P < 0.005; ^*** P < 0.01; unpaired Student's t test and Mann–Whitney test.

Analysis of action potential waveform using phase plane plots was also performed. For this purpose, the first time derivative of the action potential voltage (dV/dt) was plotted against the voltage to measure the action potential threshold and the maximal sodium conductance. These parameters were identified and quantified using the same criteria previously described by Gudes et al. (2015). Phase plot graphs revealed that SH shifted the action potential threshold to hyperpolarized values (–52 ± 0.8 mV vs. –55 ± 0.9 mV; P = 0.04), meaning that these cells have a more negative threshold for excitation (Fig. 7 E–G). An additional difference was observed in the maximal sodium conductance (dV/dt _max). SH produced a significant increase in the dV/dt _max (184 ± 11 mV ms^–1 vs. 213 ± 7 mV ms^–1; P = 0.033) suggesting two possibilities: (i) an increase in the number of sodium channels, and/or (ii) changes in its conductance after SH (Fig. 7 H).

Finally, correlation analyses between resting membrane potential (V _m) and R _n or the threshold values were caried out for each experimental group (control and SH). R _n was not significantly correlated with V _m either in control or after SH (Fig. 7 I and K). However, as shown in Fig. 7 J and L, a positive correlation was found between threshold and V _m of tonically active cells (Pearson test; Control: r = 0.88 and P = 0.0043, n = 16; Hypoxia: r = 0.84, P < 0.0001, n = 16). In both cases, the threshold value is less negative in depolarized cells.

Ion channels expression in tonically active RA motoneurons

Because SH affects several properties of the action potential, we investigated whether these changes could have been the result of differential expression of ion channels in tonically active motoneurons. For this, the cytoplasm of these cells from control and rats submitted to SH was collected at the end of the electrophysiological recordings and then single‐cell qRT‐PCR was performed to analyse the mRNA expression of Kv 1.3, Kv 4.2 and Kv 4.3, as well as Nav 1.2 and Nav 1.3, which are implicated in the control of membrane potential and cellular function. Large variability in the expression of mRNA was seen between the motoneurons. Even though Kv 1.3 was the most expressed potassium channel in tonically active motoneurons from rats submitted to SH, there was no difference between groups (Table 2). Thus, the changes observed in the kinetics of action potential waveform, threshold and sodium conductance were not a result of changes in the expression of the ion channels evaluated.

Table 2.

Single‐cell qRT‐PCR of tonically active RA motoneurons

The proportion of potassium and sodium voltage‐dependent channels, Kv 1.3, Kv 4.2, Kv 4.3, Nav 1.2 and Nav 1.3, expressed in single tonically active motoneurons in the control condition (left) and after sustained hypoxia (right). Nine cells were analysed in each experimental condition. Grey squares indicate the presence of mRNA of ion channels specified on the right and the number within it represents the ΔCt value. White squares indicate the absence of ion channels expression.

SH increases excitatory synaptic transmission to tonically active RA motoneurons

Because most of the changes produced by SH were abolished by blockers of synaptic transmission, we next investigated whether this metabolic challenge alters the synaptic transmission to tonically active RA motoneurons. Thus, spontaneous EPSCs were recorded in cells held at –55 mV, and isolated from inhibitory currents by addition of picrotoxin to the bath solution. As shown in Fig. 8, SH increased the frequency (2.2 ± 0.6 Hz vs. 6.0 ± 1; P = 0.04) and the amplitude of EPSCs (31 ± 5 pA vs. 96.7 ± 15 pA; P = 0.007). These findings indicate that excitatory inputs, maintained in the slice and probably from bulbospinal neurons (Kirkwood, 1995), are responsible for driving the enhanced activity of tonically active RA motoneurons from SH animals.

Figure 8. Hypoxia increases the excitatory post‐synaptic currents to tonically active RA motoneurons.

Representative traces of excitatory postsynaptic current frequency and amplitude (A and B) of one tonically active motoneuron of each experimental condition. Individual values of each cell for frequency (C) and the amplitude of events (D) are shown. Data are represented as the mean ± SEM. ^* P < 0.05; ^** P < 0.005; ^*** P < 0.01; Mann–Whitney test.

SH changes the morphology of tonically active RA motoneurons

Three‐dimensional reconstruction of 12 tonically active RA motoneurons was also performed to investigate whether SH affects the morphological properties of these cells. As shown in Fig. 9, SH markedly increased the somatic surface area (349.4 ± 81 μm² vs. 1119 ± 115 μm²; P = 0.013) and volume (922.5 ± 109 μm³ vs. 4606 ± 1115 μm³, P = 0.022). We did not detect changes in the length of dendrites (662.2 ± 51 μm vs. 897 ± 101 μm; P = 0.099). However, all tonically active motoneurons from SH animals showed highly branched dendritic trees, resulting in an increase in the branching complexity (9892 ± 3542 vs. 45800 ± 11760; P = 0.034) and in the number of nodes (3.25 ± 1 vs. 13.6 ± 3; P = 0.03). Interestingly, we also observed the presence of dendrite spines in motoneurons from SH rats, not seen in the control (Fig. 9 A and B). These data suggest the occurrence of structural plasticity in tonically active RA motoneurons, which may have contributed to the changes observed in the electrophysiological properties of these cells after SH.

Figure 9. SH increases the complexity of tonically active RA motoneurons.

Representative three‐dimensional reconstructed tonically active motoneurons from a control rat (A) and rat submitted to 24 h of sustained hypoxia (B). Black dotted squares indicate the location of zoom area for the control (right) and for hypoxia (left). Note the presence of dendritic spines in the tonically active motoneurons from hypoxic rats (red arrows and white squares). Quantification area and volume of soma (C and D), number of nodes (E) and complexity (F) are shown in the box plots. Data are represented as the mean ± SEM. ^** P < 0.01; unpaired t test. Control, n = 6; SH, n = 6. [Color figure can be viewed at wileyonlinelibrary.com]

SH selectivity modulates one subpopulation of RA motoneurons in in situ preparation of rats

The observation that SH selectively modulated tonically active RA motoneurons in vitro and induced active expiration in vivo led us to perform electrophysiological recordings of these cells using an integrative approach. For this, experiments were performed using the brainstem–spinal cord in situ preparation of rats, which maintains intact connections between the central pattern generator for breathing and the RA motoneurons. As shown in Fig. 10, two populations of physiologically identified RA motoneurons (antidromic stimulation) were found in this preparation: (i) motoneurons firing during stage 1 expiration or post‐inspiration and (ii) motoneurons firing during stage 2 expiration. SH significantly increased the firing frequency (22.9 ± 1 Hz vs. 96.5 ± 3 Hz; n = 6; P < 0.0001, unpaired t test) of RA motoneurons firing during stage 2 expiration (Fig. 10 A–C). By contrast, RA motoneurons firing during post‐inspiration presented a significant reduction in their activity (87.9 ± 3.8 Hz vs. 49.7 ± 2.5 Hz; n = 6; P < 0.0001, unpaired t test) (Fig. 10 D–F).

Figure 10. SH modulates selectivity one subpopulation of RA motoneurons in in situ preparation of rats.

Representative traces of abdominal projection expiratory motoneuron frequency in the post‐inspiration (E1, n = 6) and second half phase of expiration (E2, n = 6) of a rat in normoxia (A and D, respectively) and after 24 h of sustained hypoxia (E1 phase, n = 6; E2 phase, n = 6) (B and E, respectively). The average data are shown in (C) and (F). Representative traces of the EPSCs of the E1 and E2 phases of expiration under control conditions (G and I) and after 24 h of sustained hypoxia (H and J). Individual values for each expiratory motoneurons in the control (K and M) and after sustained hypoxia (L and N) are plotted. The expiratory phase was identified by the period of absence of phrenic nerve activity (PN, index of inspiration) and the identification of the expiratory motoneuron was made via antidromic stimulation (T12). ^* P < 0.05; unpaired t test.

To test our hypothesis that bulbospinal expiratory neurons are also important for driving active expiration in response to SH, by enhancing the activity of RA motoneurons, EPSCs were also recorded in RA motoneurons. Thus, GABAergic and glycinergic receptors antagonists were locally applied in the spinal cord, and EPSCs were recorded by setting the membrane potential to –55 mV. SH produced a significant increase in the frequency (47.8 ± 3.4 Hz vs. 113 ± 3.5 Hz; P < 0.0001), as well as in the amplitude (75.3 ± 4.1 pA vs. 138 ± 3 pA; P < 0.0001, unpaired t test) of EPSCs in the motoneurons firing during stage 2 expiration, which may also explain the changes observed in their firing frequency (Fig. 10 I, J, M and N). By contrast, SH produced a reduction in the same parameters (Frequency = 111 ± 4 Hz vs. 57.7 ± 4 Hz; P < 0.0001; Amplitude = 144.6 ± 4.9 pA vs. 83.5 ± 3.6 pA; P < 0.0001) of motoneurons firing at the post‐inspiration phase (Fig. 10 G, H, K and L). These data demonstrate that RA motoneurons in in situ preparations of rats are modulated by the central pattern generator of breathing and that the RA active expiration induced by SH is a result of excitation of RA motoneurons firing during stage 2 expiration.

Discussion

Previous studies have demonstrated that expiratory muscles are recruited in some diseases, such as hypertension and heart failure, as well as in hypoxic environments (Giordano, 2005; Bosc et al. 2010; Haupt et al. 2012), to generate active expiration. However, the cellular mechanisms related to how expiratory motoneurons changes their electrophysiological properties to ensure an effective muscle contraction are relevant not only to diseased states, but also to physiological conditions with respect to avoiding large respiratory variability, such as during coughing, sneezing, speaking (Hoit et al. 1988; Ono et al. 2010), physical activity (Willett et al. 2001) or even in the REM sleep epochs (Andrews & Pagliardini, 2015).

In the present study, we show that both IO and RA muscles are recruited to produce active expiration in conscious freely moving rats during acute hypoxia or after SH. Because SH rats exhibited higher fR and V _T, we speculate that the emergence of RA active expiration in response to acute hypoxia is important to promote a progressive increase in baseline V _E. In this context, and for a better understanding of how RA motoneurons integrate, amplify and transform the brainstem synaptic transmission into a motor behaviour, we characterized their electrical, molecular and morphological properties from control rats and rats submitted to SH. Initially, using slices, we showed that RA motoneurons can be grouped into two distinct populations: tonically active and silent. They differ in their resting membrane potential, R _n and cellular excitability, implying cellular heterogeneity. This heterogeneity was also observed by Jodkowski et al. (1987, 1988) in inspiratory motoneurons from cats. From a physiological perspective, the existence of two populations of RA motoneurons is expected because it has been demonstrated that expiration is composed of two distinct phases: post‐inspiration and stage 2 expiration (Zoccal et al. 2008; Richter & Smith, 2014); with each phase determined by different brainstem ventral medullary neurons (Moraes et al. 2014). In the present study, SH induced hyperexcitability in tonically active RA motoneurons in vitro and in RA motoneurons firing during stage 2 expiration in situ. Thus, we hypothesize that tonically active and silent RA motoneurons may be recruited at distinct phases of expiration in rats.

Intrinsic properties of RA motoneurons

Few studies have aimed to understand the electrophysiological properties of expiratory motoneurons (de Almeida et al. 2010; Ford et al. 2014). Although internal intercostal muscles contractions have been demonstrated at rest (Leevers & Road, 1995; de Almeida et al. 2010), most studies suggest that their excitability is solely dependent on synaptic transmission (Hoit et al. 1988; Ono et al. 2010). However, in our in vitro experiments, we verified that addition of ionotropic receptor antagonists (GABA, glutamate and glycine) to the bathing solution did not abolish the activity of most tonically active motoneurons, indicating the contribution of intrinsic properties to the excitability of these cells or the involvement of other neurotransmitters (Bayliss et al. 1997; Rekling et al. 2000). Therefore, our findings support the concept that synaptic inputs and intrinsic conductance shaped the firing pattern of tonically active motoneurons.

The intrinsic electrical properties of cells lie on the subtypes of ion channels expressed in the cell membrane, especially the voltage‐gated ones. In nonspherical cells, such as motoneurons, the membrane potential is not clamped distally to the voltage clamp electrode, and the recorded current can be severely distorted as a result of the lack of space clamp (Jackson, 1992; Bar‐Yehuda & Korngreen, 2008). To circumvent this problem, we decided to analyse the ion channels involved in the spontaneous activity of tonically active RA motoneurons using ion channels blockers rather than recording ion currents. In the absence of synaptic connection, low voltage‐activated calcium channels and HCN channels have been described as candidates for the generation of spontaneous action potentials (Raman et al. 2000; Zhang et al. 2009a; Astori et al. 2011). Regarding tonically active motoneurons, we show that blockade of low voltage‐activated channels did not change their activity. On the other hand, blockade of HCN channels produced a significant change in the resting membrane potential, R _n, excitability and action potential AHP. Because the resting membrane potential critically determines the cellular excitability, HCN channels are shown to be important to the firing pattern of tonically active RA motoneurons.

Although the number of ion channels comprised in several families is quite large, in the present study, we show that at least voltage‐gated and delayed rectifier potassium, voltage‐gated sodium, HCN and background potassium channels have a significant contribution to the excitability and action potential waveform generation in tonically active RA motoneurons. As expected, voltage‐gated potassium and sodium channels have an important contribution to motoneurons excitability.

Hyperexcitability of RA motoneurons induced by SH

Studies have shown that exposure of rats to hypoxia produces active expiration (Janczewski & Feldman, 2006; Zoccal et al. 2008; Moraes et al. 2014). In the present study, 24 h of SH selectively modulated the activity of tonic RA motoneurons as a result of the depolarization of their resting membrane potential with a consequent increase in the firing frequency, in addition to being responsible for changes in the kinetic parameters of the action potential, hyperpolarization of the threshold and an increase in the maximum sodium conductance. In in vitro preparations, we also observed that SH leads to a significant increase in the frequency and amplitude of EPSCs in tonically active RA motoneurons, supporting previous studies suggesting that glutamate receptors are a target for hypoxia (Bickler et al. 2003; Bickler et al. 2004; Zhang et al. 2009b). We must take into consideration the experimental evidence indicating that glial cells can influence the neurotransmission in respiratory centres, and also that hypoxia can modify the activity of these cells (Oliet et al. 2001; Eltzschig & Carmeliet, 2011; Accorsi‐Mendonca et al. 2013; Mukandala et al. 2016; Stokes et al. 2017). Therefore, we cannot rule out the possibility that SH may alter glial cells from bulbospinal neurons and spinal cord, which may lead to the increased synaptic drive describe here. To investigate this interesting possibility, further studies are required.

Although most of the changes produced by SH were abolished by synaptic blockers, the resting membrane potential of tonically active motoneurons remained at more depolarized values, indicating a direct effect of SH on ion channels. Indeed, several ion channels have been described as O₂ sensors, such as Kv 1.3, Kv 4.2 and Kv 4.3. Furthermore, Nav 1.2 and Nav 1.3 are considered to be involved in the control of membrane potential and cellular excitability. However, the mRNA expression of these channels in tonically active RA motoneurons was not significantly different between groups, such that other ion channels not analysed in the present study may be modulated by this metabolic challenge (Buckler, 1997; Kemp & Peers, 2007; Peers & Wyatt, 2007; Trapp et al. 2008; Cao et al. 2009). Measurement of mRNA expression can be taken as an estimative of protein synthesis (Veys et al. 2012; Temporal et al. 2014), although post‐transcriptional regulatory mechanisms may disrupt this correlation. Thus, to avoid misinterpretation, protein quantification, rather than mRNA, should be performed to confirm the effects of hypoxia on voltage‐gated sodium and/or potassium ion channel expression. Our data allow us to suggest that SH induces a remodelling of the electrogenic elements of tonically active RA motoneurons to ensure appropriate RA contraction and that the changes in the EPSCs, probably arising from bulbospinal neurons (Kirkwood, 1995; Road et al. 2013), are the main causes for the observed alterations.

Although experiments with slices represent a powerful research approach in neuroscience, some synaptic connections can be absent. Thus, we also characterized respiratory modulation and the effects of SH on RA motoneurons using the brainstem–spinal cord in situ preparation of rats, which maintains connectivity between the central pattern generator for breathing in the brainstem and spinal motoneurons. We also described two populations of RA motoneurons in this preparation, firing at post‐inspiration and stage 2 of expiration. Differently from the in vitro experiments, silent motoneurons were not identified. This divergence may also be explained by the elimination of some synaptic connections. The presence of low amplitude expiratory‐related activity in abdominal muscles during baseline conditions has been demonstrated in several studies investigating in vivo preparations of anaesthetized rats (Pagliardini et al. 2011; Huckstepp et al. 2015). However, studies from our laboratory (Zoccal et al. 2008; Moraes et al. 2014; de Britto & Moraes, 2017; Magalhães et al. 2018) and others (Pagliardini et al. 2011; Huckstepp et al. 2015) have considered active expiration only when the activity (amplitude) of the abdominal nerve/muscle during stage 2 expiration is higher than that observed in the post‐inspiration using in situ and in vivo preparations of rats. Therefore, under baseline conditions, it is not surprising that RA motoneurons present expiratory‐related activity in situ, firing during post‐inspiration with similar or even higher frequency than those firing during stage 2 expiration. Regarding the effects of SH on RA motoneurons in situ, we observed decreases in the activity of motoneurons firing at post‐inspiration and increases in the activity of motoneurons firing during stage 2 expiration. The latter was also followed by increases in the frequency and amplitude of EPSCs, reinforcing the concept of hypoxic modulation of glutamatergic synaptic transmission. In this context, we speculate that changes in RA motoneurons firing pattern after SH are a consequence of an increase in the activity of augmenting‐expiratory neurons and depression in the activity of post‐inspiratory neurons in the Bötzinger Complex, as described previously by Moraes et al. (2014). Therefore, we hypothesized that the tonically active RA motoneurons characterized in slices may be firing during stage 2 expiration observed in situ or in vivo preparation. We also consider that SH induces active expiration by increasing their excitability and glutamatergic synaptic transmission, as described by de Britto & Moraes (2017) in abdominal motoneurons in response to hypercapnia/acidosis. To what extent the observed changes in the activity of motoneurons will be correlated to muscle fibre shortening during contractions comprise a very important question that remains to be addressed in further studies.

Hypoxia induces morphological changes in tonically active RA motoneurons

It has been reported that hypoxia induces changes in the morphology of different cells types, resulting in a remodelling of neuronal excitability and function (Pokorny et al. 1982; Langmeier et al. 1989; Kolb & Whishaw, 1998; Wallace et al. 2007). In the present study, we observed that tonically active RA motoneurons exhibit large soma and complex dendritic tree. Increasing evidence shows that dendritic complexity affects the electrical properties of cells (Barrett & Crill, 1974a, b ; Purves & Hume, 1981; Krishnan, 1983; van der Velden et al. 2012; Zhu et al. 2016). In pyramidal cells, for example, dendritic hypercomplexity is associated with altered neuronal activity as a result of a reduction in spike frequency adaptation episodes (van der Velden et al. 2012). On the other hand, increased soma area was associated with decreases in R _n in hippocampal neurons from mice (Santos et al. 2017), as well as in motoneurons from the oculomotor nucleus of rats (Torres‐Torrelo et al. 2014). In the present study, we observed that SH increases the morphological complexity and somatic surface area and also decreases the R _n of tonically active RA motoneurons. Thus, assuming that the density of ion channels responsible for resting conductance does not change, we speculate that changes in R _n may be correlated with morphological changes of these cells. Furthermore, increases in the dendritic complexity and number of nodes are associated with dendritic spine formation, as present in SH rats and absent under control conditions. Dendritic spines represent specialized compartments where excitatory synapses are located, and their function is to modify the synaptic strength, increasing the excitability of cells (Hering & Sheng, 2001). This phenomenon may explain the increase in EPSCs frequency and amplitude in tonically active motoneurons after SH. In this context, we propose that increases in the synaptic strength are also necessary to change the activity of tonically active RA motoneurons under SH conditions.

In conclusion, we present a body of experimental evidence demonstrating that, in response to acute hypoxia in conscious freely moving control rats and SH rats, RA contributes to active expiration, although to a lesser extent than IO. We also demonstrate that RA motoneurons comprise a heterogeneous population with distinct electrophysiological properties. SH induces active expiration by selectively changing the synaptic inputs of tonically active motoneurons in vitro and motoneurons firing during stage 2 expiration in situ. Morphological reconstruction of the neurons provide insightful information concerning the plastic alterations induced by this metabolic challenge, which may contribute to the electrical changes observed in these groups of cells.

Physiological implications

Functionally, we observed that RA contributes to active expiration by increasing its magnitude and incidence in response to acute hypoxia. We speculate that this is an essential and additional mechanism for recruiting the expiratory reserve volume and enhancing V _T (Jenkin & Milsom, 2014) and V _E during a high metabolic drive, at least in conscious freely moving rats. The presence of delayed rhythmic expiratory activity in RA compared to IO was also described in humans during progressive exercise (Abraham et al. 2002), suggesting that this muscle is also recruited to improve active expiration in face of other metabolic challenges. The present study also advances knowledge regarding the RA motor activity in the level of spinal motoneurons, shedding new light on the electrophysiological, molecular and morphological properties of these cells after SH and at rest. Furthermore, our results are important for a better understanding of cellular mechanisms involved in RA expiratory function during physiological conditions, such as coughing, sneezing, speaking, physical activity and REM sleeping, as well as during respiratory and motor disorders, such as Rett syndrome, amyotrophic lateral sclerosis and obstructive sleep apnoea, comprising conditions in which motoneuronal functions are affected.

Additional information

Competing interests

The authors declare that they have no competing interests.

Author contributions

MPDS and LGHB contributed to experiments involving fluorescent microsphere microinjections. MPDS and DJAM contributed to experiments involving in vivo, in vitro and in situ recordings and 3D motoneuron reconstruction. MPDS and ADSM performed the qRT‐PCR experiments. MPDS, DJAM, WAV and BHM contributed to the conception, experimental design, data analyses and interpretation of the findings and the preparation of the manuscript. 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 ‘Fundação de Amparo à Pesquisa do Estado de São Paulo’ (FAPESP – postdoctoral fellowship to MPDS 2015/01073‐3 and Thematic Project to BHM, 2013/06077‐5; Young Investigator Grant to DJAM, 2013/10484‐5).

Biographies

Melina P. da Silva is postdoctoral fellow at the School of Medicine of Ribeirão Preto, University of São Paulo, Brazil. She holds her PhD in science at the same Institution (2015). Her research is focused on Physiology, with an emphasis on Cellular Biophysics and Neurophysiology, mainly concerning the electrophysiology of excitable cells.

Benedito H. Machado is a Professor of Physiology at the School of Medicine of Ribeirão Preto, University of São Paulo, Brazil. The main current focus of his laboratory is related to possible changes in the synaptic transmission of the chemoreflex pathways at the brainstem and spinal cord of rodents submitted to hypoxic challenges.