Amino acid modulation of the carotid body selectively modulates peripheral chemoreceptor respiratory reflex

Abstract

Abstract

Cardiorespiratory homeostasis is maintained by carotid body (CB) peripheral chemoreceptors that monitor blood oxygen and stimulate breathing. While proportionality between peripheral chemoreflex sensitivity and respiratory response is relative to hypoxia severity, the fundamental regulatory mechanisms remain unknown. We hypothesized that intercellular amino acid transmission between glomus cells modulates CB sensitivity. We identified a full complement of glutamate and gamma‐aminobutyric acid (GABA) synaptic signalling machinery in the CB. During hypoxia stimulation, glutamate acting via NMDA receptors (NMDA‐R) stimulated CB afferents while releasing GABA, which acted on GABA_A receptors to rapidly attenuate afferent drive mediated by purinergic P2X receptors. We propose that this intrinsic ‘accelerator–brake’ mechanism regulates CB sensitivity to hypoxia. Further, glutamate and NMDA‐R were essential for long‐term facilitation (LTF) of respiratory chemoreflex responses induced by repeated hypoxia or glutamate exposures. The accelerator–brake mechanism establishes the set‐point of chemoreflex gain and provides plasticity for modulation, explaining ventilatory LTF to repeated bouts of hypoxia.

Key points

• An intra‐carotid body interaction between glutamate and GABA transmission is crucial for the generation of the biphasic carotid body afferent nerve response when stimulated with low doses of sodium cyanide.

• Glutamate stimulates the carotid body chemoreflex to increase inspiratory drive without affecting sympathetic activity, heart rate or perfusion pressure, indicating selective activation of a carotid body respiratory reflex arc.

• Repetitive application of glutamate to the carotid body sensitizes chemoreflex phrenic nerve responses producing long‐term facilitation.

Abstract figure legend When the carotid body (CB) detects low oxygen levels (hypoxia), this leads to: (1) depolarization of the chemosensory cells (glomus cells) within the CB – the resulting increase in intracellular Ca²⁺ stimulates the exocytotic co‐release of vesicles containing ATP and glutamate; (2) synaptic ATP acts on postsynaptic petrosal neuron P2X2/3 receptors, and activation initiates the depolarization of petrosal afferent fibres; (3) glutamate acts on ionotropic NMDA receptors on adjacent glomus cells (paracrine) and back on the cell from which it was released (autoreceptor); and (4) increases in intracellular Ca²⁺ due to NMDA activation trigger the mobilization of GABA vesicles. GABA binds to GABA_A receptors on adjacent glomus cells, inhibiting their excitation and dampening the CB chemoreflex response. This intrinsic ‘accelerator–brake’ mechanism sets the chemoreflex gain and modulates breathing.

Introduction

The carotid body is the primary sensor of arterial oxygenation in mammals. Through reflex pathways, it protects cardio‐respiratory homeostasis in critical situations of hypoxia as experienced during parturition, high altitude and sleep apnoeas. Its ability to stimulate breathing is unique and profound (Conde et al., 2014; López‐Barneo, 2022; Nurse, 2014; Pijacka et al., 2016; Smith et al., 2015; Wilson & Teppema, 2016). Although the molecular mechanisms regulating carotid body sensitivity, as defined by the transfer function of hypoxia severity to increased ventilation, remain poorly understood, modulation of its sensitivity has provided a therapeutic opportunity for many sympathetically mediated diseases in animals and humans (Hering et al., 2007; Lataro et al., 2023; Narkiewicz et al., 1999a; Sacramento et al., 2017; Schultz et al., 2015). Both preclinical and clinical studies have shown that carotid body sensitivity is potentiated in cardiometabolic diseases (Chua et al., 1997; Cunha‐Guimaraes et al., 2020; Narkiewicz et al., 1999b; Ponikowski et al., 2001; Sun et al., 1999). Moreover, carotid body denervation/resection or pharmaco‐inhibition attenuates sympathetic tone in hypertension and heart failure, improves insulin resistance and glucose tolerance in diabetes, and improves respiratory stability in heart failure (Conde et al., 2018; Del Rio et al., 2016; McBryde et al., 2013; Paton & Spyer, 2013). These findings suggest that therapeutically adjusting the set point of carotid body sensitivity could offer significant clinical benefits in human patients with different cardiometabolic and respiratory diseases (Narkiewicz et al., 2016; Niewinski et al., 2021). Thus, an understanding of the intrinsic mechanisms regulating sensory sensitivity is vital for exploiting potential putative therapeutic approaches. In this regard, we have considered both the excitatory and inhibitory amino acid transmitter systems and their plasticity within the carotid body.

The carotid body demonstrates sensory plasticity following repeated hypoxia stimulation, where long‐term adaptations cause heightened chemosensitivity (Peng et al., 2003). In the brain, neural plasticity in the form of synaptic long‐term potentiation is mediated by glutamate and associated receptors (Luján et al., 2005; Zhou & Danbolt, 2014). Comparable to the CNS, the carotid body also contains glutamatergic and GABAergic machinery such as receptors, and uptake and release proteins (Fearon et al., 2003; Igarashi et al., 2009; Li et al., 2021; Oomori et al., 1994) which influence both autocrine and paracrine signalling pathways (Li et al., 2021; Nurse, 2005; Zhang et al., 2009). Activation of glutamate receptors has been shown to both increase and decrease carotid sinus nerve activity in response to hypoxia (Li et al., 2021, 2023; Liu et al., 2009; Zhao et al., 2022, 2023). Regarding GABAergic transmission, experimental data indicate that glomus cells express functional GABA_A receptors (Igarashi et al., 2009) and metabotropic GABA_B receptor subunits appear functional on glomus cells (Fearon et al., 2003). Additionally, GABA acting through GABA_A receptors has been shown to attenuate afferent drive mediated by purinergic P2X receptors (Zhang et al., 2009). However, the precise role of glutamate and GABA in modulating the chemoreflex and whether they interact is not completely understood.

Herein, we hypothesize that signalling between glutamatergic and GABAergic mechanisms within the carotid body is inter‐linked to govern its afferent sensitivity and chemoreflex control of respiration.

Methods and materials

Ethical approval

All experiments were performed in accordance with the biomedical research guidelines for animal welfare and approved by the University of Auckland Animal Ethics Committee (AEC No. 2274, 22280 and 26798). All animal procedures performed were in accordance with the guidelines from Directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes.

Animals

The experiments were conducted in male Wistar rats (12–14 or 4–6 weeks; n = 277) bred by the Vernon Jansen Unit (VJU) of the University of Auckland. We used 4–6‐week‐old rats for the in situ working heart–brainstem preparation (WHBP), in vitro isolated carotid sinus nerve recordings and single‐cell calcium imaging, as older animals are incompatible with the WHBP. Rats aged 12–14 weeks were used for digital droplet (dd)PCR and immunohistochemistry. Older animals were included to align with future planned in vivo studies. Animals were housed in conventional Tecniplast 1500 rat cages with controlled temperature (21 ± 2°C), humidity (55 ± 10%) and 12 h light–dark cycle. Animals had unlimited access to food and water.

Transcriptome data mining

We utilized two distinct RNA‐sequencing (RNA‐seq) datasets to investigate glutamatergic signalling machinery in the carotid body. Rat carotid body (GEO: GSE178504; Pauza et al., 2022) and mouse carotid body transcriptomes were accessed as previously published (GEO: GSE72165; Chang et al., 2015). Gene annotation information was retrieved using the org.Mm.eg.db (version 3.16.0) and org.Rn.eg.db (version 3.16.0) packages in R. Read counts for orthologue genes expressed as reads per kilobase million (RPKM) were used to compare expression between species.

Carotid bifurcation surgical extraction

Wistar rats were deeply anaesthetized with 5% isoflurane in O₂ (1 L/min) until loss of paw and tail withdrawal reflexes, and then killed by cervical dislocation. The left and right intact carotid artery bifurcations containing the carotid body were exposed via a longitudinal incision on the ventral surface of the neck; the salivary glands, and sternomastoid (STM) and sternohyoid (SHM) muscles were cut away and retracted. Under a stereoscopic microscope, the bifurcations were surgically resected and subsequently transferred onto a Sylgard plate filled with specified solution according to the following protocol (see below) for further dissection.

Digital droplet polymerase chain reaction

Glutamate and GABA gene expression analysis in the carotid bodies was performed on 12–14‐week‐old, male Wistar rats. After extraction, carotid bodies were micro‐dissected from the common carotid bifurcation, placed into homogeniser tubes (Cat. # 9790B, Takara, Shiga, Japan) and snap‐frozen in liquid nitrogen. Samples were stored at −80°C until batch processed. Carotid bodies were homogenized in RLT Plus lysis buffer from the RNeasy Plus Micro Kit (Cat. # 74034, Qiagen, Valencia, CA, USA), supplemented with β‐mercaptoethanol. Genomic DNA was removed using gDNA Eliminator spin columns, and total RNA was purified according to the manufacturer's instructions. RNA was eluted in 14 µL of RNase‐free water. RNA concentration was determined using the Qubit™ RNA High Sensitivity Assay Kit (Cat. # Q32852, Thermo Fisher Scientific, Waltham, MA, USA). RNA integrity was assessed using the Agilent 2100 Bioanalyzer (Cat. # G2939BA, Agilent, Santa Clara, CA, USA) with the RNA 6000 Pico Kit (Cat. # 5067‐1513, Agilent). Only samples with RIN ≥7 were included in downstream analyses.

Reverse transcription was performed using the iScript™ gDNA Clear cDNA Synthesis Kit (Cat. # 1725035, Bio‐Rad, Hercules, CA, USA). The cDNA dilution factor used for the following experiments was 1:5. ddPCR was carried out using the QX200 droplet digital PCR system (Bio‐Rad). For the ddPCR experiments, samples were prepared using the QX200 ddPCR Supermix for Probes without dUTP (Cat. # 1863025, Bio‐Rad). We prepared the reaction mixture by diluting 3 µL of cDNA into 19 µL of master mix. This master mix included 2× ddPCR supermix, UltraPure™ distilled water (Cat # 10977015, Invitrogen, Carlsbad, CA, USA), and the following TaqMan probes: Rn06323759_m1 (Gria1), Rn00568514_m1 (Gria2), Rn00680474_m1 (Grin2B), Rn00561359_m1 (Grin2C), Rn01413643_m1 (Gabra2), Rn00567055_m1 (Gabra3), Rn00589846_m1 (Gabra4), Rn00568803_m1 (Gabra5), Bt03212885_g1 (Ubb), dRnoCPE5187991 (Actb) and dRnoCPE5167180 (B2m). Each biological replicate was prepared in duplicate. After 15 min at room temperature, 20 µL of the PCR reaction mix was loaded into the middle rows of the DG8 droplet generation cartridge (Cat. # 1864008, Bio‐Rad) alongside 70 µL of droplet generation oil (Cat. # 1863004, Bio‐Rad) in the bottom of the wells. Droplet portioning was achieved by the QX 200 Droplet Generator (Bio‐Rad), whilst PCR amplification was performed using the C1000 Touch Thermal Cycler (Bio‐Rad). Finally, the concentration of target gene droplets was quantified using the QX200 Droplet Reader (Bio‐Rad), which classifies them as either positive or negative depending on their fluorescence intensity. Acquired ddPCR data were analysed using the QuantaSoft Analysis Pro software (Bio‐Rad, version 1.0). For comparison between groups, transcript abundance was calculated relative to three housekeeping genes: β‐actin (Actb), β₂ macroglobulin (B2m) and ubiquitin B (Ubb). These three values were averaged using the geometric mean to obtain a stable normalization factor. This reference mean was then used to calculate the relative abundance of the target transcripts across groups, enabling accurate comparisons between conditions.

Immunohistochemistry

Wistar rats were deeply anaesthetized with 5% isoflurane in O₂ (1 L/min) until loss of paw and tail withdrawal reflexes, and then killed by cervical dislocation. The common carotid artery bifurcations were then dissected as per carotid body surgical extraction and placed in 4% paraformaldehyde solution (Sigma‐Aldrich, St Louis, MO, USA) in PBS for an additional 4 h of fixation at 4°C. Samples were cryoprotected by overnight immersion in 30% (w/v PBS) sucrose (S9378; Sigma‐Aldrich) solution in PBS at 4°C. Samples were embedded in Tissue‐Tek OCT compound (Cat. # 4586, Tissue‐Tek^®, Sakura Finetek, Torrance, CA, USA), frozen and cryo‐sectioned in 18 µm transverse sections from rostral to caudal direction. This was done by inspecting the presence of the carotid body tissue every 100 µm in a ‘pilot’ section stained with 0.1% (w/v) toluidine blue (89640; Sigma‐Aldrich). Once the first glomus cell clusters were identified, all subsequent tissue sections were mounted on SuperFrost Plus™ (10149870; Thermo Fisher Scientific) slides. The collection was stopped once the absence of the carotid body was confirmed by contrast illumination under a benchtop brightfield microscope. During the collection, sections were mounted onto the SuperFrost® Plus microscopy slide (LBS4951+, LabServ, Auckland, New Zealand) in chronological order, representing the rostral‐to‐caudal progression.

Sections were rinsed in PBS and incubated in 5% (v/v) normal donkey serum (Cat. # D9663, Sigma‐Aldrich), 5% (v/v) normal goat serum (Cat. # 16210064, Sigma‐Aldrich) and 0.3% (v/v) Triton X‐100 solution in PBS for 2 h to block non‐specific binding and for tissue permeabilization. Sections were rinsed in PBS (3× 10 min) and incubated in primary antisera solution in 5% (v/v) normal serum at 4°C overnight. Primary antibodies were: anti‐tyrosine hydroxylase (TH) (1:500; MAB318; Sigma‐Aldrich); anti‐glutamate receptor 1 (gria1) (1:1000; Ab1504; Sigma‐Aldrich); anti‐NMDA receptor 2A (grin2c) (1:100; acg‐018; Alomone, Jerusalem, Israel). Next, sections were rinsed in PBS (3× 10 min), followed by secondary antisera incubated for 4 h at room temperature. Secondary antibodies were: goat anti‐mouse IgG H&L, Alexa Fluor® 488 (1:500; Ab150113; Abcam, Cambridge, MA, USA); and goat anti‐rabbit IgG H&L, Alexa Fluor™ 594 (1:500; A11037; Invitrogen). Next, sections were rinsed in PBS and incubated in Lycopersicon esculentum (tomato) lectin (1:100; DL11741; Vector Laboratories, Burlingame, CA, USA) for 15 min at room temperature. Sections were rinsed in PBS and incubated in 4′,6‐diamidino‐2‐phenylindole (DAPI) (1:10,000; Cat. # 62248; Invitrogen) for 10 min at room temperature. Lastly, sections were mounted using VectaShield^® Antifade Mounting Medium (Cat. # H‐1000‐10, Vector Laboratories) and kept in the dark at 4°C until imaged.

Signal specificity in target tissue was confirmed in no primary antibody control reactions. Samples were imaged using an Olympus FV3000 confocal microscope (Olympus, Japan) using a 10× objective (UPLAXAPO10X, 0.4 NA) to locate the carotid body initially. Then, a 60× oil immersion objective (UPLlanSApo 1.35 NA) was used to image carotid body cell clusters acquired at 2048 × 2048 pixels. Excitation wavelengths were 280, 493 and 353 nm, with respective emission wavelengths of 618, 517 and 465 nm; the emission filters were 580–700, 400–585 and 400–580 nm, respectively. Image processing was performed using ImageJ/Fiji (version 2.8.0/1.53t).

Single cell calcium imaging

Wistar rats were deeply anaesthetized with 5% isoflurane in O₂ (1 L/min) until loss of paw and tail withdrawal reflexes, and then killed by cervical dislocation. Following this, carotid bodies were chilled in O₂‐equilibrated HEPES‐Tyrode (HT) solution of the following composition (in mM): 140 NaCl, 5 KCl, 1.1 MgCl₂, 5 glucose and 10 HEPES. Following excision, they were enzymatically digested in Ca²⁺/Mg⁺‐free HBSS (Cat. # 14170120, Thermo Fisher Scientific) containing 2 mg/mL collagenase type IV (Cat. # 17104019, ThermoFisher), 0.4 mg/mL trypsin (Cat. # T4799, Sigma‐Aldrich) and 0.03 mg/mL DNase II (Cat. # D4527, Sigma‐Aldrich) for 60 min at 37°C. At 15, 30, 45 and 60 min, carotid bodies were mechanically agitated for 2 min by repeated aspiration through a fire‐polished Pasteur pipette, followed by the addition of 50 µM bovine serum albumin (pH 100, pH Scientific) to terminate digestion. Cells were subsequently pelleted (500 rcf, 7 min, 4°C) and transferred to Poly‐D‐Lysine (Cat. # A3890401, Thermo Fisher Scientific) coated 13 mm round coverslips (Cat. # MAR0111530, Marienfeld Superior, Lauda‐Königshofen, Germany) for 1 h before Ca²⁺ imaging or immunocytochemistry.

Glomus cells were loaded with Fluo‐4 AM (5 µM, ab241082, Abcam) for 15 min at room temperature and imaged using an upright Olympus FV3000 confocal microscope (Olympus, Japan) using a 40× water dipping objective (LUMPLFLN40 XW, NA 0.8). Cells were superfused via a mechanized syringe pump (1.5 mL/min), delivering Ca²⁺‐containing (2 mm) HT solution at 33°C. Each set of recordings were 5 min in duration, comprising 93 frames at 3.22 s per frame and a frame size of 1024 × 1024 pixels. Each cell was subjected to two or three sequential stimuli of 1 mm KCN, or 15 mm glutamate and 15 mm glycine, and/or 100 µM MK‐801 in a randomized order.

From the recordings, only KCN‐responsive glomus cells were considered as viable chemosensory cells and used for subsequent quantification. Ca²⁺ response events were analysed using custom‐written FIJI macros whereby a region of interest is drawn around each cell. Event parameters were measured by amplitude (ΔF/F ₀, change in fluorescence above background), frequency, duration and time to peak fluorescence. Event mass was taken as the product of event amplitude and duration. Ca²⁺ response proportion was determined as a fraction of the cells responding to a given stimulus to the total number of KCN‐responsive cells.

Carotid sinus nerve recording to investigate glutamate and GABA modulation in vitro

Wistar rats were deeply anaesthetized with 5% isoflurane in O₂ (1 L/min) until loss of paw and tail withdrawal reflexes, and then killed by cervical dislocation. After extraction (see ‘Carotid body surgical extraction’), the bifurcations of 4–6‐week‐old male Wistar rats were subsequently placed in a recording chamber and superfused continuously with Ringer solution (composition in mmol/L as follows: NaCl, 125; NaHCO₃, 24; KCl, 3.75; CaCl₂, 2.5; MgSO₄, 1.25; KH₂PO₄ 1.25 and d‐glucose 10; Sigma‐Aldrich). The superfusate was gassed with carbogen (5% CO₂ in 95% O₂), heated to 36–37°C (TC‐324C, Warner Instruments, Hamden, CT, USA), filtered (nylon mesh, 25 µM; Millipore, Billerica, MA, USA) and circulated at 5 mL/min (Gilson, Minipuls^® 3). Under a dissecting stereo microscope, the common carotid artery (CCA) was cannulated using polyethylene tubing (PE‐10) so that the preparation was arterially perfused via the CCA. After cannulation, the carotid sinus nerve (CSN) was carefully isolated, and its activity was recorded using a bipolar glass suction electrode (Cat # 1B150F‐4, World Precision Instruments Inc., Sarasota, FL, USA). The signal was amplified ×10 000 (AM systems, 1700), band‐pass filtered (100 Hz to 1 kHz), and digitized (20 kHz, Micro1401‐3, CED, Cambridge, UK). Acquisition and analysis were performed using Spike2 software (CED, version 10.20a).

Potassium cyanide (KCN; 0.08% w/v, 100 µL) was used as an experimental tool to chemically induce histotoxic hypoxia. KCN was injected as a bolus using a 1 mL syringe through a side port in the perfusion line near the CCA to elicit a control carotid body sensory response. This approach chemically mimics the effects of hypoxia by inhibiting cytochrome c oxidase, thereby triggering similar downstream signalling events to those induced by hypoxic hypoxia (Cabello‐Rivera et al., 2022). The agent was fundamental in teasing apart mechanisms of intra‐carotid body signalling by glutamate and GABA. Additionally, for hypoxic hypoxia stimulation, we gassed the Ringer solution with 10% O₂ and 5% CO₂ equilibrated in N₂. Hereafter, the term hypoxia will refer to hypoxic hypoxia. The carotid body was challenged with excitatory and inhibitory agonists and antagonists as detailed below; all drugs were perfused through the common carotid artery. Between each stimulus, whether drug or KCN, a 5 min wait period was allowed for carotid body recovery. We quantified the CSN discharge in two ways: by calculating the percentage increase in the area under the curve (AUC) and maximum peak relative to the baseline immediately before the stimulus; the period of baseline used for this calculation was approximately the same time‐length as the chemoreflex response (e.g. 5 s).

Working heart–brainstem preparation

We modified the WHBP (Paton, 1996) to enable vascular isolation of a single common carotid bifurcation, creating what we refer to as the in situ double‐perfused WHBP (dpWHBP; Fig. 1). This differs from the approach used by Day & Wilson (2005, 2007), as only one carotid body was isolated and stimulated, while the contralateral carotid body remained intact and perfused via the systemic circulation (Day & Wilson, 2005, 2007). Juvenile rats (3–6 weeks old, 50–80 g) were deeply anaesthetized with isoflurane (5% in O₂, 1 L/min, via inhalation) until the loss of paw and tail withdrawal reflexes, and then heparinized (350 UI i.p.; Pfizer). Subsequently, animals were killed via exsanguination following bisection below the diaphragm. After cooling the upper body in Ringer solution (composition in mmol/L as follows: NaCl, 125; NaHCO₃, 24; KCl, 3.75; CaCl₂, 2.5; MgSO₄, 1.25; KH₂PO₄ 1.25 and d‐glucose 10; Sigma‐Aldrich), animals were decerebrated pre‐collicularly, the lungs were removed, and the descending aorta was isolated and cannulated with a double‐lumen catheter. Retrograde perfusion of the thorax and head restored viability indicated by the return of a ramp‐like phrenic nerve (PN) discharge pattern characteristic of eupnoea and adequate brain oxygenation (Paton, 1996; Paton et al., 2022). The perfusate was Ringer solution supplemented with an oncotic agent (1.5%, polyethylene glycol, 95172–250G‐F, Sigma‐Aldrich), gassed with carbogen (5% CO₂, 95% O₂), warmed to 31–32°C, filtered with nylon mesh (25 µm; Millipore, Sigma‐Aldrich) and recirculated. The second lumen of the cannula was connected to a Neurolog pressure transducer (NL108T2, Digitimer, Welwyn Garden City, UK) and amplifier (NL108A, Digitimer) to monitor perfusion pressure (PP) in the aorta. The PP was maintained at 55–90 mmHg adjusting the peristaltic pump flow (20–25 mL/min; Watson‐Marlow 530s, Falmouth, UK) and the addition of vasopressin (2–2.5 nmol/L, V9879‐5MG, Sigma‐Aldrich) into the perfusate. Neuromuscular blockade was achieved with vecuronium bromide (10 mg/mL, Mylan, New Zealand) added into the reservoir to arrest respiratory‐related movement.

Figure 1. Schematic of the double perfused working heart–brainstem preparation (dpWHBP) with a perfused isolated common carotid artery bifurcation.

A, the systemic circulation (light blue arrows) is achieved by cannulating the descending aorta with a double‐lumen catheter: one lumen for arterial perfusion with Ringer solution (gassed with 5% CO₂/95% O₂) and the other for recording perfusion pressure. To isolate the right carotid body (CB) circulation, the CCA, ICA and ECA were cannulated. The CCA (green arrows) perfuses the bifurcation, the ICA is connected to a pressure transducer and the ECA (purple arrows) drains the perfusate. Two roller pumps circulate perfusate to each circulation with bubble traps and filters. CB activation was induced with potassium cyanide (KCN) injections via the CCA side port, while electrocardiogram (ECG), thoracic sympathetic (tSNA) and phrenic nerve (PNA) activities were recorded using bipolar suction electrodes. CB perfusion steps (1–10): steps 1–3 use a roller pump to circulate perfusate; step 4 diverts flow via a three‐way stopcock into the CCA (step 5) or a shunt (step 5b), where a Starling resistor adjusts CCA perfusion pressure. Steps 6–9 drain perfusate (purple arrows) to the outflow line connected to the roller pump. A sideline at step 1 allows fixed‐volume drug perfusion into the CB. Outflow during drug perfusion (steps 9–10) is temporarily discarded to prevent systemic drug effects. B, in the dpWHBP, the systemic and CB circulations are fully isolated. To confirm no significant leakage between them, we split the circulations into separate reservoirs. The CB circulation was infused with Fast Green dye, producing a distinct blue–green colour (see ‘Perfusate Reservoirs’ column). If mixing occurred, the systemic perfusate would change colour over time. Images show the WHBP, carotid artery bifurcation and perfusate reservoirs at time ‘0’ (before the dye reached the bifurcation) (i), and 30 min (ii) and 1 h:30 min (iii) after dye addition. Abbreviations: CCA, common carotid artery; ECA, external carotid artery; ICA, internal carotid artery; OCA, occipital carotid artery.

To isolate circulation to the right carotid body, a second peristaltic pump from the same reservoir perfused the right CCA (Fig. 1A ). In this circuit, the perfusate was also warmed (31–32°C), filtered and passed through a bubble trap. The trachea and larynx were removed to access the right CCA and its branches. The CCA and internal carotid artery (ICA) were cannulated with PE tubing (OD 0.61 mm, ID 0.28 mm and OD 0.50 mm, ID 0.2 mm, respectively). The ICA cannula was connected to a pressure transducer (NL108T2, Digitimer) to monitor the bifurcation perfusion pressure, which was adjusted to match aorta pressure (60–100 mmHg). The external carotid artery (ECA) and occipital carotid artery (OCA) were clamped to achieve vascular isolation of the carotid body. A small ECA branch proximal to the clamp remained patent, allowing perfusate drainage and return to the reservoir for re‐circulation.

Simultaneous recordings of the PN and thoracic sympathetic chain (tSNA; between T13 and L3) were obtained using bipolar glass suction electrodes. Heart rate (HR) was derived from inter R‐wave intervals of the electrocardiogram (ECG) recorded through two electrodes from each forelimb of the preparation. Signals were amplified ×10,000 (A‐M Systems model 1700, Carlsborg, WA, USA), filtered (bandwidth 10 Hz to 5 kHz, A‐M Systems), digitized (20 kHz, Micro1401‐3, Cambridge Electronic Design), and recorded using Spike2 software (Cambridge Electronic Design). Background noise was determined 15 min after stopping the peristaltic pump and subtracted from the tSNA signal.

To prove there was no significant leaking of drugs into the systemic circulation through the CCA–cannula connection, we carried out the following control experiment. Instead of having one perfusate reservoir for both the systemic circulation and the isolated right CCA bifurcation, we split it into two reservoirs dedicated to each compartment. In the isolated bifurcation, we added Fast Green dye sufficient to give the solution an intense colour; therefore, if there were leaks, we would observe changes in colour in the solution, perfusing the systemic circulation (Fig. 1B ). For over 1 h we did not observe significant changes in the colour of solution perfusing the systemic circulation.

Protocol in the dpWHBP

In the dpWHBP, a KCN bolus (0.08%; 50 µL) was injected using a Hamilton syringe via a side port on the perfusion line connected to the CCA to generate a control response. Using the sideline, the carotid body was then perfused with glutamate (10 mL, 15 mm) and glycine (10 mL, 15 mm), followed by a second KCN injection. After a 5 min recovery period, the carotid body was perfused with MK‐801 (100 µM, 10 mL) to antagonize NMDA receptors, followed by a solution containing glutamate with glycine and MK‐801. At the end of the latter perfusion, the carotid body was challenged with a third KCN injection 5 min thereafter. To prevent the systemic effects from drug recirculation, outflow from the ECA line was temporarily discarded in a separate container once perfusion began. In all dpWHBP figures, glutamate and glycine co‐administration are labelled ‘glutamate’.

The chemoreflex consisted of increased PN activity, bradycardia, sympathoexcitation and increased PP. We quantified the chemoreflex in two ways. First, we calculated the percentage increase in respiratory rate (i.e. tachypnoea) and sympathoexcitation relative to the baseline immediately before the stimulus with noise subtracted; the period of baseline used for this calculation was the same time‐length approximately as the chemoreflex response (e.g. 5 s). Second, the maximum bradycardia and increase in PP were calculated as the change (Δ) in HR (bpm) and PP (mmHg) from baseline. At least 5 min elapsed between consecutive KCN doses. The central inspiratory drive was quantified by calculating the area under the curve (AUC) of the phrenic nerve activity and then normalizing it to the inspiratory time (T _i).

Chemicals and reagents

This study exposed the carotid body to a series of pharmacological agents in Ringer solution. Glutamate (15 mm, Cat. # ab120049‐1002, Abcam) and glycine (15 mm, Cat. # 50046, Sigma‐Aldrich) were co‐administered in 5 mL of ACSF. Co‐binding of glutamate and glycine at distinct sites is required for NMDA receptor activation (Johnson & Ascher, 1987; Kleckner & Dingledine, 1988), enabling ion channel opening and calcium influx (Benveniste & Mayer, 1991; Clements & Westbrook, 1991). To avoid repetition, note that in all figures, each glutamate stimulus (labelled ‘glutamate’) was co‐administered with 15 mm glycine. Additionally, we used NMDA (Cat. # M3262, Sigma‐Aldrich; 15 mm; 5 mL ACSF) and AMPA (Ca. # 324817, Sigma‐Aldrich; 15 mm; 5 mL ACSF) receptor agonists (15 mm; 5 mL ACSF). We used an exploratory dose of 15 mm based on its prior use in patch‐clamp experiments (Lonart & Zigmond, 1991; Normann et al., 1986). Inhibitors including MK‐801 (Cat. # ab120027‐1001, Abcam; 100 µM; 10 mL ACSF); dl‐AP5 (Cat. # ab120004, Abcam; 1 mm; 10 mL ACSF); NBQX (Cat. # ab120046‐1001, Abcam; 100 µM; 10 mL ACSF); PPADS (Cat. # RDS062510, In Vitro Technologies, Noble Park North, Australia; 100 µM; 5 mL ACSF); Sulpiride (1 mm in 10 mL ACSF; Cat. # S7771‐5G, Sigma‐Aldrich) were used based on previous studies (Liang et al., 2024; Péczely et al., 2024); bicuculline (Cat. # 14340, Sigma‐Aldrich; 500 µM; 10 mL ACSF); CGP 55845 (Cat. # SML0594, Sigma‐Aldrich; 50 µM; 10 mL ACSF); and GABA (Cat. # AB120359‐1001, Abcam; 15 mm; 5 mL ACSF) were al;so used. While concentrations were informed by previous studies (Frankiewicz et al., 1996; Igarashi et al., 2009; Zhang et al., 2007), we applied higher doses to account for limitations in drug delivery inherent to our preparation. Because the bolus was administered via the common carotid artery, and the carotid body is both small and highly resistant, only a fraction of the perfusate enters the target tissue. Most of the flow is diverted into the internal and external carotid arteries, making the effective concentration at the site of action lower than the applied dose. Bicuculine, CGP 55845 and sulpiride were dissolved in DMSO to make stock solutions. All the other drugs were dissolved in ACSF.

Statistical analysis

Results are expressed as the mean ± standard deviation (SD). Paired Student's t‐test, one‐way ANOVA and one‐way repeated measures ANOVA were used as appropriate. Post hoc tests included Tukey's test for one‐way ANOVA and Bonferroni correction for repeated measures ANOVA. Data distribution was assessed by identifying outliers (defined as values above Q3 + 1.5 × IQR or below Q1 − 1.5 × IQR) and extreme outliers (above Q3 + 3 × IQR or below Q1 − 3 × IQR), evaluating normality with the Shapiro–Wilk test, and visually inspecting frequency histograms, density plots and Q–Q plots. Homogeneity of variance was tested using Levene's test. When this assumption was violated, Welch's ANOVA with Games–Howell post hoc test was applied. Differences were considered to be statistically significant at P < 0.05. Statistical analysis was carried out using the ‘stats’ (v4.2.0) and ‘rstatix’ (v0.7.2) packages in R. Data visualization was carried out using the ‘ggplot2’ (v3.4.4) package in RStudio.

Results

To catalogue glutamatergic signalling machinery in the rodent carotid body, we mined public transcriptome data highlighting the prevalence of ionotropic AMPA receptor (AMPAR) (Gria1,2,3) and NMDAR (Grin2c, Grina) channels and sparse expression of metabotropic glutamate receptors (mGluRs) (Fig. 2A ) (Chang et al., 2015; Pauza et al., 2022). Published single‐cell RNA‐seq data further confirmed the expression of AMPARs (Gria1‐4, Grid1‐2) (Fig. 2B ) and NMDARs (Grin1, Grin2a‐d, Grin3a‐b, Grina) in isolated murine glomus cells (Fig. 2C ) (Zhou et al., 2016). Consequently, we focused on ionotropic channels as the predominant subtype of glutamatergic transmission within the carotid body.

Figure 2. Transcriptomic expression of glutamatergic machinery in the carotid body.

A, components of the glutamatergic system expressed in the rat carotid body from bulk RNA sequencing (RNA‐seq) (Pauza et al., 2022) and in the mouse RNA‐seq data set (Chang et al., 2015). Coloured dots indicate glutamate receptors, transports and scaffolding proteins. Data are presented as the log₂ values of reads per kilobase of transcript, per million aligned reads (RPKM). Expression values were log₂ normalized to account for expression differences between abundant and sparsely expressed genes. The term ‘reads’ refers to short sequencing of RNA corresponding to a particular gene that is expressed in target tissue. B, α‐amino‐3‐hydroxy‐5‐methyl‐4‐isoxazolepropionic acid (AMPA) receptor subunits Gria1–4 and Grid1–2 expression from RNA‐seq data (Zhou et al., 2016) using isolated murine glomus cells. C, N‐methyl‐d‐aspartate (NMDA) receptor subunits Grin1, Grin2a–d, Grin3b and Grina expression from single cell RNA‐seq data (Zhou et al., 2016) using isolated murine glomus cells. In B and C data are presented as read counts generated. Mean ± SD (n = 8 glomus cells).

First, we tested the effects of exogenous glutamate on carotid body chemosensory discharge using an ex vivo carotid body CSN preparation (Fig. 3A ). Potassium cyanide (KCN) was used to induce experimental histotoxic hypoxia as this treatment evokes a sub‐maximal, temporally controlled and highly replicable CSN response permitting investigations into changes in carotid body chemosensitivity (Fig. 3B ). We observed the CSN response to KCN consisted of two distinct components: (1) an initial high‐amplitude, short‐interval response followed by (2) a secondary, low‐amplitude discharge (Fig. 3C ). Glutamate led to a 2‐fold amplification of the KCN‐evoked chemo‐afferent volley (Fig. 3D and E ) whilst NMDAR blockade (MK‐801) resulted in further prolongation of the initial transient component and an elimination of the low‐amplitude tail (Fig. 3C ). This resulted in a 2.6‐fold increase of the overall KCN‐evoked response (Fig. 3D ; P = 0.007) and demonstrated that glutamate initiates both a transient excitatory with subsequent inhibitory signal within the carotid body's response to KCN. In the presence of a hypoxic gas mixture (2 min perfusion; 5% CO₂ and 10% O₂ in N₂; Fig. 3F ), we observed the absence of the biphasic response reported with a KCN bolus injection. Instead, there was a gradual increase in carotid body activity during hypoxia, which declined upon the onset of hyperoxia (Fig. 3G ). Glutamate enhanced this hypoxic response, similar to the effects reported with glutamate and KCN (Fig. 3C ). In parallel to Fig. 3C , the hypoxia‐evoked response was further amplified 6‐fold following glutamate with MK‐801 administration (Fig. 3H ; P = 0.0051).

Figure 3. Direct and indirect glutamate effect on the carotid body response to KCN.

A, diagram depicting the isolated arterially perfused carotid body preparation used to record carotid body (CB) afferent activity. Drugs were administered via the common carotid artery (CCA) to evoke a chemoafferent response. B, experimental protocol for testing the CB chemoafferent sensitivity in response to glutamate (Glu; 15 mM), glycine (15 mM; included as a co‐agonist to facilitate NMDA receptor activation), MK‐801 (100 µM) and potassium cyanide (KCN; 0.08%) to mimic hypoxia. C, representative trace of integrated carotid sinus nerve (∫CSN) activity following a control KCN bolus injection, and then in the presence of glutamate, and glutamate with MK‐801. D and E, area under the curve (AUC; D) and maximum peak (E) following a KCN bolus with glutamate and MK‐801 administration. F, experimental protocol for testing the CB chemoafferent sensitivity in response to glutamate (Glu; 15 mM), glycine (15 mM) and MK‐801 (100 µM) and a hypoxic gas mixture (10% O₂ in N₂; $P_{{\mathrm{O}}_2}$ > 80 Torr). G, representative traces of integrated carotid sinus nerve (∫CSN) activity following a control (CTRL) hypoxia exposure with glutamate, and in the presence of glutamate with MK‐801. H and I, AUC (H) and maximum peak (I) following a hypoxia exposure in the presence of glutamate and MK‐801 administration. D (n = 10, n = 5 CBs), E (n = 10, n = 5 CBs), H–I (n = 3 CBs). Values are expressed as percentage increase compared to the CB discharge prior to chemoafferent stimulation. Mean ± SD. Analysed by paired Student's t‐test and one‐way ANOVA with Tukey post hoc test. Abbreviations: ECA, external carotid artery; ICA, internal carotid artery; MK‐801, dizocilpine; NMDA, N‐methyl‐d‐aspartic acid.

We hypothesized that the competing excitatory and inhibitory signals initiated by glutamate transmission originated from a heterogeneous population of carotid body chemosensory (glomus) cells expressing AMPAR and NMDAR. To this end, we validated the predominance of Gria1 over Gria2 (Glutamate Ionotropic Receptor AMPA Type Subunits 1 and 2) and Grin2c over Grin2b (Glutamate Ionotropic Receptor NMDA Type Subunit 2C and 2B) subtypes in the Wistar carotid body by ddPCR (Fig. 4A and B ). Using live cell confocal imaging, we found glutamate evoked [Ca²⁺]i transients in dissociated glomus cells (Fig. 4C ) that were blocked by MK‐801 – an NMDAR antagonist (Fig. 4D and E ), indicating this response to be predominantly NMDAR‐mediated. Approximately 65% of the glomus cells displayed an intracellular Ca²⁺ response to glutamate. In those cells, MK‐801 effectively suppressed the response, reducing the fluorescence intensity of the calcium signal from 3.6 ± 2.8 to 0.5 ± 0.7 (Fig. 4E ). We found immunolabelled GRIN2C was predominantly localized to the cell membrane (Fig. 4G ) (Zhu & Gouaux, 2017), whereas, interestingly, translated GRIA1 primarily displayed nuclear localization in glomus cells (Fig. 4F ), possibly indicating active endosomal recycling as reported previously (Shi et al., 2001).

Figure 4. NMDA (N‐methyl‐d‐aspartic acid) and AMPA (α‐amino‐3‐hydroxy‐5‐methyl‐4‐isoxazolepropionic acid) receptors are expressed in carotid body glomus cells.

A and B, expression of NMDA receptor subunits 2b and 2c (Grin2b, Grin2c) and AMPA receptor subunits 1 and 2 (Gria1 and Gria2) in left and right carotid bodies from 16‐week‐old Wistar rats (n = 31). ddPCR data are presented as normalized copies/mL (mean ± SD). C, confocal images of glomus cells illustrating calcium dynamics at three stages: baseline, during glutamate exposure (Glu; 15 mM) and during the recovery phase. Peanut agglutinin is used as a glomus cell marker. Scale bar = 10 µm. D, [Ca]i response of isolated glomus cells loaded with Fluo‐4 (ΔF/F ₀) during glutamate exposure (blue line) and glutamate with MK‐801 exposure (pink line). E, the normalized fluorescence intensity (ΔF/F ₀) for isolated glomus cells exposed to glutamate and glutamate with MK‐801. Data represent the mean ± SD, analysed by paired Student's t‐test (n = 8 CBs). In Fi, Gria1 (magenta) does not co‐localize with type II cells (green, GFAP) in the carotid body. Magnified views are shown in Fii and Fiii. Arrowheads indicate type II cells, asterisks mark cell nuclei and thin arrows point to Gria1. Gi, Grin2C (magenta) co‐localizes with chemosensory glomus cells (green, TH). Magnified views are shown in Gii and Giii. Arrowheads indicate chemosensory type I cells, asterisks marking cell nuclei and thin arrows pointing to Grin2C. F and G, all images are representative of Wistar rats (n = 4 CBs). Abbreviations: DAPI, 4′,6‐diamidino‐2‐phenylindole; GFAP, glial fibrillary acidic protein; LEL, Lycopersicon esculentum lectin; MK‐801, dizocilpine; TH, tyrosine hydroxylase.

To better understand ionotropic glutamate transmission at the integrated carotid body level, we investigated both the non‐selective and selective agonism of NMDAR and AMPAR using the ex vivo carotid body–CSN preparation (Fig. 5A ). Arterial administration of glutamate alone evoked a transient burst of CSN chemoafferent activity (Fig. 5A ) that was abolished by non‐selective P2 purinergic blockade (PPADS) (Fig. 5C and D ). CSN activity evoked by arterially administered NMDA (synthetic agonist) and AMPA (synthetic agonist) evoked a response that when summed was comparable to that evoked by glutamate (Fig. 5G and I ). MK‐801 partially reduced the glutamate‐evoked excitation of basal discharge by ∼70% (Fig. 5H and J ; P = 0.019) whereas this activity was completely blocked by a combination of MK‐801 and NBQX (selective NMDA and AMPA receptor antagonists, respectively) (Fig. 5J ; P = 0.007). Collectively, these data suggest that glutamate's effect in the carotid body is mediated primarily via NMDA and AMPA receptors in a 70:30 ratio, respectively, where glutamate stimulates CSN activity via secondary release of vesicular ATP.

Figure 5. The effect of NMDA (N‐methyl‐d‐aspartic acid) and AMPA (α‐amino‐3‐hydroxy‐5‐methyl‐4‐isoxazolepropionic acid) versus glutamate‐evoked carotid body activation.

A and B, the experimental protocol (A) tested carotid body (CB) chemoafferent sensitivity in response to glutamate (15 mM), glycine (15 mM) and PPADS (100 µM), with representative traces (B) of integrated and raw carotid sinus nerve (∫CSN) activity to glutamate and glutamate with PPADS. C and D, the area under the curve (AUC) and maximum peak following administration of glutamate and PPADS, analysed by paired Student's t‐test. E–H, experimental protocols (E, F) tested CB chemoafferent sensitivity in response to glutamate (15 mM) with glycine (15 mM), NMDA (N; 15 mM), AMPA (A; 15 mM), MK‐801 (100 µM) and NBQX (100 µM), with representative traces (G, H) of ∫CSN activity. I and J, the AUC following administration of glutamate, NMDA, AMPA, MK‐801 and NBQX, analysed by one‐way ANOVA with Tukey post hoc test. Values in B, G and H are based on n = 5 CBs, and values in C, D, I and J are expressed as percentage increase compared to the pre‐stimulus CB discharge (n = 5 CBs). Mean ± SD. Abbreviations: BL, baseline; MK‐801, dizocilpine; and NBQX, 2,3‐dioxo‐6‐nitro‐7‐sulfamoyl‐benzo[f]quinoxaline.

To better understand the inhibitory component of CSN response to hypoxia, we next focused on GABA and dopamine receptors for their established role in inhibitory neurotransmission in the carotid body (Iturriaga et al., 2009). We hypothesized that carotid body releases inhibitory transmitters to restrain and regulate its sensitivity. Where the loss of this ‘brake’ would potentiate the observed secondary component of the biphasic response to KCN to the point it would disappear (i.e. both phases overlapping in one big response) is shown in Fig. 6A . Consistent with a role for GABAergic signalling in the carotid body (Igarashi et al., 2009), GABRA3 expression was detected in the carotid body but was not uniformly distributed across glomus cells. Instead, intense punctate staining was observed in small regions surrounding certain glomus cell nuclei (Fig. 6C ). The administration of glutamate with bicuculline (GABA_A receptor antagonist; 500 µM) potentiated the area under the curve (AUC) of the KCN‐evoked response by 6‐fold from the control response (Fig. 6J ; P < 0.001) and increased the maximum peak by 5‐fold (Fig. 6K ; P = 0.001). The effects of bicuculline mostly eliminated the low‐amplitude tail of the biphasic response (Fig. 6G ). We found that sulpiride, a dopamine D2 receptor antagonist (1 mm), resulted in a 1.5‐fold amplification of the initial component of the carotid body response to KCN without affecting the secondary component (Fig. 6G ). The AUC exhibited a 1.5‐fold increase from baseline (Fig. 6H ; P = 0.003), and the maximum peak surged 4‐fold from baseline (Fig. 6I ; P = 0.015). However, we did not find any modulation of the low‐amplitude tail.

Figure 6. The accelerator and brake of arterial chemosensitivity.

A, expression of GABA_A receptor (GABA‐R) subunits Gabra2–5 from ddPCR validation using 16‐week‐old Wistar rat CBs. Data presented as the normalized number of copies per µL. Mean ± SD (n = 31 CBs). B, expression of dopamine receptor (DA‐R) subunits Drd1–2 from RNA sequencing data (Pauza et al., 2022) using 14–16‐week‐old Wistar Kyoto rat CB. Data presented as read counts generated. Mean ± SD (n = 12 CBs). C, GABRA3 (magenta) was co‐localized with glomus cell marker, TH (tyrosine hydroxylase; green) and cell nuclei marker, DAPI staining (blue). Ci and Cii, panels highlighting glomus cell clusters containing GABRA3‐immunoreactivity. No co‐localization with LEL (blood vessel marker) was observed. D, no primary antibody control for GABRA3, showing only TH and DAPI in the carotid body section. Images are representative of n = 4 Wistar rats. E–G, experimental protocols (E, F) tested CB chemoafferent sensitivity in response to glutamate (15 mM), glycine (15 mM), sulpiride (1 mM) and bicuculline (bicuc; 500 µM), with representative traces (G) of integrated carotid sinus nerve (∫CSN) activity following a control potassium cyanide (KCN; 0.08%) bolus injection, and KCN bolus injection in the presence of glutamate, glutamate with sulpiride, and glutamate with bicuculline (n = 11 CBs). H and I, area under the curve (AUC; H) and maximum peak (I) following a KCN bolus with glutamate and sulpiride administration. J and K, AUC (J) and maximum peak (K) following a KCN bolus in the presence of glutamate and bicuculline administration. In H, I (n = 6 CBs) and J, K (n = 5 CBs), values are expressed as mean ± SD percentage increase compared to the CB discharge prior to chemoafferent stimulation. One‐way ANOVA, Tukey post hoc test. Abbreviations: BL, baseline; DAPI, 4,6‐diamidino‐2‐phenylindole; GABA, gamma aminobutyric acid; LEL, Lycopersicon esculentum (tomato) lectin.

Similarly, co‐application of CGP, a GABA_B receptor antagonist, with glutamate produced a 1.6‐fold increase in the AUC of the KCN‐evoked response compared to the control response (Fig. 7C ; P = 0.014), and elevated the maximum peak by 1.4‐fold (Fig. 7D ; P = 0.026). However, this treatment had no effect on the low‐amplitude tail (Fig. 7B ), unlike the response observed with bicuculline (Fig. 6G ).

Figure 7. Modulation of the carotid body's biphasic response to potassium cyanide (KCN) by dopamine D2 and gamma‐aminobutyric acid (GABA) receptors in vitro .

A, experimental protocol for testing the carotid body chemoafferent sensitivity in response to glutamate (Glu; 15 mM) with glycine (15 mM), CGP (50 µM) and KCN (0.08% w/v, bolus). B, representative traces of reflex‐evoked integrated carotid sinus nerve (∫CSN) activity following injection of a control KCN, in the presence of glutamate, and glutamate with CGP. C and D, area under the curve (AUC; C) and maximum peak responses (D). C, D (n = 5 CBs). E, experimental protocol for testing the KCN‐evoked carotid body chemoafferent responses to sulpiride (sulp; 1 mM), bicuculline (bic; 500 µM) and sulpiride/bicuculline together. F, representative traces of ∫CSN activity evoked by KCN, and in the presence of sulpiride, bicuculline, and bicuculline with sulpiride. G and H, AUC (G) and maximum peak response (H). G, H (n = 5 CBs). C, D, G and H values are expressed as percentage increase compared to the CB discharge prior to chemoafferent stimulation. Statistical test: one‐way ANOVA with Tukey post hoc test. Mean ± SD. BL, baseline.

To investigate whether two inhibitory transmitters could fully eliminate the latent low‐amplitude tail of the late phase, we co‐administered the GABA_A receptor antagonist (bicuculline) with the dopamine D2 receptor antagonist (sulpiride). This co‐administration amplified the KCN‐evoked response compared to the control. However, we did not observe further elimination of the low‐amplitude tail beyond what was achieved with bicuculline alone (Fig. 6G ). The AUC increased from 41.8 ± 8.4% under control conditions to 112 ± 9.1% with the combined GABA_A and dopamine D2 receptor blockers (Fig. 7G ; P = 0.0009). Similarly, the maximum peak response rose from 176 ± 12.7 to 322 ± 11.9% (Fig. 7H ; P = 0.0007).

Following the discovery of the role of NMDAR in arterial chemoafferent sensitivity to hypoxia, we tested how glutamate and NMDAR blockade affects basal cardiorespiratory parameters. For this, we recorded phrenic nerve (PN), heart rate (HR) and thoracic sympathetic nerve activity (tSNA) using a modified WHBP in situ (Fig. 1A ) (Paton, 1996). We devised a double‐perfused system (dpWHBP) where the common carotid artery was perfused in isolation from the systemic circulation (Figs 1A and 8A ). This permitted focally applied glutamate to the carotid body chemoreceptors avoiding drugs entering the cerebral circulation while recording of chemoreflex‐evoked motor responses. We found that local arterial infusions of glutamate to the carotid body increased basal PN activity (Fig. 8C ). Glutamate administration increased respiratory frequency by 1.4‐fold (Fig. 8D ; P = 0.049) and amplitude by 1.1‐fold (Fig. 8F ; P = 0.014), with consequent 1‐fold increase in central inspiratory drive (i.e. PN AUC/T _i) (Fig. 8E ; P = 0.009). In stark contrast, glutamate administration had no significant effect on tSNA (Fig. 8G ; P > 0.05), HR (Fig. 8H ; P > 0.05) or PP (Fig. 8I ; P > 0.05). Blockade of NMDARs using MK‐801 abolished the excitatory effect of glutamate on basal PN activity (Fig. 8C–F ), which was concordant with CSN chemoafferent activity data (Fig. 5H–J ).

Figure 8. Glutamate receptors in the carotid body modulate basal and chemoreflex‐evoked phrenic nerve activity selectively.

A, diagram of the in situ double perfused working heart–brainstem preparation (dpWHBP). Green arrows indicate the isolated carotid body (CB) circulation, while black arrows represent the systemic perfusion circuit. Recordings were made ipsilateral to the CB being modulated. B, experimental protocol for testing the CB chemoafferent discharge; 5 min intervals are indicated. C, representative trace showing the chemoreflex‐evoked integrated phrenic nerve (∫PN), integrated thoracic sympathetic chain activity (∫SNA) and heart rate (HR) response to glutamate and co‐administration of glutamate and MK‐801 in a Wistar rat (n = 7). D–I, data showing the effect of glutamate (Glu) and glycine with co‐administration of MK‐801 (n = 7 animals) (D) on respiratory frequency (f _R), i.e. derived from PN, (E) on respiratory drive (AUC/T _i), (F) on phrenic nerve amplitude (AMP), (G) on thoracic sympathetic chain (tSNA), (H) on HR and (I) on perfusion pressure (PP). Data were analysed with a paired Student's t‐test. J, representative traces showing the chemoreflex‐evoked ∫PN, ∫SNA and HR response to a control KCN bolus injection, then KCN bolus injection in the presence of glutamate, and glutamate with MK‐801. K–P, data showing the effect of KCN‐evoked response to glutamate and glutamate with co‐administration of MK‐801 (n = 7 animals) (K) on AUC/T _i, (L) on f _R, (M) on ∫PN AMP, (N) on tSNA, (O) on HR and (P) on PP. Data analysed with repeated measures ANOVA with Bonferroni post hoc tests. Abbreviations: BL, baseline; ECA, external carotid artery; ECG, electrocardiogram; ICA, internal carotid artery; KCN, potassium cyanide; MK‐801, dizocilpine.

Next, we tested how glutamate would affect the sensitivity of the peripheral chemoreflex motor responses. Glutamate sensitized a component of the KCN‐evoked response as seen by a significant increase in the reflex central inspiratory drive (Fig. 8K ; P = 0.035). Blocking the NMDARs with MK‐801 (in the presence of glutamate) further amplified the chemoreflex central inspiratory drive by 1.3‐fold (Fig. 8K ; P = 0.002), an effect of magnitude similar to the in vitro data of CSN recordings (Fig. 3C ). The increase in central inspiratory drive can be explained by increases in both respiratory frequency (Fig. 8L ; P = 0.035) and amplitude (Fig. 8M ; P = 0.008). Interestingly, neither glutamate nor MK‐801 administration affected the chemoreflex‐evoked sympathoexcitation response (Fig. 8N ; P > 0.05), bradycardic response (Fig. 9O ; P > 0.05) or PP response (Fig. 8P ; P > 0.05). These data show that glutamate sensitizes preferentially the carotid body respiratory reflex with little or no effect on sympathoexcitation or HR responses.

Figure 9. Effects of excitatory amino acids on carotid body activity.

Five protocols tested carotid body (CB) chemoafferent discharge: A, four 2 min hypoxia (H) exposures; B, five 2 min glutamate (G; 15 mm) exposures. C and D, representative raw carotid sinus nerve (CSN) discharge at baseline (BL) and 60 min after the (C) fourth hypoxia or (D) fifth glutamate application. E and F, mean firing frequency for A and B. G and I, protocols testing glutamate with d‐AP5 (500 µM), glutamate with MK‐801 (M; 10 µM) at the end, or hypoxia with MK‐801. J and L, mean firing frequency for G and I. M, representative KCN‐evoked responses (control vs. 60 min). N–R, data showing area under the curve (AUC) for response to KCN bolus, respective of each protocol. Data: mean ± SD. Statistics: one‐way repeated‐measures ANOVA with Bonferroni post hoc test (E, F, J, K, L); paired Student's t‐test (N–R). C, E, N, O, Q (n = 5 CBs); D, F, K (n = 7 CBs); J, L, P, R (n = 3 CBs). In A, B, G, H and I, intervals are indicated between stimuli, coloured dashed lines indicate the time for data analysis and thick black arrows indicate KCN bolus. Abbreviations: MK‐801, dizocilpine.

After showing that the carotid body is responsive to glutamate, where it acts to modulate the ventilatory chemoreflex, we hypothesized a role in carotid body sensitization following repeated bouts of hypoxia (Peng et al., 2003; Sforza & Roche, 2016). Here, we compared the effects on CSN activity between multiple exposures to hypoxia and repeated glutamate stimulation such as would occur in clinical conditions such as sleep apnoea. The number of hypoxic/glutamate exposures was based on previously established protocols used to induce long‐term facilitation (LTF) in phrenic nerve recordings (Hoffman & Mitchell, 2013; Peng et al., 2011). Although KCN was used to induce histotoxic hypoxia in previous experiments, here we instead used Ringer solution equilibrated with a hypoxic gas mixture (i.e. 10% O₂, 5% CO₂ equilibrated in N₂).

Repeated hypoxia (four 2 min exposures at 5 min intervals; Fig. 9A ) increased the tonic basal discharge of the carotid body from 6.6 ± 0.9 to 11.3 ± 2.3 spikes s⁻¹ 60 min after final exposure (Fig. 9E ; P = 0.041), which we identify as LTF. In parallel, the increase in carotid body's tonic discharge, in response to repeated glutamate exposure (five 2 min applications of 15 mm glutamate at 5 min intervals; Fig. 9B ), was proportional to that of hypoxia (Fig. 9C and D ). Following glutamate applications, basal discharge progressively increased over the subsequent 60 min (Fig. 9F ; P = 0.014). The co‐administration of NMDA receptor antagonists, d‐2‐amino‐5‐phosphonopentanoate (d‐AP5) and MK‐801, effectively blocked both glutamate‐ and hypoxia‐induced LTF (Fig. 9J, L ; P > 0.05). This finding confirms that the observed effects are mediated via NMDA receptors. A lower dosage of NMDA antagonists was used in this experiment compared to previous CSN studies, given that MK‐801 has shown to potentiate the KCN response (Fig. 3C ). This adjustment allowed us to test whether a lower dose could effectively block the LTF response, as suggested by existing literature (McGuire et al., 2005). Administration of the low dose MK‐801 once LTF was established had no effect on CSN discharge frequency (Fig. 9K ; P < 0.05) suggesting NMDAR to be necessary for the development but not maintenance of LTF.

Next, we explored chemoafferent sensitization to glutamate and hypoxia LTF induction. We found carotid body entrainment by either repeated exposures to glutamate or hypoxia augmented carotid body sensitivity to KCN (Fig. 9N and  O ; P = 0.046 and P = 0.015, respectively) which could not be attenuated by MK‐801 (Fig. 9Q ; P > 0.05). This effect was linked to the LTF status and was absent in the presence of NMDAR blockade (Fig. 9P, R ; P > 0.05). Vehicle control experiments revealed no change in carotid body basal discharge over 90 min (Fig. 11H ; P > 0.05).

Figure 11. Effects of repeated GABA exposure on carotid body afferent activity.

A, three 1 min GABA (GB; 15 mm) exposures were delivered at 5 min intervals. B, mean carotid sinus nerve (CSN) firing frequency was measured at baseline (BL) and 10, 30 and 60 min after the final GABA exposure (n = 10 CBs; one‐way repeated measures ANOVA with Bonferroni post hoc test). C and D, representative traces showing basal CSN discharge at BL and 60 min after GABA (C), and carotid body (CB) responses to potassium cyanide (KCN; 0.08%) at BL and 60 min after GABA (D). E, area under the curve (AUC) for KCN‐evoked activity (n = 10 CBs; paired Student's t‐test). F–H, time‐matched vehicle control protocol (F), with representative CSN traces at BL and 60 min after vehicle (G) and corresponding AUC data (H; n = 5 CBs; paired Student's t‐test). AUC values in E and H are expressed as percentage increase relative to baseline CB discharge prior to stimulation. The vehicle was the circulating perfusate, consistent with previous experiments in which drugs were mixed directly into the perfusate. Data are presented as mean ± SD. Abbreviation: GABA, gamma‐aminobutyric acid.

Interestingly, alternative protocols informed by previous studies (Bach & Mitchell, 1996; Baker & Mitchell, 2000; Fuller et al., 2000; Hoffman & Mitchell, 2013; Ling et al., 1997; McGuire et al., 2005; Peng et al., 2011), with 2 or 3 min recovery periods between each glutamate exposure, did not sustain LTF for the 60 min duration in the carotid body (Fig. 10; P > 0.05). These data suggested that LTF of CSN discharge requires a specific carotid body entrainment protocol.

Figure 10. Carotid body afferent activity following different glutamate stimulation frequencies.

Graph showing the mean firing frequency of the carotid sinus nerve (CSN) basal discharge response following repetitive exposures to glutamate at 2 min spacing (pink line), 5 min spacing (orange line) and 3 min spacing (blue line) protocols. Data presented as raw CSN activity at baseline (BL), and 10, 30 and 60 min following the fifth glutamate administration. One‐way repeated measures ANOVA with Bonferroni post hoc test. Pink and blue groups n = 3 CBs, orange group n = 7 CBs. Thick line represents the mean values. Note that long term facilitation (LTF) of carotid body afferent activity was sensitive to the interval time between repeated glutamate exposures with 5 min intervals being effective only.

In contrast to glutamate, repeated exposures of 15 mm GABA (three times for 1 min at 5 min intervals; Fig. 11A ) resulted in a progressive depression of carotid body basal activity 60 min following stimulation (Fig. 11B ; P = 0.0002). This progressive depression of CSN basal discharge (Fig. 11C ) suggested long‐term depression (LTD) of carotid body activity. Importantly, again, no change in firing rate of CSN discharge was observed in the time‐matched control experiments (Fig. 11H ). Repeated exposures to GABA suppressed the carotid body chemosensitivity evident 60 min following GABA administration (Fig. 11D and E ; P = 0.015). Collectively, these findings highlight glutamate and GABA as pivotal and opposing contributors regulating carotid body sensitivity and tonic discharge.

Having established that the carotid body exhibits LTF through NMDAR and that glutamate potentiates the chemoreflex‐evoked tachypnoea in the dpWHBP, we next investigated whether the mechanism underlying our glutamate‐induced LTF protocol regulates the set point of peripheral chemoreceptor reflex ventilatory sensitivity. Using four 2 min glutamate applications with 5 min between (Fig. 12A ), we observed a significant and progressive increase in basal PN activity, increasing from 17 ± 5 to 24 ± 6 bursts per minute (Fig. 12B and C ; P = 0.004). Repeated glutamate applications also sensitized a component of the chemoreflex response, as evidenced by a significant increase in the PN AUC following KCN exposure (Fig. 12D ). The PN AUC increased from delta response of 15.7 ± 7 to 33.2 ± 17% 10 min after the final glutamate application (Fig. 12E ; P = 0.031). Each successive glutamate exposure gradually elevated PN burst frequency (Fig. 12G and H ; P = 0.049), and the PN AUC increased from 5.1 ± 2% during the first application to 28.9 ± 4% by the final exposure (Fig. 12G, I ; P = 0.001). These findings indicate that repeated glutamate exposures selectively enhance ventilatory sensitivity, consistent with a form of respiratory LTF mediated at the level of the peripheral chemoreceptors.

Figure 12. Glutamate‐induced LTF in the carotid body modulates the sensitivity of chemoreflex ventilatory control.

A, experimental protocol for testing the double‐perfused working heart–brainstem preparation (dpWHBP) motor reflex responses: four 2 min glutamate (G; 15 mm) exposures; 5 min intervals are indicated. Coloured dotted lines represent the time points used for analysis of basal activity. B, representative trace showing the basal activity of the integrated phrenic nerve (∫PN) at baseline (BL) and 10 and 30 min. C, data showing the effect of repeated glutamate on basal activity, on respiratory frequency (f _R), i.e. derived from PN, expressed as raw values (n = 5 animals; paired Student's t‐test). D, representative trace showing potassium cyanide (KCN; 0.08%) evoked chemoreflex‐responses in ∫PN activity. E, area under the curve (AUC) of PN response to KCN (n = 5 animals; paired Student's t‐test). F, experimental protocol. Coloured boxes indicate the glutamate responses for data analysis. G, representative traces showing the motor reflex‐evoked phrenic nerve (PN) discharge rate (dr; breaths per minute), and integrated (∫) PN activity. H and I, data showing (H) f _R and (I) ∫PN AUC following glutamate, expressed as percentage increase relative to BL (n = 5 animals; one‐way repeated measures ANOVA with Bonferroni post hoc test). Data: mean ± SD. Abbreviations: ECG, electrocardiogram; bpm, breaths per minute; and bpm, beats per minute.

In contrast, repeated glutamate exposures had minimal impact on sympathetic or cardiac components of the chemoreflex. Baseline thoracic sympathetic nerve activity (tSNA) did not show significant changes across animals, with some showing an upward trend and others either no change or a downward trend (Fig. 13C ; P > 0.05). HR also remained stable, with no significant difference before and after glutamate exposure (Fig. 13D ; P > 0.05). Additionally, glutamate application did not affect the chemoreflex‐evoked sympathoexcitation (Fig. 13G ; P > 0.05) or the bradycardic response (Fig. 13H ; P > 0.05). Across repeated applications, tSNA showed no consistent changes (Fig. 13J and K ; P > 0.05), and HR responses were highly variable without a consistent pattern (Fig. 13L ; P > 0.05). Together, these data suggest that glutamate‐induced LTF selectively enhances the respiratory limb of the chemoreflex while sparing sympathetic and cardiac responses, highlighting a distinct role for glutamate in modulating ventilatory sensitivity.

Figure 13. Sympathetic activity and heart rate responses to repeated glutamate exposures.

A, protocol showing four 2 min glutamate (Glu; 15 mm) exposures at 5 min intervals in the double‐perfused working heart–brainstem preparation (dpWHBP); dashed lines indicate analysis time points. B, representative trace of integrated sympathetic nerve activity (∫SNA) at baseline (BL), and 10 and 30 min after glutamate. C and D, data showing tonic ∫SNA and heart rate (HR) after glutamate (n = 5 animals; paired Student's t‐test). E, chemoreflex protocol. F, representative KCN‐evoked (0.08% bolus) tSNA response at CTRL and 10 min after glutamate. G and H, data showing mean tSNA and HR responses (n = 5 animals; paired Student's t‐test), expressed as percentage change from BL. I, protocol showing four glutamate exposures. J, representative traces of chemoreflex‐evoked ∫SNA across exposures. K and L, data showing mean tSNA and HR responses across exposures (n = 5 animals; repeated measures ANOVA with Bonferroni post hoc test). Data: mean ± SD. Abbreviations: bpm, beats per minute; HR, heart rate; tSNA, thoracic SNA; ∫SNA, integrated sympathetic nerve activity.

Discussion

We describe the integration of glutamatergic and GABAergic signalling within the carotid body and demonstrate their functional role in modulating both chemoafferent sensitivity and the peripheral chemoreflex‐mediated phrenic nerve response. Our findings indicate that hypoxia‐evoked excitation of carotid body cells engages both NMDA and GABA_A receptors. These transmitters appear to signal between distinct cellular populations: hypoxia triggers glutamate release that acts in a paracrine fashion on NMDARs located on neighbouring glomus or type II cells, which in turn release GABA to provide inhibitory feedback on the glutamate‐releasing cells (Fig. 14). This mechanism probably contributes to the biphasic response observed with KCN (Fig. 3C ). Glutamate also stimulates ATP release, activating P2X receptors on petrosal afferents and creating an intrinsic inhibitory feedback loop that counterbalances sensory afferent activity (Fig. 14). Repetitive glutamate exposure increases basal chemoafferent activity and augments the chemoreflex respiratory response evoked by hypoxia, producing LTF that depends on NMDAR activation during induction but not for maintenance. Collectively, these findings support a novel ‘accelerator and brake’ mechanism in the carotid body, where glutamate and GABA fine‐tune chemoafferent set‐point sensitivity and selectively modulate respiratory but not sympathetic outputs.

Figure 14. Proposed signalling mechanism of ATP, glutamate and GABA in the carotid body (CB).

1: reductions in blood concentrations of O₂ (hypoxia) are detected by CB glomus cells. This signal is thought to raise the membrane potential (V _m), resulting in the opening of voltage‐gated calcium (Cav) channels. The increase in intracellular calcium (Ca²⁺) stimulates the exocytotic release of vesicles containing ATP and glutamate from the ‘Accelerator’ cell. 2: synaptic ATP acts on postsynaptic petrosal neuron P2X2/3 receptors, and activation initiates the depolarization of petrosal afferent fibres that transmit signals to the nucleus of the solitary tract via the carotid sinus nerve. VGLUT is also expressed in the petrosal neuron, where increases in intracellular Ca²⁺ could trigger vesicle mobilization and glutamate release from the petrosal neuron endings. 3: Glutamate acts on NMDA receptors on adjacent glomus or type II cells (paracrine) that act as the ‘Brake’ and releases GABA. 4: GABA inhibits the cell from which glutamate was released (autoreceptor). GABA binds to GABAA receptors on adjacent glomus cells that inhibits excitation and (5) quenches the chemoafferent response. Abbreviations: CB, carotid body; ATP, adenosine triphosphate; NMDA, N‐methyl‐d‐aspartate; GABA, gamma‐aminobutyric acid.

Our data reveal a comprehensive glutamatergic signalling system within the carotid body, extending beyond the identification of individual receptors or transporters (Fig. 2A ). Analysis of transcriptomic data from both mouse and rat support this, showing conserved expression of glutamate and GABA receptor genes across species (Fig. 2A ). Consistent with this, human carotid body studies have reported expression of GABA_A receptor subunits (Fagerlund et al., 2010) and AMPA receptor subunits GluA1–3 (Liu et al., 2018), supporting the translational relevance of our findings. This builds on previous localization studies showing NMDAR and AMPAR expression in chemosensory cells (Liu et al., 2009; Liu et al., 2018), although their physiological role in carotid body afferent firing and chemoreflex function was previously unclear. We found membrane localization of GRIN2C in glomus cells, consistent with its role as a key NMDAR subunit involved in synaptic plasticity and intracellular signalling (Paoletti et al., 2013). NMDAR activation, triggered by glycine binding to GluN1 and glutamate binding to GluN2, permits ion influx and activates intracellular pathways that regulate synaptic strength and neuronal excitability (Zhu & Gouaux, 2017). This distribution suggests that GRIN2C contributes to sensory transduction in the carotid body, probably responding to membrane potential changes driven by oxygen levels. Our findings further indicate that these receptors are essential for setting chemoafferent sensitivity and can amplify sensory transduction through LTF.

Our in vitro findings show that glutamate activates ∼65% of glomus cells, increasing intracellular Ca²⁺ primarily via NMDARs, as this response was attenuated after NMDAR blockade (Fig. 4E and F ). Our RNA‐seq data show that Grina, an NMDAR subunit, is the most highly expressed glutamate receptor (Fig. 2A ), consistent with our functional findings that NMDARs predominate over AMPARs in the carotid body. This predominance suggests a specific role for NMDA‐mediated Ca²⁺ signalling in modulating carotid body responses, in agreement with previous reports (Liu et al., 2009; Zhou et al., 2016). The elimination of the glutamate response with PPADS (Fig. 5B ) suggests that glutamate receptors are unlikely to be located on petrosal afferent terminals. Instead, these receptors are probably located elsewhere in the carotid body, such as auto‐receptors on glomus cells or neighbouring cells (Fig. 14). Type II cells are another plausible source of glutamate. In astrocytes, ATP triggers glutamate release via P2X7 and P2Y1 receptors and, in some cases, Panx1 channels (Fellin et al., 2006; Malarkey & Parpura, 2008; Pascual et al., 2012). Similarly, carotid body type II cells release ATP via Panx1 channels following P2Y2 receptor activation (Nurse et al., 2018; Zhang et al., 2012). Given their expression of purinergic receptors, Panx1 channels and glutamate transporters, ATP‐induced glutamate release from type II cells cannot be ruled out.

Blocking NMDARs with MK‐801 amplified the KCN‐evoked carotid body response, producing a single high‐amplitude discharge and eliminating the low‐amplitude tail (Fig. 3C ) – an effect also observed with GABA_A receptor blockade (Fig. 6G ). Glutamate similarly augmented the response to hypoxia, although this lacked the distinct biphasic pattern seen with KCN (Fig. 3G ). Regarding KCN and hypoxia as stimuli; KCN was used as a potent experimental tool to resolve time‐dependent components of the chemosensory response. The differing response profile evoked by hypoxia probably reflects the properties of the stimulus and slower kinetics of activation of the CB. Current models of CB O₂ sensing place mitochondrial metabolism at the centre of stimulus transduction, integrating ion channels (Buckler, 2007, 2015; Ortega‐Sáenz et al., 2003), and metabolic (Chang et al., 2015) and gasotransmitter signalling (Prabhakar et al., 2018; Prabhakhar & Joyner, 2015) to regulate glomus cell excitability. We propose that stimulus kinetics underlies the distinct response patterns observed. The gradual development of hypoxia in our preparation accounts for the slower rise in afferent activity, recruiting excitatory and inhibitory mechanisms in parallel and resulting in overlapping ‘accelerator’ and ‘brake’ influences that manifest as a single sustained response. In contrast, the rapid and defined activation produced by KCN may transiently outpace inhibitory feedback, thereby unmasking a distinguishable secondary inhibitory phase. The absence of a biphasic pattern during moderate hypoxia therefore reflects stimulus dynamics rather than mechanism unique to KCN. Importantly, glutamate receptor blockade during KCN and hypoxia produced comparable modulatory effects, supporting commonality of glutamatergic signalling independent of stimulus type.

These findings indicate that KCN is particularly useful for revealing the balance between excitatory and inhibitory influences on carotid body output. In our data, the ‘brake’ appears to involve NMDAR‐mediated GABA release, restraining afferent firing. Additional experiments using a D2 receptor antagonist showed that dopamine contributes to transient inhibition but does not determine the secondary phase of the KCN response (Fig. 6G ). Together, these results point to a central role for NMDAR–GABA interactions in chemosensory regulation, resembling the NMDAR‐driven GABA release described in the olfactory bulb, which fine‐tunes excitatory output (Halabisky et al., 2000).

Remarkably, glutamate stimulation of NMDARs in the carotid body selectively enhanced both basal PN activity and the PN component of the chemoreflex, while having minimal impact on sympathetic activity or heart rate (Fig. 8). This preferential modulation of respiratory output supports the idea that carotid body pathways can differentially regulate motor outflows, consistent with the ‘ribbon cable’ hypothesis (Zera et al., 2019), which proposes that phenotypically diverse glomus cells respond to specific stimuli by releasing distinct neurotransmitters to activate selective subsets of petrosal afferents. In line with this model, our in vitro cellular data indicate that glutamate activates a subpopulation of glomus cells (Fig. 4C–E ), suggesting that NMDAR expression is functionally restricted and providing a basis for pathway‐selective recruitment. Such cellular diversity is further supported by recent evidence from Spiller et al. (2025), which demonstrated that glomus cell subpopulations possess distinct electrophysiological properties and differential sensitivities to sustained hypoxia (Spiller et al., 2025). We interpret this difference as reflecting heterogeneity among glomus cells, with glutamatergic and GABAergic signalling systems not uniformly expressed. This suggests that a glutamate–NMDAR–ATP pathway may exist only within a subset of glomus cells that interface to a subset of petrosal afferent fibres, potentially explaining the selective respiratory effects observed. Given this selective glutamatergic influence on respiratory outputs, we next asked whether repeated glutamate exposure could drive long‐term adaptations in carotid body function similar to those seen with repeated hypoxia, which sensitizes carotid body activity and can induce LTF of chemoafferent discharge (Peng et al., 2003; Sforza & Roche, 2016).

In our study, repeated glutamate and hypoxia exposures produced sustained increases in chemoafferent activity (Fig. 9), consistent with LTF. Induction of this plasticity required NMDAR activation, whereas its maintenance did not (Fig. 9J–L ), suggesting a trigger role for NMDA‐mediated Ca²⁺ influx. In the dpWHBP preparation, repeated glutamate further elevated basal PN discharge and amplified the PN chemoreflex response (Fig. 12), without altering tonic sympathetic nerve activity, heart rate or cardiovascular chemoreflex components (Fig. 13). This selective respiratory facilitation indicates that glutamate‐dependent plasticity preferentially enhances phrenic motor output via defined carotid body–nucleus tractus solitarii (NTS) pathways (Zera et al., 2019). These findings suggest that peripheral carotid body mechanisms contribute to PN LTF and may participate in the ventilatory LTF observed following intermittent hypoxia in vivo (McGuire et al., 2003).

This form of plasticity parallels LTP in the hippocampal CA1 region, where NMDAR‐dependent Ca²⁺ entry initiates intracellular signalling that strengthens synaptic transmission (Malenka & Bear, 2004; Malinow & Malenka, 2002). In the carotid body, high expression of Gria1 and Gria2 (Fig. 4B ) suggests that AMPAR‐mediated depolarization could facilitate relief of the NMDAR Mg²⁺ block, promoting Ca²⁺‐dependent signalling cascades. Whether canonical LTP‐associated pathways operate in glomus cells remains to be determined, but our data support a similar principle of activity‐dependent, NMDAR‐triggered plasticity.

The effectiveness of spaced stimulation (Fig. 10) further supports this interpretation. Intermittent activation is known to preferentially engage intracellular signalling pathways compared with continuous stimulation, as demonstrated in phrenic LTF models (Baker & Mitchell, 2000) and other neural systems (Tully et al., 1994; Yanow et al., 1998; Yin et al., 1995). In the carotid body, periodic Ca²⁺ transients may similarly drive signalling events required for sustained facilitation, reinforcing the concept that peripheral chemoreceptors possess intrinsic, pattern‐dependent plasticity mechanisms.

Our findings indicate that hypoxia‐induced LTF in the carotid body may contribute to the heightened chemoreflex sensitivity observed in human conditions involving intermittent hypoxia, such as sleep apnoea. Sleep apnoea is characterized by recurrent breathing interruptions during sleep, often associated with exaggerated carotid body responses that lower $P_{\mathrm{aC}{\mathrm{O}}_2}$ and reduce respiratory drive, precipitating apnoeic episodes (Marcus et al., 2014; Narkiewicz et al., 1999b). Similar mechanisms may also contribute to the periodic hyperventilation and apnoea seen in Cheyne–Stokes respiration (Cheyne, 1818), indicating that peripheral chemoreceptor plasticity can profoundly shape central breathing patterns. Moreover, carotid body neurotransmitter profiles change with age, as chemoreceptor cells in aged mammals become hypofunctional, reducing peripheral ventilatory drive (Conde et al., 2006). These age‐related changes in excitatory–inhibitory balance may help explain reduced chemoreflex sensitivity and may have wider consequences for ventilatory control in health and disease.

Our findings should be considered in light of several experimental limitations. As discussed above, KCN differs from physiological hypoxia in potency and kinetics; it was used here as a temporally precise experimental tool. While we examined AMPA, NMDA and GABA receptors using immunohistochemistry, we did not directly quantify protein expression with techniques such as Western blotting, but instead complemented immunohistochemistry with pharmacological blockade to assess functional outcomes. The precise cellular sources of glutamate and GABA in the carotid body remains unresolved. Our data support glutamate release from glomus cells, consistent with VGLUT expression, yet the potential contribution of type II cells remains uncertain. Type II cells, which share astrocytic features, may shape glutamatergic signalling through excitatory amino acid transporter (EAAT)‐mediated uptake or release, as seen in the NTS (Matott et al., 2017). Inhibitory mechanisms also remain incompletely defined. GABA_A receptors contributed to inhibition, but bicuculline alone did not abolish the entire late‐phase KCN response, suggesting involvement of other inhibitory modulators. Blocking dopamine D2 and GABA_B receptors had little effect, raising the possibility of alternative mechanisms such as adrenergic receptor‐mediated inhibition (e.g. pre‐synaptic alpha2 receptors, Almaraz et al., 1997). The limited effect of dopamine was unexpected given its established inhibitory role in the carotid body (Gonzalez et al., 1994; Iturriaga et al., 2009; López‐Barneo et al., 2016), where its release scales with hypoxic intensity and CSN activity in response to hypoxia (Fidone et al., 1982). This may reflect distinct dopamine‐driven pathways that influence hypoxia responses independently of glutamate signalling. A further limitation is that our study relied solely on pharmacological approaches to investigate glutamatergic and GABAergic signalling. Although these methods provide valuable functional evidence, complementary strategies such as genetic knockdown or knockout experiments may help to confirm the roles of endogenous glutamate and GABA in carotid body signalling. Finally, our in situ chemoreflex analysis measured only inspiratory activity from the phrenic nerve. Recording expiratory motor outputs, such as laryngeal or abdominal nerve activity, would clarify how transmitter‐specific mechanisms shape post‐inspiratory and late expiratory activities.

In sum, our discovery of a novel interactive ‘accelerator’ and ‘brake’ mechanism (Fig. 14) indicates fundamental signalling governing the chemoafferent set‐point for regulating breathing. We propose that NMDA and GABA_A receptor signalling allows plasticity in the carotid body to increase and decrease its sensitivity, respectively, preferentially modulating chemoreflex control of breathing. The presence of LTF and pattern‐dependent plasticity suggests that carotid body sensory circuits are capable of adaptive learning‐like processes. Such plasticity may underlie enhanced chemoreflex sensitivity in humans exposed to intermittent hypoxia and could represent a target for therapeutic strategies aimed at stabilizing breathing by modulating peripheral chemoreceptor function.

Competing interests

All authors declare they have no competing interests.

Author contributions

O.M.S.G. performed the in vitro CB‐CSN experiments, the RNA extraction, ddPCR and the IHC experiments. O.M.S.G. performed the analysis, prepared drugs and protocols. I.S.A.F. performed the dpWHBP experiments. X.S. performed the calcium imaging experiments. A.G.P. and J.F.R.P advised on experimental design and wrote the manuscript with O.M.S.G. J.F.R.P provided funding.

Funding

This study was funded by the Health Research Council of New Zealand (programme grant 19/687) and the Sidney Taylor Trust. O.M.S.G. is funded by a University of Auckland Doctoral scholarship; J.F.R.P. is an inaugural Partridge research laureate. A.G.P. is funded by the New Zealand Heart foundation Grant No. 2021 supported by Ernest Hyam Davis & Ted & Mollie Carr Legacies and No. 2026 supported by G.R. Winn Trust. I.F. is also funded by the New Zealand Heart foundation Grant No. 2025 and 2016 supported by Kenneth Ronald Boyce Foundation.

Supporting information

Peer Review History

Biography

Olivia Gold is a Research Fellow at the University of Auckland working with Professor Julian Paton. She recently completed her PhD in the Paton Translational Cardiorespiratory Research Laboratory, where she investigated the fundamental mechanisms underlying carotid body sensitivity. Her current research focuses on modulating carotid body sensitivity to adjust cardiorespiratory control in conditions characterized by suppressed breathing.

Data availability statement

All individual data points supporting the findings of this study are presented within the figures of the manuscript. Raw data and analysis code used for statistical analysis and figure generation are available from the corresponding author upon reasonable request.