Gene expression analyses reveal metabolic specifications in acute O₂‐sensing chemoreceptor cells

Abstract

Key points

• Glomus cells in the carotid body (CB) and chromaffin cells in the adrenal medulla (AM) are essential for reflex cardiorespiratory adaptation to hypoxia. However, the mechanisms whereby these cells detect changes in O₂ tension are poorly understood.

• The metabolic properties of acute O₂‐sensing cells have been investigated by comparing the transcriptomes of CB and AM cells, which are O₂‐sensitive, with superior cervical ganglion neurons, which are practically O₂‐insensitive.

• In O₂‐sensitive cells, we found a characteristic prolyl hydroxylase 3 down‐regulation and hypoxia inducible factor 2α up‐regulation, as well as overexpression of genes coding for three atypical mitochondrial electron transport subunits and pyruvate carboxylase, an enzyme that replenishes tricarboxylic acid cycle intermediates.

• In agreement with this observation, the inhibition of succinate dehydrogenase impairs CB acute O₂ sensing. The responsiveness of peripheral chemoreceptor cells to acute hypoxia depends on a ‘signature metabolic profile’.

Abstract

Acute O₂ sensing is a fundamental property of cells in the peripheral chemoreceptors, e.g. glomus cells in the carotid body (CB) and chromaffin cells in the adrenal medulla (AM), and is necessary for adaptation to hypoxia. These cells contain O₂‐sensitive ion channels, which mediate membrane depolarization and transmitter release upon exposure to hypoxia. However, the mechanisms underlying the detection of changes in O₂ tension by cells are still poorly understood. Recently, we suggested that CB glomus cells have specific metabolic features that favour the accumulation of reduced quinone and the production of mitochondrial NADH and reactive oxygen species during hypoxia. These signals alter membrane ion channel activity. To investigate the metabolic profile characteristic of acute O₂‐sensing cells, we used adult mice to compare the transcriptomes of three cell types derived from common sympathoadrenal progenitors, but exhibiting variable responsiveness to acute hypoxia: CB and AM cells, which are O₂‐sensitive (glomus cells > chromaffin cells), and superior cervical ganglion neurons, which are practically O₂‐insensitive. In the O₂‐sensitive cells, we found a characteristic mRNA expression pattern of prolyl hydroxylase 3/hypoxia inducible factor 2α and up‐regulation of several genes, in particular three atypical mitochondrial electron transport subunits and some ion channels. In addition, we found that pyruvate carboxylase, an enzyme fundamental to tricarboxylic acid cycle anaplerosis, is overexpressed in CB glomus cells. We also observed that the inhibition of succinate dehydrogenase impairs CB acute O₂ sensing. Our data suggest that responsiveness to acute hypoxia depends on a ‘signature metabolic profile’ in chemoreceptor cells.

Key points

• Glomus cells in the carotid body (CB) and chromaffin cells in the adrenal medulla (AM) are essential for reflex cardiorespiratory adaptation to hypoxia. However, the mechanisms whereby these cells detect changes in O₂ tension are poorly understood.

• The metabolic properties of acute O₂‐sensing cells have been investigated by comparing the transcriptomes of CB and AM cells, which are O₂‐sensitive, with superior cervical ganglion neurons, which are practically O₂‐insensitive.

• In O₂‐sensitive cells, we found a characteristic prolyl hydroxylase 3 down‐regulation and hypoxia inducible factor 2α up‐regulation, as well as overexpression of genes coding for three atypical mitochondrial electron transport subunits and pyruvate carboxylase, an enzyme that replenishes tricarboxylic acid cycle intermediates.

• In agreement with this observation, the inhibition of succinate dehydrogenase impairs CB acute O₂ sensing. The responsiveness of peripheral chemoreceptor cells to acute hypoxia depends on a ‘signature metabolic profile’.

Abbreviations

7‐AAD

7‐aminoactinomycin D

Acacb

acetyl‐CoA carboxylase b

Acly

ATP citrate lyase

AM

adrenal medulla

Cacna1d/Cav1.3

calcium channel, voltage‐dependent, L type, alpha 1D subunit

Cacna1h/Cav3.2

calcium channel, voltage‐dependent, T type, alpha 1H subunit

CB

carotid body

Chga

chromogranin A

Cox4i2

cytochrome c oxidase subunit IV isoform 2

Cox8b

cytochrome c oxidase subunit VIIIb

DAPI

4’,6’‐diamidino‐2‐phenylindole

DMEM

Dulbecco's modified Eagle's medium

DMM

dimethyl malonate

ETC

electron transport chain

ETF

electron transport flavin/quinone oxidoreductase

FFA

free fatty acid

Gdf10

growth differentiation factor 10

Gdnf

glial cell line derived neurotrophic factor

GFP

green fluorescent protein

Gls

glutaminase

Hif1β/Arnt

aryl hydrocarbon receptor nuclear translocator

Hif2α/Epas1

hypoxia inducible factor 2α/endothelial PAS domain protein 1

Hif2β/Arnt2

aryl hydrocarbon receptor nuclear translocator 2

HVR

hypoxic ventilatory response

Idh1

isocitrate dehydrogenase 1 (NADP+), soluble

Idh3

isocitrate dehydrogenase 3 (NAD+)

Igfbd3

insulin‐like growth factor binding protein 3

Kcnh5

potassium voltage‐gated channel, subfamily H (eag‐related), member 5

Kcnh7

potassium voltage‐gated channel, subfamily H (eag‐related), member 7

Kcnip3

Kv channel interacting protein 3, calsenilin

Kcnj3

potassium inwardly rectifying channel, subfamily J, member 3

Kcnma1

potassium large conductance calcium‐activated channel, subfamily M, alpha member 1

Kcnmb1

potassium large conductance calcium‐activated channel, subfamily M, beta member 1

Kcnmb2

potassium large conductance calcium‐activated channel, subfamily M, beta member 2

Kcnn2

potassium intermediate/small conductance calcium‐activated channel, subfamily N, member 2

Kcnq3

potassium voltage‐gated channel, subfamily Q, member 3

Kcnq5

potassium voltage‐gated channel, subfamily Q, member 5

Kcnt2

potassium channel, subfamily T, member 2

Kv/Kcn

potassium voltage gated channel

Kv3/Kcnc

potassium voltage gated channel, Shaw‐related subfamily

Kv4/Kcnd

potassium voltage‐gated channel, Shal‐related family

L‐15

Leibowitz medium

Ldh

lactate dehydrogenase

MCI, MCII, MCIII, MCIV

mitochondrial complex I, II, III, IV, respectively

Ndufa4l2

NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 4‐like 2

Ntrk1

neurotrophic tyrosine kinase receptor

Olfr78

olfactory receptor 78

Pcx

pyruvate carboxylase

Pdha1

pyruvate dehydrogenase E1 alpha 1

Pdk4

pyruvate dehydrogenase kinase, isoenzyme 4

pFDR

P values adjusted with the false discovery rate

Phd3/Egln3

prolyl hydroxylase 3/egl‐9 family prolyl hydroxylase 3

Pnmt

phenylethanolamine‐N‐methyltransferase

Ppia

peptidylpropyl isomerase A

QH₂

ubiquinol/reduced ubiquinone

Rgs5

regulator of g‐protein signalling 5

RIN

RNA integrity number

ROS

reactive oxygen species

SCG

superior cervical ganglion

Scn7a

sodium channel, voltage‐gated, type VII, alpha

Scn9a/Nav1.7

sodium channel, voltage‐gated, type IX, alpha

SDHD

succinate dehydrogenase complex, subunit D, integral membrane protein

Slc1a5

solute carrier family 1 (neutral amino acid transporter), member 5

Slc18a1

solute carrier family 18 (vesicular monoamine), member 1

Slc7a5

solute carrier family 7 (cationic amino acid transporter, y+ system), member 5

SST‐RMA

signal space transformation‐robust multiarray average

Task1/Kcnk3

potassium channel, subfamily K, member 3

Task3/Kcnk9

potassium channel, subfamily K, member 9

TCA

tri‐carboxylic acid

TH

tyrosine hydroxylase

Trpc5

transient receptor potential cation channel, subfamily C, member 5

Ucp2

uncoupling protein 2

Vegfa

vascular endothelial growth factor A

Vegfc

vascular endothelial growth factor C

Introduction

Acute oxygen (O₂) sensing is essential for individuals to survive in environmental or pathological conditions that result in low O₂ tension ($P_{{\mathrm{O}}_2}$) in the blood (hypoxaemia). The carotid body (CB), strategically located at the carotid bifurcation, is the main arterial chemoreceptor that responds to hypoxia by triggering fast (in seconds) adaptive cardiorespiratory reflexes (hyperventilation and sympathetic activation) to compensate for the decrease in $P_{{\mathrm{O}}_2}$ (for a recent review see López‐Barneo et al. 2016a). Together with other O₂‐sensitive organs, the CB forms part of the homeostatic acute O₂ sensing system (Weir et al. 2005). It has close developmental and functional links with the adrenal medulla (AM), which is innervated by sympathetic nerves and, although less potently than the CB, also has, particularly in the neonate, intrinsic, non‐neurogenic O₂ sensitivity (Adams et al. 1996; Mochizuki‐Oda et al. 1997; Mojet et al. 1997; Thompson et al. 1997, 2002; García‐Fernández et al. 2007a). Recently, the physiology of the CB–AM axis has attracted medical interest due to the fact that its over‐activation can contribute to the exaggerated sympathetic outflow underlying hypertension and other comorbidities associated with highly prevalent human diseases (McBryde et al. 2013; Ribeiro et al. 2013; Marcus et al. 2014; del Río et al. 2016).

The mechanisms of acute O₂ sensing have been studied in greatest detail in the CB, which is composed of clusters of O₂‐sensitive glomus cells. These neurosecretory, presynaptic‐like elements release transmitters that activate sensory fibres, which impinge upon brainstem neurons involved in the control of respiration and autonomic function. Glomus cells contain a variety of K⁺ channels, which are inhibited during hypoxia to produce depolarization and Ca²⁺‐dependent secretory vesicle exocytosis (see Lopez‐Barneo et al. 2016a). O₂‐regulated K⁺ channels have also been described in AM cells as well as in other cells of the acute O₂‐sensing system (for reviews see Weir et al. 2005; Nurse et al. 2009). However, the precise molecular processes underlying the detection of changes in O₂ by chemoreceptor cells, and the nature of the signals that link O₂ sensing to membrane ion channels have remained unclear and a matter of debate (Peers, 2015; for an updated review see López‐Barneo et al. 2016b). Recently, we have shown that ablation of the mouse Ndufs2 gene, which encodes a component of the ubiquinone biding site in mitochondrial complex I (MCI) (see Baradaran et al. 2013), results in selective abolition of both the hypoxic ventilatory response (HVR) and sensitivity to hypoxia in single glomus and AM chromaffin cells. The data suggest that normoxic peripheral chemoreceptor cells possess special metabolic features, which result in the accumulation of reduced ubiquinone (QH₂). Slow‐down of the mitochondrial electron transport chain (ETC) during hypoxia may further increase the QH₂ pool, thereby enhancing the production of reactive oxygen species (ROS) and reduced pyridine nucleotides in MCI to signal membrane ion channels (Fernández‐Agüera et al. 2015; Gao et al. 2017).

To advance our knowledge of the metabolic specifications characteristic to acute O₂‐sensing cells, we performed a comparative analysis of the gene expression profile in the mouse CB, AM and superior cervical ganglion (SCG). A previous gene expression study, carried out on the CB and AM from mice subjected to either normoxia or sustained hypoxia, focused on ion channels and considered the AM as an O₂‐insensitive tissue, the gene expression profile of which was used for background subtraction (Ganfornina et al. 2005). In another study, the gene expression profile of the CB from two different mouse strains with variable responsiveness to hypoxia was analysed. This work reported differences in genes encoding ion channels or related to neurotransmitter metabolism, synaptic vesicles and the development of neural crest‐derived cells (Balbir et al. 2007). More recently, the human CB transcriptome has been studied with attention to the expression of channels and receptors relevant for anaesthesia and the up‐regulation of CB genes involved in the inflammatory response (Fagerlund et al. 2010; Mkrtchian et al. 2012). While the current investigation was in progress, a transcriptomic analysis, comparing mouse CB with olfactory or vomeronasal sensory neurons using single‐cell RNA sequencing, was reported (Zhou et al. 2016). This study identified abundant G protein‐coupled receptor signalling, various types of ion channels and hypoxia inducible factor 2α (Hif2α) in neonatal [postnatal day 4 (P4)–P5] glomus cells. In addition, two atypical mitochondrial ETC subunits were among the most specifically expressed genes identified in CB cells (Zhou et al. 2016). In the current study, we used adult mice to compare the transcriptomes of three cell types with the same embryological origin (neural crest‐derived sympathoadrenal progenitors) but variable responsiveness to acute hypoxia: CB glomus cells and AM chromaffin cells, which are O₂‐sensitive (glomus cells > chromaffin cells), and SCG neurons, which are O₂‐insensitive. We expected this experimental approach to facilitate the identification of genes relevant to acute O₂ sensing in comparison with genes related to other cellular functions, developmental specifications or age. Our results reveal a characteristic mRNA expression pattern of prolyl hydroxylase 3 (Phd3)/Hif2α in O₂‐sensitive cells and expand previous studies regarding the abundance of atypical mitochondrial ETC subunits in these cells. In addition, we found that several metabolic enzymes, in particular pyruvate carboxylase (Pcx), which is fundamental to tricarboxylic acid (TCA) cycle anaplerosis (see Owen et al. 2002), are differentially expressed in CB chemoreceptor cells compared to the SCG. Finally, we show that pharmacological or genetic inhibition of succinate dehydrogenase impairs CB acute O₂ sensing. Our data support the concept that responsiveness to acute hypoxia depends on a ‘signature metabolic profile’ in peripheral chemoreceptor cells.

Methods

Ethical approval

All procedures were approved by the Institutional Committee of the University of Seville for Animal Care and Use (2012PI/LB02 and 22‐09‐15‐332). Handling of the animals was conducted in accordance with the European Community Council directives 86/609/EEC, and 2010/63/EU for the Care and Use of Laboratory Animals. The experiments comply with the principles of animal research established by the Journal of Physiology (Grundy, 2015).

Animals

TH‐GFP transgenic mice were originally obtained from GENSAT (RRID: MMRRC_000292‐UNC) on a mixed background (Gong et al. 2003) and back‐crossed to C57/B6 background in our laboratory. This line was generated by random insertion of a bacterial artificial chromosome containing regulatory sequences of tyrosine hydroxylase (TH) expression followed by EGFP reporter gene. TH‐SDHD mice, in which the mitochondrial complex II (MCII) subunit D (Sdhd) was deleted in TH+ cells, were generated previously in our laboratory (Díaz‐Castro et al. 2012; Platero‐Luengo et al. 2014). Transgenic and wild‐type (C57/B6) mice were housed at regulated temperature (22 ± 1°C) in a 12 h light/dark cycle with ad libitum access to food and drink. Both male and female mice were used in the current study. Mice were killed via intraperitoneal administration of a lethal dose of sodium thiopental (120–150 mg kg⁻¹) before tissue dissection. Dissected tissues were either fast‐frozen with liquid N₂ and stored at −80 °C for RNA isolation, or processed for immunohistochemical analysis, cell sorting, or functional analyses, as described below.

Microarray analysis

Total RNA was isolated from CB, AM and SCG of wild‐type adult (∼2 months old) mice using RNeasy Micro kit (Qiagen, Valencia, CA, USA). Due to the small tissue size, each CB replicate was pooled from 10 mice, whereas each AM and SCG replicate was pooled from three mice to obtain sufficient RNA. The RNA quality was determined using an Agilent 2100 Bioanalyzer (Agilent, Santa Clara, CA, USA). RNA samples with RNA integrity number (RIN) ≥ 7.8 were further processed for microarray analysis.

RNA was amplified and labelled using the GeneChip WT PLUS Reagent Kit (Affymetrix, Santa Clara, CA, USA). Amplification was performed with 50 ng of total RNA input following procedures described in the WT PLUS Reagent Kit user manual. The amplified cDNA was quantified, fragmented and labelled in preparation for hybridization to GeneChip Mouse Transcriptome 1.0 Array (Affymetrix) using 5.5 μg of single‐stranded cDNA product and following protocols outlined in the user manual. Washing, staining (GeneChip Fluidics Station 450, Affymetrix) and scanning (GeneChip Scanner 3000, Affymetrix) were performed following protocols outlined in the user manual for cartridge arrays. Data were processed for gene‐level background subtraction, normalization and signal summarization (SST‐RMA, signal space transformation‐robust multi‐array average) using an Affymetrix Expression Console. Gene‐level differential expression analysis was then performed using Transcriptome Analysis Console 3.0 (Affymetrix). One‐way between‐subject (unpaired) ANOVA was used and P values adjusted with the false discovery rate (pFDR) were calculated. Gene expression was considered different between groups with pFDR <0.05 and fold change >2 or <−2. In addition, after the SST‐RMA normalization, data were analysed with Bioconductor to determine the similarity of gene expression profiles among samples using hierarchical clustering analysis, to evaluate the difference of gene expression patterns among the samples using principal component analysis, and to visualize differentially expressed genes using a volcano plot (Huber et al. 2015). Microarray raw data are deposited in the NCBI GEO database (accession number: GSE99593).

Flow cytometry

Freshly dissected CB and AM from TH‐GFP mice were quickly placed in ice‐cooled modified Tyrode solution (in mm: 148 NaCl, 2 KCl, 3 MgCl₂, 10 Hepes, 10 glucose, pH 7.4) for enzymatic dispersion. Dispersion of CB glomus cells and AM chromaffin cells was performed following the procedures described by our laboratory (Piruat, et al. 2004; Muñoz‐Cabello et al. 2005; Levitsky and López‐Barneo, 2009). SCGs dissected from TH‐GFP mice were collected in ice‐cooled Leibowitz medium (L‐15), followed by enzymatic cell dispersion as described previously (Alberola‐Die et al. 2013). Dispersed cells from each tissue were incubated with 7‐aminoactinomycin D (7‐AAD, 1:200 dilution, BD Biosciences, Franklin Lakes, NJ, USA) in FACS buffer (L‐15 supplemented with 1% penicillin/streptomycin, 0.2% bovine serum albumin, 10 mm Hepes, 5 mm EDTA) to label the non‐viable cells. Green fluorescent protein positive (GFP+) cells were sorted using a FACSJazz cell sorter (BD Biosciences). GFP+ cells were either collected in PBS for immunocytochemical analysis or in Buffer RLT (Qiagen) and immediately frozen at −80 °C for RNA isolation.

Real‐time quantitative PCR

To validate the results of microarray analysis, total RNA was isolated using an RNeasy Micro kit (Qiagen) from CB, AM, and SCG of wild‐type adult mice, which were different from those used for microarray analysis. Each CB replicate was pooled from 15 mice, whereas each AM and SCG replicate was pooled from four mice. In addition, total RNA was also isolated from GFP+ cells of each tissue from TH‐GFP mice, which were sorted by flow cytometry as mentioned above. The RNA quality was determined using an Agilent 2100 Bioanalyzer (Agilent) and cRNA was amplified using either a GeneChip WT PLUS Reagent Kit (Affymetrix) from whole tissues or GeneChip WT Pico kit (Affymetrix) from GFP+ cells.

In total, 500 ng of cRNA was copied to cDNA using a QuantiTect Reverse Transcription Kit (Qiagen) in a final volume of 20 μl. Taqman mouse Endogenous Control Array (Applied Biosystems, Carlsbad, CA, USA) was used in a Viia7 Real‐Time PCR system (Applied Biosystems) to select a housekeeping gene among three tissues. Real‐time quantitative PCRs were performed in a 7500 Fast Real Time PCR System (Life Technologies). PCRs were performed in duplicate in a total volume of 20 μl containing 1–4 μl of cDNA solution and 1 μl of Taqman probe of the specific gene (ThermoFisher Scientific Inc., Waltham, MA, USA). Peptidylprolyl isomerase A (Ppia) was also estimated in each sample to normalize the amount of cRNA input in order to perform relative quantifications.

Immunohistological analysis

GFP+ cells sorted by flow cytometry were plated on poly‐l‐lysine‐treated coverslips and incubated for 2 h in either Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin, 1% l‐glutamine and 84 mU ml⁻¹ insulin for CB and SCG, or DMEM supplemented with 10% fetal bovine serum, 2 mm l‐glutamine, 1% penicillin/streptomycin for AM. For immunohistochemical studies, mice were perfused first with PBS and then with 4% paraformaldehyde in PBS before scarification and tissue dissection. Carotid bifurcation and adrenal gland were fixed with 4% paraformaldehyde in PBS for 2 h, cryoprotected overnight with 30% sucrose in PBS and embedded in OCT (Tissue‐Tek). Tissue sections of 10 μm were cut with a cryostat (Leica, Wetzlar, Germany). Cells and tissue sections were incubated with primary antibodies overnight at 4°C: Cox4i2 (1:100 dilution, 11463‐1‐1AP, Proteintech, Chicago, IL, USA); Ndufa4l2 (1:50 dilution, 16480‐1‐AP, Proteintech); Pcx (1:100 dilution, ab115579, Abcam, Cambridge, MA, USA); TH (1:5000 dilution, NB300‐109, Novus Biological, Inc., Littleton, CO, USA); or TH (1:100 dilution, AB1542, Millipore, Billerica, MA, USA). This was followed by incubation with fluorescent secondary antibodies: Alexa 568 or Alexa 488 (1:500 dilution, A11057, A11011, A11008, A11015, Invitrogen, Carlsbad, CA, USA). Nuclei were labelled with 4’,6’‐diamidino‐2‐phenylindole (DAPI) staining. Immunofluorescence images were obtained using Nikon A1R+ confocal microscopy (Nikon).

Amperometric recording of catecholamine secretion

Dissected CBs were placed in ice‐cooled modified Tyrode solution and CB slices were prepared as described previously (Ortega‐Sáenz et al. 2010). Briefly, CB slices (150 μm) were sectioned using a vibratome (VT1000, Leica). After a brief enzymatic digestion, the slices were incubated in DMEM supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin, 1% l‐glutamine, 1.2 U ml⁻¹ erythropoietin and 84 mU ml⁻¹ insulin at 37°C in 5% CO₂ for 24 h. In some experiments, 2 mm dimethyl malonate (DMM) was added 12 h before recording. Catecholamine secretion was recorded perfusing CB slices continuously with a recording solution (in mm: 117 NaCl, 4.5 KCl, 23 NaHCO₃, 1 MgCl₂, 2.5 CaCl₂, 5 glucose and 5 sucrose, pH 7.4) using an upright microscope (Axioscope, Zeiss, Oberkochen, Germany). The ‘normoxic’ solution was bubbled with 5% CO₂, 20% O₂ and 75 % N₂ ($P_{{\mathrm{O}}_2}$ = 150 mmHg). The ‘hypoxic’ solution was bubbled with 5% CO₂ and 95% N₂ ($P_{{\mathrm{O}}_2}$ ∼10–20 mmHg). In the high K⁺ solution, NaCl was replaced by KCl equimolarly. DMM (2 mm) was added to the recording solution when necessary. All experiments were carried out at ∼36°C. Secretory events were recorded with a 10 μm polarized carbon fibre electrode. Amperometric currents were recorded with an EPC‐8 patch‐clamp amplifier (HEKA Electronics, Lambrecht/Pfaltz, Germany). Data acquisition and analysis were carried out with an ITC‐16 interface (Instrutech Corporation, Longmont, CO, USA) and PULSE/PULSEFIT software (HEKA Electronics). The secretion rate (fC min^–1) was calculated as the amount of charge transferred to the recording electrode during a given period of time and cumulative secretion was calculated as the integral on time of the amperometric recording.

Cytosolic Ca²⁺ measurements by microfluorimetry

CB glomus cells were dispersed as described previously (Piruat et al. 2004), seeded on glass poly‐l‐lysine‐treated glass coverslips, and kept in DMEM supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin, 1% l‐glutamine and 84 mU ml⁻¹ insulin. To measure changes in cytosolic Ca²⁺ concentration, glomus cells were first incubated with 4 μm 2‐AM Fura (F1225, Thermo Scientific) in DMEM for 30 min at 37°C in a 5% CO₂ incubator. Loaded cells were then recorded using an inverted microscope (Eclipse Ti, Nikon) equipped with epifluorescence and photometry (Fernández‐Agüera et al. 2015). Alternating excitation wavelengths of 340 and 380 nm were used, and background fluorescence was subtracted before calculating the ratio of fluorescence intensity between 340 and 380 nm. Data were digitized at a sampling interval of 1 s. All the experiments were performed at 36°C.

Statistics

Data were presented as mean ± SEM with the number (n) of experiments indicated and analysed using Sigmaplot v12.0. Normality was tested with the Shapiro–Wilk test. When necessary, a log transformation was performed to normalize the data distribution prior to parametric analyses using a t test. A P value <0.05 was considered statistically significant. The statistical analysis of the microarray data is described in the ‘Microarray analysis’ section.

Results

Gene expression profiles in the CB, AM and SCG

Gene expression profiles were compared among cells in the CB, AM, and SCG from adult mice using Affymetrix GeneChip Mouse Transcriptome Assay 1.0 Array, which includes >23000 protein coding genes and >55000 non‐protein coding genes, such as non‐coding RNA, pseudogenes or rRNA. Hierarchical clustering analysis was first performed to analyse the similarity of the gene expression profiles between the samples. As shown in Fig. 1 A, replicates from the same tissue were clustered together and appeared distant from replicates of other tissues. Principal component analysis was performed to determine the difference in gene expression patterns among the samples, revealing that the three tissues were clearly separated with two principal components (Fig. 1 B). These data demonstrate the existence of a global difference in gene expression among CB, AM and SCG cells.

Figure 1. Microarray analysis of transcriptomes of the carotid body (CB), adrenal medulla (AM), and superior cervical ganglion (SCG) from adult mice.

A, hierarchical clustering analysis. Numbers indicate separate tissue samples. B, principal component (PC) analysis. C, volcano plots demonstrating differentially expressed genes (DE, light red, fold change > 2, pFDR < 0.05) compared to genes without differential expression (non‐DE, light blue). Left, CB versus SCG; right, AM versus SCG.

Differential gene expression between the CB and SCG

Gene expression in the CB compared to the SCG was studied based on cutoff values of fold change >2 and pFDR <0.05. Up‐ and down‐regulated genes can be visualized in Fig. 1 C (light red). The genes with the largest changes in expression level (top 20 up‐regulated and top 20 down‐regulated genes) are listed in Table 1. Among the most differentially expressed genes, a number related to G‐protein signalling [e.g. regulator of g‐protein signalling 5 (Rgs5)] were up‐regulated in the CB. Genes related to trophic factors were also up‐regulated in the CB [growth differentiation factor 10 (Gdf10) and insulin‐like growth factor binding protein (Igfbd3)] or the SCG [neurotrophic tyrosine kinase receptor (Ntrk1)]. Notably, several cyto/chemokines and extracellular matrix proteins were overexpressed in the CB, which may be related to the abundance of immune cells in this organ (Mkrtchian et al. 2012) and the organization of various cell types in the CB glomeruli. In this unbiased general analysis, three genes putatively involved in O₂ sensing were among the most highly expressed in the CB: NADH dehydrogenase 1 alpha subcomplex, 4‐like 2 (Ndufa4l2), endothelial PAS domain protein 1 (Hif2α) and the Kcnk9 K⁺ channel (Task3) (see below).

Table 1.

Top 20 up‐ and down‐regulated genes in the CB versus the SCG of adult mice by microarray analysis

Gene symbol	Description	Fold change (linear)	pFDR
Up‐regulated
Gdf10	growth differentiation factor 10	760	1.49E‐03
Ndufa4l2	NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 4‐like 2	721	3.90E‐04
Mgp	matrix Gla protein	485	4.55E‐03
Acta2	actin, alpha 2, smooth muscle, aorta	388	2.20E‐04
Dpt	dermatopontin	164	1.26E‐03
Igfbp3	insulin‐like growth factor binding protein 3	160	2.29E‐04
Rgs5	regulator of G‐protein signalling 5	142	8.02E‐04
Cfh	complement component factor h	119	2.29E‐04
Ccl21a	chemokine (C‐C motif) ligand 21A (serine)	104	1.39E‐03
Epas1(Hif2a)	endothelial PAS domain protein 1	102	2.49E‐03
Cpa3	carboxypeptidase A3, mast cell	100	1.39E‐03
Myh11	myosin, heavy polypeptide 11, smooth muscle	98	6.71E‐04
Itga8	integrin alpha 8	96	2.74E‐03
Cytl1	cytokine‐like 1	85	4.35E‐03
Kcnk9(Task3)	potassium channel, subfamily K, member 9	83	2.19E‐03
Ccl21a/Ccl21b/Ccl21c	chemokine (C‐C motif) ligand, 21A (serine), 21B (leucine), 21C (leucine)	81	1.49E‐03
Slc9a2	solute carrier family 9 (sodium/hydrogen exchanger), member 2	71	8.22E‐04
Adm	adrenomedullin	64	1.99E‐03
Myl9	myosin, light polypeptide 9, regulatory	62	2.23E‐03
Mcpt4	mast cell protease 4	62	3.36E‐03
Down‐regulated
Cxcr4	chemokine (C‐X‐C motif) receptor 4	−41	7.12E‐03
Napb	N‐ethylmaleimide sensitive fusion protein attachment protein beta	−39	2.34E‐02
Tubb4a	tubulin, beta 4A class IVA	−38	4.66E‐02
Mapk11	mitogen‐activated protein kinase 11	−36	1.29E‐02
Tmem179	transmembrane protein 179	−35	1.06E‐02
Htr3a	5‐hydroxytryptamine (serotonin) receptor 3A	−35	2.19E‐02
Ntrk1	neurotrophic tyrosine kinase, receptor, type 1	−30	3.42E‐02
Nrip3	nuclear receptor interacting protein 3	−29	4.13E‐03
Sult4a1	sulfotransferase family 4A, member 1	−29	1.03E‐02
Htr3b	5‐hydroxytryptamine (serotonin) receptor 3B	−28	1.83E‐02
Nefl	neurofilament, light polypeptide	−27	3.10E‐02
Vwc2l	von Willebrand factor C domain‐containing protein 2‐like	−27	3.00E‐03
Maob	monoamine oxidase B	−27	1.71E‐02
Gpr158	G protein‐coupled receptor 158	−25	1.64E‐02
Ehd3	EH‐domain containing 3	−25	3.06E‐02
Ttll1	tubulin tyrosine ligase‐like 1	−24	2.16E‐02
Cyp2j12	cytochrome P450, family 2, subfamily j, polypeptide 12	−24	1.01E‐02
Stk32a	serine/threonine kinase 32A	−23	9.87E‐03
Slc6a15	solute carrier family 6 (neurotransmitter transporter), member 15	−22	7.08E‐03
Maoa	monoamine oxidase A	−21	1.92E‐02

pFDR, P value adjusted with the false discovery rate.

We focused our analysis on genes implicated in the O₂‐sensing pathway or related to cellular functions that could be relevant to acute responsiveness to hypoxia. In addition to the up‐regulation of Hif2α, we found significantly increased expression of the constitutively active Hifβ subunit (Arnt2) and Hif‐dependent angiogenic genes (Vegfa and Vegfc). Among the Phd enzymes, which hydroxylate Hifα protein isoforms for degradation by the proteasome, we found selective mRNA down‐regulation of Phd3 (Egln3) (Table 2). As MCI integrity seems to be essential for acute O₂ sensing by peripheral chemoreceptors (Fernández‐Agüera et al. 2015), we also studied the expression of genes encoding ETC subunits. The level of mRNA expression of most of the ETC subunits was similar between the CB and SCG. However, several genes, which code for subunits of MCI to MCIV, were slightly down‐regulated in the CB compared to the SCG (Table 2). In contrast, the mRNAs of three ETC subunits were markedly overexpressed in CB cells: the Ndufa4l2 subunit and the genes encoding for cytochrome c oxidase subunit IV isoform 2 (Cox4i2) and cytochrome c oxidase subunit VIIIb (Cox8b).

Table 2.

Hypoxia‐related differential gene expression in the CB compared to the SCG of adult mice by microarray analysis*

Gene symbol	Description	Fold change (linear)	pFDR
Phd/Hif pathway and targets
Epas1(Hif2a)	endothelial PAS domain protein 1	102.0	2.49E‐03
Arnt2	aryl hydrocarbon receptor nuclear translocator 2	2.9	1.47E‐02
Egln3(Phd3)	egl‐9 family hypoxia‐inducible factor 3	−4.1	1.89E‐02
Vegfa	vascular endothelial growth factor A	4.2	1.86E‐02
Vegfc	vascular endothelial growth factor C	3.9	7.34E‐03
Mitochondria
Ndufa4l2	NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 4‐like 2	720.9	3.90E‐04
Ndufa8	NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 8	−2.1	4.01E‐02
Ndufa9	NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 9	−3.4	5.17E‐03
Ndufa10	NADH dehydrogenase (ubiquinone) 1 alpha subcomplex 10	−2.4	1.28E‐02
Ndufaf7	NADH dehydrogenase (ubiquinone) 1 alpha subcomplex assembly factor 7	−3.1	1.20E‐02
Ndufb6	NADH dehydrogenase (ubiquinone) 1 beta subcomplex, 6	−2.1	3.16E‐02
Ndufs8	NADH dehydrogenase (ubiquinone) Fe‐S protein 8	−3.9	1.09E‐02
Sdhd	succinate dehydrogenase complex, subunit D, integral membrane protein	‐4.0	1.03E‐02
Uqcrc1	ubiquinol‐cytochrome c reductase core protein 1	−2.6	1.07E‐02
Cyb561	cytochrome b‐561	−5.0	2.78E‐02
Cyb561d2	cytochrome b‐561 domain containing 2	−2.7	3.33E‐03
Cyb5b	cytochrome b5 type B	−5.0	6.07E‐03
Cyb5d1	cytochrome b5 domain containing 1	−3.6	3.94E‐02
Cyba	cytochrome b‐245, alpha polypeptide	3.2	1.49E‐02
Cox4i2	cytochrome c oxidase subunit IV isoform 2	10.1	6.03E‐03
Cox8b	cytochrome c oxidase subunit VIIIb	11.8	1.30E‐02
Cox5a	cytochrome c oxidase subunit Va	−2.2	4.54E‐02
Coa3	cytochrome C oxidase assembly factor 3	−3.0	2.21E‐02
Cox15	cytochrome c oxidase assembly protein 15	−3.6	1.79E‐02
Slc25a27	solute carrier family 25, member 27	−2.8	2.33E‐03
TCA cycle/anaplerosis/biotin‐related
Pcx	pyruvate carboxylase	4.9	1.31E‐02
Pdha1	pyruvate dehydrogenase E1 alpha 1	−2.5	1.90E‐02
Pdk4	pyruvate dehydrogenase kinase, isoenzyme 4	9.1	3.90E‐03
Clybl	citrate lyase beta like	−2.7	9.23E‐03
Acacb	acetyl‐Coenzyme A carboxylase beta	2.7	9.57E‐03
Slc7a5	solute carrier family 7 (cationic amino acid transporter, y+ system), member 5	−3.6	2.24E‐02
Idh1	isocitrate dehydrogenase 1 (NADP+), soluble	−3.5	1.34E‐02
Idh3a	isocitrate dehydrogenase 3 (NAD+) alpha	−3.1	5.50E‐03
Idh3b	isocitrate dehydrogenase 3 (NAD+) beta	−4.6	3.83E‐02
Ogdhl	oxoglutarate dehydrogenase‐like	−2.6	2.09E‐02
Ogfod1	2‐oxoglutarate and iron‐dependent oxygenase domain containing 1	−3.2	9.12E‐03
Mdh1	malate dehydrogenase 1, NAD (soluble)	−2.4	8.89E‐03
Others
Gdnf	glial cell line derived neurotrophic factor	2.1	9.36E‐03
Pparg	peroxisome proliferator activated receptor gamma	2.4	8.63E‐03

^*pFDR (P value adjusted with the false discovery rate) < 0.05 and fold change >2 or <−2.

CB glomus cells (and, to a lesser degree, AM chromaffin cells) contain large amounts of biotin, a coenzyme of carboxylases (Ortega‐Sáenz et al. 2016). In addition, CB cells contain high levels of succinate, which could be involved in acute O₂ sensing (Fernández‐Agüera et al. 2015). These facts led us to explore the status of genes involved in pyruvate metabolism and the TCA cycle. Notably, our analysis showed a clear induction of Pcx with decreased expression of a pyruvate dehydrogenase subunit (Pdha1), suggesting pyruvate‐mediated TCA anaplerosis (see Owen et al. 2002). We also found increased mRNA expression of pyruvate dehydrogenase kinase (Pdk4 isoform), which phosphorylates Pdha1 and inhibits the conversion of pyruvate to acetyl‐CoA (Table 2). However, Pdk4 overexpression was not confirmed in real‐time quantitative PCR analyses of sorted tyrosine hydroxylase‐positive (TH+) cells (see below). In contrast, the mRNA level of citrate lyase (Acly), a classical cataplerotic enzyme that converts cytosolic citrate to acetyl‐CoA and oxaloacetate (see Owen et al. 2002), was down‐regulated in sorted CB cells (see below). Acetyl‐CoA carboxylase b (Acacb) was slightly up‐regulated in CB cells, whereas several mitochondrial and cytosolic isoforms of isocitrate dehydrogenase (Idh1 and Idh3) were down‐regulated in CB samples. In parallel, we observed a decrease in the mRNA expression of the cationic amino acid transporter (Slc7a5), which could be related to decreased glutamate uptake and cytosolic production of α‐ketoglutarate (Table 2). Among other relevant genes, we found that the neurotrophic factor Gdnf, which is necessary for the maintenance of glomus cells (Villadiego et al. 2005; Pascual et al. 2008), was significantly overexpressed in the CB (Table 2).

Peripheral chemoreceptor cells contain several subtypes of O₂‐regulated K⁺ channels, which play a central role in cellular responsiveness to hypoxia. In addition, other channel types mediate the cell excitability and Ca²⁺ influx that are necessary for neurotransmitter release. The ion channels which were differentially expressed in the CB relative to the SCG are listed in Table 3. Task3 (Kcnk9, see above) was the most highly up‐regulated K⁺ channel gene in CB cells, although Task1 (Kcnk3) and maxi‐K⁺ β‐subunits (Kcnmb1 and Kcnmb2) were also induced. CB cells expressed a broad variety of voltage‐gated K⁺ channel α‐subunits, including members of the Kv3 (Kcnc) and Kv4 (Kcnd) families, which are regulated by hypoxia in mouse glomus cells (Pérez‐García et al. 2004). However, several subclasses of Kv (Kcn) α‐subunits were markedly down‐regulated in the CB (see below). The robust up‐regulation of a Na⁺‐activated K⁺ channel (Kcnt2) seen in the microarray analysis was not validated by quantitative PCR studies (see below). Interestingly, calsenilin (Kcnip3), a K⁺ channel‐interacting Ca²⁺‐binding protein involved in the regulation of gene expression and cell excitability (Spreafico et al. 2001), was up‐regulated in CB tissue. Kcnq3 channels, which mediate the acetyl choline‐activated M‐current characteristic of sympathetic neurons (see Brown & Passmore, 2009), were markedly up‐regulated in SCG tissue (Table 3), further supporting the proposition that the differences in the microarray expression profile reflect genes that are differentially expressed in CB versus SCG cells. Similar to Kv channels, most Na⁺ and Ca²⁺ channel subunits were up‐regulated in the SCG, which is compatible with the need for a high density of voltage‐gated ion channels to support the electrical excitability of large SCG sympathetic neurons in comparison with small CB glomus cells (Table 3). Notable exceptions to this general trend were the CB expression of Cacna1h (T‐type Ca²⁺ channel α‐1H subunit), which is induced by hypoxia in an HIF2α‐dependent manner (del Toro et al. 2003; Carabelli et al. 2007), and Scn7a, an atypical Na⁺ channel activated by extracellular Na⁺ that is involved in the regulation of salt intake behaviour (Hiyama et al. 2002). CB overexpression of Scn7a, however, was not confirmed in our PCR validation studies. Interestingly, the Scn9a gene, which encodes the Na⁺ channel α‐subunit (Nav1.7) that is involved in transmission of pain sensation (Cox et al. 2006), was markedly down‐regulated in the CB relative to the SCG. In agreement with the single‐cell sequencing study of Zhou et al. (2016), the gene encoding the cation‐permeable transient receptor potential channel Trpc5 was among the most highly expressed in CB cells.

Table 3.

Differential expression of ion channel genes in the CB compared to the SCG of adult mice by microarray analysis*

Gene symbol	Description	Fold change (linear)	pFDR
Potassium channels
Kcnab1	potassium voltage‐gated channel, shaker‐related subfamily, beta member 1	−5.0	1.09E‐02
Kcnab2	potassium voltage‐gated channel, shaker‐related subfamily, beta member 2	−5.3	6.56E‐03
Kcnb2	potassium voltage gated channel, Shab‐related subfamily, member 2	−4.0	6.99E‐03
Kcnc4	potassium voltage gated channel, Shaw‐related subfamily, member 4	−10.3	1.66E‐02
Kcnd1	potassium voltage‐gated channel, Shal‐related family, member 1	−2.7	1.64E‐02
Kcnd2	potassium voltage‐gated channel, Shal‐related family, member 2	−11.6	6.27E‐03
Kcne4	potassium voltage‐gated channel, Isk‐related subfamily, gene 4	2.3	1.88E‐02
Kcnh1	potassium voltage‐gated channel, subfamily H (eag‐related), member 1	−6.7	4.35E‐03
Kcnh5	potassium voltage‐gated channel, subfamily H (eag‐related), member 5	−13.3	3.20E‐03
Kcnh6	potassium voltage‐gated channel, subfamily H (eag‐related), member 6	−2.5	3.48E‐02
Kcnh7	potassium voltage‐gated channel, subfamily H (eag‐related), member 7	−4.1	1.20E‐02
Kcnip3	Kv channel interacting protein 3, calsenilin	5.7	1.54E‐03
Kcnip4	Kv channel interacting protein 4	−4.8	3.64E‐02
Kcnj3	potassium inwardly rectifying channel, subfamily J, member 3	−7.0	6.51E‐03
Kcnj10	potassium inwardly rectifying channel, subfamily J, member 10	−2.9	2.30E‐02
Kcnk3(Task1)	potassium channel, subfamily K, member 3	2.5	2.17E‐02
Kcnk9(Task3)	potassium channel, subfamily K, member 9	83.5	2.19E‐03
Kcnk10	potassium channel, subfamily K, member 10	−3.7	5.80E‐03
Kcnk18	potassium channel, subfamily K, member 18	−6.3	9.87E‐03
Kcnma1	potassium large conductance calcium‐activated channel, subfamily M, alpha member 1	−2.1	3.64E‐02
Kcnmb1	potassium large conductance calcium‐activated channel, subfamily M, beta member 1	4.6	5.68E‐03
Kcnmb2	potassium large conductance calcium‐activated channel, subfamily M, beta member 2	3.1	1.45E‐02
Kcnmb4	potassium large conductance calcium‐activated channel, subfamily M, beta member 4	−2.3	1.96E‐02
Kcnn3	potassium intermediate/small conductance calcium‐activated channel, subfamily N, member 3	−3.8	1.87E‐02
Kcnq2	potassium voltage‐gated channel, subfamily Q, member 2	−3.5	1.01E‐02
Kcnq3	potassium voltage‐gated channel, subfamily Q, member 3	−20.1	1.40E‐02
Kcnq5	potassium voltage‐gated channel, subfamily Q, member 5	−5.6	1.75E‐02
Kcnt1	potassium channel, subfamily T, member 1	−4.7	4.81E‐03
Kcnt2	potassium channel, subfamily T, member 2	7.4	2.03E‐03
Kctd9	potassium channel tetramerization domain containing 9	−2.3	9.29E‐03
Calcium channels
Cacna1a	calcium channel, voltage‐dependent, P/Q type, alpha 1A subunit	−3.0	1.40E‐02
Cacna1b	calcium channel, voltage‐dependent, N type, alpha 1B subunit	−5.5	4.54E‐02
Cacna1c	calcium channel, voltage‐dependent, L type, alpha 1C subunit	2.2	3.95E‐03
Cacna1h	calcium channel, voltage‐dependent, T type, alpha 1H subunit	2.7	7.70E‐04
Cacna1i	calcium channel, voltage‐dependent, alpha 1I subunit	2.6	9.87E‐03
Cacna2d3	calcium channel, voltage‐dependent, alpha2/delta subunit 3	−4.1	1.38E‐02
Cacnb1	calcium channel, voltage‐dependent, beta 1 subunit	−2.2	2.01E‐02
Cacnb3	calcium channel, voltage‐dependent, beta 3 subunit	−6.4	1.07E‐02
Cacnb4	calcium channel, voltage‐dependent, beta 4 subunit	−3.6	1.63E‐03
Cacng2	calcium channel, voltage‐dependent, gamma subunit 2	−6.1	1.29E‐02
Cacng3	calcium channel, voltage‐dependent, gamma subunit 3	−12.2	1.05E‐02
Sodium channels
Scn1a	sodium channel, voltage‐gated, type I, alpha	−8.8	6.04E‐03
Scn2a1	sodium channel, voltage‐gated, type II, alpha 1	−4.0	4.71E‐03
Scn2b	sodium channel, voltage‐gated, type II, beta	−11.6	5.63E‐03
Scn3a	sodium channel, voltage‐gated, type III, alpha	−4.9	1.98E‐02
Scn3b	sodium channel, voltage‐gated, type III, beta	−7.7	1.23E‐02
Scn7a	sodium channel, voltage‐gated, type VII, alpha	7.5	4.71E‐03
Scn9a	sodium channel, voltage‐gated, type IX, alpha	−14.9	1.38E‐02
Trp channels
Trpc5	transient receptor potential cation channel, subfamily C, member 5	30.1	2.74E‐03
Trpc6	transient receptor potential cation channel, subfamily C, member 6	−3.0	2.08E‐02
Trpc7	transient receptor potential cation channel, subfamily C, member 7	−4.1	1.01E‐02
Trpv2	transient receptor potential cation channel, subfamily V, member 2	−2.3	9.96E‐03
Trpv4	transient receptor potential cation channel, subfamily V, member 4	2.2	4.47E‐03

^*pFDR (P value adjusted with the false discovery rate) < 0.05 and fold change >2 or <−2.

Differential gene expression between the AM and SCG

Up‐ and down‐regulated AM genes can be visualized in Fig. 1 C (light red). The genes with the largest changes in expression level (top 20 up‐regulated and down‐regulated genes) are listed in Table 4. Among the most differentially up‐regulated genes in the AM were some related to steroid metabolism, which suggests contamination from adrenocortical cells that in the mouse can be embedded in the AM region. However, within this group we also found several highly induced genes, such as phenylethanolamine‐N‐methyltransferase (Pnmt), vesicle monoamine transporter 1 (Slc18a1) and chromogranin A (Chga), which are characteristic of AM chromaffin cells. In addition, this unbiased analysis revealed genes relevant to CB acute O₂ sensing (mitochondrial subunit Ndufa4l2 and Task3 and Task1 K⁺ channels; see Table 1), which were also highly expressed in AM cells (Table 4). Moreover, a number of genes that were more highly expressed in the SCG than the CB, such as Htr3a, Htr3b, Napb, Nefl, Tubb2a and Nrip3 (see Table 1), appeared in the list of the top 20 down‐regulated genes in AM cells (Table 4).

Table 4.

Top 20 up‐ and down‐regulated genes in the AM versus the SCG of adult mice by microarray analysis

Gene symbol	Description	Fold change (linear)	pFDR
Up‐regulated
Cyp11b1	cytochrome P450, family 11, subfamily b, polypeptide 1	22442	3.01E‐07
Srd5a2	steroid 5 alpha‐reductase 2	21977	2.07E‐07
Cyp11a1	cytochrome P450, family 11, subfamily a, polypeptide 1	19438	4.11E‐07
Hsd3b1	hydroxy‐delta‐5‐steroid dehydrogenase, 3 beta‐ and steroid delta‐isomerase 1	10373	2.00E‐06
Star	steroidogenic acute regulatory protein	9955	2.07E‐07
Cyp21a1	cytochrome P450, family 21, subfamily a, polypeptide 1	8609	1.00E‐06
Adh1	alcohol dehydrogenase 1 (class I)	1340	2.00E‐06
Kcnk9(Task3)	potassium channel, subfamily K, member 9	1244	1.60E‐05
Abcb1b	ATP‐binding cassette, sub‐family B (MDR/TAP), member 1B	966	2.00E‐06
Akr1cl	aldo‐keto reductase family 1, member C‐like	913	2.00E‐06
Akr1b7	aldo‐keto reductase family 1, member B7	850	5.10E‐05
Pnmt	phenylethanolamine‐N‐methyltransferase	500	3.70E‐05
Dlk1	delta‐like 1 homolog (Drosophila)	438	5.90E‐07
Ndufa4l2	NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 4‐like 2	233	4.00E‐06
Slc18a1	solute carrier family 18 (vesicular monoamine), member 1	156	2.40E‐05
Scarb1	scavenger receptor class B, member 1	148	4.00E‐06
Chga	chromogranin A	116	3.90E‐05
Gdf10	growth differentiation factor 10	114	8.00E‐05
Tacr2	tachykinin receptor 2	106	7.30E‐05
Kcnk3(Task1)	potassium channel, subfamily K, member 3	105	3.20E‐05
Down‐regulated
Htr3a	5‐hydroxytryptamine (serotonin) receptor 3A	−1920	3.46E‐07
Tubb3	tubulin, beta 3 class III	−831	8.00E‐06
Htr3b	5‐hydroxytryptamine (serotonin) receptor 3B	−828	2.00E‐06
Sv2c	synaptic vesicle glycoprotein 2c	−627	1.00E‐05
Tspan8	tetraspanin 8	−624	2.00E‐06
Prph	peripherin	−622	2.00E‐05
Ddah1	dimethylarginine dimethylaminohydrolase 1	−598	2.00E‐06
Tubb2b	tubulin, beta 2B class IIB	−378	2.00E‐05
Ret	ret proto‐oncogene	−313	1.90E‐05
Sncg	synuclein, gamma	−289	2.40E‐05
Avil	advillin	−276	3.20E‐05
Napb	N‐ethylmaleimide sensitive fusion protein attachment protein beta	−257	2.00E‐06
Ngfr	nerve growth factor receptor (TNFR superfamily, member 16)	−244	3.00E‐06
Ppp1r1c	protein phosphatase 1, regulatory (inhibitor) subunit 1C	−212	1.00E‐06
Nefl	neurofilament, light polypeptide	−195	1.60E‐05
Fxyd7	FXYD domain‐containing ion transport regulator 7	−188	2.00E‐06
Tubb2a	tubulin, beta 2A class IIA	−173	2.60E‐05
Nrip3	nuclear receptor interacting protein 3	−156	2.00E‐06
Areg	amphiregulin	−146	1.60E‐05
Rab6b	RAB6B, member RAS oncogene family	−134	1.00E‐05

pFDR, P value adjusted with the false discovery rate.

When we focused our analysis on genes related to O₂ sensing or potentially relevant to acute responsiveness to hypoxia, we found a qualitative profile similar to that previously revealed in the comparison between CB and SCG cells. We observed high expression of Hif2α and Hifβ (Arnt) isoforms as well as down‐regulation of Phd3 in the AM versus SCG. Moreover, in parallel with a slight decrease in the expression of some subunits of mitochondrial ETC complexes, the three atypical mitochondrial subunits up‐regulated in the CB (Ndufa4l2, Cox4i2 and Cox8b) were also enriched in AM cells (Table 5). In addition, our analysis identified a clear induction of Pcx, with down‐regulation of Pdha1 and decreased expression of the neutral amino acid transporter (Slc1a5), which can also transport glutamine (van Geldermalsen et al. 2016). The data further revealed a clear decrease in lactate dehydrogenase (Ldh) and glutaminase (Gls) expression in AM cells, which was not detected in CB tissue.

Table 5.

Hypoxia‐related differential gene expression in the AM compared to the SCG of adult mice by microarray analysis*

Gene symbol	Description	Fold change (linear)	pFDR
Phd/Hif pathway and targets
Hif1a	hypoxia inducible factor 1, alpha subunit	2.1	7.46E‐03
Epas1(Hif2a)	endothelial PAS domain protein 1	6.6	1.31E‐03
Arnt	aryl hydrocarbon receptor nuclear translocator	2.8	7.19E‐04
Arnt2	aryl hydrocarbon receptor nuclear translocator 2	15.5	5.10E‐05
Egln3(Phd3)	egl‐9 family hypoxia inducible factor 3	−4.6	3.44E‐03
Vegfa	vascular endothelial growth factor A	5.1	4.64E‐03
Mitochondria
Ndufa4	NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 4	−4.6	3.67E‐03
Ndufa4l2	NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 4‐like 2	232.6	4.00E‐06
Ndufaf5	NADH dehydrogenase (ubiquinone) complex I, assembly factor 5	−2.0	1.29E‐03
Ndufb6	NADH dehydrogenase (ubiquinone) 1 beta subcomplex, 6	−2.5	6.00E‐03
Ndufs4	NADH dehydrogenase (ubiquinone) Fe‐S protein 4	−2.1	1.05E‐02
Sdhd	succinate dehydrogenase complex, subunit D, integral membrane protein	−2.6	3.54E‐03
Cybb	cytochrome b‐245, beta polypeptide	3.8	4.11E‐04
Cox4i2	cytochrome c oxidase subunit IV isoform 2	2.1	1.72E‐03
Cox5a	cytochrome c oxidase subunit Va	−4.4	2.58E‐04
Cox5b	cytochrome c oxidase subunit Vb	−2.2	1.35E‐02
Cox6b1	cytochrome c oxidase, subunit VIb polypeptide 1	−3.1	4.10E‐03
Cox7a2	cytochrome c oxidase subunit VIIa 2	−2.5	1.14E‐02
Cox8b	cytochrome c oxidase subunit VIIIb	6.4	5.48E‐04
Coa3	cytochrome C oxidase assembly factor 3	−2.8	4.69E‐03
Ucp1	uncoupling protein 1 (mitochondrial, proton carrier)	−3.7	3.11E‐02
Ucp2	uncoupling protein 2 (mitochondrial, proton carrier)	10.1	5.54E‐04
Ucp3	uncoupling protein 3 (mitochondrial, proton carrier)	2.1	1.14E‐03
Slc25a27	solute carrier family 25, member 27	−5.0	7.00E‐05
TCA cycle/anaplerosis/biotin‐related
Pcx	pyruvate carboxylase	6.1	6.21E‐04
Pdha1	pyruvate dehydrogenase E1 alpha 1	−3.0	4.32E‐04
Pdk4	pyruvate dehydrogenase kinase, isoenzyme 4	−2.6	1.12E‐02
Slc1a5	solute carrier family 1 (neutral amino acid transporter), member 5	−3.4	1.60E‐02
Acaca	acetyl‐coenzyme A carboxylase alpha	−3.7	8.01E‐04
Ldha	lactate dehydrogenase A	−2.8	2.40E‐03
Ldhb	lactate dehydrogenase B	−9.3	7.70E‐05
Gls	glutaminase	−4.4	1.24E‐03
Tgm2	transglutaminase 2, C polypeptide	2.1	4.65E‐03
Aco1	aconitase 1	2.2	4.74E‐03
Idh3a	isocitrate dehydrogenase 3 (NAD+) alpha	−2.1	1.09E‐03
Suclg2	succinate‐coenzyme A ligase, GDP‐forming, beta subunit	3.9	1.52E‐04
Others
Ppara	peroxisome proliferator activated receptor alpha	−2.4	2.61E‐04
Ppargc1a	peroxisome proliferative activated receptor, gamma, coactivator 1 alpha	−6.0	3.96E‐04
Olfr78/Olfr560	olfactory receptor 78; olfactory receptor 560	−2.3	9.02E‐03

^*pFDR (P value adjusted with the false discovery rate) < 0.05 and fold change >2 or <−2.

The general pattern of ion channel gene expression in the AM versus SCG was, despite some notable exceptions, qualitatively similar to that observed in the comparison between the CB and SCG. The K⁺ channels with more pronounced differential expression were Task3 (Kcnk9) and Task1 (Kcnk3). As in the CB, maxi‐K⁺ channel subunits (Kcnma1 and Kcnmb2) and calsenilin (Kcnip3) were up‐regulated in AM cells (Table 6). A member of the small/intermediate conductance Ca²⁺‐activated K⁺ channel family (Kcnn2), which could mediate part of the O₂‐sensitive K⁺ current (Keating et al. 2001), was also slightly overexpressed in AM cells. Interestingly, the level of expression of the Kcnj11 subunit, which encodes a K_ATP channel (Kir 6.2) highly relevant to the developmental decrease of hypoxia sensitivity in chromaffin cells (Buttigieg et al. 2009; Salman et al. 2014), was practically the same in AM in comparison with CB or SCG cells (1.09‐fold change AM vs. SCG and 1.06‐fold change CB vs. SCG). As in the CB, we also observed a generalized decrease in the expression of voltage‐gated K⁺ channel subunits in the AM, particularly in the case of the genes encoding eag‐related (Kcnh5 and Kcnh7) and G protein‐regulated inward rectifier (Kcnj3) channels. K⁺ channels mediating the M‐current (Kcnq3 and Kcnq5) were down‐regulated in AM cells. Similar to the CB, several subunits of voltage‐gated Ca²⁺ and Na⁺ channels were down‐regulated in the AM in comparison with SCG cells. The largest difference was seen in the Scn9a gene, which encodes the Na⁺ channel α‐subunit (Nav1.7) that mediates pain sensation in the peripheral nervous system (Cox et al. 2006). In contrast, the Cacna1d and Cacna1h genes which encode, respectively, the α‐subunit of low voltage‐activated L‐type Ca²⁺ channels (Cav1.3) (Vandael et al. 2015) and the O₂‐sensitive component of the T‐type current (del Toro et al. 2003; Carabelli et al. 2007) were overexpressed in chromaffin cells. As in the CB, the Trpc5 gene was also highly induced in AM cells.

Table 6.

Differential expression of ion channel genes in the AM compared to the SCG of adult mice by microarray analysis*

Gene symbol	Description	Fold change (linear)	pFDR
Potassium channels
Kcna1	potassium voltage‐gated channel, shaker‐related subfamily, member 1	−28.4	4.27E‐04
Kcna2	potassium voltage‐gated channel, shaker‐related subfamily, member 2	−12.3	1.60E‐03
Kcna6	potassium voltage‐gated channel, shaker‐related, subfamily, member 6	−13.9	1.94E‐04
Kcnab1	potassium voltage‐gated channel, shaker‐related subfamily, beta member 1	−10.4	1.57E‐04
Kcnab2	potassium voltage‐gated channel, shaker‐related subfamily, beta member 2	−8.2	5.00E‐05
Kcnb2	potassium voltage gated channel, Shab‐related subfamily, member 2	−5.0	2.81E‐04
Kcnc4	potassium voltage gated channel, Shaw‐related subfamily, member 4	−4.5	3.40E‐04
Kcnd1	potassium voltage‐gated channel, Shal‐related family, member 1	−3.5	8.80E‐04
Kcnd2	potassium voltage‐gated channel, Shal‐related family, member 2	−21.8	1.95E‐04
Kcng4	potassium voltage‐gated channel, subfamily G, member 4	−2.6	1.16E‐02
Kcnh1	potassium voltage‐gated channel, subfamily H (eag‐related), member 1	−11.8	4.60E‐05
Kcnh5	potassium voltage‐gated channel, subfamily H (eag‐related), member 5	−24.4	1.60E‐05
Kcnh7	potassium voltage‐gated channel, subfamily H (eag‐related), member 7	−22.7	2.40E‐05
Kcnh8	potassium voltage‐gated channel, subfamily H (eag‐related), member 8	−6.8	7.70E‐05
Kcnip3	Kv channel interacting protein 3, calsenilin	16.2	5.00E‐05
Kcnip4	Kv channel interacting protein 4	−9.2	6.94E‐04
Kcnj3	potassium inwardly rectifying channel, subfamily J, member 3	−56.3	4.50E‐05
Kcnj6	potassium inwardly rectifying channel, subfamily J, member 6	2.2	8.71E‐04
Kcnj8	potassium inwardly rectifying channel, subfamily J, member 8	2.3	8.72E‐03
Kcnj10	potassium inwardly rectifying channel, subfamily J, member 10	−5.9	7.14E‐04
Kcnj13	potassium inwardly rectifying channel, subfamily J, member 13	−9.4	2.20E‐05
Kcnj16	potassium inwardly rectifying channel, subfamily J, member 16	−2.9	7.69E‐04
Kcnk2	potassium channel, subfamily K, member 2	12.1	1.41E‐04
Kcnk3(Task1)	potassium channel, subfamily K, member 3	105.4	3.20E‐05
Kcnk9(Task3)	potassium channel, subfamily K, member 9	1244.1	1.60E‐05
Kcnk10	potassium channel, subfamily K, member 10	−6.1	9.80E‐05
Kcnk18	potassium channel, subfamily K, member 18	−12.2	5.30E‐04
Kcnma1	potassium large conductance calcium‐activated channel, subfamily M, alpha member 1	3.5	8.40E‐05
Kcnmb2	potassium large conductance calcium‐activated channel, subfamily M, beta member 2	6.0	5.09E‐04
vKcnmb4	potassium large conductance calcium‐activated channel, subfamily M, beta member 4	−3.6	9.57E‐04
Kcnn2	potassium intermediate/small conductance calcium‐activated channel, subfamily N, member 2	2.0	2.29E‐03
Kcnn3	potassium intermediate/small conductance calcium‐activated channel, subfamily N, member 3	−6.1	1.19E‐04
Kcnn4	potassium intermediate/small conductance calcium‐activated channel, subfamily N, member 4	−3.6	1.74E‐04
Kcnq2	potassium voltage‐gated channel, subfamily Q, member 2	−3.2	4.39E‐04
Kcnq3	potassium voltage‐gated channel, subfamily Q, member 3	−16.4	3.12E‐04
Kcnq5	potassium voltage‐gated channel, subfamily Q, member 5	−19.8	8.40E‐05
Kcnt1	potassium channel, subfamily T, member 1	−7.7	5.00E‐05
Kctd9	potassium channel tetramerization domain containing 9	−2.6	5.76E‐04
Kctd16	potassium channel tetramerization domain containing 16	−2.7	6.83E‐04
Calcium channels
Cacna1a	calcium channel, voltage‐dependent, P/Q type, alpha 1A subunit	−2.6	5.76E‐04
Cacna1b	calcium channel, voltage‐dependent, N type, alpha 1B subunit	−6.2	3.67E‐04
Cacna1c	calcium channel, voltage‐dependent, L type, alpha 1C subunit	5.5	3.63E‐04
Cacna1d	calcium channel, voltage‐dependent, L type, alpha 1D subunit	14.3	7.00E‐05
Cacna1h	calcium channel, voltage‐dependent, T type, alpha 1H subunit	6.4	8.67E‐04
Cacna2d1	calcium channel, voltage‐dependent, alpha2/delta subunit 1	3.2	3.68E‐03
Cacna2d3	calcium channel, voltage‐dependent, alpha2/delta subunit 3	−18.5	8.50E‐05
Cacnb1	calcium channel, voltage‐dependent, beta 1 subunit	−2.7	2.09E‐03
Cacnb2	calcium channel, voltage‐dependent, beta 2 subunit	4.6	1.08E‐04
Cacnb3	calcium channel, voltage‐dependent, beta 3 subunit	−5.8	2.90E‐04
Cacnb4	calcium channel, voltage‐dependent, beta 4 subunit	−6.2	1.90E‐05
Cacng2	calcium channel, voltage‐dependent, gamma subunit 2	−10.0	6.20E‐05
Cacng3	calcium channel, voltage‐dependent, gamma subunit 3	−4.9	7.60E‐04
Cacng4	calcium channel, voltage‐dependent, gamma subunit 4	−2.4	4.31E‐04
Sodium channels
Scn1a	sodium channel, voltage‐gated, type I, alpha	−13.7	4.80E‐05
Scn2a1	sodium channel, voltage‐gated, type II, alpha 1	−6.0	1.96E‐04
Scn2b	sodium channel, voltage‐gated, type II, beta	−13.6	9.90E‐05
Scn3a	sodium channel, voltage‐gated, type III, alpha	−2.3	4.34E‐03
Scn3b	sodium channel, voltage‐gated, type III, beta	−6.8	7.21E‐04
Scn7a	sodium channel, voltage‐gated, type VII, alpha	−6.2	4.05E‐04
Scn9a	sodium channel, voltage‐gated, type IX, alpha	−42.2	1.30E‐05
Scn11a	sodium channel, voltage‐gated, type XI, alpha	−2.1	5.84E‐03
Trp channels
Trpc1	transient receptor potential cation channel, subfamily C, member 1	−2.4	4.09E‐03
Trpc5	transient receptor potential cation channel, subfamily C, member 5	32.5	4.80E‐05
Trpc6	transient receptor potential cation channel, subfamily C, member 6	−3.0	2.59E‐03
Trpc7	transient receptor potential cation channel, subfamily C, member 7	−3.2	1.82E‐03
Trpm3	transient receptor potential cation channel, subfamily M, member 3	−2.9	4.37E‐03
Trpv2	transient receptor potential cation channel, subfamily V, member 2	−3.3	3.72E‐04

^*pFDR (P value adjusted with the false discovery rate) < 0.05 and fold change >2 or <−2.

Expression of selected genes in tissues and sorted TH‐positive cells

To validate our microarray results, real‐time quantitative PCR was performed using a new set of biologically independent samples. As shown in Table 7, similar results were obtained for most of the genes tested in the two comparative analyses (CB vs. SCG and AM vs. SCG), thereby supporting the results of the microarray study. A limitation of using whole tissues is the contamination from non‐neuronal cells that occurs, particularly in the case of the CB, which, in addition to TH+ glomus cells, contains numerous capillaries and other cell types. To circumvent this issue, we used TH‐GFP mice in which the expression of GFP was under the control of the TH promoter. This allowed us to isolate TH+ (GFP+) cells from each tissue (Fig. 2 A). We confirmed that the sorted cells were TH+ by immunofluorescence staining (Fig. 2 B). The percentage of TH+/GFP+ cells counted in random samples was 87% (377/434), 95% (124/131) and 83% (80/97) from the CB, SCG and AM, respectively. Genes analysed by real‐time PCR using the sorted TH+ cells showed similar relative expression to that seen in the analysis of whole tissues (Table 7). The up‐regulation of Hif2α and the ETC subunits Ndufa4l2, Cox4i2 and Cox8b, and the down‐regulation of Phd3 in the CB and, less potently, in the AM versus SCG were validated with high significance in the PCR analyses on isolated TH+ cells. Overexpression of Pcx and down‐regulation of Acly, Slc7a5 and Idh1 were also confirmed in the comparison between TH+ CB and SCG cells. With the exception of Pcx and Cox4i2, this metabolic gene profile was also clearly seen in AM cells (Table 7). All the ion channel‐related genes overexpressed in the CB and AM were also validated with high significance by quantitative PCR of sorted TH+ cells, with the exception of Trpc5 in the AM. In general, high variability among the replicates of sorted TH+ cells from each tissue was observed compared to that of whole tissue analysis, which could explain the lack of statistical significance obtained using TH+ cells. Among other genes tested, uncoupling protein 2 (Ucp2) was up‐regulated in TH+ CB and AM cells, which also validates the microarray data. In addition, the quantitative PCR analysis of whole tissue or sorted TH+ cells showed that the Olfr78 gene, which encodes an atypical olfactory receptor expressed in several tissues outside the nasal epithelium, and particularly in the ganglia of the autonomic nervous system (Weber et al. 2002), was up‐regulated in CB glomus cells relative to sympathetic neurons, but markedly down‐regulated in AM chromaffin cells (Table 7).

Table 7.

Differential gene expression in the CB, AM and SCG analysed by real‐time quantitative PCR in adult mice

	CB vs. SCG						AM vs. SCG
	A. Microarray		B. Whole tissue		C. TH+ cells		A. Microarray		B. Whole tissue		C. TH+ cells
Gene symbol	Fold change	pFDR	Fold change†	P *	Fold change†	P *	Fold change	pFDR	Fold change†	P *	Fold change#	P *
Phd/Hif pathway
Epas1(Hif2a)	102.0	0.00249	190.2 ± 38.5	*	914.2 ± 270.3	*	6.6	0.00131	6.9 ± 0.7	*	16.4 ± 5.2	*
Hif1a	1.8	0.05001	6.4 ± 0.6	*	1.9 ± 0.5		2.1	0.00746	2.9 ± 0.4	*	2.6 ± 0.7	*
Egln3(Phd3)	−4.1	0.01886	−2.2 ± 0.5	*	CB NE		−4.6	0.00344	−2.9 ± 0.2	*	−21.1 ± 16.7	*
Mitochondrial ETC subunits
Ndufa4l2	720.9	0.00039	1003.8 ± 264.0	*	4015.5 ± 849.8	*	232.6	0.00000	257.6 ± 41.1	*	512.8 ± 184.5	*
Ndufa4	−1.9	0.21813	1.0 ± 0.1		1.5 ± 0.2		−4.6	0.00367	1.0 ± 0.1		1.2 ± 0.1
Cox4i2	10.1	0.00603	360.0 ± 69.1	*	1334.4 ± 226.9	*	2.1	0.00172	25.4 ± 5.1	*	15.2 ± 12.4
Cox4i1	−1.5	0.09280	1.5 ± 0.2	*	1.1 ± 0.1		−1.7	0.01749	1.3 ± 0.1		−1.5 ± 0.2
Cox8b	11.8	0.01295	512.9 ± 143.9	*	4699.9 ± 1024.0	*	6.4	0.00055	255.3 ± 29.4	*	8672.1 ± 2019.0	*
Metabolic enzymes and transporters
Pcx	4.9	0.01312	20.7 ± 3.8	*	23.5 ± 2.8	*	6.1	0.00062	8.7 ± 0.8	*	−1.3 ± 0.5
Pdha1	−2.5	0.01900	−1.4 ± 0.2		−2.1 ± 1.2		−3.0	0.00043	−1.2 ± 0.2		−4.7 ± 1.4	*
Slc7a5	−3.6	0.02241	−2.4 ± 0.4	*	−3.5 ± 0.6	*	1.7	0.08917	1.9 ± 0.3	*	−3.9 ± 1.9	*
Idh1	−3.5	0.01335	−1.8 ± 0.2	*	−6.7 ± 2.1	*	1.6	0.07287	2.6 ± 0.4	*	−3.3 ± 0.7	*
Idh3a	−3.1	0.00550	−1.2 ± 0.1	*	1.9 ± 0.4	*	−2.1	0.00109	−1.3 ± 0.1	*	−1.3 ± 0.4
Idh3b	−4.6	0.03826	−1.6 ± 0.2	*	1.2 ± 0.2		−1.5	0.00257	−1.2 ± 0.2		−1.4 ± 0.4
Acly	−4.4	0.07742	−1.5 ± 0.3		−4.6 ± 0.7	*	1.5	0.19180	1.3 ± 0.1		−4.8 ± 1.1	*
Acacb	2.7	0.00957	18.9 ± 3.7	*	‡		1.2	0.37025	3.8 ± 0.9	*	6.3 ± 3.1	*
Ion channels
Kcnk3(Task1)	2.5	0.02174	13.8 ± 2.6	*	74.8 ± 12.0	*	105.4	0.00003	23.1 ± 2.4	*	19.7 ± 5.4	*
Kcnk9(Task3)	83.5	0.00219	2371.4 ± 522.9	*	18431.8 ± 5917.5	*	1244.1	0.00002	6527.6 ± 1268.1	*	6647.1 ± 1736.0	*
Kcnip3	5.7	0.00154	32.3 ± 3.7	*	75.7 ± 10.8	*	16.2	0.00005	18.6 ± 1.3	*	30.1 ± 4.5	*
Trpc5	30.1	0.00274	4258.8 ± 628.5	*	132974.6 ± 48426.0	*	32.5	0.00005	1014.1 ± 77.6	*	1100.1 ± 799.9
Cacna1h	2.7	0.00077	41.6 ± 5.5	*	60.9 ± 11.3	*	6.4	0.00087	21.8 ± 2.8	*	8.7 ± 3.5	*
Others
Ucp2	2.4	0.06361	7.9 ± 1.3	*	10.4 ± 1.5	*	10.1	0.00055	13.1 ± 2.5	*	22.1 ± 2.9	*
Olfr78 §	1.3	0.55079	8.5 ± 1.6	*	5.3 ± 1.0	*	−2.3	0.00902	−22.6 ± 4.0	*	−141.4 ± 11.6	*

A. For comparison, the results from microarray analysis are also listed. B. Real‐time quantitative PCR analysis using RNA isolated from whole tissue. C. Real‐time PCR analysis using RNA isolated from TH⁺ cells from each tissue.

^* P < 0.05 compared to the SCG.

^†Mean ± SEM, n = 3–5 per group.

^‡Not available (highly variable, no conclusion).

^§Microarray detects Olfr78 and Olfr560, whereas real‐time PCR detects Olfr78 only; CB NE, no expression was detected from CB TH+ cells.

Abbreviations: Epas1(Hif2α), endothelial PAS domain protein 1; Hif1α, hypoxia inducible factor 1, alpha subunit; Egln3(Phd3), egl‐9 family hypoxia inducible factor 3; Ndufa4l2, NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 4‐like 2; Ndufa4, NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 4; Cox4i2, cytochrome c oxidase subunit IV isoform 2; Cox4i1, cytochrome c oxidase subunit IV isoform 1; Cox8b, cytochrome c oxidase subunit VIIIb; Pcx, pyruvate carboxylase; Pdha1, pyruvate dehydrogenase E1 alpha 1; Slc7a5, solute carrier family 7 (cationic amino acid transporter, y+ system), member 5; Idh1, isocitrate dehydrogenase 1 (NADP+), soluble; Idh3a, isocitrate dehydrogenase 3 (NAD+) alpha; Idh3b, isocitrate dehydrogenase 3 (NAD+) beta; Acly, ATP citrate lyase; Acacb, acetyl‐coenzyme A carboxylase beta; Kcnk3(Task1), potassium channel, subfamily K, member 3; Kcnk9(Task3), potassium channel, subfamily K, member 9; Kcnip3, Kv channel interacting protein 3, calsenilin; Trpc5, transient receptor potential cation channel, subfamily C, member 5; Cacna1h, calcium channel, voltage‐dependent, T type, alpha 1H subunit; Ucp2, uncoupling protein 2 (mitochondrial, proton carrier); Olfr78, olfactory receptor 78.

Figure 2. Sorting of tyrosine hydroxylase (TH) positive cells from adult wild‐type and TH‐GFP mice by flow cytometry.

A, isolation of green fluorescent protein positive (GFP+) cells from the carotid body (CB), superior cervical ganglion (SCG) and adrenal medulla (AM) of TH‐GFP mice by flow cytometry. 7‐AAD, 7‐aminoactinomycin D. Plots from wild‐type animals are shown as control. B, immunofluorescent staining demonstrating that the GFP+ cells express TH. Scale bar, 5 μm.

To investigate whether the differential gene expression observed in our microarray analysis was reflected at the protein level, we performed immunofluorescence experiments on carotid bifurcations and adrenal glands. We focused this analysis on the mitochondrial ETC subunits for which antibodies are available (Ndufa4l2 and Cox4i2) and Pcx. As shown in Fig. 3 A, the immunoreactive signal against Ndufa4l2 was higher in CB glomus cells than SCG neurons, as demonstrated by Ndufa4l2/TH co‐localization. A similar result was observed when comparing the AM with the adrenal cortex (Fig. 3 B). Strong immunostaining against the Cox4i2 subunit was observed in both CB and AM TH+ cells (Fig. 4). Pcx was more highly expressed in the CB than the SCG (Fig. 5). However, the immunostaining signal against this protein in the AM was indistinguishable from that in the adrenal cortex (data not shown). This is in agreement with our quantitative PCR data, in which no up‐regulation of Pcx mRNA was observed in sorted AM chromaffin cells (Table 7).

Figure 3. Immunohistochemical analysis of NADH dehydrogenase 1 alpha subcomplex 4‐like 2 (Ndufa4l2) expression in adult mice.

A, upper panels: representative section of the carotid bifurcation demonstrating high Ndufa4l2 immunoreactivity in the carotid body (CB) compared to superior cervical ganglion (SCG) and the co‐localization of Ndufa4l2 (red) and tyrosine hydroxylase (TH, green). Scale bar, 50 μm. Lower panels: magnification of SCG and CB regions indicated in the upper right panel. Scale bar, 10 μm. B, upper panels: representative section of the adrenal gland demonstrating high Ndufa4l2 immunostaining in the adrenal medulla (AM) compared to adrenal cortex (AC) and the co‐localization of Ndufa4l2 (red) and TH (green). Scale bar, 100 μm. Lower panel: magnification of AC and AM regions indicated in the upper right panel. Scale bar, 20 μm.

Figure 4. Immunohistochemical analysis of cytochrome c oxidase subunit IV isoform 2 (Cox4i2) expression in adult mice.

A, upper panels: representative section of the carotid bifurcation demonstrating high Cox4i2 immunoreactivity in the carotid body (CB) compared to superior cervical ganglion (SCG) and the co‐localization of Cox4i2 (red) and tyrosine hydroxylase (TH, green). Scale bar, 50 μm. Lower panels: magnification of SCG and CB regions indicated in the upper right panel. Scale bar, 10 μm. B, upper panels: representative section of the adrenal gland demonstrating high Cox4i2 immunostaining in the adrenal medulla (AM) compared to adrenal cortex (AC) and the co‐localization of Cox4i2 (red) and TH (green). Scale bar, 100 μm. Lower panel: magnification of AC and AM regions indicated in the upper right panel. Scale bar, 20 μm.

Figure 5. Immunohistochemical analysis of pyruvate carboxylase (Pcx) expression in adult mice.

Upper panels: representative section of the carotid bifurcation demonstrating high Pcx immunoreactivity in the carotid body (CB) compared to superior cervical ganglion (SCG) and the co‐localization of Pcx (red) and tyrosine hydroxylase (TH, green). Scale bar, 50 μm. Lower panels. Magnification of SCG and CB regions indicated in the upper right panel. Scale bar, 10 μm.

Impairment of acute O₂ sensing by genetic and pharmacological inhibition of succinate dehydrogenase

In a previous study, we suggested that CB glomus cells contain high levels of QH₂ due to succinate‐dependent metabolism, and that a further increase in the QH₂ pool during hypoxia leads to the generation of ROS and reduced pyridine nucleotides to signal membrane ion channels (Fernández‐Agüera et al. 2015). This proposal, which is compatible with the up‐regulation of Pcx in glomus cells described here (see Discussion), predicts that inhibition of succinate dehydrogenase should decrease acute responsiveness to low $P_{{\mathrm{O}}_2}$. We have generated an MCII‐null mouse (TH‐SDHD) carrying a floxed Sdhd allele that encodes the membrane anchoring subunit D of succinate dehydrogenase. This allele was deleted in CB glomus cells and other catecholaminergic cells by the transgenic expression of a Cre recombinase under the control of the TH promoter (Díaz‐Castro et al. 2012). The analysis of responsiveness to hypoxia in the glomus cells of TH‐SDHD mice is not straightforward because these cells enter a degenerative process that leads to their death (Díaz‐Castro et al. 2012; Platero‐Luengo et al. 2014). However, we were able to demonstrate that responsiveness to hypoxia disappears in Sdhd‐deficient glomus cells before they stop responding to high extracellular K⁺ or hypoglycaemia (Fig. 6 A and B). These data, which suggest that succinate dehydrogenase activity is required for normal acute O₂ sensing, were confirmed by experiments using glomus cells incubated overnight with dimethyl malonate (DMM), a membrane‐permeant competitive inhibitor of succinate dehydrogenase (see Gutman, 1978). DMM‐treated (for ∼12 h) glomus cells showed a drastic decrease in hypoxia‐induced catecholamine secretion, which was partially recovered during washout of malonate from the extracellular solution. Incubation with DMM did not significantly affect the secretory response to high extracellular K⁺ (Fig. 6 C–E).

Figure 6. Impairment of acute O₂ sensing by genetic and pharmacological inhibition of succinate dehydrogenase.

A, representative recordings of the ratiometric increase in cytosolic [Ca²⁺] elicited in Fura‐2‐loaded dispersed glomus cells from wild‐type and TH‐SDHD animals in response to hypoxia, 0 glucose and 40 mm K⁺. B, top: percentage of the number of cells that responded to a specific stimulus compared to the number of cells that responded to high potassium (Hypoxia: wild‐type, n = 69 from 7 mice; TH‐SDHD, n = 32 from 9 animals. 0 glucose: wild‐type, n = 19 from 3 mice; TH‐SDHD, n = 14 from 6 animals). Bottom: quantification of the increase in cytosolic [Ca²⁺] in glomus cells from wild‐type and TH‐SDHD animals that responded to the specific stimulus (40 mm K⁺: wild‐type, n = 69 from 7 mice; TH‐SDHD, n = 32 from 9 mice. Hypoxia: wild‐type, n = 61 from 7 mice; TH‐SDHD, n = 5 from 2 mice. 0 glucose: wild‐type, n = 19 from 3 mice; TH‐SDHD, n = 14 from 6 mice). C, representative amperometric recordings of responses to hypoxia and 40 mm K⁺ of glomus cells in wild‐type mouse CB slices incubated overnight with or without 2 mm dimethyl malonate (DMM). D, quantification of the secretion rate elicited by hypoxia (left) and 40 mm K⁺ (right) in control and DMM‐treated CB glomus cells from wild‐type animals (Hypoxia: control, n = 8 from 6 mice; DMM, n = 7 from 5 animals. 40 mm K⁺: control, n = 7 from 5 mice; DMM, n = 6 from 4 animals.) E, representative recording demonstrating the recovery of hypoxia‐induced secretory activity of glomus cells in wild‐type mouse slices after the washout of DMM. Cumulative secretion rate (red line) is represented at the bottom. ^* P < 0.05 compared to wild‐type or control.

Discussion

Differential gene expression profiles in the CB or AM versus SCG

In this study, we analysed the gene expression pattern in three sympathoadrenal tissues (CB, AM and SCG) with variable sensitivity to hypoxia to identify genes that are relevant to acute O₂ sensing in peripheral chemoreceptors. Our results reveal several genes potentially related to acute O₂ sensing that are modulated (induced or repressed) in the CB (and, to a lesser extent, in the AM) in comparison with the SCG. We then focused on genes that encode proteins related to the Phd/Hif pathway, mitochondrial ETC, metabolic enzymes and transporters, and ion channels. Some of the genes described here were also investigated in previous microarray analyses of CB tissue (Ganfornina et al. 2005; Balbir et al. 2007; Fagerlund et al. 2010; Mkrtchian et al. 2012) and particularly in a recent work that used single CB cell RNA sequencing (Zhou et al. 2016). However, our experimental approach, based on the comparison of adult tissues of the same developmental origin but variable sensitivity to hypoxia, allowed us to focus on genes with a similar qualitative expression pattern in O₂‐sensitive tissues (CB and AM) in comparison with O₂‐insensitive SCG cells.

Phd3 and Hif2α

Hif2α is one of the most highly expressed genes in CB cells versus SCG neurons, whereas Hif1α is only slightly up‐regulated. This finding confirms previous reports of the constitutive expression of Hif2α in adult mouse CB tissue (Tian et al. 1998) and the large number of Hif2α transcripts in neonatal mouse CB cells (Zhou et al. 2016). In contrast, we found a robust and previously undetected down‐regulation of Phd3 mRNA in CB versus SCG cells. Indeed, Phd3 mRNA expression was negligible in sorted glomus cells. Phd3 has relatively higher influence on Hif2α hydroxylation leading to its degradation (Appelhoff et al. 2004). Therefore, in addition to other functions (see below), down‐regulation of Phd3 probably contributes to the maintenance of a high Hif2α protein level in normoxic CB glomus cells. Hif2α mRNA overexpression and Phd3 down‐regulation were also clearly observed in AM cells. Hif2α has an important role in CB homeostasis (Peng et al. 2011), and is necessary for a normal HVR and CB growth in response to chronic hypoxia (Hodson et al. 2016). On the other hand, the Phd3/Hif2α pathway is involved in the regulation of sympathoadrenal development (Bishop et al. 2008; Macías et al. 2014). The CB and AM are slightly hypertrophied in Phd3‐null animals, and glomus cells exhibit a robust responsiveness to hypoxia (Bishop et al. 2008; Macias et al. 2014; our own unpublished observations). Therefore, our findings suggest that the pattern ‘Phd3 down‐regulation/Hif2α overexpression’ is essential for the metabolic specialization that confers acute O₂ sensitivity upon CB and AM cells (see below).

Atypical mitochondrial subunits

We found three atypical mitochondrial ETC subunits, Ndufa4l2, Cox4i2 and Cox8b, the mRNA of which was highly up‐regulated in CB and AM versus SCG cells. The protein levels of Ndufa4l2 and Cox4i2 were also higher in CB and AM chemoreceptor cells in comparison with SCG neurons. High levels of Ndufa4l2 and Cox4i2 transcripts have also been observed in neonatal CB glomus cells (Zhou et al. 2016). It has previously been reported that Ndufa4l2 and Cox4i2 mRNAs are strongly up‐regulated by hypoxia (Fukuda et al. 2007; Brown et al. 2010; Tello et al. 2011; Aras et al. 2013), but, to our knowledge, the regulation of Cox8b expression by hypoxia has not been documented. However, the mouse Cox8b gene contains two putative Hif binding sites in the promoter, one of them near the transcription initiation site (data not shown). Therefore, it is possible that the three atypical mitochondrial subunits are overexpressed in glomus and chromaffin cells due to the constitutively high Hif2α expression in these cells. However, the function of these atypical ETC subunits is unclear and their role in CB or AM cell physiology remains to be determined. Ndufa4l2 is a paralogue of the more ubiquitous subunit Ndufa4, which was thought to be a component of MCI (Carroll et al. 2006), although recent studies have suggested that it is associated with MCIV (Balsa et al. 2012; Kadenbach & Hüttemann, 2015). Ndufa4l2 expression seems to decrease O₂ consumption and ROS production by mitochondria (Tello et al. 2011). On the other hand, Cox4i2, the expression of which is restricted to lung and other highly oxygenated tissues, is an isoform of the broadly expressed subunit Cox4i1. In some mammalian cells exposed to hypoxia, Cox4i2 is induced to replace Cox4i1 thereby increasing the efficiency with which MCIV facilitates the transfer of electrons to O₂ and decreasing oxidative stress (Fukuda et al. 2007). In lung, Cox4i2 knockout reduces MCIV activity to 50%, despite the presence of the Cox4i1 isoform (Hüttemann et al. 2012). There are three tissue‐specific Cox8 isoforms (a, b and c), although to date no specific function has been identified for any of these (Kadenbach and Hüttemann, 2015). In several studies, however, Cox8b has been associated with thermogenic differentiation (‘browning’) of the white adipose tissue in rodents (Fisher et al. 2012; García et al. 2016). How the combination of the three subunits (Ndufa4l2, Cox4i2 and Cox8b) could render glomus cell mitochondria highly O₂‐sensitive is unknown. Cox4i2 and Cox8b are integral proteins with parallel single transmembrane α‐helices that run in close proximity (see Tsukihara et al. 1996). One could therefore speculate that these subunits interact to regulate O₂ diffusion through the inner membrane, in order to reach heme a3/CuB, which is buried inside MCIV (Michel et al. 1998). Low O₂ accessibility to the catalytic site would make the rate of cytochrome a3/CuB oxidation highly sensitive to decreases in $P_{{\mathrm{O}}_2}$. Together, the data discussed here suggest the existence of mitochondrial specializations that are characteristic of acute O₂‐sensing cells. Several decades ago a low affinity cytochrome C oxidase was proposed to exist in the CB, although located in type II rather than O₂‐sensitive glomus (type I) cells (Mills and Jöbsis, 1970). More recently, the involvement of mitochondria in acute O₂ sensing has been suggested by numerous studies of CB and AM cells (Duchen and Biscoe, 1992a, b; Ortega‐Sáenz et al. 2003; Wyatt & Buckler, 2004; Buttigieg et al. 2008; Buckler & Turner, 2013, Fernández‐Agüera et al. 2015). It is also worth noting that the existence of a mitochondrial based redox acute O₂ sensor is a long‐standing hypothesis, which was postulated to explain acute hypoxic pulmonary artery vasoconstriction (Archer et al. 1993; Waypa et al. 2001).

Pyruvate carboxylase and other metabolic enzymes and transporters

Our results demonstrate that Pcx is highly expressed in CB glomus cells (at the level of mRNA and protein) in comparison with SCG neurons. This occurs in parallel with a smaller decrease in pyruvate dehydrogenase (Pdha1) mRNA levels. Therefore, these data suggest that pyruvate is preferentially used by Pcx in glomus cell mitochondria to generate oxaloacetate, which represents a classical anaplerotic reaction that results in replenishment of the pool of TCA cycle intermediates (see Owen et al. 2002). These observations explain the extraordinarily high levels of biotin (a cofactor of carboxylases) recently found in glomus cells (Ortega‐Sáenz et al. 2016) and the need for HCO₃ ⁻/CO₂ buffers (which are the source of the carbon atoms required by carboxylases) to maintain a robust chemosensory function in CB in vitro preparations (Iturriaga & Lahiri, 1991; our own unpublished observations). Replenishment of TCA cycle intermediates is also compatible with the high levels of succinate in the CB and the postulated accumulation of QH₂ in acutely responding O₂‐sensitive cells (Fernández‐Agüera et al. 2015). If pyruvate is preferentially used to generate oxaloacetate in glomus cell mitochondria, the main source of acetyl‐CoA necessary for the TCA cycle could be free fatty acid (FFA) β‐oxidation. Although we do not have any direct evidence that FFA catabolism is activated in chemoreceptor cells, this idea is compatible with the existence of abundant adipose tissue in the CB (Ortega‐Sáenz et al. 2013) and the low Phd3 level characteristic of glomus cells, as it is known that Phd3‐dependent hydroxylation of Acacb activates the synthesis of malonyl‐CoA, a potent inhibitor of mitochondrial FFA uptake (German et al. 2016). Enhanced FFA catabolism (facilitated by Phd3 down‐regulation) would also significantly contribute to increasing the QH₂ pool in chemoreceptor cells, as each β‐oxidation cycle produces FADH₂, which is directly converted to QH₂ by the electron transport flavin/quinone oxidoreductase (ETF). Numerous cell types (in particular proliferating cells) rely on glutamine oxidation for TCA cycle anaplerosis. Glutamine is deaminated to glutamate, which is converted into α‐ketoglutarate, a TCA cycle intermediate (Yang et al. 2014) and a substrate of the O₂‐regulated prolyl hydroxylases (Epstein et al. 2001). Interestingly, in glomus cells, we observed a significant decrease not only in the expression of the cationic amino acid transporter Slc7a5, but also in the mRNA levels of the soluble isoform of isocytrate dehydrogenase (Idh1), which is necessary for the synthesis of α‐ketoglutarate in the cytosol. These metabolic adaptations, which help to maintain low levels of α‐ketoglutarate, are probably required to ensure low Phd3 activity, as it is known that under conditions in which Phd3 is inhibited (e.g. during hypoxia) the enzyme can be reactivated when cells are exposed to daily α‐ketoglutarate administration (Tennant & Gottlieb, 2010). The metabolic specializations of CB glomus cells, which may be relevant to acute O₂ sensing, were also qualitatively present in AM cells with the exception of Pcx which was not up‐regulated in our sorted TH+ AM cells. However, immunocytochemical analyses have directly shown high levels of biotin in the AM, which suggests the presence of high carboxylase activity in chromaffin cells (Ortega‐Sáenz et al. 2016).

Ion channels and other genes

Our comparative microarray analysis has shown that mouse CB and AM cells express a broad variety of calcium‐ and/or voltage‐gated ion channel genes, including some of the subunits which have been proposed to form the O₂‐sensitive voltage‐gated K⁺ channels (Wyatt & Peers, 1995; Pérez‐García et al. 2004). However, we also found that numerous voltage‐dependent K⁺, Na⁺ and Ca²⁺ channel subunits were down‐regulated in the CB or AM in comparison with the SCG. This is probably a consequence of the high density of voltage‐gated ion channels that is needed to sustain the electrical excitability of sympathetic neurons, which have a large somatodendritic arbor and profuse axonal branching. Task3 (Kcnk9) was the most highly up‐regulated K⁺ channel gene in CB and AM cells, although Task1 (Kcnk3) was also overexpressed. High expression of Task1 has previously been reported in the CB; however, Task3 was not detected in either a human CB microarray study (Mkrtchian et al. 2012) or single neonatal mouse CB cell transcriptomes (Zhou et al. 2016). This suggests that the expression of Task subunits may differ between species or during development. Task‐like channels appear to be the major channels responsible for the hypoxia‐induced depolarization of CB chemoreceptor cells (Buckler et al. 2000; Kim et al. 2009; Kobayashi & Yamamoto 2010). In addition, Task1/Task3 heteromers have been proposed to be the channels that mediate the O₂‐sensitive background current in adult rat (Kim et al. 2009) and mouse (Turner & Buckler, 2013) glomus cells. Whereas Task1‐deficient glomus cells have normal electrical parameters, cells from double Task1/Task3 knockout mice exhibit a clear depolarization, thereby supporting the role of Task3 channels in setting the resting potential of glomus cells (Ortega‐Sáenz et al. 2010). However, it must be noted that responsiveness to hypoxia is fully maintained in cells from Task1‐ or Task1/Task3‐null mice. Therefore, Task1 and Task3 channels do not seem to be indispensable for acute O₂ sensing, with other channels appearing to mediate sensitivity to changes in $P_{{\mathrm{O}}_2}$ in their absence (Ortega‐Sáenz et al. 2010). Two additional ion channel genes potentially related to acute responsiveness to hypoxia were overexpressed in glomus and chromaffin cells: Cacna1h (T‐type Cav3.2) and Trpc5. Ca²⁺ currents mediated by T‐type channels have previously been recorded from mouse CB cells (Ortega‐Sáenz et al. 2010), and it has recently been suggested that the Cav3.2 channel subtype mediates the CB over‐activation induced by chronic intermittent hypoxia (Makarenko et al. 2016). On the other hand, Cav3.2 mRNA is up‐regulated by chronic hypoxia in an Hif2α‐dependent manner in PC12 and chromaffin cells (del Toro et al. 2003; Carabelli et al. 2007). Cav3.2 expression is high in neonatal AM chromaffin cells but, as this occurs with acute sensitivity to hypoxia, it decreases with postnatal maturation. However, adult chromaffin cell sensitivity to hypoxia increases after AM denervation in parallel with the re‐appearance of T‐type Ca²⁺ channels (Levitsky and López‐Barneo, 2009). Together, these data suggest that Cav3.2 channels may play a fundamental role in facilitating the responsiveness of peripheral chemoreceptor cells to hypoxia. Similarly, the strong mRNA up‐regulation of Trpc5 channels in CB and AM cells suggests that they may also have an important role in the physiology of peripheral chemoreceptors. The existence of a significant background cationic conductance has been proposed to explain the relatively depolarized resting potential of glomus cells with respect to the K⁺ equilibrium potential (Carpenter & Peers, 2001; García‐Fernández et al. 2007b). In addition, it has been suggested that cationic currents participate in hypoxic activation of glomus (Kang et al. 2014) and chromaffin (Inoue et al. 1998) cells. Moreover, a cationic conductance, possibly mediated by Trp channels, is activated by hypoglycaemia in glomus cells (García‐Fernández et al. 2007b). As there are several subtypes of Trp channels in the CB (Buniel et al. 2003), Trpc homo‐ or heteromers could mediate the cationic currents mentioned above. Trpc channels are promiscuously activated by stretch, phospholipids and other variables (Beech, 2007). Interestingly, they are also inhibited by anaesthetics such as chloroform, propofol or halothane (Bahnasi et al. 2008), which could help to explain the strong respiratory depression induced by these drugs (Teppema et al. 2002).

Among other genes studied, we found that Ucp2 was systematically up‐regulated in CB and AM cells. Ucp2 and other uncoupling proteins are associated with thermogenesis and appear to be co‐expressed with Cox8b mRNA in some studies of adipose tissue thermogenic differentiation (Wu et al. 1999; Fisher et al. 2012). However, the significance of these observations in the context of CB physiology is currently unknown. Although it has been reported that Ucp2 knockout mice present pseudohypoxic pulmonary vascular remodelling and hypertension (Dromparis et al. 2013), preliminary experiments performed in this mouse model indicate that CB responsiveness to hypoxia is not significantly affected (our own unpublished observations). Olfr78 is an atypical olfactory receptor that is expressed outside the nasal mucosa and particularly in CB glomus cells (Chang et al. 2015; Zhou et al. 2016). In our analyses, Olfr78 mRNA was up‐regulated in the CB relative to the SCG but markedly down‐regulated in the AM. It has been reported that Olfr78 is a lactate receptor that is required for acute O₂ sensing, given that Olfr78‐deficient mice lose the HVR and glomus cell responsiveness to hypoxia. Based on these data, an endocrine model of acute O₂ sensing has been proposed in which lactate released from tissues during hypoxia activates glomus cells to produce hyperventilation (Chang et al. 2015). However, this model is incompatible with abundant data indicating that acute O₂ sensitivity is a cell‐autonomous phenomenon that can be observed in isolated cells bathed in lactate‐free solutions (see López‐Barneo et al. 2016a, b). In addition, two independent groups have found that, in contrast to the findings of Chang et al. (2015), Olfr78 is a poor lactate receptor (Aisenberg et al. 2016; Zhou et al. 2016). The fact that Olfr78 mRNA expression is much higher in the SCG than in the AM argues against any direct involvement of this receptor in acute O₂ sensing. However, ongoing experiments in several laboratories using various strains of Olfr78‐null mice should clarify the role of Olfr78 in CB physiology.

Signature gene‐expression profile in chemoreceptor cells and the mechanism of acute O₂ sensing

Together, the data available on the gene expression profile of chemoreceptor cells versus O₂‐insensitive SCG neurons suggest that a mix of genes encoding mitochondrial subunits, metabolic enzymes/transporters and ion channels is characteristic of cells that are acutely responding to decreases in $P_{{\mathrm{O}}_2}$. Our study suggests that down‐regulation of Phd3 and up‐regulation of Hif2α, Ndufa4l2, Cox4i2, Cox8b and Pcx confer CB glomus cells with their special sensitivity to hypoxia. Other genes that may also contribute to the ‘acute O₂‐sensing signature metabolic profile’ of chemoreceptor cells are Pdha1, Idh1, Acacb and Slc7a5. The absence of frank Pcx overexpression is probably one of the reasons why AM cells are less O₂‐sensitive than glomus cells. In addition to these ten ‘metabolic’ genes, four other genes which encode ion channels (Task3, Task1, Trp5, Cacna1h) are characteristic of CB and AM O₂‐sensitive cells. Although it is possible that none of these genes is absolutely required for acute O₂ sensing, their concerted action could result in a metabolic status that renders the cells sensitive to hypoxia. These data fit quite well with the MCI signalling model of acute O₂ sensing, which is based on high succinate content and QH₂ accumulation in glomus cells (Fernández‐Agüera et al. 2015), as well as the abundant evidence suggesting that special mitochondrial properties could contribute to hypoxia responsiveness in glomus cells (Mill & Jöbsis, 1970; Duchen & Biscoe, 1992a, b; Ortega‐Sáenz et al. 2003; Wyatt & Buckler, 2004; Buckler & Turner, 2013, Fernández‐Agüera et al. 2015).

The metabolic features of chemoreceptor cells, in particular glomus cells, are schematically summarized in Fig. 7. Phd3 has a central position in this scheme, as its low level of expression probably permits Hif2α stabilization and the subsequent induction of the genes that encode the three atypical mitochondrial subunits (Ndufa4l2, Cox4i2 and possibly Cox8b). The suppression of Phd3 activity, which is favoured by maintaining a low level of cytosolic α‐ketoglutarate, also decreases malonyl‐CoA synthesis and therefore enables FFA‐dependent mitochondrial metabolism and the generation of both acetyl‐CoA for the TCA cycle and large amounts of QH₂. In parallel, the high levels of Pcx and relative down‐regulation of Pdh provide abundant oxaloacetate to replenish TCA cycle intermediates and to further increase the QH₂ pool through the activity of MCI and MCII. The presence of Cox4i2, Cox8b and Ndufa4l2 could make cytochrome c oxidase activity highly sensitive to decreases in $P_{{\mathrm{O}}_2}$, such that even relatively mild hypoxia causing accumulation of an extra amount of QH₂ leads to the slow‐down or even reversal of MCI and the production of molecules (reduced pyridine nucleotides and ROS) that signal via ion channels in the plasma membrane (see Fernández‐Agüera et al. 2015).

Figure 7. Model of the metabolic features of chemoreceptor glomus cells relevant to O₂ sensing based on their gene expression profile.

Proteins (or subunits of an enzyme complex) highly expressed in the CB in comparison to the SCG, as suggested by the microarray analysis, are highlighted in red, whereas proteins (or subunits of an enzyme complex) with relatively low expression in the CB are represented in dark blue. Arrows attached to continuous lines indicate one‐step chemical reactions, whereas arrows attached to discontinuous lines represent multistep chemical modifications. Acac, acetyl‐coenzyme A carboxylase; Cox4i2, cytochrome c oxidase subunit IV isoform 2; Cox8b, cytochrome c oxidase subunit VIIIb; ETF, electron transport flavin/quinone oxidoreductase; FFA, free fatty acid; Hif2α, endothelial PAS domain protein 1 (Epas1); Idh, isocitrate dehydrogenase; MCI, MCII, MCIII, MCIV, mitochondrial complex I, II, III, IV, respectively; Ndufa4l2, NADH dehydrogenase (ubiquinone) 1 alpha subcomplex, 4‐like 2; Pcx, pyruvate carboxylase; Pdh, pyruvate dehydrogenase; Phd3, egl‐9 family prolyl hydroxylase 3 (Egln3); Q, ubiquinone; QH₂, ubiquinol/reduced ubiquinone; Slc7a5, solute carrier family 7 (cationic amino acid transporter, y+ system), member 5; TCA, tri‐carboxylic acid. See text for detailed explanation.

For the past three decades, the identification of the ‘O₂ sensor’ in glomus cells has been a matter of much discussion and investigation. However, numerous studies, in many cases using genetically modified animals with selective ablation of specific genes, have suggested that, in line with the model proposed here, O₂ sensing is a multifactorial process that depends on the biophysical and metabolic properties of chemoreceptor cells rather than on the function of a specific O₂‐sensing molecule (for recent reviews see Lopez‐Barneo et al. 2016a, b). Interestingly, recent work on mice with ablation of MCI genes (Fernandez‐Agüera et al. 2015) and the gene expression analyses discussed here suggest the compatibility of the two classical ‘models’ of CB acute O₂ sensing: the ‘membrane model’, based on the modulation of ion channels by hypoxia and the ‘metabolic hypothesis’ which claims a fundamental role for mitochondria in this process. A major advantage of the scheme in Fig. 7 is that it allows predictions that can be experimentally tested. It has already been shown that pharmacological or genetic abolition of MCI selectively abolishes responsiveness to hypoxia in glomus and chromaffin cells (Ortega‐Saenz et al. 2003; Thompson et al. 2007; Fernández‐Agüera et al. 2015). Similarly, it is also known that CB cells are highly dependent on MCII for survival (Díaz‐Castro et al. 2012; Platero‐Luengo et al. 2014) and that genetic or pharmacological MCII dysfunction causes inhibition of responsiveness to hypoxia in pulmonary myocytes (this paper; see also Paddenberg et al. 2012). It can therefore be expected that future experiments, focused on the genes discussed here, will provide further understanding of the mechanisms of acute O₂ sensing by chemoreceptor cells, and their modifications in variable pathophysiological or developmental conditions.

Additional information

Competing interests

The authors have no conflicts of interest to declare

Author contributions

L.G., V.B.‐H., P.G.‐F., I.A.‐M. and P.O.‐S. performed the experiments and participated in the interpretation of data. L.G., P.O.‐S and J.L.‐B. designed the figures and wrote the first draft of the paper. J.L.B. wrote the final draft of the paper and supervised the project. All authors read and approved the final paper.

Funding

This research was supported by the Botín Foundation, the Spanish Ministry of Economy and Innovation (SAF2012‐39343, SAF2016‐74990‐R) and the European Research Council (ERC Advanced Grant PRJ201502629).