Research ArticlePhysiology

Acute O2 sensing through HIF2α-dependent expression of atypical cytochrome oxidase subunits in arterial chemoreceptors

See allHide authors and affiliations

Science Signaling  21 Jan 2020:
Vol. 13, Issue 615, eaay9452
DOI: 10.1126/scisignal.aay9452

Breathing faster with HIF2α

In response to hypoxia, glomus cells in the carotid body trigger an increase in ventilation. At the molecular level, hypoxia slows down the electron transport chain in mitochondria, resulting in the accumulation of ROS and NADH, which ultimately activate glomus cells. HIF2α is a hypoxia-induced transcription factor that is highly abundant in glomus cells. Moreno-Domínguez et al. found that HIF2α was necessary for acute responses to hypoxia (see the Focus by Bishop and Ratcliffe). HIF2α mediated the expression of three atypical electron transport chain subunits that were necessary for a rapid buildup of ROS and NADH under hypoxic conditions in glomus cells. Similar to mice deficient in HIF2α in glomus cells, mice that lacked at least one of these atypical electron transport chain subunits in glomus cells also failed to increase ventilation in response to hypoxia. The authors propose that HIF2α target gene expression may set a tissue’s acute O2-sensing ability.


Acute cardiorespiratory responses to O2 deficiency are essential for physiological homeostasis. The prototypical acute O2-sensing organ is the carotid body, which contains glomus cells expressing K+ channels whose inhibition by hypoxia leads to transmitter release and activation of nerve fibers terminating in the brainstem respiratory center. The mechanism by which changes in O2 tension modulate ion channels has remained elusive. Glomus cells express genes encoding HIF2α (Epas1) and atypical mitochondrial subunits at high levels, and mitochondrial NADH and reactive oxygen species (ROS) accumulation during hypoxia provides the signal that regulates ion channels. We report that inactivation of Epas1 in adult mice resulted in selective abolition of glomus cell responsiveness to acute hypoxia and the hypoxic ventilatory response. Epas1 deficiency led to the decreased expression of atypical mitochondrial subunits in the carotid body, and genetic deletion of Cox4i2 mimicked the defective hypoxic responses of Epas1-null mice. These findings provide a mechanistic explanation for the acute O2 regulation of breathing, reveal an unanticipated role of HIF2α, and link acute and chronic adaptive responses to hypoxia.


Oxygen (O2) is essential for life, with mammals being particularly susceptible to low O2 environments, even for brief periods. Decreases in blood O2 tension (PO2) trigger rapid adaptive cardiorespiratory reflexes (hyperventilation and sympathetic activation) that, within a period of seconds, increase O2 uptake and its distribution to the most vulnerable organs such as the brain or heart (1, 2). These systemic acute responses to O2 deficiency (hypoxia) are primarily mediated by the carotid body (CB), an arterial chemoreceptor located in the carotid bifurcation. The CB contains neurosensory glomus cells (also called type I cells) with O2-sensitive K+ channels that are inhibited by hypoxia, thus leading to cell depolarization, Ca2+ influx, and the release of transmitters that activate afferent nerve fibers impinging upon the brainstem respiratory and autonomic centers (3). The molecular mechanisms by which glomus cells acutely detect hypoxia have remained elusive (4, 5). Although several O2-sensing mechanisms have been postulated [see (6, 7) for recent reviews], none have proven to be essential for CB function, because mice with ablation of genes coding for the relevant enzymes or receptors exhibit normal CB responses to hypoxia (812).

A study using genetically modified mice lacking the NDUFS2 subunit, a component of mitochondrial complex I (MCI) essential for its assembly and for ubiquinone and rotenone binding (13), has shown that MCI function is required for CB acute O2 sensing. These data suggest that acute hypoxia slows down the mitochondrial electron transport chain (ETC) in glomus cells, consequently producing an accumulation of NADH (reduced form of nicotinamide adenine dinucleotide) and reactive oxygen species (ROS), which modulate ion channels to elicit membrane depolarization (13, 14). Gene expression analyses have also shown that adult CB glomus cells have a signature metabolic profile, characterized by an elevated constitutive expression of Epas1 [coding for hypoxia-inducible factor 2α (HIF2α)], as well as genes coding atypical mitochondrial ETC subunits (NDUFA4L2, COX4I2, and COX8B) (15, 16). Because Ndufa4l2 (17) and possibly Cox4i2 (18, 19) are transcriptionally regulated by hypoxia, and because the Cox8b promoter contains putative HIF binding motifs (16), we hypothesized that the gene expression profile and O2-sensing properties of glomus cells could depend on HIF2α (16). Abolition of Epas1 in embryonic life results in CB atrophy (20), and HIF2α deficiency in adulthood results in reduction in amplitude of the hypoxic ventilatory response (HVR) and inhibition of the CB growth necessary for acclimatization to sustained hypoxia (21, 22). Here, we report that, apart from anemia (23) and left ventricular hypertrophy, adult mice with conditional deletion of the Epas1 gene appeared normal without other alterations in body weight, organismal functions, or metabolic regulation. However, these animals exhibited strong suppression of glomus cell acute mitochondrial and cellular responses to hypoxia and inhibition of the HVR. Epas1-deficient mice showed decreased expression of Ndufa4l2, Cox4i2, and Cox8b in the CB. Genetic ablation of Cox4i2 in CB glomus cells mimicked the effects of Epas1 deficiency. These data provide an explanation for the acute regulation of breathing by O2, a fundamental biological process of pathophysiological relevance.


Selective loss of cellular and mitochondrial responses to hypoxia in conditional Epas1-deficient glomus cells

We studied mice carrying normal (+), null (−), or floxed Epas1 alleles and that ubiquitously expressed tamoxifen-inducible Cre recombinase. Two-month-old mice were treated with a tamoxifen-containing diet for 4 weeks, and experiments were performed ~2 months later (in ~5-month-old mice) to allow for complete degradation of HIF2α and HIF2α-induced proteins (Fig. 1A, inset). Epas1-deficient mice (Epas1f/−,Cre) showed a marked decrease in Epas1 mRNA with unaltered Hif1α mRNA levels in the CB (Fig. 1A) and adrenal medulla (AM) (fig. S1A), which are catecholaminergic tissues with high HIF2α mRNA expression under normoxic conditions (16, 24). Similar results were obtained for kidney (Fig. 1B) and lung (fig. S1B). About 50% decrease in mRNA levels was observed in heterozygous Epas1+/− mice (Fig. 1, A and B, and fig. S1A). Epas1-null mice showed a marked decrease in hematocrit (Fig. 1C), further supporting the essential role of HIF2α in adult erythropoiesis (23, 25). These mice had normal body and organ weights, with the exception of an appreciable left ventricular hypertrophy that was likely secondary to anemia (fig. S1, C to F). The size and histological appearance of the CB was similar in wild-type (Epas1+/+) and Epas1-null mice (fig. S2A). However, the responsiveness of glomus cells to hypoxia, as determined by amperometric measurement of quantal dopamine release with a carbon fiber electrode (13), was markedly inhibited in Epas1f/−,Cre mice in comparison with wild-type mice used as control (Fig. 1, D and E). The percentage of cells that responded to hypoxia decreased from 96% in wild-type mice (Epas1+/+) to 66% in heterozygous (Epas1+/−) mice and to 42% in Epas1-null (Epas1f/−,Cre) mice (Fig. 1F). In addition, Epas1-deficient cells that exhibited some secretory activity in hypoxia also had a decreased average secretion rate (~40% control value in the case of cells from Epas1f/−,Cre mice) (Fig. 1G, left). In contrast, responsiveness to hypercapnia and depolarization with high extracellular K+ were unaffected (Fig. 1, D to G, and fig. S2, B and C). When a mild stimulus was applied, we could resolve individual secretory events, and in these conditions, we noted that the distribution of quantal charge (26, 27) in recordings from HIF2α-deficient glomus cells shifted toward lower values, which resulted in a reduction in the mean charge per event with respect to controls (fig. S2D). However, this reduction had only a minor effect on the global secretory activity of the Epas1-null cells, because the secretion rate induced by CO2 and high K+ was not significantly altered (Fig. 1G and fig. S2C). In agreement with these results, single Epas1-deficient glomus cells loaded with Fura-2 exhibited inhibition of the hypoxia-induced increase in cytosolic [Ca2+] without changes in the response to CO2 or high K+ (Fig. 1, H to L). Electrical parameters (cell capacitance and resting input resistance) and voltage-gated K+ and Ca2+ currents were similar in wild-type and Epas1-null glomus cells recorded by patch clamp (fig. S3, A to E), indicating that the loss of O2 sensitivity due to HIF2α deficiency is selective and occurs in healthy cells with normal electrical properties and expression of voltage-gated ion channels.

Fig. 1 Loss of responsiveness to acute hypoxia in glomus cells from conditional Epas1-null mice.

(A and B) Inset: Experimental mouse model and time course of tamoxifen (TMX) treatment. Epas1 (Hif2α) and Hif1α mRNA levels in CBs and kidney of Epas1+/+, Epas1+/−, and Epas1f/−,Cre mice [n = 5 to 6 replicates per group, with each replicate consisting of five (CB) or one (kidney) mice of the corresponding genotype]. (C) Hematocrit values in Epas1+/+ (n = 65), Epas1+/− (n = 35), and Epas1f/−,Cre (n = 68) mice. (D and E) Amperometric recordings of catecholamine release from glomus cells in CB slices from Epas1+/+ and Epas1f/−,Cre mice exposed to hypoxia (Hx; ~10 to 15 mmHg), hypercapnia (CO2; 20% CO2), and high K+ (40K; 40 mM KCl). Cumulative secretion signals are in red. The vertical discontinuous lines indicate resetting of the integrator. (F) Percentage of cells with a secretory response to hypoxia and hypercapnia in Epas1+/+ (n = 25/16 and 21/12 slices/mice, respectively), Epas1+/− (n = 12/3 and 12/3 slices/mice, respectively), and Epas1f/−,Cre (n = 21/9 and 20/9 slices/mice, respectively) mice. (G) Average secretion rate (picocoulombs per minute) measured during exposure to hypoxia and 40 mM KCl in control (hypoxia, n = 24/15 slices/mice; 40K, n = 25/16 slices/mice), Epas1+/− (hypoxia, n = 8/3 slices/mice; 40K, n = 12/3 slices/mice), and Epas1-null (hypoxia, n = 9/4 slices/mice; 40K, n = 21/9, slices/mice) mice. (H and I) Changes in cytosolic [Ca2+] elicited by hypoxia, hypercapnia, and high K+ in dispersed glomus cells from Epas1+/+ and Epas1-null mice. (J) Percentage of responding cells (exhibiting an increase in cytosolic [Ca2+] in response to hypoxia or hypercapnia) relative to high K+ responding cells from Epas1+/+ (n = 28/6 and 15/3 cells/mice, respectively) and Epas1f/−,Cre (n = 25/6 and 23/3 cells/mice, respectively) mice. (K and L) Quantification of the increase in cytosolic [Ca2+] elicited by hypoxia or hypercapnia (K) and 40 mM K+ (L) in dispersed cells from Epas1+/+ (n = 24/6, 13/3, and 28/6 cells/mice, respectively) and Epas1-null (n = 3/3, 13/3, and 25/6 cells/mice, respectively) mice. Data are represented as mean ± SEM. Statistically significant differences compared to Epas1+/+ (*) or Epas1+/− (#) values are indicated. *P < 0.05; **P < 0.001; #P < 0.05. a.u., arbitrary units.

Exposure of O2-sensing glomus cells to decreased PO2 elicits NADPH (nicotinamide adenine dinucleotide phosphate) accumulation (13, 28, 29) and production of ROS by MCI and possibly MCIII due to slowdown of the mitochondrial ETC and the backlog of electrons (Fig. 2A) (14). NADH and ROS are hypoxic mitochondrial signals that can modulate membrane ion channels to induce cell depolarization and transmitter release (13, 14). Basal levels of NADH autofluorescence were similar in Epas1+/+ and Epas1f/−,Cre glomus cells, indicating normal activity of MCI and Kreb’s cycle dehydrogenases (Fig. 2B). However, the hypoxia-induced increase in NADH that is characteristic of these cells was inhibited after Epas1 deletion (Fig. 2, C and D). The percentage of cells that responded to hypoxia was lower in Epas1-null cells (28%) compared to Epas1+/+ (84%) or Epas1+/− (81%) cells (Fig. 2D, left). The average amplitude of the NADH signal in Epas1-null and Epas1+/− cells that exhibited some NADH accumulation under hypoxia was less than half the control value (Fig. 2D, right). Rotenone (a selective blocker of MCI NADH/quinone oxidoreductase) induced a robust NADH signal in all hypoxia-unresponsive cells tested, further supporting the notion that HIF2α-deficient glomus cells had normal mitochondrial metabolism (Fig. 2C). We also monitored the compartmentalized mitochondrial ROS production during hypoxia using a redox-sensitive green fluorescent protein probe (roGFP) genetically targeted to either the intermembrane space (IMS) or the mitochondrial matrix (14, 30). As shown before (14), hypoxia induced a fast and reversible increase in ROS in the IMS and a decrease in matrix ROS. However, these responses were differentially affected by Epas1 deficiency (Fig. 2, E to H). The IMS ROS signal, which is blocked by rotenone and inhibited in O2-sensitive glomus cells with MCI dysfunction (14), was diminished in Epas1-null cells (Fig. 2E), with a decrease in the number of responding cells and in the amplitude of responses (Fig. 2F). In contrast, the decrease in matrix ROS, which is unaffected by rotenone and unrelated to acute O2 sensing by glomus cells (14), was not altered (Fig. 2, G and H). In all the hypoxia-unresponsive Epas1-deficient cells tested, rotenone elicited an increase in IMS or matrix ROS, indicating that the mitochondrial ETC was functioning normally (Fig. 2, E and G). Together, these results demonstrate that inactivation of Epas1 in adult CB glomus cells does not appreciably alter the ETC but leads to inhibition of mitochondrial signaling and cellular responsiveness to hypoxia.

Fig. 2 Inhibition of hypoxia-induced mitochondrial signaling in glomus cells in conditional Epas1-null mice.

(A) Scheme illustrating the accumulation of reduced ubiquinone (CoQH2), NADH, and ROS during exposure to hypoxia. (B) Basal NADPH autofluorescence in glomus cells from Epas1+/+ (n = 23/4 cells/mice) and Epas1f/−,Cre (n = 23/4 cells/mice) mice. (C) Hypoxia (~10 to 15 mmHg)–induced changes in NADPH autofluorescence (AF) recorded in glomus cells from Epas1+/+ (top) and Epas1f/−,Cre (two separate recordings in the middle) mice. Bottom: Rotenone (1 μM)–induced increase in NADPH in an Epas1-deficient glomus cell (n = 5 similar experiments performed with the same result). (D) Left: Percentage of cells with a hypoxia-induced increase in NADPH autofluorescence from Epas1+/+ (n = 49/6 cells/mice), Epas1+/− (n = 21/3 cells/mice), and Epas1f/−,Cre (n = 25/4 cells/mice) mice. Right: Increase in NADH autofluorescence in Epas1+/+ (41/6 cells/mice), Epas1+/− (17/3 cells/mice), and Epas1f/−,Cre (7/4 cells/mice) responding cells. (E) Changes in ROS in the mitochondrial intermembrane space (IMS ROS) induced by hypoxia in Epas+/+ (top) and Epas1f/−,Cre (middle) mice. The response to 0.1 mM H2O2 was also tested to ensure correct functioning of the mitochondrial roGFP probes. Rotenone (1 μM; bottom) was used to show that the MCI was functioning normally. (F) Left: Percentage of glomus cells that exhibited a hypoxia-induced increase in IMS ROS from Epas1+/+ (n = 25/5 cells/mice) and Epas1f/−,Cre (n = 31/6 cells/mice) mice. Right: Hypoxia-induced increase in IMS ROS from hypoxia-responsive glomus cells from Epas1+/+ (n = 20/5 cells/mice) and Epas1-null (n = 9/5 cells/mice) mice. (G) Changes in ROS in the mitochondrial matrix (Matrix ROS) induced by hypoxia in control (top) and Epas1f/−,Cre (middle) mice. The response to H2O2 (0.1 mM) was also tested. Bottom: Response to rotenone (1 μM). (H) Left: Percentage of glomus cells that exhibited a hypoxia-induced decrease in matrix ROS from Epas1+/+ (n = 17/3 cells/mice) and Epas1f/−,Cre (n = 20/3 cells/mice) mice. Right: Hypoxia-induced decrease in matrix ROS in hypoxia-responsive glomus cells from Epas1+/+ (n = 14/3 cells/mice) and Epas1-null (n = 18/3 cells/mice) mice. Data are presented as mean ± SEM. *P < 0.05.

Inhibition of the HVR in conditional Epas1-null mice

Although we did not attempt a detailed study of the systemic respiratory function in Epas1-null mice, we observed that ablation of Epas1 alleles led to changes in the HVR. In agreement with a previous report (21), heterozygous Epas1+/− mice showed a small decrease in the HVR (Fig. 3, A to E), although they had a normal hematocrit and were apparently indistinguishable from Epas1+/+ mice (Fig. 1C and fig. S1, A to D). However, the effect of HIF2α deficiency was most obvious in Epas1f/−,Cre animals, in which the HVR (measured once O2 levels in the plethysmographic chamber had reached 10%) was almost fully suppressed, but the ventilatory response to hypercapnia was unaltered (Fig. 3, A to D). We noted an increase of ~15% in the basal respiratory frequency in Epas1-null mice compared to wild-type mice, which probably reflects respiratory alterations due to anemia. Epas1-null mice showed a characteristic small transient hyperventilatory response at the onset of exposure to hypoxia (Fig. 3, A and E), suggesting incomplete extinction of HIF2α-dependent signals (21). However, once a steady-state level of hypoxia was reached, these mice showed an absence of HVR and, in some cases, a depression of breathing frequency to levels below normoxic values (Fig. 3, A and D). HIF2α-deficient mice typically exhibited another transient hyperventilatory phase during the return to normoxia (21% O2), which could reflect the reactivation of central respiratory neurons that had been depressed during the preceding uncompensated pronounced hypoxemia (31).

Fig. 3 Selective inhibition of the HVR in Epas1-null mice.

(A and B) Average time course of plethysmographic recordings illustrating the increase in respiratory frequency induced during exposure of mice to hypoxia (~10% O2) (A) and hypercapnia (5% CO2) (B) in Epas1+/+, Epas1+/−, and Epas1f/−,Cre mice studied 2 months after finishing tamoxifen treatment. (C) Respiratory frequencies measured during normoxia (Nx), hypoxia (Hx), and hypercapnia (CO2) in Epas1+/+ (Nx, n = 23; Hx, n = 23; and CO2, n = 20), Epas1+/− (Nx, n = 14; Hx, n = 14; and CO2, n = 12), and Epas1f/−,Cre (Nx, n = 26; Hx, n = 26; and CO2, n = 24) mice. (D) Time-dependent decrease in the systemic HVR (n = 7 mice per group) of Epas1f/−,Cre mice compared to their Epas1+/+ and heterozygous littermates before (60 days, Pre-TMX) and after (90, 120, and 150 days, Post-TMX) tamoxifen treatment. Note that the decrease in the HVR is already evident in heterozygous mice with partial Epas1 deficiency. (E) Average increase (basal to peak value) in respiratory frequency measured during hypoxia (10% O2 and 0.04% CO2), normalized to the values in normoxia, in wild-type (blue, Epas1+/+, n = 23), heterozygous (green, Epas1+/−, n = 14), and homozygous (red, Epas1f/−,Cre, n = 26) mice. Data are presented as mean ± SEM. Statistically significant differences compared to Epas1+/+ (*) or Epas1+/− (#) values are indicated. *P < 0.05; **P < 0.001; #P < 0.05; ##P < 0.001.

Selective inhibition of the expression of atypical mitochondrial electron transport subunits in HIF2α-deficient glomus cells

To further investigate the mechanisms underlying the loss of acute responsiveness to hypoxia in Epas1f/−,Cre mice, we measured mRNA levels of genes that are part of the defined signature metabolic profile of glomus cells (16). In addition to Epas1 and Hif1α (Fig. 1, A and B), we studied genes encoding several mitochondrial ETC subunits (Ndufa4, Ndufa4l2, Cox4i1, Cox4i2, and Cox8b), enzymes (Phd3, Pdha1, Pgk1, Pcx, and Th), and ion channels (Kcnk3, Kcnk9, Cacna1h, and Trpc5). We also measured the mRNA levels of other genes (Ndufs2, Sod1, and Sod2) that are not highly expressed in the CB but are relevant for O2 sensing and superoxide metabolism (Fig. 4A). Of the genes studied, Ndufa4l2, Cox4i2, and Cox8b, which encode the atypical mitochondrial ETC subunit isoforms that are highly expressed in the CB (15, 16), were decreased in Epas1-null CB in comparison with wild-type CB (Fig. 4A). The expression of these subunits (in particular, Ndufa4l2 and Cox4i2 mRNAs) was also partially reduced in heterozygous (Epas1+/−) mice in comparison with wild-type mice, although with the number of samples studied the differences were not statistically significant (Fig. 4B). Ndufa4 and Cox4i1 mRNAs, the most broadly expressed isoforms of Ndufa4l2 and Cox4i2, respectively (17, 32), were also present in the CB, but their levels did not change in HIF2α-deficient glomus cells. Among the enzymes studied, pyruvate carboxylase (Pcx) mRNA, encoding an anaplerotic enzyme highly expressed in CB glomus cells (16), also decreased in Epas1-deficient CBs (Fig. 4A). We found high levels of Pgk1 mRNA, a typical HIF1α-dependent gene, whose expression increases in liver (25) and lung (33) after Epas1 ablation. Using immunofluorescence with commercially available antibodies (16), we showed a decrease in the protein levels of NDUFA4L2 and COX4I2 in tyrosine hydroxylase (TH)–positive CB glomus cells from Epas1-null mice (Fig. 4, C and D). However, Epas1 inactivation did not alter the steady-state activity of MCI or MCIV in normoxic kidney cells (Fig. 4E). Cytochrome oxidase activity was also normal in lung tissue from Epas1-null mice (Fig. 4E), although Epas1 (3436), Ndufa4l2 (17), and Cox4i2 (37) mRNAs are highly expressed in this tissue.

Fig. 4 Changes in the expression of selected genes in CB cells from Epas1-null mice.

(A) mRNA levels of CB signature metabolic profile genes from Epas1+/+ and Epas1f/−,Cre mice (n = 6 replicates per group, with each replicate consisting of five mice for each genotype). (B) mRNA levels of CB atypical mitochondrial subunits from Epas1+/+ and Epas1+/− mice (n = 5 replicates per group, with each replicate consisting of five mice for each genotype). (C and D) Representative sections of the CB from Epas1+/+ (top) and Epas1f/−,Cre (bottom) mice demonstrating immunoreactivity for NDUFA4L2 (green, C) and COX4I2 (green, D) mitochondrial subunit isoforms (similar immunocytochemical studies performed in n = 3 mice for each genotype). The CB was identified by its expression of tyrosine hydroxylase (TH, red). Note the decrease in NDUFA4L2 and COX4I2 immunoreactivity in the CB from Epas1f/−,Cre mice. Scale bar, 10 μm. (E) Activity of mitochondrial complex I (MCI) and complex IV (MCIV) in kidneys of Epas1+/+ (blue, n = 10 for MCI and n = 6 for MCIV) and Epas1-null (red, n = 10 for MCI and n = 6 for MCIV) mice. MCIV activity from lungs of Epas1+/+ (blue, n = 7) and Epas1-null (red, n = 8) mice. Data are presented as mean ± SEM. *P < 0.05.

Differential effects of cytochrome oxidase subunits on acute O2 sensing

To study the role of HIF2α-dependent mitochondrial ETC subunits in CB acute O2 sensing, we used genetically modified Ndufa4l2f/− or Cox4i2f/− adult mice expressing Cre recombinase under control of the Th promoter, a gene that is highly expressed in glomus and other catecholaminergic cells. These two mice strains developed normally without decreases in weight or other overt phenotypes. Ndufa4l2f/−,Th-Cre mice (which we named TH-NDUFA4L2 mice) showed an almost complete loss of NDUFA4L2 immunostaining in CB glomus cells (fig. S4A); however, they responded normally to hypoxia, with secretion rates even slightly higher than those recorded in cells from wild-type mice (fig. S4, B to D). TH-NDUFA4L2 mice exhibited ventilatory responses to hypoxia and hypercapnia that were also similar to those of wild-type mice (fig. S4, E and F). In contrast, Cox4i2f/−,Th-Cre mice (which we named TH-COX4I2 mice) showed a decrease in Cox4i2 mRNA level in catecholaminergic tissues (CB and superior cervical ganglion; Fig. 5A) and almost complete loss of COX4I2 immunostaining in CB glomus cells (Fig. 5B). Thus, these mice exhibited functional alterations that resembled those seen in Epas1-null mice. The number of glomus cells that showed an increase in cytosolic Ca2+ (Fig. 5, C to F) and catecholamine release (fig. S5, A to E) in response to hypoxia, as well as the amplitude of the individual responses, was decreased in Cox4i2-null mice in comparison with controls. In addition, mitochondrial responses to hypoxia (an increase in NADH and IMS ROS levels) were also inhibited in COX4I2-deficient glomus cells (Fig. 5, G to L). In all COX4I2-null cells that were not responsive to hypoxia, cyanide, a selective cytochrome oxidase inhibitor and powerful CB stimulant (38), induced the accumulation of NADH and an increase in IMS ROS, indicating that MCIV was functioning normally (Fig. 5, G and J). In agreement with the single-cell data, TH-COX4I2 mice showed inhibition of the HVR, with similar features to the altered response as seen in Epas1-null mice (Fig. 6, A to C). The ventilatory response of TH-COX4I2 mice to hypercapnia was normal (Fig. 6, B and C).

Fig. 5 Loss of acute responsiveness to hypoxia in Cox4i2-deficient glomus cells.

(A) Cox4i2 mRNA levels in the CB and superior cervical ganglion (SCG) of wild-type and TH-COX4I2 mice (n = 5 replicates per group, with each CB or SCG replicate consisting of one mice of the corresponding genotype). (B) Representative immunostaining of COX4I2 and TH in CB sections from wild-type and TH-COX4I2 mice and colocalization of COX4I2 and TH. Scale bars, 10 μm. Similar immunocytochemical studies performed in n = 4 (wild type) and n = 6 (TH-COX4I2) mice. (C) Changes in cytosolic [Ca2+] elicited by hypoxia, hypercapnia, and high K+ in dispersed glomus cells from COX4I2-null mice. (D) Percentage of responding cells (exhibiting an increase in cytosolic [Ca2+] in response to hypoxia or hypercapnia) relative to the high potassium responding cells from wild-type (n = 11/4 cells/mice) and TH-COX4I2 (n = 13/6 cells/mice) mice. (E and F) Quantification of the increase in cytosolic [Ca2+] elicited by hypoxia or hypercapnia (E) and 40 mM K+ (F) in dispersed cells from wild-type (n = 9/4, 10/4, and 11/4 cells/mice, respectively) and TH-COX4I2 (n = 4/3, 11/6, and 13/6 cells/mice, respectively) mice. (G) Left: Representative recording of NADPH autofluorescence during exposure to hypoxia in dispersed glomus cells from COX4I2-null mice. Right: Response of a COX4I2-deficient glomus cell to cyanide (500 μM), indicating that the MCIV was functioning normally (n = 3 similar experiments performed with the same result). (H) Percentage of cells with a hypoxia-induced increase in NADPH autofluorescence from wild-type (n = 16/4 cells/mice) and COX4I2-null (n = 15/5 cells/mice) mice. (I) Increase in NADPH autofluorescence in wild-type (13/4 cells/mice) and TH-COX4I2 (4/3 cells/mice) responding cells. (J) Left: Representative recording of IMS ROS in COX4I2-deficient glomus cells during exposure to hypoxia. Middle: Response of a COX4I2-deficient glomus cell to cyanide (500 μM), indicating that the MCIV was functioning normally (n = 23 similar experiments performed with the same result). Right: Response to 0.1 mM H2O2 as a positive control. (K) Percentage of glomus cells that exhibited a hypoxia-induced increase in IMS ROS from wild-type (n = 20/4 cells/mice) and TH-COX4I2 (n = 33/4 cells/mice) mice. (L) Hypoxia-induced increase in IMS ROS from hypoxia-responsive glomus cells from wild-type (n = 18/4 cells/mice) and COX4I2-null (n = 7/4 cells/mice) mice. Data are presented as mean ± SEM. *P < 0.05; **P < 0.001.

Fig. 6 Selective inhibition of the HVR in COX4I2-null mice.

(A and B) Average time course of the plethysmographic recordings illustrating the increase in respiratory frequency induced during hypoxia (21 to ~10% O2) and hypercapnia (0.04 to 5% CO2) in wild-type and COX4I2-null mice. (C) Average respiratory frequency measured during normoxia (Nx), hypoxia (Hx), and hypercapnia (CO2) in wild-type (Nx, n = 12; Hx, n = 12; CO2, n = 8) and COX4I2-null (Nx, n = 12; Hx, n = 12; CO2, n = 8) mice. Data are presented as mean ± SEM. *P < 0.05.


Here, we showed that ablation of Epas1, the gene that encodes HIF2α, virtually abolished the sensitivity of CB glomus cells to acute hypoxia in adult mice. Our data suggest that HIF2α, which is constitutively expressed in the normoxic CB (15, 16, 24), is required for inducing the transcriptional program that confers acute responsiveness to changes in PO2 to glomus cells. Transgenic HIF2α overexpression elicits CB hypertrophy (39), and embryonic ablation of Epas1 inhibits CB development (20). Moreover, Epas1f/f,ESR-Cre mice receiving short-term tamoxifen treatment show reduction of HVR by ~40% (21) and inhibition of CB cell proliferation required for acclimatization to sustained hypoxia (22). These mice also have alterations in the shape of dense core vesicles in glomus cells (22), which could be related to the decrease in the size of secretory events reported here (fig. S2D). However, an absolute requirement of HIF2α in determining the acute O2-sensing properties of glomus cells and the HVR was unanticipated. Heterozygous (+/−) HIF2α mice have been studied before with conflicting results. Whereas these mice have been reported to show exaggerated CB sensitivity to hypoxia (40) and increased CB levels of Hif1α (41), subsequent studies have shown that heterozygous HIF2α (+/−) mice have partial inhibition of the HVR (21). Although we performed only few experiments on heterozygous HIF2α (+/−) mice, we did not detect any change in CB HIF1α expression (studied in CB, AM, and kidney) or alteration in organ and body weight. However, we confirmed the previously described partial inhibition of HVR in this mouse model (21). The lack of HIF1α-dependent compensation for HIF2α deficiency is consistent with previous cellular and animal studies indicating that HIF1α is not essential for CB function (20, 21, 42). Our findings further support the concept that HIF1α and HIF2α are not functionally redundant but that HIF2α is selectively expressed in some tissues to serve specialized adaptive functions (23, 3436, 43, 44).

A fundamental finding here is that ablation of the Epas1 gene resulted in the selective abolition of acute cellular and mitochondrial responses to hypoxia, whereas glomus cells maintained normal morphological and electrophysiological properties, basal NADH levels, and responses to rotenone, thereby reflecting unaltered mitochondrial metabolism and cell homeostasis. Moreover, Epas1 deficiency resulted in selective inhibition of the hypoxic mitochondrial IMS ROS signal without changes in the hypoxia-induced decrease in matrix ROS. These results support the view that mitochondrial ROS compartmentalization is specifically associated with physiological responsiveness to acute hypoxia (13, 14). The mitochondrial IMS ROS signal induced by hypoxia has been suggested to be mainly generated at the ubiquinone-binding site in MCI due to increased reduced ubiquinone/ubiquinone (QH2/Q) ratio, and it disappears in NDUFS2-deficient mice in which MCI is not assembled (13, 14).

We also identified here that the potential mechanism by which HIF2α determined the acute O2-sensing properties of glomus cells was through the transcriptional regulation of genes encoding specific atypical mitochondrial ETC subunits (NDUFA4L2, COX8B, and COX4I2). We showed that embryonic genetic ablation of at least one of these subunits (COX4I2) in glomus cells mimicked the inhibition of acute O2 sensing by the CB that is observed in HIF2α-deficient mice. However, it remains to be studied whether transgenic COX4I2 expression in HIF2α-deficient mice rescues the O2-sensing phenotype in glomus cells. The role of COX8B in glomus cell O2 sensing could not be tested because a Cox8b knockout mouse model is not yet available. Future studies must also determine whether and how conditional ablation in adulthood of one or several genes encoding the atypical ETC subunits (Ndufa4l2, Cox4i2, and Cox8b) alters acute CB O2 sensing. Ndufa4l2 is induced by hypoxia in a HIF-dependent manner (17); the Cox8b gene contains putative HIF binding sites in the promoter region (16); and Cox4i2 expression has been suggested to be increased by low PO2 (18, 19). However, the precise function of these subunit isoforms and how they can render glomus cell mitochondria highly O2 sensitive are unknown. NDUFA4L2 is a paralog of the more ubiquitously expressed NDUFA4 subunit, which was originally assigned to MCI (45), although it may be associated with MCIV and participate in the interaction between complexes (46, 47). On the other hand, COX8B and COX4I2 are tissue-specific isoforms of the ubiquitously expressed COX8A and COX4I1 MCIV subunits (32). COX8 and COX4 are integral proteins with adjacent single-transmembrane α helices running in parallel near the periphery of MCIV, and which are present at the surface of MCIV in its monomeric or dimeric forms (48). COX8 participates in hydrophobic MCI-MCIV protein-protein interactions in the MCI/MCIII2/MCIV mammalian respirasome (4951). Therefore, it is likely that COX8B and COX4I2 isoforms influence the association of MCIV with other ETC complexes and its enzymatic activity, making the rate of heme a3/CuB oxidation highly sensitive to decreases in PO2, and thus explain the function of MCIV as an O2 sensor at physiologically relevant O2 concentrations (Fig. 7). Under these circumstances, even relatively mild hypoxia would cause a backlog of electrons along the ETC and the accumulation of QH2. These effects would lead to a slowdown or even reversal of MCI (the main effector of hypoxic mitochondria) and the production of signaling molecules (NADH and ROS) that modulate ion channels (Fig. 7) (13, 14). This signaling role of MCI explains the suppression of glomus cell responsiveness to hypoxia by rotenone (38) or by genetic disruption of MCI function (Ndufs2 knockout mice) (13, 14). Several decades ago, a low-affinity cytochrome oxidase was proposed to exist in the CB, although in type II rather than in the O2-sensitive glomus cells (52), and since then, the involvement of glomus cell mitochondria in CB acute O2 sensing has been suggested in numerous studies (13, 28, 29, 38, 5355). Our “mitochondria to membrane signaling” model of CB acute O2 sensing combines concepts (the metabolic and membrane hypotheses) that had thus far evolved in parallel and provides an explanation for a fundamental biological process that has remained unknown until the present time.

Fig. 7 Model for acute O2 sensing by glomus cells.

Schematic representation of the mitochondrial-to-membrane signaling model of chemotransduction in CB glomus cells. MCIV is the oxygen sensor; MCI is the effector; and NADH and ROS are signaling molecules that modulate ion channels in the plasma membrane. IMS, intermembrane space; MCI, MCII, MCIII, and MCIV, mitochondrial complexes I, II, III, and IV, respectively; COX4I2, cytochrome c oxidase subunit IV isoform 2; COX8B, cytochrome c oxidase subunit VIIIb. The size of MCIV relative to other complexes is enlarged to facilitate explanation of the model.

An additional point worth discussing is the potential relevance of our findings to an understanding of the function or dysfunction of other organs that respond acutely to hypoxia. Although each tissue may have specific molecular adaptations, it is likely that all cells with acute O2-sensing abilities also share similar principles. For example, chromaffin cells in the adult AM, which have much less intrinsic O2 sensitivity than CB cells (13, 56), have a HIF2α and atypical mitochondrial subunit expression profile that is not as increased as in the CB (16). These cells are affected very little in embryonic Epas1-null mice (20). On the other hand, hypoxic pulmonary vasoconstriction, a sensitive acute response to hypoxia that leads to pulmonary hypertension and that, as the CB secretory response, depends on ROS production and modulation of “O2-sensitive” K+ channels (30, 57, 58), is also altered after decreased Ndufs2 expression (59) or in Cox4i2-null mice (60). However, it must be stressed that the hypoxia-induced changes in ROS in pulmonary myocytes cannot be directly compared with glomus cells because whether hypoxia increases or decreases mitochondrial ROS production in pulmonary myocytes is still under debate (30, 57, 59, 60). It is possible that the differences in the data reported are due to the different probes and methodologies used in the various settings, which may detect ROS changes in distinct cell compartments. Our data have confirmed the compartmentalization of mitochondrial ROS signals in glomus cells during hypoxia (14) and showed that they were affected differentially by HIF2α deficiency. Regardless of the nature of the hypoxic ROS signals in vascular myocytes, HIF2α dysfunction has been repeatedly associated with pulmonary hypertension (6165). Together, the data discussed here suggest that expression of defined HIF2α-regulated genes in specific organs determines their acute O2-sensing functions. The requirement of HIF2α for the correct functioning of the O2-sensing pathway in glomus cell reveals an unanticipated role of this transcription factor linking acute and chronic adaptive responses to hypoxia.


Mouse models

Mice carrying Epas1 floxed allele (The Jackson Laboratory, stock no. 008407) (23) were bred with mice carrying tamoxifen-inducible Cre recombinase transgene (66) to generate Epas1 conditional knockout mice (Epas1f/−,Cre). To obtain Ndufa4l2 conditional knockout mice, a targeting vector was designed in which exon 2 of Ndufa4l2 was flanked by two LoxP sites (fig. S6). A neomycin (Neo) cassette, containing the PGK-gb2 promoter and flanked by two flippase recognition target (FRT) sites, was introduced between exon 2 and the 3′-loxP site for selection of recombinant clones. The vector was electroporated in G4 embryonic stem (ES) cells (67), and homologous recombinant clones were screened by Southern blot. Positive clones were microinjected in C57BL/6BrdCrHsd-Tyrc blastocysts for chimera generation. Subsequently chimeras were crossed with Tg.CAG-Flp females (68) for Flp-mediated excision of the selection cassette. The resulting mice carrying Ndufa4l2 floxed allele were further propagated and mated with mice expressing Th-IRES-Cre (69) to obtain the conditional knockout model (TH-NDUFA4L2). The genotype of each mouse was determined by polymerase chain reaction (PCR). For Ndufa4l2 wild-type and flox alleles, the primers 5′-CTCTTGTCACCCTGCCTCTC-3′ (NDUFA4L2_LOXP-F) and 5′-ACTCAAGGTCTGATGCCACC-3′ (NDUFA4L2_LOXP-R) were used. For the Ndufa4l2 excised (−) allele, the primers 5′-CTCTTGTCACCCTGCCTCTC-3′ (NUOMS-REC-F) and 5′-TGTCGCCATGGCAACCCTGT-3′ (NUOMS-REC-R) were used.

Mice carrying Cox4i2 floxed allele were generated using the Cox4i2 conditional targeting vector described in fig. S7. Briefly, a 10.7-kb genomic DNA fragment flanking exons 1 and 2 of Cox4i2 was retrieved from a bacterial artificial chromosome clone (RP23:109I1) into backbone cloning vector pSP72 (Promega). Next, the loxP/FRT flanked Neo cassette was inserted 222 base pairs (bp) downstream of exon 2, and the single LoxP site, containing engineered Hind III site for Southern blot analysis, was inserted 410 bp upstream of exon 1. The targeting vector was linearized by Not I and then transfected into C57BL/6 (BF1) ES by electroporation. G418 antibiotic was used to select the positive recombinant ES clones. Afterward, the Neo cassette was removed by Flp-FRT recombination. Correctly targeted ES cells were microinjected into Balb/c blastocysts to create chimeric embryos. The resulting mice carrying Cox4i2 floxed allele were mated with mice containing Cox4i2-null allele (37) and Th-IRES-Cre (69) to generate the Cox4i2 conditional knockout model (TH-COX4I2). The presence of floxed allele in each mouse was determined by PCR (Cox4i2 wild-type and flox alleles: 3-LOX1, 5′-GAGCCAGAGCTTCTGCTTAGAGGTC-3′; SDL2, 5′-ACCTAGATCTTTGGGGCCCTCTTG-3′).

Mouse maintenance and treatments

Mice were housed at a regulated temperature (22 ± 1°C) in a 12-hour light/12-hour dark cycle with ad libitum access to food and drink. Both male and female mice were used in the current study. The strain of TH-NDUFA4L2 mice used in the current study was in C57BL/6 genetic background, whereas the other two strains were on a mixed genetic background (TH-COX4I2, 129Sv:C57BL/6:BALB/c; Epas1f/−,Cre, 129Sv:C57BL/6). To induce Cre-mediated recombination, 2-month-old Epas1 wild-type and knockout mice were fed a tamoxifen-containing diet (TAM400/CreER; tamoxifen citrate, 400 mg/kg; Envigo) for a month followed by 1.5 to 2 months on a normal diet before experiments. We followed this protocol to allow for complete degradation of HIF2α and HIF2α-induced proteins. For in vitro studies, animals were sacrificed with intraperitoneal administration of a lethal dose of sodium thiopental (120 to 150 mg/kg) before tissue dissection. As described below, dissected tissues were used for functional or immunohistochemical analyses. For RNA isolation or mitochondrial complex activity measurements, tissues were flash-frozen in liquid N2 and stored at −80°C until use. The animals were maintained before, during, and after the experiments according to European Directive 2010/63/EU regarding the use of experimental animals and other scientific purposes (Royal Decree 53/2013, 8 February). Likewise, the evaluation and authorization were carried out by the Ethics Committee of Animal Experimentation (CEEA/CEI) of Hospital Virgen del Rocío/Institute of Biomedicine of Seville.

Preparation of CB slices and dispersed cells

CB slices and dispersed glomus cells were prepared as described previously (7072). Briefly, animals were sacrificed by intraperitoneal administration of a lethal dose of sodium thiopental (120 to 150 mg/kg), and carotid bifurcations were removed and placed in Tyrode 0 Ca2+ solution containing 148 mM NaCl, 2 mM KCl, 3 mM MgCl2, 10 mM Hepes, and 10 mM glucose (pH 7.4) at 4°C. CBs were dissected and differently processed depending on the experiment. To prepare CB slices, CBs were embedded in 1% (w/v) low–melting point agarose in phosphate-buffered saline (PBS) and sectioned in 150-μm slices with a vibratome (VT1000S, Leica). Slices were then enzymatically dispersed for 5 min at 37°C with PBS containing 50 μM CaCl2, collagenase type II (0.6 mg/ml), trypsin (0.3 mg/ml), and porcine elastase (1.25 U/ml). Last, slices were washed with PBS and cultured in Dulbecco’s modified Eagle’s medium (DMEM; 0 glucose)/DMEM-F12 medium (3:1) supplemented with penicillin (100 U/ml), streptomycin (10 μg/ml), 2 mM l-glutamine, 10% fetal bovine serum (FBS), insulin (84 U/liter), and erythropoietin (1.2 U/ml) at 37°C in a 5% CO2 and 21% O2 incubator for 24 to 48 hours before use. To prepare dispersed CB cells, dissected CBs were incubated in the same enzymatic solution used for slices for 20 min at 37°C. Then, CBs were mechanically stretched with needles and incubated at 37°C for 5 min. Last, after mechanical dispersion using a pipette tip, the digestion was stopped with cold (4°C) culture medium and centrifuged for 5 min at 300g. Cells were plated on glass coverslips treated with poly-l-lysine and kept in the same culture medium as for slices (except without erythropoietin) and maintained at 37°C in a 5% CO2 and 21% O2 incubator for 24 hours.


Plethysmography used to monitor ventilation in conscious unrestricted rodents has been previously described (71). Mice were placed inside plethysmographic chambers (emka TECHNOLOGIES) and continuously perfused at a constant flow rate (1 liter/min) with a gas mixture containing 21% O2, 0.04% CO2, and 78.96% N2 (normoxia); 10% O2, 0.04% CO2, and 89.96% N2 (hypoxia); or 21% O2, 5% CO2, and 74% N2 (hypercapnia). Data were acquired with the software Iox2 (emka TECHNOLOGIES). The hermetic chambers were provided with O2 and CO2 sensors to monitor the gas composition in parallel with changes in respiratory frequency recorded by a pressure sensor during the experiment. Each animal was exposed to 2 cycles of hypoxia (7 min) followed by a final exposure to hypercapnia (2 min).

Real-time quantitative PCR

Total RNA was isolated from Epas1 wild-type and knockout mice, ~2 months after the completion of tamoxifen treatment (~5 months of age), and from COX4I2 wild-type and knockout mice (TH-COX4I2) at 2 months of age. Total RNA was isolated from CB, AM, and superior cervical ganglion using an RNeasy Micro kit (Qiagen). For the analysis of CB gene expression profile in Epas1 wild-type and knockout mice, each CB replicate was pooled from five mice to obtain enough amount of mRNA. The RNA quality was determined using an Agilent 2100 Bioanalyzer. Complementary RNA (cRNA) was then amplified from CB and AM total RNA using the GeneChip WT PLUS Reagent Kit (Affymetrix). Total RNA from lungs and kidney was isolated using TRIzol (Ambion, Life Technologies).

Total RNA (500 ng; or amplified cRNA in the case of CB and AM) was copied to complementary DNA (cDNA) using the QuantiTect Reverse Transcription Kit (Qiagen) in a final volume of 20 μl. Real-time quantitative PCRs were performed in the 7500 Fast Real Time PCR System (Life Technologies). PCRs were performed in duplicate in a total volume of 20 μl containing 1 to 4 μl of cDNA solution and 1 μl of TaqMan probe of the specific gene (Thermo Fisher Scientific). Gapdh or Ppia was also estimated in each sample to normalize the amount of total RNA (or cRNA) input to perform relative quantifications.


For immunohistochemical studies, mice were perfused first with PBS and then with 4% paraformaldehyde in PBS before tissue dissection. Carotid bifurcation was fixed with 4% paraformaldehyde in PBS for 2 hours, 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), TH (1:2500 dilution, NB300-109, Novus Biological Inc., Littleton, CO, USA), or TH (1:100 to 1:200 dilution, AB1542, Millipore, Billerica, MA, USA). This was followed by incubation with fluorescent secondary antibodies: Alexa Fluor 568 or Alexa Fluor 488 (1:500 to 1:1000 dilution, A11011 and A11015, Invitrogen, Carlsbad, CA, USA). Nuclei were labeled with 4′,6′-diamidino-2-phenylindole (DAPI). Immunofluorescence images were obtained using Nikon A1R+ confocal microscopy.

Measurement of MCI and MCIV activities

Mitochondria were isolated from kidneys and lungs of Epas1 wild-type and knockout mice, and maximum enzymatic activities of MCI and MCIV were measured using a spectrophotometer as previously described with minor modifications (13, 7375). Mitochondria were freeze-thawed three times to disrupt membranes. The resulting submitochondrial particles (10 to 40 μg) from kidneys were used to estimate NADH–ubiquinone oxidoreductase MCI activity, which was determined by measuring the rotenone-sensitive decrease in absorbance at 340 nm due to NADH oxidation at 30°C. To estimate cytochrome c oxidase MCIV activity, 4 to 12 μg of submitochondrial particles from kidneys or lungs were added to the reaction buffer containing 10 mM potassium phosphate (pH 7.0) and 1% reduced cytochrome c. The MCIV activity was determined by measuring the decrease in absorbance at 550 nm at 38°C due to the oxidation of cytochrome c. Both MCI and MCIV activities were normalized with the protein amount, as determined by Bradford assay (Bio-Rad).

Amperometric recording of single-cell catecholamine secretion in slices

Catecholamine secretion from glomus cells in CB slices was performed as described previously in our laboratory (70, 71). To test responsiveness to hypoxia, hypercapnia, or hypoglycemia, slices were transferred to a recording chamber and continuously perfused with different recording solutions (see recording solutions). Secretory events were recorded with a 10-μm carbon fiber electrode. Amperometric currents were recorded with an EPC-8 patch-clamp amplifier (HEKA Elektronik, Lambrecht/Pfaltz, Germany), filtered at 100 Hz, and digitized at 250 Hz before storage on computer. Data acquisition and analysis were performed with an ITC-16 interface (InstruTECH Corporation, NY, USA) and PULSE/PULSEFIT software (HEKA Elektronik). The secretion rate (femtocoulombs per minute) was calculated as the amount of charge transferred to the recording electrode during a given period of time. The cumulative secretion signal was the sum of charges of successive amperometric events during a given time period.

Microfluorimetric measurements

For the cytosolic Ca2+ measurements, CB dispersed cells were incubated in DMEM/F-12 (without FBS) containing 4 μM Fura 2-AM (F-1221, Molecular Probes, Bleiswijk, The Netherlands) for 30 min at 37°C in a 5% CO2 incubator. To perform the experiments, a coverslip with Fura 2-AM loaded cells was placed on a recording chamber mounted on the stage of an inverted microscope (Nikon Eclipse Ti) equipped with epifluorescence and photometry. Alternating excitation wavelengths of 340 and 380 nm with emission wavelength of 510 nm were used to obtain the F340/F380 ratio (13, 76). Cytosolic Ca2+ signals were digitized at a sampling interval of 500 ms. Experiments were performed at ~35°C. NADPH microfluorimetric measurements were performed in CB dispersed cells using a nonratiometric protocol. NADH was excited at 360 nm and measured at 460 nm.

To measure mitochondria ROS production in CB slices, different roGFP probes were used, being targeted to either the mitochondrial IMS or the mitochondrial matrix [reviewed in (14, 30) and references therein]. The constructs were incorporated into adenoviral vectors used to transfect CB slices in 2 ml of complete culture medium containing 2 μl of viruses from a stock (1.1 × 1012 viral particles/ml; ViraQuest Inc., North Liberty, IA). Slices were kept in the incubator for >48 hours to allow the roGFP expression before the recording of ROS production (13, 14). RoGFPs have two fluorescence excitation peaks at 400 and 484 nm, with emission at 535 nm, which allow the ratiometric measurements and to examine reversible changes of the redox state.

Patch-clamp recordings in dispersed CB glomus cells

Macroscopic ionic currents were recorded from dispersed mouse glomus cells using the whole-cell configuration as adapted in our laboratory (70, 72). Patch-clamp pipettes (2 to 3 megohms) were pulled from capillary glass tubes (Kimax, Kimble Products) with a horizontal pipette puller (model P-1000, Sutter Instrument) and fire-polished with a microforge (MF-830, Narishige). Voltage-clamp recordings were obtained with an EPC-7 amplifier (HEKA Elektronik). The signal was filtered (3 to 10 kHz), digitized with an analog/digital converter (ITC-16 InstruTECH Corporation), and sent to the computer. Data acquisition and storage were performed using the PULSE/PULSEFIT software (HEKA Elektronik) at a sampling interval of 20 μs.

Recording solutions

For amperometric and microflourimetric measurements, dispersed cells and slices, which had been incubated overnight in culture medium, were transferred to a recording chamber and continuously perfused with a control solution containing 125 mM NaCl, 4.5 mM KCl, 23 mM NaHCO3, 1 mM MgCl2, 2.5 mM CaCl2, 5 mM glucose, and 5 mM sucrose at ~35°C. In 40 mM K+ solution, NaCl was replaced equimolarly with KCl. The “normoxic” solution was bubbled with a gas mixture of 20% O2, 5% CO2, and 75% N2 (O2 tension, ~145 mmHg). The “hypoxic” solution was bubbled with 5% CO2 and 95% N2 to reach an O2 tension of ~10 to 15 mmHg in the chamber. The “hypercapnic” solution was bubbled with 20% CO2, 20% O2, and 60% N2. Osmolality of solutions was ~300 mosmol/kg with pH 7.4.

Macroscopic Ca2+, Na+, and K+ currents were recorded in dialyzed glomus cells. Solutions used to record whole-cell Na+ and Ca2+ currents contained the following: external solution: 140 mM NaCl, 10 mM BaCl2, 4.7 mM KCl, 10 mM Hepes, and 10 mM glucose (pH 7.4) (osmolality, 300 mOsm/kg); internal solution: 130 mM CsCl, 10 mM EGTA, 10 mM Hepes, and 4 mM adenosine triphosphate (ATP)–Mg (pH 7.2) (osmolality, 285 mOsm/kg). The external solution used to record whole-cell K+ currents was the same control solution used for amperometric and microflourimetric analysis, and the internal solution contained 80 mM K+ glutamate, 50 mM KCl, 1 mM MgCl2, 10 mM Hepes, 4 mM ATP-Mg, and 5 mM EGTA (pH 7.2).

Quantification and statistical analysis

Unless otherwise specified, the data were presented as mean ± SEM, with the number (n) of experiments indicated. Normality was tested with the Shapiro-Wilk test. Data from two groups were analyzed with either t test or paired t test. Data with multiple groups were analyzed by either analysis of variance (ANOVA) or repeated-measures ANOVA followed by post hoc Tukey’s test. Statistical analyses were performed using SigmaPlot (12.0). P < 0.05 was considered statistically significant.


Fig. S1. Phenotypic characterization of conditional Epas1f/−,Cre mice.

Fig. S2. Structure and function of CBs from Epas1-null mice.

Fig. S3. Electrophysiological properties of CB glomus cells from wild-type and Epas1-null mice.

Fig. S4. Glomus cell responsiveness to acute hypoxia and HVR in NDUFA4L2-null mice.

Fig. S5. Loss of secretory activity in response to acute hypoxia in COX4I2-null mice.

Fig. S6. Generation of TH-NDUFA4L2 mice.

Fig. S7. Generation of TH-COX4I2 mice.


Acknowledgments: We thank IBiS staff for technical assistance. Funding: This research was supported by the Spanish Ministries of Science and Innovation and Health (SAF2012-39343 and SAF2016-74990-R) and the European Research Council (ERC Advanced Grant PRJ201502629). Author contributions: A.M.-D., P.O.-S., L.G., and J.L.-B. designed the experiments and wrote the manuscript. A.M.-D., P.O.-S., L.G., O.C., P-G.-F., and V.B.-H. performed the experiments and analyzed data. J.A., M.H., L.I.G., N.W., and N.S. provided essential experimental materials and contributed to the discussion of the data. J.L.-B. coordinated the project. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article