Oxygen regulation of breathing through an olfactory receptor activated by lactate

Abstract

Animals have evolved homeostatic responses to changes in oxygen availability that act on different timescales. Although the hypoxia-inducible factor (HIF) transcriptional pathway that controls long-term responses to low oxygen (hypoxia) has been established, the pathway that mediates acute responses to hypoxia in mammals is not well understood. Here we show that the olfactory receptor gene Olfr78 is highly and selectively expressed in oxygen-sensitive glomus cells of the carotid body, a chemosensory organ at the carotid artery bifurcation that monitors blood oxygen and stimulates breathing within seconds when oxygen declines. Olfr78 mutants fail to increase ventilation in hypoxia but respond normally to hypercapnia. Glomus cells are present in normal numbers and appear structurally intact, but hypoxia-induced carotid body activity is diminished. Lactate, a metabolite that rapidly accumulates in hypoxia and induces hyperventilation, activates Olfr78 in heterologous expression experiments, induces calcium transients in glomus cells, and stimulates carotid sinus nerve activity through Olfr78. We propose that, in addition to its role in olfaction, Olfr78 acts as a hypoxia sensor in the breathing circuit by sensing lactate produced when oxygen levels decline.

Extended Data Figure 1. Model of oxygen sensing by the carotid body and the mitochondrion

a, Anatomy and blood supply of the carotid body (CB). CB is located bilaterally at bifurcation of carotid artery (CA) in the neck. Its location can be variable as well as the source of its blood supply, which can come from branches of nearby internal and external carotid, occipital, pharyngeal arteries. Blood flows through fenestrated capillaries close to clusters of Type I glomus cells and drains from CB into jugular vein (JV) on ventral side. Panel adapted from ref. b, Cellular organization of CB. CB is composed of several cell types, including Type I glomus cells (red) that sense changes in blood oxygen and are organized in clusters, Type II sustentacular cells (blue) that resemble neuroglia and surround glomus cell clusters, carotid sinus nerve (CSN) fibers that innervate glomus cells, and endothelial (E) and smooth muscle cells (not shown) that form the tortuous vasculature. Panel adapted from ref. c, Oxygen-sensing respiratory circuit. The primary chemoreceptor for blood oxygen is the carotid body. A decrease in PaO₂ of arterial blood from normoxia (100 mmHg) to hypoxia (<80 mmHg) stimulates glomus cells to signal the carotid sinus nerve, a branch of glossopharyngeal nerve (GN) with cell bodies in petrosal ganglion (PG). Axons of the GN terminate in nucleus tractus solitarius (NTS) in brainstem, a site of many converging afferent inputs. The signal from NTS is transmitted to ventral respiratory group (VRG) that includes preBötzinger complex, a region essential for respiratory rhythm generation. From VRG, neurons project to premotor and motor neurons that innervate respiratory muscles, such as diaphragm and intercostal muscles. In addition to carotid body, vagus nerve afferents can also contribute to respiratory behaviors under specialized conditions. The vagus nerve innervates heart and lung and oxygen-sensitive cells of aortic body⁷⁰. Panel adapted from ref. d, A current model of acute oxygen sensing by carotid body. A decrease in PaO₂ in blood causes a decrease in O₂ concentration inside carotid body glomus cells. This causes a decrease in activity of mitochondrial electron transport chain (ETC) and changes in other putative oxygen-sensing pathways, such as oxygen-sensitive K⁺ channels^,, heme oxygenase, AMP kinase, and hydrogen sulfide signaling. These changes are hypothesized to converge on oxygen-sensitive K⁺ channels, which close in hypoxia and depolarize the plasma membrane. Depolarization then opens voltage-gated Ca⁺² channels, leading to an increase in intracellular calcium that stimulates transmitter release to carotid sinus nerve to increase breathing. Mitochondria of carotid body cells are highly sensitive to hypoxia compared to other tissues, as assayed by imaging of mitochondrial membrane potential, NADH levels, and spectral properties^,–. Drugs and mutations that inhibit the ETC mimic the effect of hypoxia on carotid body activity and breathing^,–. e, Regulation of lactate production by oxygen. In normoxia, pyruvate produced by glycolysis is transported into mitochondria and efficiently used in Krebs cycle to supply electrons to ETC to produce ATP. In hypoxia, lack of oxygen to act as the final electron acceptor limits electron transport, causing pyruvate to build up and become converted to lactate^,,. The ETC poison cyanide inhibits the heme a3 subunit of cytochrome c oxidase to prevent transfer of electrons to oxygen, leading to lactate accumulation even in presence of adequate oxygen. Cytosolic lactate accumulation results in transport of lactic acid (lactate and H⁺) out of the cell by monocarboxylate transporters^,. In normoxia, lactate concentrations in blood, tissue, and tissue interstitium are 1–5 mM^,.

Extended Data Figure 2. RNA sequencing and whole genome microarrays detect Olfr78 transcripts enriched in the carotid body

a, Histogram of frequency of genes for different levels of expression enrichment in carotid body (CB) relative to adrenal medulla (AM) by RNA sequencing. Log₂(CB/AM) values are shown, with data binned for every log₂ interval of 1.0 centered at integers. b, Plot of log₂ values of reads per kilobase per million (RPKM) in CB and AM of all 1,126 olfactory receptor (OR) genes annotated in RefSeq shown in alphanumerical order. The five OR genes expressed at RPKM > 2 (dashed line) are indicated. Samples that had no transcripts are plotted at a value of −7.1, just below the smallest RPKM value for ORs. Data presented in Supplementary Table 1. c, Comparison of expression levels of >34,000 genes in adult mouse CB and AM by whole genome microarrays. Plot shows log₂ of the ratio for CB relative to AM of the fluorescence intensity values for the 45,000 probe sets. The three probe sets for Olfr78 transcripts are indicated (circles). Expression of Olfr78 was significantly different between CB and AM for all three probe sets (P<0.05 by ANOVA with false discovery rate control). d, Histogram of the frequency of genes for different levels of expression enrichment in CB relative to AM in microarray data. Log₂(CB/AM) values are shown, with data binned for every log₂ interval of 1.0 centered at integers. The three probe sets detecting Olfr78 mRNA (arrows) confirmed the RNA-seq data (a, Fig. 1a, b, and Extended Data Table 1) showing Olfr78 among the mRNAs most highly enriched in carotid body. Mouse CB Olfr78 expression is consistent with previous microarray data^,. a-d, n=3 cohorts of 10 animals each. Data as mean. e, Genomic locus showing the large cluster of ~160 Class I OR genes on chromosome 7, with region encoding MOR18 subfamily (Olfr78, Olfr558, and Olfr557) expanded below. We did not detect transcripts in either tissue for Olfr557, which lies adjacent to Olfr558 in the cluster, or for the intervening (Olfr33, Olfr559) and intronic (Olfr560) ORs. Clusters of genes encoding globins, Trims, and USP proteins are also found with this OR cluster. Large box, coding sequence; arrowhead, coding orientation; small box, non-coding exons.

Extended Data Figure 3. Olfr78 and Olfr558 expression in tissues in the oxygen-sensing circuit

Expression of Olfr78 reporter in heterozygous (a) and homozygous (b-d) Olfr78-GFP-taulacZ reporter animals. a-c, Sections of carotid bifurcations stained for GFP (Olfr78 reporter; green), tyrosine hydroxylase (TH; red), and DAPI (nuclei; blue). a, Section of CB showing co-expression of reporter GFP and TH in glomus cells. Monoallelic expression would predict that only half of TH-positive cells express the reporter. Arrowheads, clusters of glomus cells expressing both GFP and TH. b, c, Sections of the same carotid bifurcation. Panels on right show close-ups of boxed region (petrosal ganglion, PG). No GFP-positive cells were found in petrosal ganglion. TH-positive nerve fibers (arrowheads) and cell bodies were found in glossopharyngeal nerve (GN) and petrosal ganglion. Dashed circle indicates vagus nerve (VN). NG/JG, nodose/jugular ganglia. d, X-gal staining of a brain sagittal section. Reporter expression (blue) was restricted to olfactory bulb (arrowhead) in this section and complete brain serial sagittal sections. Anterior, right; dorsal, up. e-h, Olfr558 expression in a knockout/reporter mouse in which the Olfr558 coding region is replaced with lacZ encoding β-galactosidase. e, Olfr558 reporter expression in blood vessels of CB and SCG by X-gal staining. Heterozygous Olfr558^+/lacZ samples showed the same pattern of staining (data not shown). f-h, CB sections immunostained for β-galactosidase (Olfr558 reporter; green), TH (red), with DAPI counterstain (blue) in f, and for β-galactosidase (green) and CD31 (red) in g or smooth muscle actin (red) in h. Scale bars, 100 µm (a, b-right, c-right, f-h), 200 µm (b-left, c-left), 500 µm (e), and 2 mm (d).

Extended Data Figure 4. Tidal volume and minute ventilation of Olfr78^−/− mutants exposed to hypoxia and hypercapnia

Whole body plethysmography of unrestrained, unanesthetized Olfr78^+/+ control and Olfr78^−/− mutant littermates (as in Fig. 2). a, b, Tidal volume (TV) and minute ventilation (MV) of animals exposed to hypoxia. n=9 (+/+), 8 (−/−) animals. c, d, TV and MV of animals exposed to hypercapnia. n=4 (+/+), 5 (−/−) animals. Data as mean ± s.e.m. *P<0.05, ***P<0.001 by paired t test.

Extended Data Figure 5. Physiological responses of Olfr78^−/− mutants to hypoxia in vivo

a-f, Arterial blood gas measurements of Olfr78^−/− control and Olfr78^−/− mutant animals exposed to hypoxia. PaO₂ (a), PaCO₂ (b), and pH (c) values of blood collected from the right carotid artery of anesthetized Olfr78^+/+ control and Olfr78^−/− mutant animals exposed to normoxia (21% O₂) and hypoxia (10% O₂) for 3 min. Oxygen saturation (sO₂,d), [HCO₃⁻] (e), and base excess of extracellular fluid (BE_ecf,f) calculated from PaO₂ (a), PaCO₂ (b), and pH (c) values. n=4 (+/+, 21% O₂), 5 (−/−, 21% O₂), 4 (+/+, 10% O₂), 6 (−/−, 10% O₂) animals. g, Body temperature of unanesthetized Olfr78^+/+ control and Olfr78^−/− mutant littermates in room air (21% O₂) and exposed to hypoxia (10% O₂) for indicated times. n=4 (+/+), 6 (−/−) animals. h-j, Metabolic values measured by indirect calorimetry of unanesthetized Olfr78^+/+ control and Olfr78^−/− mutant littermates exposed to normoxia (21% O₂) and hypoxia (10% O₂) for 10 min. n=4 (+/+), 6 (−/−) animals. Data as mean ± s.e.m. *P<0.05 by unpaired t test.

Extended Data Figure 6. Carotid body chemosensory responses assayed by carotid sinus nerve activity

a, b, Raw discharge frequency (extracellular recording) of carotid sinus nerves from Olfr78^+/+ control and Olfr78^−/− mutant animals at time 0 (a) and 9 minutes (b) after the change in gas bubbling the perfusion buffer from 95% O₂/5% CO₂ to 95% N₂/5% CO₂. c, d, Carotid sinus nerve activity of an Olfr78^+/− nerve 9 minutes after the change in gas to 95% N₂/5% CO₂ (c) and 2 minutes later after addition of 7.5 µM tetrodotoxin (TTX) while still bubbling 95% N₂/5% CO₂ (d). Scored action potentials are marked by filled circles. e-h, Time course of carotid sinus nerve activity in the Olfr78 genotypes indicated scored using Spike2 software (e, g) or by hand (f, h) and showing mean ± s.e.m (e, f) or individual (g, h) values. The residual responses of Olfr78^−/− nerves to hypoxia were more apparent when scored by hand. n=6 (3 +/+, 3 +/−), 5 (−/−) animals. *P<0.05, **P<0.01, ***P<0.001 by unpaired t test. Olfr78^+/+ and Olfr78^+/− recordings were not significantly different from each other at any time point, except for time=11 min, by unpaired t test (P>0.05). i, j, Time course of raw discharge of carotid sinus nerves from Olfr78^+/+ control and Olfr78^−/− mutant animals in response to acetate (30 mM, 5 min), propionate (30 mM, 5 min), and lactate (30 mM, 5 and 10 min), and pH 7.0 (5 min) scored using Spike2 software (i) or by hand (j). Recovery times were 15 min between acetate, propionate, and lactate, and at least 30 min between lactate and pH 7.0. To minimize the contribution of endogenous hypoxic signals, the superperfusion buffer in the chamber was maintained at hyperoxic conditions (PO₂=625 mmHg). n=5 (+/+), 5 (−/−) animals. Data as mean ± s.e.m. *P<0.05, **P<0.01 by unpaired t test.

Extended Data Figure 7. Lactate activates Olfr78 expressed in HEK293T cells and increases acutely in blood in hypoxia in vivo

a, b, HEK293T cells transfected with empty vector pCI (a) or pCI-Rho-Olfr78 (b) and RTP1S (OR transport protein) and cytoplasmic GFP (co-transfection marker) plasmids. Transfected cells were stained before fixation to detect Rho-tagged Olfr78 (anti-Rho; red) on the cell surface. GFP (transfection marker, green); DAPI (nuclei, blue). Bar, 100 µm. c, Quantitation of cells expressing GFP and Rho as percentage of DAPI-positive cells in fields shown in a and b. n=164 (pCI), 108 (pCI-Rho-Olfr78) cells. Data as percent ± standard error of percentage. d, Dose-response curves for propionate, acetate, and chloride compared to lactate in activation of Olfr78 in transfected HEK293T cells as in Fig. 4a. n=8 (propionate), 12 (acetate), 4 (chloride), and 12 (lactate) wells. Data as mean ± s.e.m. By analysis of variance (ANOVA), all chemicals except chloride (P=0.309) showed significant difference (P<0.001). e, Dose-response curves as in c except cells were transfected with empty vector (pCI). ANOVA showed no significant difference (P>0.05) for any chemical. f, EC₅₀, 95% confidence interval of EC₅₀, and relative maximal activation values from fitted curves in c. ND, not determined due to lack of curve fitting to data. g, Structures of the short-chain fatty acids. h, Lactate concentrations in blood collected from tail artery of restrained, unanesthetized Olfr78^+/+ control and Olfr78^−/− mutant littermates exposed to hypoxia (10% O₂) for 4–5 min. Values for animals in normoxia (21% O₂) are likely to be an overestimate of baseline concentrations due to greater restraint required to immobilize animals in normoxia. n=5 (+/+), 6 (−/−) animals. Data as mean ± s.e.m. *P<0.05 by paired t test.

Extended Data Figure 8. Calcium imaging of responses of carotid body glomus cells to chemosensory stimuli

a, Carotid body (CB) of a Th-Cre; ROSA-tdTomato adult immunostained for the Cre-dependent reporter tdTomato (red) and TH (green) to show glomus cells^,, and counterstained with DAPI (nuclei, blue). tdTomato labeled glomus cells. b-e, Tissue preparations for calcium imaging of CBs from TH-Cre; ROSA-GCaMP3 animals that express the calcium indicator GCaMP3 selectively in glomus cells^–. b, DIC image of whole mount carotid bifurcation with GCaMP3 fluorescence pseudocolored green. c, High magnification, two-photon image of boxed region in b. d, DIC image of CB tissue slice with GCaMP3 fluorescence pseudocolored green. e, Two-photon image of CB slice in d. Inset shows glomus cell marked by asterisk at higher magnification. GCaMP3 fluorescence was seen in cytoplasm and excluded from nucleus of glomus cells. Bars, 100 µm (a), 200 µm (b), 50 µm (c-e). f-i Time course of calcium responses of individual glomus cells to hypoxia, lactate, and cyanide. Whole mount CBs were exposed sequentially to hypoxia (40–50 mmHg), lactate (30 mM), and cyanide (2 mM). Interval between data points is ~2 minutes, the time required to acquire a stack of images through the CB, excluding the 2 minute ramp times between stimuli. All glomus cells analyzed (n=42 cells) responded strongly to cyanide. Fluorescence traces shown are the 29 individual glomus cells that responded to both hypoxia and lactate, arranged in order of decreasing initial fluorescence intensity. The other 13 glomus cells responded to either hypoxia (9 cells) or lactate (4 cells). Multiple data points for buffer or stimuli were averaged to generate the data presented in Fig. 4c. Background colors match bar colors in Fig. 4d.

Figure 1. Olfr78 is expressed in carotid body glomus cells

a, Expression of 26,728 genes in adult mouse carotid body (CB) and adrenal medulla (AM) by RNA sequencing. Log₂ values of number of aligned reads per 10⁷ aligned reads generated. b, OR genes highly expressed in CB and/or AM. X, fold enrichment (CB/AM). a, b, n=3 cohorts of 10 animals each. Data as mean (a) or mean ± standard error of the mean (s.e.m., b). *P<0.05, **P<0.01 by paired t test by cohort. c-l, Expression of Olfr78 knock-in reporter mouse. c-f, X-gal staining (blue) detects taulacZ (β-galactosidase) reporter activity. c, Adrenal gland showing adrenal medulla (AM). Reporter not expressed. d, Carotid bifurcation (dorsal view, superior cervical ganglion (SCG) removed). Reporter expressed in CB (dashed circle) and sporadic blood vessels (arrowhead). e, f, Transverse section of carotid bifurcation (e) and close-up (f). g, h, Immunostaining of CB sections. Olfr78 reporter expression (GFP; green) co-localized with CB glomus cell marker (tyrosine hydroxylase, TH; red; g) but not endothelial cell marker (CD31; red; h). TH is also expressed in nerve fibers and SCG. DAPI (blue), nuclei. i-k, X-gal stained carotid bifurcations (ventral view). i, CB (dashed circle) innervated by carotid sinus nerve (filled arrowhead), a branch of glossopharyngeal nerve (open arrowhead). j, “Miniglomerulus” (MG; dashed circle) innervated by glossopharyngeal nerve (arrowhead). k, Petrosal ganglion (arrow), nodose/jugular ganglia (arrowhead). Reporter not expressed. l, Olfr78 reporter expression (X-gal staining) during CB development. Filled circles, robust expression; open circles, not detected. Bars, 500 µm (c-e, i-k) and 100 µm (f-h).

Figure 2. Ventilatory responses of Olfr78 null mutants to hypoxia and hypercapnia

Respiratory rate (RR), tidal volume (TV), and minute ventilation (MV=RR*TV) by whole body plethysmography of unrestrained, unanesthetized Olfr78^+/+ and Olfr78^−/− littermates exposed to hypoxia (a, b) or hypercapnia (c, d). a, b, Respiratory rate in hypoxia (a) and hypoxic response (b) as percent change in hypoxia (10% O₂) versus control (21% O₂). n=9 (+/+), 8 (−/−) animals. c, d, Respiratory rate in hypercapnia (c) and hypercapnic response (d) as percent change in hypercapnia (5% CO₂) versus control (0% CO₂). n=4 (+/+), 5 (−/−) animals. Data as mean ± s.e.m. *P<0.05, **P<0.01, ***P<0.001 by paired t test (a, c) or unpaired t test (b, d).

Figure 3. Olfr78 mediates carotid body oxygen sensing

a, b, CB sections from Olfr78^+/+ control (a) and Olfr78^−/− knockout allele (b) in which GFP-IRES-taulacZ replaces Olfr78 coding region. GFP (green), tyrosine hydroxylase (TH; red), and DAPI (blue). Mutant CB shows normal organization. c, Quantification of CB TH-positive cells. n=8 (+/+), 14 (−/−) CBs. Data as mean ± s.e.m. P=0.454 by unpaired t test. d-g, Transmission electron micrographs of Olfr78^+/+ (d, e) and Olfr78^−/− (f, g) CBs. e,g, close-ups of boxed regions. Both wild type and mutant glomus cells have large nuclei (asterisks), large dense core vesicles (open arrowheads), and small clear core vesicles (filled arrowheads). Bars, 100 µm (a, b), 600 nm (d, f), and 200 nm (e, g). h, i, CB responses to hypoxia (h) and low pH (i) assayed by carotid sinus nerve discharge frequency (impulses/sec) of Olfr78^+/+ and Olfr78^+/− controls (blue) and Olfr78^−/− mutants (red). h, Hypoxia response as superperfusate changed from bubbling 95% O₂/5% CO₂ to 95% N₂/5% CO₂ (t=0 min) and back to 95% O₂/5% CO₂ (t=8 min, arrow). Gray line, representative time course of PO₂ in recording chamber. Discharge frequency of control nerves began increasing at PO₂=80 mmHg (t=6 min) and peaked at PO₂=60 mmHg (t=9 min). n=6 (3 +/+, 3 +/−), 5 (−/−) animals. i, Shift from pH 7.4 to pH 7.0. n=5 (+/+), 5 (−/−) animals. Data as mean ± s.e.m. *P<0.05, **P<0.01, ***P<0.001 by unpaired t test.

Figure 4. Lactate activates Olfr78 and carotid body sensory activity

a, Lactate activation of Olfr78 expressed in HEK293T cells detected by dual reporter assay (see Methods). Cells transfected with pCI (empty vector, gray) or pCI-Rho-Olfr78 (epitope-tagged Olfr78, black). n=12 (4 transfected wells per plate on 3 plates performed on separate days). ANOVA for pCI data, P=0.478. b, Arterial blood lactate of anesthetized Olfr78^+/+ and Olfr78^−/− littermates exposed to hypoxia for 3 min. n=4 (+/+, 21% O₂), 4 (+/+, 10% O₂), 4 (−/−, 21% O₂), 6 (−/−, 10% O₂) animals. c, d, Calcium response of GCaMP3-expressing glomus cells exposed to hypoxia (PO₂=40–50 mmHg), lactate (30 mM), and cyanide (2 mM), in whole mount and slice. c, Time course of stimuli. Every data point differs from previous point (P<0.001 by paired t test). d, GCaMP3 fluorescence change (F₁-F₀)/F₀ in percent. n=42 (whole mount, all stimuli), 29 (slice, hypoxia and cyanide), 22 (slice, lactate) cells. All changes differ from pre-stimulus (P<0.001 by paired t test). e, Olfr78^+/+ and Olfr78^−/− CB response to 30 mM lactate for 10 min, assayed by carotid sinus nerve activity. n=5 (+/+), 5 (−/−) animals. a-e, Data as mean ± s.e.m. *P<0.05, **P<0.01 by unpaired t test. f, Model of oxygen sensing by Olfr78. In normoxia, pyruvate is efficiently used in Krebs cycle, supplying electrons to mitochondrial electron transport chain (ETC) to produce ATP. In hypoxia, lack of oxygen as final electron acceptor slows ETC, causing pyruvate to accumulate. Pyruvate is converted to lactate, which is secreted and binds Olfr78 on CB glomus cells, increasing intracellular calcium and transmitter release to afferent nerves to stimulate breathing.