Introduction

The carotid body (CB) is a paired sensory organ located at the carotid bifurcation. It is composed of heterogenous cell clusters known as glomeruli, within which glomus (or type I) cells are the most abundant. Glomus cells are in close contact with a dense network of fenestrated capillaries as well as glial (type II) cells, and profusely innervated by fibers of the glossopharyngeal nerve. These neuron-like excitable cells are multimodal chemoreceptors activated by changes in several physicochemical parameters in arterial blood, such as hypercapnia, acidemia, hypoglycemia or lactate. However, their unique sensitivity to changes in oxygen (O2) tension distinguishes them as essential sensory receptors for physiological adaptation to reduced O2 levels (hypoxia), caused by environmental conditions or pathological states that impair gas exchange in the lungs1,2,3,4. In response to hypoxia, glomus cells rapidly release neurotransmitters thereby activating sensory fibers that impinge on brainstem centers to induce hyperventilation and increased heart output. Acute O₂ sensing is an intrinsic property of chemoreceptive glomus cells, though autocrine and paracrine interactions can modulate CB’s output to the brain5,6,7.

The CB originates from neural crest progenitors and shares many histological and neurochemical characteristics with the adrenal medulla and sympathetic ganglia8. However, unlike these or other organs of the peripheral nervous system, the CB demonstrates extraordinary morphological and physiological plasticity. It can undergo severalfold reversible enlargement, increasing the number of O₂-sensitive glomus cells during sustained hypoxia9,10,11. This remarkable adaptability enables most mammals, particularly humans, to acclimatize to high altitudes where O₂ tension is significantly lower than at sea level12,13. The adult CB functions as a stem cell niche, housing not only mature cells but also multipotent progenitors and immature neuroblasts14,15. Neurotransmitters and neuropeptides released by O₂-sensitive glomus cells under hypoxia stimulate not only sensory fibers mediating the acute cardiorespiratory reflexes, but also the proliferation and differentiation of progenitor cells. This process leads to the generation of ecto-mesenchymal cells, such as endothelial cells and vascular smooth muscle cells, as well as maturation of neuroblasts into new neuron-like glomus cells15,16,17,18.

While the basic electrophysiological and neurosecretory properties supporting the sensory function of glomus cells are well established, the molecular mechanisms underlying O₂ sensing have only recently been elucidated. Studies over the past few years have shown that responsiveness of glomus cells to hypoxia relies on the constitutive expression of HIF2α and a set of HIF2α-dependent genes, which define a mitochondrial-based pathway for acute O2 sensing and signaling19,20,21,22,23,24,25. However, the manner in which the O₂-sensing genetic profile is expressed within the chemoreceptor cell lineage, as compared to the various cell types present within the adult CB, remains largely unexplored. Understanding the molecular mechanisms responsible for glomus cell activation by hypoxia and CB plasticity is of significant physiological and medical relevance. CB overactivation has been associated with excessive sympathetic outflow, a hallmark of prevalent conditions such as sleep apnea, type II diabetes, or chronic heart failure26,27,28,29,30. Furthermore, investigating the mechanisms underlying CB growth could provide valuable insights into the pathogenesis of paragangliomas, which arise from the proliferation and malignant transformation of cells within the O₂-sensing lineage31,32,33,34,35,36,37. In this study, we present a transcriptomic map of the various cell types within the CB. This analysis confirms the presence of multipotent, neural crest-derived progenitor cells and provides a detailed characterization of the O₂-sensing genetic profile in the neuronal glomus cell lineage. Additionally, we describe changes in the CB gene expression pattern during acclimatization to chronic hypoxia.

Results

Genetic cell clustering in the adult carotid body stem cell niche

We performed a droplet-based single-cell RNA sequencing (scRNA-seq) on enzymatically dispersed CB cells from adult (3-month-old) rats. After quality control of the sequencing data, cells were clustered using the Seurat workflow and clusters were assigned to major cell types. This unbiased genetic map of the adult CB revealed nine distinct cell types —glomus cells, neuroblasts, neural crest (NC)-derived progenitor cells, Schwann cells, pericytes, fibroblasts, smooth muscle cells, endothelial cells, and blood cells— each of them identified by the expression of unique, highly selective, markers (Fig. 1A and B; Fig. S1A; Table S1). For example, the glomus cell cluster, a dopaminergic neuron-like population, was characterized by the specific expression of Slc18a1 (encoding vesicular monoamine transporter 1, VMAT1)38, Cdh9 (encoding synaptic-specific cadherin 9)39, and Apoc4 (encoding the lipid-processing enzyme apolipoprotein C4)40 (Fig. 1B). Glomus cells, and to a lesser extent neuroblasts, expressed more genes and a larger number of transcripts than other cell classes in the CB. Notably, the percentage of mitochondrial DNA-encoded transcripts expressed by glomus cells was also significantly higher than in other cell types (Fig. 1C). These findings are consistent with the classical view of the CB as one of the tissues with relative higher blood perfusion and O2 consumption in the body41,42,43. They also support the O2 sensing and signaling role of mitochondria in glomus cells (see below).

Fig. 1
figure 1

Single-cell transcriptomic analysis of the normoxic carotid body. (A) UMAP embedding of 9,658 single carotid body cells from rats maintained in standard normoxic conditions (21% O2 tension atmosphere). Transcriptionally similar cell clusters are color-coded and labeled by cell type. (B) Violin plots displaying the expression levels of three of the most highly expressed gene markers for each cell type within the normoxic carotid body. (C) Violin plots displaying transcriptomic generic features for each cluster: top, number of genes being expressed; center, total number of transcripts; bottom, percentage of transcripts corresponding to mitochondrial genes. (D) Violin plots showing the expression of specific sympathoadrenal markers in carotid body cells. Schwann cell and blood cell clusters have been excluded for clarity. (E) UMAP plots depicting the expression of selected sympathoadrenal markers in mesenchymal, neuronal, and progenitor cell clusters within the normoxic carotid body.

Full size image

To confirm the accuracy of the cluster classification and identify markers distinguishing progenitors from mature cells in the two main sympathoadrenal cell lineages —neuronal and mesenchymal14,16—, we analyzed the expression of genes known to be associated with NC-derived sympathoadrenal cells44. Markers specific to progenitors and each cell lineage are highlighted in Fig. 1D. The dot plots in Fig. 1E illustrate the separation of the three cell classes with representative genes for each: Plp1 for multipotent, undifferentiated progenitors, Phox2b for neuronal cells, and Prrx1 for mesenchymal cells44. Genes encoding transcription factors critical for sympathoadrenal neuron development, such as Phox2b and Hand244, were selectively expressed in glomus cells and neuroblasts. In contrast, genes characteristic of mesenchymal derivatives in the sympathoadrenal lineage, such as Prrx1 and Ebf144, were uniquely expressed in the mesenchymal-differentiated cells, including smooth muscle cells, pericytes and fibroblasts, within the CB (Fig. 1D and E).

Multipotent progenitors

To further characterize the multipotent NC-derived CB progenitors, we identified the markers most prominently expressed in this cell population. In addition to the previously mentioned Plp1 gene, which encodes proteolipid protein 1, a key structural component of myelin sheaths, we observed high expression of additional markers predominantly restricted to the progenitor population (Fig. S1B). These markers included the cell adhesion molecule Chl145, cadherin 19 (Cdh19)46, the sodium channel Scn7a47,48, and the growth factor Egfl849. The 100 most significantly expressed genes in the NC progenitor population, compared to other CB cell types, were associated with glial and mesenchymal-like phenotypes, as well as neuron-supporting functions (Fig. S1C). Interestingly, the Gfap transcript, whose protein product has classically been used to identify glia-like sustentacular type II cells in the CB14, although expressed at relatively low levels, appeared almost exclusively in the progenitor cell cluster (Fig. S1D). Similarly, S100b, another specific type II cell marker50, was also preferentially expressed in progenitor cells (Fig. 1D, Fig. S1E). These observations indicate that, as suggested in previous studies14,18, type II cells in the adult CB express a transcriptome compatible with multipotent NC-derived progenitors.

We performed in vitro differentiation studies to demonstrate the ability of genetically identified CB progenitors to generate mature cells of the sympathoadrenal lineage. To this end, we examined the expression of cell surface markers in progenitor cells (Fig. 2A). Among these, cell adhesion molecule 2 (Cadm2) and HNK-1, a surface epitope synthesized by the beta-1,3-glucuronyltransferase 2 (B3gat2) enzyme (Fig. 2A), were selected for sorting by flow cytometry (Fig. 2B). Antibodies against these surface epitopes, CADM2 and HNK-1, enabled the sorting of NC progenitors using distinct gates (Fig. 2B). Sorted cells were subsequently used to generate neurospheres (NS) in floating conditions (Fig. 2C). As previously shown14, these neurospheres contained a large core with blebs of differentiating cells (white arrowheads in Fig. 2C). Differentiation assays on these NS replated onto adherent substrate demonstrated the ability of CADM2 + and HNK-1 + progenitors to produce neurospheres with neuronal (tyrosine hydroxylase-positive; TH +) or mesenchymal (smooth muscle actin-positive, SMA +) derivatives, as well as multipotent neurospheres with both (Fig. 2, D and E). It has been shown previously that the TH + neuronal CB derivatives obtained in vitro exhibit an O2-sensing glomus cell phenotype, characterized by Ca2+ and K+ currents and the ability to release dopamine in response to hypoxia14. Taken together, the transcriptomic data provide a detailed characterization of the diverse cell classes within the CB, including the identification of NC-derived sympathoadrenal progenitors and their differentiated derivatives in both mesenchymal and O2-sensing neuronal lineages.

Fig. 2
figure 2

Identification of multipotent neural crest-derived progenitor cells in the carotid body. (A) Dot plot showing the expression of progenitor cell surface markers at the population level. (B) Flow cytometry plot illustrating the expression of two cell surface markers (CADM2 and HNK-1) in carotid body cells, including the gating strategy used for cell sorting. (C) Neurosphere (NS) formation by progenitor cells. Left: Percentage of sorted cells, based on each surface marker, from the total carotid body cell preparation. Center: Light microscopy images of NS formed after 10 days in culture using each cell surface marker. Arrowheads indicate typical surface differentiating blebs characteristic of carotid body NS14. Scale bar: 100 µm. Right: Efficiency of NS formation, calculated as the percentage of sorted cells that form NS. (D) Immunocytochemistry on NS cultured in adherent substrate, using antibodies against tyrosine hydroxylase (TH) and smooth muscle actin (SMA) to classify NS types. Mesenchymal NS contained only SMA + cells, neuronal NS contained only TH + cells, and multipotent NS contained both. Cells were counterstained with DAPI. Scale bars: 50 µm. Insets: 25 µm. (E) Quantification of NS subtypes obtained by sorting with each cell surface marker.

Full size image

O2-sensing cell lineage: Neuroblasts and mature glomus cells

In addition to the in vitro experiments demonstrating the generation of TH + and O2-responsive glomus cells from multipotent CB progenitors, in vivo lineage tracing studies have identified stem cells in the adult CB capable of producing dopaminergic glomus cells14. Additionally, functional studies have suggested the presence of immature, mildly TH + and weakly O2-sensitive, neuroblasts in the adult CB that can divide once or twice and differentiate in response to hypoxia, thereby rapidly increasing the population of non-dividing mature glomus cells15,51. We focused on the genetic characterization of progenitors, neuroblasts, and glomus cells within the O2-sensing CB cell lineage. As shown previously (see Fig. 1), these three cell types were clearly defined in our unbiased CB cell map and could be distinctly separated by specific markers (Fig. 3A; Fig. S2A and B). Although neuroblasts and glomus cells were genetically closer to each other compared to progenitor cells, they were further distinguished by the expression of specific genes. For example, Adora2a (encoding adenosine receptors) and Chga (encoding chromogranin, a chromaffin granule protein) were almost exclusively expressed in glomus cells. In contrast, genes associated with autonomic development (Ascl1)52 or with cell proliferation/differentiation (Epcam)53 were selectively expressed in neuroblasts (Fig. 3B).

Fig. 3
figure 3

Characterization of the sympathoadrenal neuronal lineage in the carotid body. (A) UMAP plot highlighting the three cell clusters that define the sympathoadrenal neuronal lineage within the carotid body. (B) UMAP plots showing the expression of selected genes to distinguish glomus cells from neuroblasts. (C) UMAP embedding plot illustrating the dopaminergic differentiation score for each cell. The color gradient represents the expression intensity of the differentiation gene signature. Progenitor cells exhibit the lowest differentiation scores, while glomus cells show the highest, with neuroblasts in an intermediate state. (D) Violin plot visualization of the dopaminergic differentiation score, with significant p-values indicated. (E) UMAP plots displaying genes highly correlated with the differentiation trajectory, either positively (upregulated; top) or negatively (downregulated; bottom). (F) Correlation analysis between the dopaminergic differentiation score and other gene expression signatures, including "Neural Crest Cell Differentiation" (from the Gene Ontology database) and "O₂-sensing" (defined in this work).

Full size image

Subsequently, we defined a “differentiation score” for each cell, based on the similarity between expressed genes and a global “dopaminergic gene ontology signature” (see Table S2), utilizing the UCell algorithm54. As expected, the most differentiated cell cluster corresponded to glomus cells, the least differentiated to progenitor cells, and neuroblasts occupied an intermediate position (Fig. 3C and D). This result is consistent with our previous functional studies demonstrating that electrophysiological and neurosecretory properties of glomus cells begin to manifest in neuroblasts15. We also used the differentiation score to identify genes that positively or negatively correlate with the differentiation axis (Fig. S2C). For example, the Rgs5 or Scg2 genes, associated with G-protein signaling or chromaffin vesicle formation, respectively55,56, were highly correlated with the differentiation score and preferentially expressed in glomus cells. Interestingly, Scn7a and Grik2 genes, encoding Na+ channel and glutamate ionotropic receptor subunits, respectively, had negative correlation with the differentiation score and were almost absent in glomus cells but highly expressed in progenitor cells (Fig. S2C; Fig. 3E). Correlation analyses using comprehensive gene signatures (see Table S2) demonstrated a highly significant negative correlation of the differentiation score with the category “Neural crest cell differentiation”, indicating that the high differentiation potential of NC progenitors diminishes along the differentiation axis (Fig. 3F, left).

O2-sensing gene signature in carotid body cells

Once the progenitor-to-glomus cell differentiation pathway was established, we focused on the transcriptomic characterization of the O2 sensing properties of CB cells. We defined an “O2-sensing gene signature” (see Table S2) composed of a selected group of nine genes known to be either essential for acute CB O2 sensing or involved in the activation and neurosecretory response to hypoxia in glomus cells20,22,23,24,25. The O2-sensing gene signature included Epas1 (coding Hif2α) and Hif2α-dependent mitochondrial genes Cox4i2, Ndufa4l2, Higd1c, and Pc20,23. Additionally, it encompassed genes encoding K+ channels, either background (Kcnk3 and Kcnk9) or voltage-gated (Kcnq3), that are highly expressed in glomus cells (22,25; this work), and contribute to their resting potential and responsiveness to hypoxia57,58,59,60. The O2-sensing gene signature also included the Nalcn gene, encoding a Na+ channel that determines the resting potential and excitability of glomus cells61,62. The expression levels of each of these genes in different CB cell types are illustrated in Fig. 4A, where Hif1a expression is shown for comparison. The analysis of the correlation between the differentiation score and the O2-sensing gene signature demonstrates a positive and highly significant correlation between the O2-sensing function and the differentiation axis (see Fig. 3F, right), indicating that the acquisition of sensitivity to hypoxia is a key feature of dopaminergic neuronal differentiation in the CB.

Fig. 4
figure 4

Transcriptional characterization of O₂-sensing capacity in carotid body cells. (A) Dot plot showing the expression of genes identified as particularly relevant for O₂-sensing and the neurosecretory response to hypoxia in glomus cells (O₂-sensing gene signature) across cell clusters in the carotid body. Hif1a is included for comparison with Epas1/Hif2α. (B) Violin plot representing the O₂-sensing score for each carotid body cell cluster, with a black line indicating the average signature expression in each population. Significant differences in O₂-sensing scores are highlighted by p-values. (C) UMAP plot displaying the O₂-sensing score across neuronal, mesenchymal, and progenitor cell clusters. (D) UMAP plots illustrating the expression of selected genes previously implicated in O₂-sensing. Pearson correlation coefficients (Cc) with the O₂-sensing score are indicated for each gene. Genes in the lower row show little to no correlation with the O₂-sensing transcriptional signature.

Full size image

To further evaluate the potential O2-sensing capacity of CB cells, we utilized also here the UCell algorithm54, to calculate an “O2-sensing score” which is proportional to the combined level of expression of the genes comprising the O2-sensing signature in the CB cell populations (see Materials and Methods). Our analysis revealed that while glomus cells have the highest O2-sensing score, followed by neuroblasts, this parameter is very low in NC progenitors and the other cell types in the CB (Fig. 4A–C). An exception to this general trend was smooth muscle cells, which displayed a relatively high and statistically significant O2-sensing score in comparison with progenitors (Fig. 4B and C). The expression in smooth muscle cells of genes associated with glomus cell acute O2 sensing, such as Cox4i2 or Ndufa4l2 (Fig. 4A) or the relatively high signature score of this cell class (Fig. 4B and C) aligns with the O2-sensing role of myocytes in acute hypoxic responses of the vasculature63,64,65,66.

Next, we conducted a linear correlation analysis between levels of expression of individual genes in the CB cell population and the O2-sensing score previously assigned to each CB cell type. In this analysis, a higher correlation coefficient signifies an increased probability that the corresponding gene contributes to the cell’s O2-sensing capacity and its responsiveness to hypoxia. We selected the genes with the highest correlation coefficient, all exceeding 0.5, and grouped them into four main categories representing well-established functions of glomus cells: “Ion Channels and Transporters”, “Mitochondrial Genes”, “G Protein-Related Genes”, and “Calcium Signaling Genes” (see Tables S3–S6). The 10 genes in each category with the highest correlation coefficients are shown in Fig. S3, A-H. Although the specific roles of many of these genes in the context of CB O2 sensing are unstudied, it is remarkable that many of them are associated with the acquisition of a mature neuronal phenotype in glomus cells—a critical feature for hypoxia responsiveness. Notably, the ten genes identified as having high correlation coefficients to differentiation scores (Fig. S2C), being two of them (Higd1c and Ndufa4l2) part of the O2-sensing signature group (see Fig. 4A), were found to be highly expressed in O2-sensing glomus cells (Fig. S3I). In agreement with generalized high expression of mitochondrial DNA-encoded genes in glomus cells (see Fig. 1C), Mt-nd1, Mt-co1, Mt-co2, —necessary for maintaining an active oxidative phosphorylation and a mitochondrial-based O2 sensing mechanism— were expressed in glomus cells at much higher level than in any other cell type (Fig. S3C, Table S4). These genes were also highly correlated with the O2 sensing score (Fig. S3D; Table S4). Furthermore, a plot comparing gene correlation coefficients to differentiation score vs. to O2-sensing score revealed a strong positive linear correlation (r = 0.88) (Fig. S3J). These findings underscore a close association between neuronal differentiation and the acquisition of the O2-sensing phenotype.

Remarkably, components of the O2-sensing signature, such as Epas1 (Hif2α), Cox4i2, Higd1c or Pc, not only appeared preferentially expressed in glomus cells and neuroblasts, but their expression levels also showed a high correlation coefficient with the O2-sensing score to which they contributed (Fig. 4D, top row). In contrast, genes coding enzymes, such as cystathionine-g-lyase (Cth), hemeoxygenase 2 (Hmox2), NADPH oxidase 1 (Nox1) or AMP kinase (Prkaa1)—which are not directly related with mitochondrial function but have been proposed to be involved in CB O2 sensing in previous studies67,68,69,70—were not preferentially expressed in the O2-sensing cell lineage and showed no correlation with the O2-sensing score. (Fig. 4D, bottom row; Fig. S4). Taken together, our genetic analyses demonstrate that glomus cells are transcriptionally specialized for O2 sensing, while neuroblasts exhibit significant transcriptional similarity, suggesting they are on the path to acquiring full hypoxia sensitivity. The O2-sensing properties of glomus cells appear closely linked to their maturation into a neuronal phenotype. Additionally, smooth muscle cells express components of the O2-sensing signaling pathway, highlighting their role in regulating the diameter of CB vessels in response to hypoxia.

Cellular and genetic modifications induced by chronic hypoxia in the carotid body niche

In CB samples from rats exposed to hypoxia (10% O2 atmosphere) for 10 days, we obtained 18,586 cells which were processed for transcriptomic analysis after quality control. This cell number almost nearly double that in the normoxic sample (9,658), likely due to higher cell density and larger size of hypoxic CBs. Using the same clustering procedure as applied to the normoxic sample, we constructed a genetic map of the cell types present in the hypoxic CB parenchyma. The hypoxic CB contained the same cell types identified in normoxic samples, with the sole exception of neuroblasts, which completely disappeared from the genetic map, suggesting their conversion into differentiated mature glomus cells (Fig. 5A and B). While the absolute numbers of all CB cell types increased under hypoxia, only glomus cells and blood cells showed a significant increase in their proportions within the tissue (Fig. 5B).

Fig. 5
figure 5

Single-cell transcriptomic analysis of the hypoxic carotid body. (A) UMAP embedding of 18,586 single cells from hypoxic rat carotid bodies. Transcriptionally similar cell clusters are color-coded and labeled by cell type. (B) Comparison of cell cluster frequencies between carotid bodies obtained from rats maintained in standard conditions (21% O2 atmosphere: normoxia, Nx) or after 10 days in a chamber with 10% O2 tension (hypoxia, Hx). Note how neuroblasts disappear in hypoxia, while glomus and blood cells clearly increase their proportions within the tissue. (C) Violin plots showing the expression of specific sympathoadrenal markers in hypoxic carotid body cells. Schwann cell, endothelial cell and blood cell clusters have been excluded for clarity. (D) UMAP plots depicting the expression of selected sympathoadrenal markers (see Fig. 1B) in mesenchymal, neuronal, and progenitor cell clusters within the hypoxic carotid body.

Full size image

The markers used in normoxic samples to identify NC progenitors and mature cells of the two main cell lineages—neuronal and mesenchymal— (see Fig. 1D and E) were also selectively expressed in the corresponding cell populations of the hypoxic CB (Fig. 5C and D). Furthermore, the most highly expressed markers in hypoxic NC progenitor cells (Fig. S5A and B) were similar to those observed in their normoxic counterparts (Fig. 2A and Fig. S1B). Gene ontology analysis of enriched markers revealed similar categories in both conditions, including glia-like and mesenchymal-like phenotypes, as well as neuron-supporting function (Figs. S1C and S5C). Therefore, while hypoxia induced substantial changes in the genetic map of the CB, the cell types and sympathoadrenal phenotype of the CB multipotent niche were preserved. The notable exception was the conversion of neuroblasts into mature glomus cells.

To facilitate comparison between normoxic and hypoxic CB cells, we generated a genetic map merging the two data sets. The embedded genetic maps revealed that only hypoxic glomus cells exhibited relevant and sufficient qualitative transcriptomic changes to be clearly distinguishable in the UMAP representation from their normoxic counterparts (Fig. 6A–C). Under hypoxia, all CB cell types (including glomus cells) displayed generalized reduction in overall gene expression (Fig. S6A) and fewer transcripts (Fig. S6B) compared to normoxic conditions. This observation suggests the activation of an energy conservation mechanism in response to hypoxia. A notable exception to this trend was the upregulation of DNA-encoded mitochondrial genes, which exhibited increased expression levels in glomus cells and, to a lesser extent, in endothelial cells (Fig. S6C).

Fig. 6
figure 6

Merged analysis of O2-sensing capacity between normoxic and hypoxic carotid body cells. (A) UMAP representation of normoxic cell clusters. (B) UMAP representation of hypoxic cell clusters. Note that a few glomus cells and neuroblasts did not complete maturation and their transcriptomic profiles remained as in the normoxic condition. (C) Merged representation of normoxic and hypoxic cell clusters, highlighting glomus cells as the only cell class that significantly changes its transcriptome in response to hypoxia. (D) Violin plot representing the O₂-sensing score for each cell cluster, with a black line indicating the average signature expression in each population. Selective significant differences in O₂-sensing capacity are highlighted by p-values. (E) Violin plots depicting an increased expression in O2-sensing relevant genes under hypoxia, such as Cox4i2, Higd1c and Pc. (F) Dot plot displaying the expression of Olr59 in carotid body cells. Note how the expression of Olr59 decreases along the O2-sensing neuronal lineage, specially under hypoxia. (G) Violin plot depicting the expression of Olr59 in the neuronal lineage.

Full size image

Maturation of glomus cells under hypoxia

Although chronic hypoxia induced important transcriptomic changes in glomus cells (Fig. 6A–C, see below), the expression of O2-sensing signature genes in the ranking of most expressed genes was only slightly diminished in comparison with the normoxic situation (Fig. 6D; Fig. S6D). In fact, most O2-sensing genes, particularly Cox4i2, Higd1c, and Pc, were clearly upregulated by hypoxia (Fig. 6E; Fig. S6E). Remarkably, the expression of the gene encoding the olfactory receptor 59 (Olr59), the rat ortholog for the mouse Olfr78 gene, which is high in neuroblasts and lower in glomus cells, was drastically reduced in hypoxic glomus cells (Fig. 6F). This olfactory receptor is selectively expressed in the neuronal cell lineage (Fig. 6F and G) in a Hif2α-dependent manner20,71. Reduction of Olr59 expression along the differentiation pathway further supports its role in driving neuroblast differentiation71 (see Discussion).

A detailed description of the general transcriptomic changes induced by chronic hypoxia in glomus cells (see Fig. S7, A-C) is beyond the scope of this study. However, it is worth noting that both increases and decreases in gene expression levels were observed (see Ref.72). Interestingly, among the most overexpressed genes in hypoxic glomus cells compared to normoxic counterparts were Hbb-a2 and Hbb-b1, encoding hemoglobin alpha and beta chains, respectively, which exhibited elevated expression across all hypoxic cell types (Fig. S7C). Gene ontology analysis of differentially expressed genes in hypoxic versus normoxic glomus cells revealed that both upregulated and downregulated categories were predominantly associated with the acquisition of a neuronal phenotype (Fig. S7D and E). Indeed, a “differentiation gene signature”—defined by eight selected genes representative of mature glomus cells associated with their O2-sensing and neurosecretory phenotype (see Table S2)— was markedly upregulated in hypoxic glomus cells, especially when compared to immature neuroblasts (Fig. S7F and G). Together, these findings strongly suggest that the transcriptional changes in hypoxic glomus cells further enhance their neuronal phenotype and hypoxia-responsiveness.

Discussion

In this work, we have performed a comprehensive transcriptomic characterization of the cell types present in the adult CB, the main arterial chemoreceptor and acute O₂-sensing organ in the body. We have genetically identified the various cell classes present within the organ and have specified a population of multipotent progenitors which includes the glia-like type II cells typically observed in the adult CB. As suggested by previous studies, these CB progenitors can form neurospheres and differentiate into mature cells of the two main sympathoadrenal lineages (neuronal and mesenchymal)14. Our cell sorting studies suggest that adult CB progenitors likely have a heterogeneous phenotype and origin. However, this aspect is not further analyzed here and will be explored in detail in future work. The current study focuses on two key features of the CB that are essential for understanding its function under standard physiological conditions and during acclimatization to hypoxia. These are the characterization of CB neuroblasts, which support rapid adaptation to hypoxia, and the definition of the genetic profile that sustains hypoxia responsiveness in cells of the O2-sensing chemoreceptor lineage.

An important outcome of the current study is the genetic identification of CB neuroblasts, whose existence was suggested by our previous histological and functional data15,51. In adults, CB neuroblasts are quiescent, immature TH + cells, with a genetic profile, in terms of dopaminergic differentiation, mitochondrial gene expression, and O2 sensitivity, intermediate between multipotent CB progenitors and glomus cells. The transcriptomic data indicate that responsiveness to hypoxia and the completion of the dopaminergic neurosecretory phenotype are the two major functions that distinguish CB glomus cells from neuroblasts. It is remarkable that some genes with known functional relevance are highly expressed in neuroblasts and down-regulated in glomus cells. In this regard, it is notable the high level of expression of Chrna7 (encoding an α-subunit of a nicotinic cholinergic receptor) in neuroblasts (see Fig. S3, A and B), which explains the pro-differentiation effect of acetylcholine (ACh) on neuroblasts maintained in culture15. Interestingly, a high density of ACh receptors in “small” glomus cells—renamed later as CB neuroblasts15—was described in early morphometric CB studies73. ACh is a transmitter synthesized by glomus cells and released during hypoxia74, which may activate neighboring neuroblasts to promote their proliferation and differentiation into new glomus cells15. Similarly, the Olr59 gene, which encodes an atypical G-coupled olfactory receptor expressed outside the olfactory epithelium, is a highly abundant transcript in neuroblasts. Olr59 has lower expression level in mature glomus cells and is further down-regulated by chronic hypoxia (see Fig. 6F and G). This is an unexpected finding since, at least in mouse CB, Olr59 (or in this case its ortholog Olfr78) gene expression is markedly reduced in Hif2α-deficient mice in comparison with controls71. Given that the level of Hif2α mRNA (Epas1) in glomus cells —already high under normoxic conditions— remains practically unchanged during hypoxia (see Fig. S6E), it appears that other hypoxia-inducible factors counteract the effect of Hif2α and repress Olr59 expression in chronically hypoxic glomus cells. While Olr59 is not directly involved in O2-sensing71,75, it may act as a driver for neuroblast maturation/differentiation, and its activity might be spared once neuroblasts reach a differentiated state (see Ref.71). A pro-differentiation effect of Olr59 expression has also been recently described for enterochromaffin cells in the mouse colon76.

The genetic map of CB cells shows that the population of neuroblasts completely disappears after 10 days in hypoxia in parallel with a marked increase in the number of mature glomus cells. This finding supports our previous cell cytometry experiments using HNK-1 glycoepitope to sort neuroblasts, in which we showed that hypoxia can induce neuroblast proliferation and differentiation to glomus cells in less than 48 h15. Similar neuroblast to glomus cell conversion has been observed during repeated exposure to intermittent hypoxia as it occurs in cases of sleep apnea51. Taken together, these new observations explain previous histological studies using cell cycle markers (such as PCNA or Ki67) and anti-TH antibodies, which suggested that some TH + CB cells could proliferate77,78,79. Our data demonstrate the existence in the CB of quiescent, pre-differentiated TH + neuroblasts, which can rapidly differentiate into glomus cells in response to hypoxia. This “fast neurogenesis” is a process without precedent in any other structure of the adult peripheral nervous system. It has probably evolved to increase CB responsiveness to lowering O2 tension before hypertrophy of the organ takes place, thus accelerating acclimatization to hypoxia11,12,13,15.

Another main finding in this paper is the characterization of the genetic profile underpinning O2-sensing and responsiveness to hypoxia in mature CB glomus cells. Previous studies have analyzed CB cells at the gene expression level to establish a genetic profile linked to their O2-sensing capability (summarized in Ref.80). Most of these studies have concentrated on the expression of ion channels and receptors to identify markers that distinguish glomus cells from O₂-insensitive tissues81,82,83,84. More recent work employing single-cell RNA sequencing has revealed abundant expression of G protein-coupled receptors, various types of ion channels, and hypoxia-inducible factor 2α (Epas1) in neonatal glomus cells25. Furthermore, a comparative transcriptomic analysis between adult glomus cells, adrenal medulla chromaffin cells, and sympathetic neurons of the superior cervical ganglion highlighted that glomus cells differ from O₂-insensitive cells in several key metabolic features, including the expression of mitochondrial enzymes and atypical subunit isoforms of the electron transport chain complexes22. Some of the genes coding these proteins were also found to be highly expressed in a transcriptomic study of neuroepithelial cells in zebrafish gills, which contain O2-sensitive cells, similar to CB glomus cells85. Functional studies on genetically modified mice have demonstrated that CB acute O2 sensing relies on mitochondrial function21,86 and that constitutive high expression of Epas1 (encoding Hif2α) and Hif2α-dependent mitochondrial proteins are essential for acute O2 sensing by glomus cells19,20,23,24.

Herein, we show that the percentage of mitochondrial DNA-encoded transcripts expressed in cells of the neuronal O2-sensing lineage is about 1.5–3 times higher than in other CB cell types. We defined a CB acute O2-sensing gene signature including Epas1, the MCIV genes Cox4i2, Higd1c, and Ndufa4l2 (coding tissue-specific atypical MCIV subunit isoforms), Pc (coding pyruvate carboxylase, highly expressed in glomus cell mitochondria), and Kcnk3, Kcnk9, Kcnq3, and Nalcn, (coding ion channels that mediate hypoxia-induced changes in glomus cell excitability). Remarkably, among the different cell types in the CB, this O2-sensing gene signature appeared almost exclusively expressed in glomus cells and to a lesser measure in neuroblasts, and showed a highly significant close correlation with dopaminergic differentiation of the O2-sensing cell lineage. It is notable that apart from glomus cells and neuroblasts, vascular myocytes —known to be intrinsically O2 sensitive and responsive to acute hypoxia (see Ref.65)—are the only additional CB cell type with a relatively high level of expression of mitochondrial DNA-encoded genes and that also expresses some components of the O2-sensing gene signature.

The transcriptomic analysis presented here strongly supports the mitochondria-to-membrane signaling pathway of CB acute O2 sensing described in recent functional studies on genetically modified mice21,23,87. Our methodological approach has also allowed us to directly investigate the expression pattern of several molecules previously proposed to be necessary for acute O2 sensing in glomus cells67,68,69,70. Among these, genes coding for enzymes involved in sulfur metabolism (Cth and Cbs), hemeoxygenase-2 (Hmox2), NADPH oxidases (Nox1, Nox4), or subunits of AMP kinase (Prkaa1, Prkab1, Prkab2) were not selectively expressed in the O2 sensing cell lineage but appeared uniformly expressed at relatively low levels across the various CB cell types. These findings make any essential role of these genes in acute O2 sensing unlikely, although they may contribute to other aspects of glomus cell function. They also explain why CB responsiveness to hypoxia is unaltered in most reports published using genetically modified mouse models lacking any of these genes88,89,90,91,92.

Genetic changes induced by chronic hypoxia (10 days) were observed in all CB cell types, with glomus cells being the most affected. Hypoxia led to a decrease in transcript count and overall gene expression, in all CB cell types, suggesting an adaptive response to reduce metabolic expenditure in cells experiencing O₂ shortage (see Ref.72,93). However, mitochondrial DNA-encoded genes—already expressed at higher levels in glomus cells compared to other cell types—were further upregulated by chronic hypoxia. This effect occurred in parallel with the upregulation of Hif2α-dependent nuclear genes encoding atypical mitochondrial complex subunit isoforms or enzymes, such as Cox4i2, Higd1c, and Pc, which are known to participate directly in O₂ sensing22,23,24. The “differentiation gene signature” —defined by eight selected genes representative of mature glomus cells primarily associated with their neurosecretory phenotype— was more highly expressed in glomus cells than in neuroblasts and was also upregulated by hypoxia. Since the genes responsible for acquiring the mature catecholaminergic phenotype in sympathoadrenal cells depend on Hif2α94,95,96, these findings further underscore the central role of Hif2α in the maturation of glomus cells, enabling them to perform their critical functions such as acute O2 sensing, catecholamine synthesis, and G-protein signaling. Taken together, our data indicate that the sustained CB afferent drive to the brain during acclimatization to hypoxia depends not only on an increased number of O₂-sensitive chemoreceptor cells due to neuroblast proliferation and differentiation but also on the enhanced mitochondria-dependent responsiveness of individual glomus cells to hypoxia.

In conclusion, this study describes the transcriptomic organization of the CB stem cell niche (see Fig. 7). It demonstrates the existence of multipotent progenitors capable of generating cells of the sympathoadrenal mesenchymal and neuronal lineages. Differentiation along the neuronal lineage—giving rise to neuroblasts and glomus cells—is characterized by a progressive increase in the number of transcribed genes and transcripts, as well as expression of mitochondrial DNA-encoded genes and nuclear genes encoding atypical mitochondrial electron transport subunits and enzymes. The genetic evolution from progenitor to chemoreceptor glomus cells is paralleled by the appearance and enhancement of the dopaminergic neurosecretory phenotype and responsiveness to hypoxia. Exposure to chronic hypoxia markedly accelerates the conversion of neuroblasts into mature glomus cells and upregulates the signature gene program responsible for mitochondrial-based acute O2 sensing and signaling (see scheme in Fig. 7). These data provide a robust genetic framework to explain the cellular and molecular physiology of the CB, particularly its essential role as an acute O₂ sensor and a key contributor to acclimatization to chronic hypoxia. These findings may also be relevant to the development of novel therapeutic strategies for prevalent cardiorespiratory metabolic diseases associated with CB-dependent sympathetic overactivation7,11, as well as paragangliomas/pheochromocytomas34,35,36.

Fig. 7
figure 7

Scheme of the transcriptomic organization of the carotid body stem cell niche. Representation of cellular relationships characterized within the adult carotid body stem cell niche, focusing on the neuronal lineage. The transcriptomic features depicted below the diagram represent the differences observed among the transcriptomes of the three principal cell types within the neuronal lineage. Specifically, these correspond to transcriptomic differences between progenitor cells, immature neuroblasts, and mature glomus cells, following a gradient of increasing intensity. These features reflect the molecular transitions associated with neuronal differentiation and maturation within the adult carotid body parenchyma. Moreover, the figure illustrates how hypoxic stimulation further promotes this differentiation process, amplifying transcriptomic divergence and thereby enhancing the acquisition of a mature glomus cell phenotype characterized by neurosecretory activity and oxygen sensitivity.

Full size image

Materials and methods

Animals

Experiments were performed using 8–12 week-old (~ 300 g weight) Wistar rats. Male and female animals were used in every experiment and no differences were found between them. Rats were housed and treated according to the animal care guidelines of the European Community Council (2010/63/EU). The Animal Research Committee at the University of Seville approved all procedures. 6 rats were used for the scRNA-seq analysis (3 normoxic and 3 hypoxic), and 6 more rats were used for the in vitro neurosphere experiments. All animals were purchased from Charles River. Rats were anesthetized via intraperitoneal administration of a ketamine (Ketolar, Pfizer; 80 mg/kg) and xylazine (Rompum, Bayer; 10 mg/kg) mixture. Euthanasia was subsequently performed through an overdose of sodium pentobarbital. Animal use is reported according to the ARRIVE guidelines.

In vivo hypoxic treatments

Rats were chronically exposed to 10% O2 environment for 8 days, using a hermetic isobaric chamber with controlled O2 levels and continuous monitoring of CO2, humidity, and temperature (Coy Laboratory Products). Animals were housed in standard rodent cages, with ad libitum access to pellet food and water, and within a 12 h light/dark cycle room. Control rats were similarly housed in ambient air outside the chamber.

Dissociation of carotid body cells

Rat carotid bodies (CBs) were dissociated by enzymatic and mechanical treatment. Briefly, CBs were dissociated in PBS solution containing 0.6 mg/ml collagenase type II (Sigma), 0.35 U/ml porcine elastase (Calbiochem), 0.3 mg/ml trypsin (Sigma), and 10 μl/ml of 5 mM CaCl2 solution, for 25 min at 37ºC, using a shaker15. Once dissociated, single cells were resuspended in a staining solution composed by (for 50 ml): 44 ml L15 medium (Gibco), 0.5 ml penicillin/streptomycin (Gibco), 0.5 ml 1 M HEPES buffer (Gibco), 0.1 g BSA (Sigma), and 5 ml distilled and deionized water. Dispersed cells were used for either single cell transcriptome analysis or for flow cytometry sorting and neurosphere formation in culture.

Flow cytometry sorting

Flow cytometry sorting was performed in a MoFlo Astrios sorter (Beckman Coulter). Normoxic rats were anaesthetized and their carotid bodies extracted and dissociated to obtain a solution of dispersed cells (see above). For flow cytometry staining, we used as primary antibodies mouse IgM anti-HNK-1 (CD57) (1:200; BD Pharmingen), and rabbit anti-Cadm2 (Cadm2) (1:100; Thermo Fisher). As secondary antibodies we used Alexa Fluor 488-conjugated donkey anti-rabbit IgG (1:500; Molecular Probes), and AlexaFluor 568-conjugated goat anti-mouse IgM (1:200; Molecular Probes). Dispersed CB cells were resuspended in 200 μl of staining solution (see above) and were incubated with the primary antibody for 30 min on ice, followed by incubation with the secondary antibody for 30 min on ice. DAPI (1:1,000, Sigma) was used as viability marker. Sorted cells were directly recollected in culture medium (see below).

Neurosphere formation in cell culture

Culture medium was previously described in our laboratory15 as complete medium, composed by: D-MEM:F-12 (Gibco), 15% FBS (Gibco), 1% N2 supplement (Gibco), 2% B27 supplement (Gibco), 1% penicillin/streptomycin (Gibco), 10 ng/ml recombinant human bFGF (R&D Systems), 20 ng/ml recombinant human IGF-1 (R&D Systems), and 20 ng/ml recombinant human EGF (R&D Systems). Neurospheres were obtained from sorted cells by placing them in ultra-low binding plates (Costar, Corning) in standard incubators with CO2 control. After 10 days in culture, neurospheres were counted, measured and re-plated onto coated culture plates (Nunc Multidish 24-well, Thermo Scientific). The coating solution contained: 78% acetic acid (17 mM, Sigma), 20% poly-D-lysine (0.5 mg/ml, Sigma) and 2% collagen type I solution (Sigma). Finally, after 48 h in adherent conditions, neurospheres were subjected to immunocytochemical staining.

Immunocytochemical methods

Attached neurospheres were fixed with 4% PFA for 20 min at room temperature, preblocked for 1 h, incubated with primary antibodies overnight at 4 °C, and with secondary antibodies for 2 h at room temperature. Blocking solution was composed by (for 50 ml) 43.5 ml PBS (Gibco), 0.1% BSA (Sigma), 1.5 ml Triton-X (10%, Sigma) and 5 ml goat serum (Gibco). Primary antibodies: rabbit anti-tyrosine hydroxylase (TH) (1:1,000, Novus Biologicals), and mouse anti-smooth muscle actin (SMA) (1:10,000, Sigma). Secondary antibodies: Phycoerythrin (PE)-conjugated goat anti-rabbit IgG (1:200, Jackson ImmunoResearch), and Alexa Fluor 488-conjugated donkey anti-mouse IgG (1:500, Molecular Probes). Nuclei were counterstained with DAPI (1:1,000, Sigma). Images were acquired in an inverted fluorescence microscope Olympus IX-71.

Single-cell RNA-seq library preparation and preprocessing

Single-cell RNA sequencing (scRNA-seq) was performed with the 10 × Genomics Chromium Single Cell 3′ Kit (v.3) according to the standard protocol. Libraries were sequenced on an Illumina NovaSeq 6000 sequencing platform. Cell Ranger v.7.0.1 (10 × Genomics) was used to align the sequenced reads to the rat genome (mRatBN7 genome build), perform quality control on cells and generate count tables of unique molecular identifiers (UMIs) for each gene and cell. Reads mapping to introns were included in the counts.

Quality control and normalization of single-cell data

The R package Seurat v.5.1.0 was used to calculate quality control metrics97. Cells with fewer than 300 or higher than 5,000 distinct genes, fewer than 1,000 counts, or more than 25% of reads mapping to mitochondrial genes were removed. Doublets were identified and filtered using the R package DoubletFinder v.2.0.4 with default settings98. Genes expressed in fewer than three cells were excluded. Counts were normalized for sequencing depth, scaled to 10,000 UMIs per cell and log-transformed using the Seurat NormalizeData function.

Feature selection, dimensionality reduction, and cluster identification

The Seurat FindVariableFeatures function was used to define highly variable genes that were used as input to principal component analysis (PCA), with the number of variable features set to 2,500. The normalized counts were scaled, and elbow plots were generated to determine the number of principal components to include in the analysis (n = 15). A Uniform Manifold Approximation and Projection (UMAP) embedding was computed using these principal components, and cells were clustered using the FindClusters function of the Seurat package, with default parameters.

Cell type annotation

Genes differentially expressed between clusters (i.e. marker genes) were identified using the Seurat FindAllMarkers function. In detail, this analysis conducts FDR-corrected Wilcoxon rank-sum tests to assess the significance of upregulation of normalized gene expression in each cluster. The clusters were annotated by comparing their marker genes to canonical cell type markers reported in the literature.

Gene expression visualizations

Gene expression at the single cell level was visualized using dot plots generated with the FeaturePlot function of the Seurat R package. Gene expression at the population level was visualized using violin plots and dot plots generated with the VlnPlot and DotPlot functions of the Seurat package, respectively. Heatmaps for gene expression visualization were obtained by using the DoHeatmap function of the Seurat package. The spider plot used to visualize the expression of differentiation genes in hypoxic versus normoxic glomus cells was generated using the radarchart function of the fmsb R package (version 0.7.6). Progenitor cell surface markers were obtained from the list of most expressed markers belonging to the Gene Ontology category ‘Cell surface’ (GO:0009986). In addition to these markers, the expression of the enzyme beta-1,3-glucuronyltransferase 2 (B3GAT2) is also analyzed, as it is responsible for the synthesis of the HNK-1 surface epitope, which was previously utilized by our group to isolate undifferentiated cells in the carotid body15.

Gene signature expression calculation

The UCell R package (v. 2.6.2) was employed to compute the expression scores for a specific gene signature at the single-cell level. This method utilizes the Mann–Whitney U test to calculate the score for each individual cell54. The resulting scores were visualized using dot plots or violin plots, with the functions described in the preceding section. Genes used in the different signatures are described in Table S2. To represent the signature score against the expression level of a specified gene, the geom_point function of the ggplot2 R package (v.3.5.1) was employed.

Evaluation of differentiation status in the neuronal lineage

The grade of differentiation of neuronal cells was determined by defining a gene signature for dopaminergic differentiation. This signature included all the genes comprised in the following Gene Ontology categories: dopaminergic synaptic transmission (GO:0001963); dopamine biosynthetic process (GO:0042416); and dopamine catabolic process (GO:0042420). The specific genes included in this signature are reported in Table S2. The resulting dopaminergic differentiation score was represented in dot plots, with the FeaturePlot function of the Seurat package, or against other signature scores, using the geom_point function of the ggplot2 R package.

Gene ontology enrichment analysis

The enrichGO function of the clusterProfiler R package (v.4.10.1) was employed to identify enriched Gene Ontology (GO) categories within a specified set of genes99. To convert the gene names between GeneSymbol, Entrezgene_id and Ensembl databases, the R package biomaRt (v.2.58.2) was used100. The genes used for GO analysis for the different cell types were obtained with the FindAllMarkers function of the Seurat package and filtered by significant p-value (adjusted p-value < 0.05). When comparing two cell types, the genes utilized for GO enrichment analysis were identified using the FindMarkers function of the Seurat package, and also filtered by p-value.

Statistical measurements

Parametric variables were compared using Student’s t-test, while non-parametric variables were analyzed using the Wilcoxon test. Student’s t-test was performed with the t.test function of the stats R package, and the Wilcoxon test with the wilcox.test function of the same package (stats, v.4.3.2). P-values are reported in the figures and were adjusted for multiple comparisons using the Benjamini–Hochberg correction, as implemented in the p.adjust function of the stats R package. Pearson correlation coefficients were computed using the cor function of the stats R package. Correlation values for each gene, calculated against O2-sensing and dopaminergic differentiation scores, were visualized using the geom_point function of the ggplot2 R package.