Non-neuronal expression of SARS-CoV-2 entry genes in the olfactory system suggests mechanisms underlying COVID-19-associated anosmia

Abstract

Altered olfactory function is a common symptom of COVID-19, but its etiology is unknown. A key question is whether SARS-CoV-2 (CoV-2) - the causal agent in COVID-19 - affects olfaction directly, by infecting olfactory sensory neurons or their targets in the olfactory bulb, or indirectly, through perturbation of supporting cells. Here we identify cell types in the olfactory epithelium and olfactory bulb that express SARS-CoV-2 cell entry molecules. Bulk sequencing demonstrated that mouse, non-human primate and human olfactory mucosa expresses two key genes involved in CoV-2 entry, ACE2 and TMPRSS2. However, single cell sequencing revealed that ACE2 is expressed in support cells, stem cells, and perivascular cells, rather than in neurons. Immunostaining confirmed these results and revealed pervasive expression of ACE2 protein in dorsally-located olfactory epithelial sustentacular cells and olfactory bulb pericytes in the mouse. These findings suggest that CoV-2 infection of non-neuronal cell types leads to anosmia and related disturbances in odor perception in COVID-19 patients.

Fig. 1. Schematic of the nasal respiratory epithelium, olfactory epithelium, and the olfactory bulb.

Sagittal view of the human nasal cavity, in which respiratory and olfactory epithelia are colored (left). For each type of epithelium, a schematic of the anatomy and known major cell types are shown (right). In the olfactory bulb in the brain (tan) the axons from olfactory sensory neurons coalesce into glomeruli, and mitral/tufted cells innervate these glomeruli and send olfactory projections to downstream olfactory areas. Glomeruli are also innervated by juxtaglomerular cells, a subset of which are dopaminergic.

Fig. 2

Coronavirus cell entry-related genes are expressed in human respiratory and olfactory epithelium but are not detected in human OSNs. (A) UMAP representation of cell types in human nasal biopsy scSeq data from Durante et al. 2020 (38). Each dot represents an individual cell, colored by cell type (HBC = horizontal basal cell, OSN = olfactory sensory neuron, SUS = sustentacular cell, MV: microvillar cell, Resp.: respiratory, OEC = olfactory ensheathing cell, SMC=smooth muscle cell). (B) UMAP representations of 865 detected immature (GNG8) and mature (GNG13) OSNs. Neither ACE2 nor TMPRSS2 are detected in either population of OSNs. The color represents the normalized expression level for each gene (number of UMIs for a given gene divided by the total number of UMIs for each cell). (C) UMAP representations of all cells, depicting the normalized expression of CoV-2 related genes ACE2 and TMPRSS2, as well as several cell type markers. ACE2 and TMPRSS2 are expressed in respiratory and olfactory cell types, but not in OSNs. ACE2 and TMPRSS2 are detected in HBC (KRT5) and sustentacular (CYP2A13) cells, as well as other respiratory epithelial cell types, including respiratory ciliated (FOXJ1) cells. (D) Percent of cells expressing ACE2 and TMPRSS2. ACE2 was not detected in any OSNs, but was observed in sustentacular cells and HBCs, among other olfactory and respiratory epithelial cell types. Olfactory and respiratory cell types are shown separately. ACE2 and TMPRSS2 were also significantly co-expressed (Odds ratio 7.088, p-value 3.74E-57, Fisher’s exact test). (E) ACE2 immunostaining of human olfactory mucosal biopsy samples (taken from a 28-year old female). ACE2 protein (green) is detected in sustentacular cells and KRT5-positive HBCs (red; white arrowhead). Nuclei were stained with DAPI (blue). Bar = 25 μm. The ACE2 and KRT5 channels from the box on the left are shown individually on the right.

Fig. 3

Coronavirus cell entry-related genes are expressed at comparable levels across respiratory and olfactory epithelial datasets. (A) Schematic of the mapping strategy used to identify similar cell types across datasets, applied to a toy example. Each cell type from “Dataset 1” dataset is mapped to cell types from the “Dataset 2”. From left to right: Each Dataset 1 cell voted on its 5 most similar cells in Dataset 2; the total number of votes cast for each Dataset 2 cell type was quantified; and vote totals were Z-scored against 1000 shuffles where cell type labels were permutated. (B) Mapping was performed bi-directionally between the Deprez et al. (41) and Durante et al. (38) datasets, and the mapping Z-scores in each direction are compared. The set of cell type correspondences with high Z-scores (>25) in both directions are colored red (top). The set of cell type correspondences with high bi-directional mappings shown in red in top panel are highlighted in yellow (bottom). (C) Gene expression across cell types and tissues in Durante et al. (top) and Deprez et al. (bottom). Each gene is normalized to its maximum value across all tissues. Gene expression from Durante et al. was normalized to that in Deprez et al. to enable comparisons (see Methods and Figure S3). The tissues correspond to the indicated positions along the airway from nasal to distal lung. ACE2 expression in olfactory HBC and sustentacular cells is comparable to that observed in other cell types in the lower respiratory tract.

Fig. 4

ACE2 is expressed in the mouse nasal epithelium but not in mature OSNs. (A) Log₂-fold change (FC) in mean across-replicate gene expression between olfactory sensory neurons (OSNs) and whole olfactory mucosa (WOM) for coronavirus (CoV)-related genes and cell type markers (HBC = horizontal basal cells, SUS = sustentacular cells), data from Saraiva et al. (46). (B) UMAP representation of scSeq data from the WOM, colored by cell types (mOSN: mature OSN, iOSN: immature OSN, INP: immediate neural precursor, GBC: globose basal cell, MV: microvillar cell, Resp.: respiratory). (C) Percent of cells expressing Ace2 and Tmprss2 in olfactory and respiratory cell types in the WOM (Drop-seq) dataset. Detection was considered positive if any transcripts (UMIs) were expressed for a given gene. Sustentacular cells (SUS) from dorsal and ventral zones are quantified separately. Ace2 is detected in dorsal sustentacular, Bowman’s gland, HBCs, as well as respiratory cell types. (D) UMAP representation of sustentacular cells, with expression of CoV-2 related genes Ace2 and Tmprss2, as well as marker genes for SUS (both pan-SUS marker Cbr2 and dorsal specific marker Sult1c1) indicated. The color represents the normalized expression level for each gene (number of UMIs for a given gene divided by the total number of UMIs for each cell; in this plot Ace2 expression is binarized for visualization purposes). Ace2-positive sustentacular cells are found within the dorsal Sult1c1-positive subset. UMAP plots for other cell types are shown in Figure S4.

Fig. 5

ACE2 protein is detected in the mouse olfactory and respiratory epithelium. (A) ACE2 immunostaining of mouse main olfactory epithelium. As shown in this coronal section, ACE2 protein is detected in the dorsal zone and respiratory epithelium. The punctate Ace2 staining beneath the epithelial layer is likely associated with vasculature. Bar = 500 μm. Arrowheads depict the edges of ACE2 expression, corresponding to the presumptive dorsal zone (confirmed in G). Dashed boxes indicate the areas shown in B and G (left). (B) ACE2 protein is detected in the dorsal zone of the olfactory epithelium. Bar = 50 μm. (C) Dorsal zone-specific expression of ACE2 in the olfactory epithelium was confirmed by co-staining with NQO1, a protein expressed in dorsal-zone OSNs. Bar = 50 μm. (D) ACE2 signal in dorsal olfactory epithelium does not overlap with the cilia of olfactory sensory neurons, as visualized by CNGA2. Bar = 50 μm. (E) High magnification image of the apical end of the olfactory epithelium reveals that ACE2 signal is localized at the tip of villi of sustentacular cells, visualized by Phalloidin (F-Actin), but does not overlap with cilia of olfactory sensory neurons, as visualized by Acetylated Tubulin (Ac. Tubulin). Bar = 10 μm. (F) Bowman’s glands, which span from the lamina propria to the apical surface (arrowheads), were positive for ACE2 staining. Bar = 50 μm. (G) ACE2 expression in the respiratory epithelium was confirmed by co-staining with TUBB4. Bar = 50 μm.

Fig. 6

ACE2 is expressed in the mouse nasal epithelium in an injury model. (A) UMAP representation of data from an scSeq HBC lineage dataset, which includes several timepoints after epithelial injury induced by methimazole (mOSN: mature OSN, iOSN: immature OSN, INP: immediate neural precursor, SUS: sustentacular cell, GBC: globose basal cell, HBC: horizontal basal cell, HBC*: activated or cycling HBCs. MV: microvillar cell, Resp.: respiratory). (B) UMAP representation of CoV-2 related genes Ace2 and Tmprss2, as well as marker genes for the HBC-derived cell types. The color represents normalized expression (number of UMIs for a given gene divided by the total number of UMIs for each cell). (C) Percent of cells expressing Ace2 and Tmprss2. Ace2 is detected in sustentacular cells, HBC, activated/cycling HBC and respiratory cells. (D) UMAP representation of all sustentacular cells, indicating the normalized expression of CoV-2 related genes Ace2 and Tmprss2, as well as sustentacular (Ermn) cell markers. Ace2-positive sustentacular cells are largely a subset of dorsal SUS cells, as identified via the expression of Sult1c1. Sult1c1-positive sustentacular cells have higher levels of Ace2 (p=1.87E-03, Mann-Whitney test) and Ace2-positive sustentacular cells have higher levels of Sult1c1 (p=8.06E-07, Mann-Whitney test). (E) ACE2 immunostaining of mouse nasal epithelium after methimazole treatment, together with cycling cell marker Ki67 and HBC marker KRT5. At 48 hours after treatment, ACE2 signal is detected in Ki67+/KRT5+ activated HBCs (top). At 96 hours after treatment, ACE2 signal is observed at the apical surface of Ki67+ cells (bottom). Some ACE2-positive cells express low levels of the HBC marker KRT5 and have immunostaining patterns similar to that of dorsal sustentacular cells, suggesting that they are sustentacular cells in the process of differentiating from their HBC precursors. Bar = 25 μm.

Fig. 7

Expression of coronavirus entry genes in mouse olfactory bulb. (A) UMAP visualization of OB scSeq highlighting the main cell classes and subtypes from two integrated scSeq datasets (see Methods). VIP, vasoactive intestinal peptide-expressing neurons; ETCs, external tufted cells; OPCs, oligodendrocyte precursor cells; IPCs, intermediate precursor cells; OECs, olfactory ensheathing cells. Cluster information is summarized in Figures S7-8. (B) UMAP representation of the vascular cell cluster showing expression of CoV-2 entry genes (Ace2 and Tmprss2) and Kcnj8, a pericyte marker. Color scale depicts log-normalized UMI counts. (C) Normalized gene expression of coronavirus entry genes and cell class markers in mouse olfactory bulb. Color scale shows scaled mean expression level per cell type, normalized by their maximum expression across cell types. Ace2 is specifically expressed in vascular cells. (D) Percent of cells expressing Ace2. “Other vascular” denotes all vascular cells excluding pericytes. Ace2 expression was only detected in vascular cell types. (E) Log₂-normalized expression (Log₂(TPM+1)) of coronavirus entry genes and dopaminergic neuron markers in manually sorted and deeply-sequenced single olfactory bulb dopaminergic neurons. (F) ACE2 immunostaining of the mouse main olfactory bulb. Left, section of olfactory bulb containing the glomerular layer (with example glomeruli circled), mitral cell layer (MCL) and granule cell layer (GCL). ACE2 protein is present in vascular mural cells but not in OB neurons or OSN axons. Boxes i and ii indicate the locations of enlarged images. Bar = 100 μm. (i) enlarged image of glomerular layer (middle). ACE2 protein staining was restricted to vascular cells. Bar = 50 μm. (ii) enlarged image of MCL (dashed line) and GCL showing the lack of ACE2 (right). Bar = 50 μm. (G) An olfactory bulb section showing ACE2 protein is detected in PDGFRB-positive mural cells, including smooth muscle cells (SMC) and pericytes. Bar = 25 μm.