Single-cell transcriptomics reveals receptor transformations during olfactory neurogenesis

See allHide authors and affiliations

Science  04 Dec 2015:
Vol. 350, Issue 6265, pp. 1251-1255
DOI: 10.1126/science.aad2456

Maturation of olfactory neurons

The sense of smell depends on neurons in the olfactory epithelium to perceive chemical scents. Each neuron specializes with one receptor. Hanchate et al. now show that the one-for-one relationship is not as simple as thought. As new neurons develop to replenish the olfactory epithelium, they initially express several different alleles of olfactory receptors. Then, as each neuron matures, they specialize to express a single receptor.

Science, this issue p. 1251


The sense of smell allows chemicals to be perceived as diverse scents. We used single-neuron RNA sequencing to explore the developmental mechanisms that shape this ability as nasal olfactory neurons mature in mice. Most mature neurons expressed only one of the ~1000 odorant receptor genes (Olfrs) available, and at a high level. However, many immature neurons expressed low levels of multiple Olfrs. Coexpressed Olfrs localized to overlapping zones of the nasal epithelium, suggesting regional biases, but not to single genomic loci. A single immature neuron could express Olfrs from up to seven different chromosomes. The mature state in which expression of Olfr genes is restricted to one per neuron emerges over a developmental progression that appears to be independent of neuronal activity involving sensory transduction molecules.

Odor detection in mammals is mediated by odorant receptors on olfactory sensory neurons (OSNs) in the nasal olfactory epithelium (1, 2). In mice, ~1000 odorant receptor genes (Olfrs) and 350 pseudogenes reside at dozens of distinct loci on 17 of 21 chromosomes (35). Each Olfr is expressed by a small subset of OSNs scattered in one epithelial spatial zone (68). Previous studies suggest that each mature OSN expresses one intact Olfr allele, but some coexpress an Olfr pseudogene (911). In a prevailing model of “OR [Olfr] gene choice,” the developing OSN selects a single Olfr allele for expression, and the encoded receptor provides feedback that prevents expression of other Olfrs (1217). OSNs are generated in a developmental progression from progenitors to precursors to immature OSNs to mature OSNs (18, 19). We investigated when and how the developing OSN selects one Olfr for expression.

We used single-cell RNA sequencing (RNA-seq) (20) to analyze the transcriptomes of single epithelial neurons during development. We first prepared cDNA libraries from single isolated cells (10) and analyzed the libraries for markers of the four stages of OSN development, using polymerase chain reaction. We then conducted Illumina sequencing (21) of libraries from multiple cells in each stage, as well as duplicate libraries from some cells. We used TopHat (22) and Cufflinks (23) to identify genes expressed in each cell and to estimate their relative mRNA abundances (see fig. S1 for technical quality metrics).

We compared 85 cell transcriptomes using Monocle, an unsupervised algorithm that determines each cell’s stage of differentiation in “pseudotime,” which represents progress through geneexpression changes during development (24). Monocle showed a linear nonbranching trajectory of development (Fig. 1A). Based on cell stage markers in individual transcriptomes, the trajectory reflects the developmental progression from progenitors through mature OSNs. The following gene markers were used: for progenitors, Ascl1 (achaete-scute complex homolog 1); for precursors, Neurog1 (neurogenin 1) and/or Neurod1 (neurogenic differentiation 1); for immature OSNs, Gap43 (growth-associated protein 43) and/or Gng8 (guanine nucleotide–binding protein gamma 8); and for mature OSNs, Omp (olfactory marker protein) and four olfactory sensory transduction molecules downstream of odorant receptors—Gnal (guanine nucleotide binding protein, alpha stimulating, olfactory type), Adcy3 (adenylate cyclase 3), Cnga2 (cyclic nucleotide gated channel alpha 2), and Cnga4 (cyclic nucleotide gated channel alpha 4) (18, 19).

Fig. 1 Olfactory neurons exhibit large-scale shifts in gene expression during development.

(A) Unsupervised analysis of single-cell gene expression profiles with Monocle revealed a linear trajectory (black line) along which cells develop in pseudotime. Coloring of cells based on the expression of developmental markers shows that the trajectory corresponds to a stepwise development from olfactory progenitors to precursors to immature OSNs to mature OSNs. (B) Global analysis of gene expression kinetics along the trajectory identified 3830 genes that vary significantly over developmental pseudotime (false discovery rate < 5%, determined by a Tobit-valued generalized linear model likelihood ratio test; supplementary materials). Hierarchical clustering of these genes via Ward’s method recovered 11 nonredundant groups that covary over the trajectory. Cluster analysis indicates that multiple large shifts in gene expression occur as neurons progress through development. The bar on top shows the locations of individual cells, colored by stage of development, along this developmental trajectory. The Expression Z score indicates changes in a gene relative to its dynamic range over pseudotime. (C) Kinetic diagrams show the expression of known markers of different developmental stages over the developmental progression. Numbers in parentheses indicate the groups in which genes are found in (B). Dots indicate individual cells and are colored according to developmental stage. Black lines indicate local polynomial regression smoothing (span, 0.75; degree, 2) of log-transformed FPKM values over developmental pseudotime.

Immature OSNs were further divided into two subsets based on their expression of olfactory sensory transduction molecules. Early immature OSNs lacked one or more olfactory transduction molecules, whereas late immature OSNs expressed all four (Fig. 2).

Fig. 2 Immature neurons can express multiple Olfrs.

(A) Neurons assigned to different developmental stages were arranged by developmental progress, as measured in pseudotime. Different developmental stages are indicated by differently colored ticks. Different Olfrs are represented by different colors in the bars. The total number of Olfr transcripts per cell shows a steady, though variable, increase during development. (B) Multiple different Olfr transcripts were detected in 12 of 25 early immature, 6 of 13 late immature, and 6 of 25 mature OSNs with Olfr transcripts. (C) The number of different Olfr transcripts per cell was highest in early immature OSNs and then declined over development. Early immature OSNs tended to express similar levels of different Olfrs. In contrast, the majority of mature OSNs expressed only one Olfr or high levels of one Olfr and low levels of one or two additional Olfrs. Each color in a bar represents a single Olfr, except gray, which represents >1 Olfr. (D) Olfrs stimulate neuronal activity via mechanisms involving sensory transduction molecules encoded by Gnal (or possible Gnas in immature OSNs), Adcy3, Cnga2, and Cnga4. Six immature and six mature neurons with >1 Olfr expressed all four genes, suggesting that neuronal activity downstream of odorant receptors is not what reduces the number of Olfrs expressed per neuron. Omp, which is highly expressed in mature OSNs, was absent from six early immature OSNs with >1 Olfr, arguing against contamination from mature OSNs. Gapdh and Actb are housekeeping genes.

A total of 3830 genes were differentially expressed over development. Clusters of genes changed in expression during specific developmental periods, suggesting sequential large and coordinated changes in gene expression during OSN development (Fig. 1B and table S1). By gene ontology, most clusters contained genes associated with transcriptional regulation and/or chromatin modification, suggesting potential regulators of development (table S1). In kinetic diagrams, markers of early and late developmental stages show peak expression early and late in the developmental progression, respectively (Fig. 1C and fig. S2).

Olfr expression first appeared at the late precursor to early immature OSN stages (Fig. 2). Olfr transcripts were found in one of nine precursors, 38 of 40 immature OSNs, and 25 of 25 mature OSNs (Fig. 2). None were seen in two non-neuronal epithelial supporting cells or in three cells of undetermined type. Overall, the number of Olfr transcripts per cell increased over OSN development (Fig. 2A). In early immature, late immature, and mature OSNs, Olfrs were detected at an average of 1998, 4146, and 8169 FPKM (fragments per kilobase of transcript per million mapped reads), respectively, with median values of 930, 2575, and 4026 FPKM. Transcripts of individual Olfrs were detected at an average of 657, 2156, and 6382 FPKM, with median values of 99, 807, and 2672 FPKM.

These studies indicate that the developing OSN can initially express multiple Olfrs (Fig. 2). Roughly half (48%, 12 of 25) of early immature OSNs with Olfrs expressed >1 Olfr. Coexpression of different Olfrs in single neurons declined as development progressed, with 46% (6 of 13) of late immature and 24% (6 of 25) of mature OSNs expressing >1 Olfr. Moreover, single early immature OSNs expressed up to 12 different Olfrs, whereas mature OSNs with >1 Olfr expressed two or at most three (Fig. 2B).

Early immature and mature OSNs with >1 Olfr also differed in the relative abundance of different Olfr transcripts (Fig. 2C). Most (10 of 12) of the early immature OSNs had similarly low levels of different Olfrs. The most abundant Olfr was detected at 55 to 396 FPKM in individual neurons and the next highest at, on average, 60.5% of this level (median, 60.1%). However, in three of six mature OSNs with >1 Olfr, the most abundant was detected at 14,557 to 18,056 FPKM, with the next highest, on average, only 3.3% as abundant (median, 0.5%).

In mature OSNs, Olfr and Omp transcripts averaged 8169 and 10,167 FPKM per cell, respectively. However, 6 of 12 early immature OSNs that expressed >1 Olfr did not express Omp (Fig. 2D), arguing against the possibility that the Olfr transcripts detected were due to contamination from mature OSNs.

Data from eight duplicate cell samples (technical replicates) were analyzed (figs. S3 and S4). The duplicates confirmed the expression of >1 Olfr in specific OSNs (table S2). The data were consistent with reported stochastic losses of low–copy number transcripts in single-cell RNA-seq data. Olfrs present in both replicates tended to be expressed at higher levels, and those present in only one replicate tended to be expressed at lower levels.

The above results indicate that early immature OSNs can express low levels of multiple Olfrs, but, during subsequent development, two changes typically occur. Expression favors one Olfr by up to 100 times or more, and the expression of additional Olfrs declines or disappears.

To validate these findings, we used RNA–dual fluorescence in situ hybridization (dual RNA-FISH) with nasal tissue sections. At postnatal day 3 (P3), a peak time of OSN neurogenesis (19, 25), 0.22 ± 0.05 to 0.22 ± 0.12% of neurons labeled for a single Olfr were colabeled with a mix of probes for other Olfrs expressed in the same nasal zone (Fig. 3A and table S3). Neurogenesis decreases as mice mature, and no colabeled cells were seen in adults. Using a highly sensitive RNA-FISH method with branched DNA signal amplification (26), 0.41 ± 0.09 to 0.60 ± 0.13% of cells labeled for one Olfr were co-labeled for another Olfr at P3, and 0.10 ± 0.02 to 0.18 ± 0.05% were co-labeled for another at P21 (Fig. 3B and table S4). Among neurons labeled for one Olfr, the percentage co-labeled for the immature OSN markers Gap43 and Gng8 also changed, respectively, from 80.1 ± 3.2% and 62.8 ± 0.9% at P3 to 19.5 ± 0.5% and 14.3 ± 1.1% at P21 (table S5). These results confirm that single OSNs can express more than one Olfr and suggest that Olfr coexpression occurs predominantly, if not exclusively, in immature OSNs.

Fig. 3 Olfrs expressed in the same neuron belong to a regional gene set.

(A) Dual RNA-FISH of P3 tissue sections using a conventional method showed a small percentage of OSNs co-labeled with an Olfr1507 probe and a mix of probes for other Olfrs expressed in the same zone (zone 4). Cell nuclei were counterstained with 4',6-diamidino-2-phenylindole (blue). Two co-labeled cells are shown, one on the left half and the other on the right half of the panel. Scale bar, 5 μm. (B) Dual RNA-FISH of P3 tissue sections using a highly sensitive method showed a small percentage of OSNs coexpressing Olfr1507 and Olfr286. Two co-labeled cells are shown [as in (A)], and a cell labeled with only one probe (red only) is also shown on the right. Scale bar, 5 μm. (C) Dual RNA-FISH shows that Olfrs coexpressed in single immature OSNs (neuron D200 or D243) are singly expressed in neurons in the same or partially overlapping zones in adult olfactory epithelium sections. This correspondence suggests that Olfr expression in the immature OSN is restricted to a spatially determined set of Olfr genes. In the upper row, colored dots indicate the locations of labeled neurons. Boxed areas in the upper row are shown at higher magnification in the lower row. Scale bars, 500 μm (upper row) and 250 μm (lower row).

To examine whether odorant receptor–induced neuronal activity might be involved in the observed developmental shift in Olfr expression, we analyzed transcriptome data for the expression of olfactory sensory transduction molecules: Gnal (or Gnas, which may substitute for Gnal), Adcy3, Cnga2, and Cnga4. All four molecules were expressed in 6 of 18 immature OSNs and 6 of 6 mature OSNs with >1 Olfr (Fig. 2D). Furthermore, one or more were absent in data from 13 of 20 immature and 3 of 19 mature OSNs with only one Olfr. These results suggest that odorant receptor–induced neuronal activity is neither necessary nor sufficient for the decline in coexpressed Olfrs during development.

We next tested whether the developing OSN is restricted to activating Olfrs expressed in a particular nasal zone. Using dual RNA-FISH, we compared the nasal expression patterns of 11 pairs of Olfrs coexpressed in six different OSNs. In every case, the paired Olfrs were expressed either in the same spatial zone or in partially overlapping zones (Fig. 3C and table S6). These results suggest that the developing neuron is restricted to the expression of a particular Olfr regional gene set, which can include Olfrs with only partially overlapping expression patterns in the adult.

To investigate whether early coexpression of multiple Olfrs could result from chromatin changes at a single genomic locus containing those Olfrs, we determined the chromosome locations of Olfrs coexpressed in individual OSNs. For OSNs expressing 4 to 12 Olfrs, coexpressed Olfrs mapped to three to seven different chromosomes and four to nine distinguishable Olfr gene loci (Fig. 4 and table S7). Thus, the immature OSN is not restricted to expressing Olfrs from a single chromosomal region.

Fig. 4 Immature neurons coexpress Olfrs from multiple chromosomal loci.

Diagrams show the chromosomal locations of Olfrs expressed in single OSNs of different stages. Each mouse chromosome is indicated by a vertical bar with its number below. The names of neurons, parenthesized number of Olfrs per neuron, and dots indicating the chromosomal locations of those Olfrs are shown in different colors for different neurons.

Odor detection in the mouse nose is mediated by 1000 different odorant receptors, each expressed by a different subset of sensory neurons. We asked when and how a neuron comes to express a single Olfr. We found that the developing neuron can express low levels of multiple Olfrs. As development proceeds, this ability declines. The mature neuron typically expresses high levels of a single Olfr. Coexpressed Olfrs tend to be expressed by other neurons in the same region of the olfactory epithelium, suggesting regional biases in Olfr gene choice, but they can reside at multiple chromosomal locations.

How does the developing OSN transition from expressing low levels of multiple Olfrs to expressing a high level of a single Olfr? One possibility is a “winner-take-all” mechanism. In this model, multiple Olfrs are initially expressed, but one becomes dominant—for example, by the capture of limiting factors required for high-level Olfr expression (fig. S5). In an alternative model, selection of a single Olfr for high-level expression occurs independently of those initially expressed. In either model, early low-level expression of other Olfrs could subside, owing to the closing of a developmental time window or to feedback signals generated by the highly expressed Olfr. OSNs expressing multiple Olfrs are probably not pruned by apoptosis, as suggested for OSNs in the nasal septal organ (27), given genetic evidence that some OSNs expressing one Olfr previously expressed another (13). This Olfr “switching” may reflect the early expression of more than one Olfr per immature OSN, as observed in this study.

Supplementary Materials

Materials and Methods

Figs. S1 to S5

Tables S1 to S7

References (2835)


  1. ACKNOWLEDGMENTS: We thank J. Delrow, A. Marty, and A. Dawson at the Fred Hutchinson Cancer Research Center (FHCRC) Genomics Facility for assistance with RNA-seq; M. Fitzgibbon and J. Davidson at the FHCRC Bioinformatics Resource for early assistance with sequence analyses; and J. Vasquez and the FHCRC Scientific Imaging Facility for help with confocal microscopy. We also thank members of the Buck laboratory for helpful discussions. This work was supported by the Howard Hughes Medical Institute (L.B.B.), NIH grants R01 DC009324 (L.B.B.) and DP2 HD088158 (C.T.), an Alfred P. Sloan Fellowship (C.T.), and a Dale F. Frey Award for Breakthrough Scientists from the Damon Runyon Cancer Research Foundation (C.T.). L.B.B. is on the Board of Directors of International Flavors & Fragrances. The supplementary materials contain additional data. N.K.H., C.T., and L.B.B. designed the research; N.K.H. and C.T. performed the research; N.K.H., C.T., K.K., Z.L., D.K., X.Y., X.Q., and L.B.B. analyzed the data; L.P. provided guidance; and N.K.H, C.T., and L.B.B. wrote the paper. Raw sequencing data related to this study have been archived in the Gene Expression Omnibus (GEO) database under accession number GSE75413 (available at
View Abstract

Stay Connected to Science

Editor's Blog

Navigate This Article