Research Article

Negative Feedback Regulation Ensures the One Receptor-One Olfactory Neuron Rule in Mouse

See allHide authors and affiliations

Science  19 Dec 2003:
Vol. 302, Issue 5653, pp. 2088-2094
DOI: 10.1126/science.1089122

Abstract

In the mouse olfactory system, each olfactory sensory neuron (OSN) expresses only one odorant receptor (OR) gene in a monoallelic and mutually exclusive manner. Such expression forms the genetic basis for OR-instructed axonal projection of OSNs to the olfactory bulb of the brain during development. Here, we identify an upstream cis-acting DNA region that activates the OR gene cluster in mouse and allows the expression of only one OR gene within the cluster. Deletion of the coding region of the expressed OR gene or a naturally occurring frame-shift mutation allows a second OR gene to be expressed. We propose that stochastic activation of only one OR gene within the cluster and negative feedback regulation by that OR gene product are necessary to ensure the one receptor–one neuron rule.

The mouse has more than 1000 OR genes clustered at multiple loci on many different chromosomes (1). The OR genes are expressed in a mutually exclusive, monoallelic manner by OSNs in the nasal olfactory epithelium (OE) (25). OSNs expressing a specific OR gene are distributed in one of four zones within the OE (68), and their axons converge on a specific set of glomeruli in the olfactory bulb (911). The one receptor–one neuron rule is essential for the conversion of olfactory signals into an accurate topographical map in the olfactory bulb. Using transgenic mice carrying a large DNA region of an OR gene cluster in which the MOR28 gene resides, we previously showed that the transgenic and endogenous alleles of MOR28 are expressed in a mutually exclusive manner (4). The exclusion was also observed for separately tagged MOR28 transgenes that were integrated in tandem into the same chromosomal location. Although these MOR28 genes have the same coding and regulatory sequences, no more than one allele of MOR28 is expressed under any circumstance. Such an unusual mode of gene regulation has previously been reported only for the antigen receptor genes of the immune system.

A cis-acting locus control region activates the OR gene cluster. Using transgenic constructs in yeast artificial chromosomes (YACs), we have studied the expression of the OR gene cluster containing the MOR28 gene (4). By comparing the DNA structures of various YAC constructs and their expression patterns, we found that the DNA region 40 to 150 kb upstream of MOR28 appears to contain a cis-acting regulatory element that is necessary for the expression of transgenic OR genes. Sequence comparison of the mouse and human genomes revealed that a 2-kb sequence 75 kb upstream of the murine MOR28 gene is highly homologous to the human sequence 32 kb upstream of the human HOR28 gene (Fig. 1A). Percent identity plotting between the mouse and human sequences (12, 13) revealed that this 2-kb homology (H) region is the only conserved sequence in the upstream region of MOR28 (Fig. 1B).

Fig. 1.

The homology (H) region that activates the MOR28 cluster. (A) Structures of seven transgenic constructs are compared. MOR28 in each construct is shown as expressed (+) or not (–). Nucleotide sequences were compared between mouse and human for the upstream regulatory region. A 2-kb homology (H) was identified (dot matrix) 75 kb upstream of MOR28. (B) Percent identity plot for the mouse MOR28 cluster against the human homolog. Repetitive sequences were masked with RepeatMasker (12, 13). Similarities higher than 50% are shown as vertical bars. (C) Deletion of the H region abolishes the expression of the MOR28 cluster. The H region was deleted from the YAC-290 construct (4), generating YAC-290dH. Three zone 4–specific OR genes in the constructs were separately tagged. OE sections of the transgenic mice were examined for the expression of the transgenes by fluorescence (EGFP) or by immunostaining (lacZ and WGA). The same results were obtained from six transgenic lines for YAC-290 and five lines for YAC-290dH. (D) Expression of the zone 1–specific MOR29. OE sections were analyzed by fluorescence (EGFP for MOR29) and by immunostaining (lacZ for MOR28). The same results were obtained from four transgenic lines for YAC-290 and five lines for YAC-290dH. Scale bars, 20 μm.

To determine whether the H region is necessary for the expression of MOR28 and other downstream OR genes in the cluster, we deleted the H region from the YAC-290 construct in which the three OR genes MOR28, MOR10, and MOR83 had been separately tagged with lacZ (expressing Escherichia coli β-galactosidase), EGFP (expressing enhanced green fluorescent protein), and WGA (expressing wheat germ agglutinin), respectively (Fig. 1C) (14). The control YAC-290, without the H deletion, expressed all three OR genes in zone 4 of the OE. When the H region was deleted (YAC-290dH), none of these transgenes were expressed. It was later found that the YAC-290 construct contained another OR gene, MOR29, that is expressed in zone 1 of the OE (15). We separately tagged MOR28 and MOR29 with lacZ and EGFP, respectively. When the H region was deleted, the expression of both MOR28 (lacZ-tagged) and MOR29 (EGFP-tagged) was abolished (Fig. 1D). These results indicate that the zone 4–specific MOR28, MOR10, and MOR83 genes, as well as the zone 1–specific MOR29 gene, are under the control of the H region.

Our previous transgenic studies showed that the shorter YAC constructs lacking the upstream region of YAC-290 do not express the MOR28 transgene (4). We examined whether the expression of these constructs could be restored by adding the H region. All of the OR transgenes in the YAC-H180 construct— MOR28 (lacZ-tagged), MOR10 (EGFP-tagged), MOR83 (WGA-tagged), and MOR29 (DsRed-tagged, expressing red fluorescent protein from the coral Discosoma)—were reactivated when the H region was attached to the upstream ends of YAC-180 (Fig. 2A). Interestingly, comparison of YAC-H140 to YAC-290 showed that the number of OSNs expressing MOR28 was greatly increased, whereas those for the downstream genes were substantially reduced (Fig. 1, C and D, and Fig. 2B). This phenomenon is likely attributed to the relocation of the H region immediately upstream of the MOR28 in the YAC-140 construct. We then examined whether the H region can activate the MOR28 minigene tagged with EYFP (expressing enhanced yellow fluorescent protein), which contains exons 1 and 2 flanked by 3-kb regions of genomic DNA (Fig. 2C). Unlike the minigenes examined by Vassalli et al. (16), our minigene could not be expressed on its own. When the H region was attached (H-MOR28), a large number of OSNs expressed the minigene in 5 of 10 mouse lines, whereas the control minigene without H was not expressed in any of the 8 lines examined. Because the transgene-positive cells were particularly dense in zone 4 of the OE, we were curious as to whether the transgene-expressing cells still showed mutually exclusive expression of ORs. In situ hybridization with the RNA probes for the zone 4–specific OR genes revealed that cells expressing both endogenous and transgenic OR genes were rarely found (less than 1%) among several hundred OSNs expressing the endogenous OR genes (17).

Fig. 2.

Addition of the H region to the truncated transgenic constructs. The H region was joined at the upstream ends of the truncated transgenic constructs YAC-180 (A), YAC-140 (B), and the MOR28 minigene (C). The constructs lacking the H region do not express the transgenes. The same results were obtained from three transgenic lines for YAC-180, eight lines for YAC-140, and eight lines for the MOR28 minigene. All OR genes in the constructs were separately tagged as illustrated. When the H region was added, expression of the MOR28 as well as other transgenes was restored. The same results were obtained from five lines for YAC-H180 and two lines for YAC-H140. The H-MOR28 minigene was expressed in 5 of 10 lines. Copy numbers of the H-MOR28 minigene in line 6 and line 9 were 2 and 1, respectively. OE sections of the transgenic mice were stained to detect the expression of the transgenes by fluorescence (EGFP and DsRed) or by immunostaining (lacZ, EYFP, and WGA). Scale bars, 20 μm [(A) and (B)], 200 μm (C).

Deleting the coding region permits expression of other OR genes. To achieve the mutually exclusive expression, we assumed that negative regulation was taking place. We postulated that once a particular OR gene is stochastically chosen, the gene product prevents further activation of the remaining genes. To test this hypothesis, we deleted the entire coding sequence of the MOR28 gene in YAC-290G (Fig. 3A). To visualize the transgene-expressing OSNs, we tagged MOR28 with IRES-tau-EGFP in the 3′-untranslated region. We analyzed two sets of transgenic mice: one carrying the deletion-type MOR28 (del-MOR28) and the other carrying the wild-type MOR28 allele.

Fig. 3.

A coding-deletion mutant of the MOR28 transgene. (A) Structure of del-MOR28. The coding sequence was deleted from exon 2 of MOR28 in the YAC-290G construct (4), generating YAC-290GdC. MOR28 was tagged with IRES-tau-EGFP. (B) Detection of EGFP fluorescence in the OE sections. Three different EGFP-tagged MOR28 genes were analyzed for their expression patterns in the OE: the endogenous MOR28 gene in the knock-in mouse (MOR28) (4), the wild-type MOR28 transgene (Tg MOR28), and the deletion-type transgene (del-MOR28). (C) Coexpression of the transgenic del-MOR28 with the endogenous MOR28 gene. A section of OE from the YAC-290GdC mouse was hybridized with the MOR28-coding probe to detect the expression of the endogenous MOR28 without the EGFP tag (stained red). The same section was immunostained with anti-EGFP to detect the expression of del-MOR28, which lacks the coding sequence (stained green). The two pictures were superimposed (merge) to detect the coexpressed cells. Enlarged photos are shown at the bottom. The MOR28 coding probe is cross-hybridizable with the MOR10 mRNA. Scale bars, 100 μm (B), 50 μm (C).

The OE sections were analyzed by fluorescence microscopy for transgene expression (Fig. 3B). Although the OSNs with activated del-MOR28 expressed only the EGFP protein, their densities and distribution in the OE were indistinguishable from those for the endogenous MOR28 and for the wild-type transgene (Fig. 3B). We then examined whether mutually exclusive expression of OR genes is violated in those OSNs expressing del-MOR28. If the expressed OR gene products inhibit further activation of other OR genes, deletion of the coding sequence in the MOR28 should allow for the expression of the second allele of MOR28 or other OR genes. In contrast, if the activation process itself is responsible for the monoallelic and mutually exclusive expression of OR genes, then secondary activation of other OR genes should not be observed even in the OSNs with an activated del-MOR28 locus. We first examined the exclusion between the endogenous MOR28 and the transgenic del-MOR28 genes (Fig. 3C). The OE sections were hybridized with an RNA probe for the MOR28 coding sequence and then treated with antibodies to EGFP (anti-EGFP). Because del-MOR28 does not contain the MOR28 coding region, the RNA probe only detects OSNs that express the endogenous MOR28 (shown in red), whereas immunostaining detects cells expressing the transgene (shown in green). Using confocal microscopy, we carefully analyzed the stained OSNs for coexpression. Yellow-colored cells were frequently found (Fig. 3C). Note that the detection of double-stained cells is not due to the overlap of red and green cells in different optical planes.

We also examined the coexpression of the transgenic del-MOR28 gene with other endogenous OR genes whose expression areas in the OE overlap those of MOR28. Among more than 200 OR genes analyzed by in situ hybridization, about 40 OR genes were found to share expression areas with MOR28. For these endogenous OR genes, coexpression was examined with del-MOR28 (Fig. 4A); wild-type transgenic MOR28 without deletion was used as a control (Fig. 4B). OE sections of the transgenic mice were first hybridized with RNA probes of the endogenous OR genes and immunostained with anti-EGFP to detect transgene expression. In Fig. 4, the OSNs expressing the transgene are shown in green, whereas those expressing the endogenous OR genes are in red. Most OR genes that share expression areas with the MOR28 gene yielded yellow-stained OSNs in the merged pictures, indicating their coexpression with del-MOR28. Examples for MOR230-1, MOR244-2, MOR248-2, and MOR246-2 are shown in Fig. 4A. In all transgenic lines examined, more than 75% of the OSNs with activated del-MOR28 coexpressed one of the 40 tested endogenous OR genes. We expect this fraction to be higher with an expanded repertoire of OR gene probes. No yellow-stained cells were detected with the control transgene, the EGFP-tagged wild-type MOR28, for all three lines examined (Fig. 4B).

Fig. 4.

Coexpression of the del-MOR28 transgene with the endogenous OR genes. (A) Coronal OE sections of the YAC-290GdC mice were hybridized with various RNA probes of endogenous OR genes (stained red). The same sections were then immunostained with anti-EGFP to detect the expression of del-MOR28 (stained green). Examples of the four endogenous OR genes are shown (MOR230-1, MOR244-2, MOR248-2, and MOR246-2). Coexpression rates of del-MOR28 with various OR genes were calculated for the YAC-290GdC line 1 mouse. The percentage of the probe-positive cells among ∼500 del-MOR28–expressing cells is shown. Because the coding-region probes are cross-hybridizable with related genes, they detect multiple OR genes belonging to the same subfamily. The total coexpression rate with 13 different probes was 76.8%. (B) The OE of the YAC-290G mouse was analyzed in parallel as a negative control. No yellow-colored cells were detected with any OR gene probes tested thus far. Scale bars, 50 μm.

We then examined the projection patterns of OSNs expressing del-MOR28. Whole-mount views of the olfactory bulbs were compared for the wild-type MOR28 and del-MOR28 transgenes (Fig. 5, A and B). Projecting axons were tagged by EGFP. OSNs expressing the endogenous MOR28 gene or the wild-type transgene projected their axons to a specific site in the posteroventral portion of the olfactory bulb (Fig. 5A). In contrast, the axons of OSNs that expressed del-MOR28 did not converge, but appeared to wander and target to multiple glomeruli near the MOR28 projection site (Fig. 5B). To examine the projection more closely, we immunostained serial olfactory bulb sections with anti-EGFP. Multiple glomerular structures were targeted by the del-MOR28–expressing OSNs (Fig. 5D). It is possible that these glomeruli represent the target sites for the secondarily chosen OR genes in the del-MOR28–expressing OSNs. In the control mouse, whose endogenous MOR28 gene had been tagged with EGFP, only one distinct glomerular structure was detected in the sections from the posteroventral portion of the olfactory bulb (Fig. 5C).

Fig. 5.

Axonal projection of OSNs expressing the EGFP-tagged MOR28. (A) Whole-mount views of the lateral aspect of the olfactory bulbs visualized by the fluorescence of EGFP. Both the endogenous MOR28 (knock-in) and the transgenic MOR28 were analyzed as positive controls. Directions: A, anterior; P, posterior; D, dorsal; V, ventral. (B) Whole-mount staining of the olfactory bulbs isolated from the del-MOR28 transgenic mice. Two independent lines were analyzed. Fluoresced axons did not converge to a specific site in both lines. (C and D) Coronal sections of the olfactory bulbs were immunostained with anti-EGFP (red) and counter-stained by 4′,6′-diamidino-2-phenylindole (DAPI) (blue). Three consecutive sections (each separated by 96 μm) are shown. Broken lines demarcate the olfactory nerve from the glomerular layers. Scale bars, 500 μm [(A) and (B)], 200 μm [(C) and (D)].

Frame-shifted pseudogenes enable concurrent OR gene expression. A substantial number of pseudogenes are present in the mouse OR gene family (1820). Because our transgenic studies indicated an inhibitory role for OR gene products, we expected that some of the pseudogenes would allow the secondary activation of other OR genes, even after the pseudogenes had been chosen for expression. To examine this possibility, we searched for OR pseudogenes in the C57BL/6J mouse genome database from the Mouse Genome Sequencing Consortium. At least 100 OR genes were confirmed to be pseudogenes; some lack the start codon, some contain a frame shift, and others have stop codons in their coding sequences. Among 33 genes examined by in situ hybridization, 3 were found to give sufficient signals for transcription in the OE.

To test for coexpression, we used the transgenic minigene H-MOR28 tagged with ECFP (expressing enhanced cyan fluorescent protein) as a partner, because of the enhanced frequency of the H-MOR28 activation and mutually exclusive mode of expression. OE sections of the H-MOR28 transgenic mouse were hybridized with the pseudogene probes and then immunostained for ECFP. Three pseudogenes, MOR171-43, MOR135-18, and MOR248-12, were sequenced and found to contain frame shifts and premature stops (Fig. 6A). Hybridization of OE sections with these probes yielded double-stained OSNs in the superimposed pictures (Fig. 6B). For example, the MOR171-43 probe gave 35 coexpressing cells (shown in yellow) among 635 cells expressing MOR171-43 (stained red). As a positive control for double staining, the OE section of the H-MOR28 transgenic mouse was hybridized with the MOR28 probe and treated with anti-EGFP. All of the transgene-expressing OSNs turned yellow (Fig. 6C). As negative controls, probes for functional OR genes, MOR171-1 and I7, were used for in situ hybridization. Although the cells expressing the partner gene (shown in green) were very crowded, the OSNs that hybridized with the OR gene probes (shown in red) were not yellow in the merged pictures (Fig. 6C). These results with the pseudogenes further confirm the notion that the OR gene products act as inhibitors to prevent the further activation of other OR genes.

Fig. 6.

Coexpression of the OR pseudogenes with the transgenic H-MOR28 and a model for the OR gene expression. (A) Structures are shown for three pseudo OR genes. The frame-shift starting points are indicated by arrows. Residue numbers are for nucleotides. Nucleotide deletions or additions that have caused the frame shift are shown in boxes. Stop codons created by the frame shift are indicated by arrowheads. Frame-shifted peptide regions are in red. (B) OE sections of the H-MOR28 transgenic mouse were subjected to in situ hybridization with pseudo OR gene probes of MOR171-43, MOR135-18, and MOR248-12 (stained red). The same OE sections were then immunostained with anti-EGFP to detect ECFP for the transgene-expressing cells (stained green). Two pictures were superimposed (merge) for analysis of coexpression. The ratios of coexpressing cells (shown in yellow) to the total red-colored cells are shown at the bottom. (C) Functional OR genes, MOR171-1 and I7, were examined as negative controls. (D) Schematic diagram of the mutually exclusive expression of the OR genes. We assume that the activation complex formed in the H region stochastically chooses one promoter (P) site by random collision, activating one particular OR gene member within the cluster. We propose that once the activated gene is expressed, the functional OR molecules transmit an inhibitory signal to block the further activation of additional OR genes or clusters.

Activation of one OR gene within the cluster. It is now well established that each OSN expresses only one member of the OR gene family in the mouse olfactory system (25). However, it is still not clear how one OR gene is chosen for expression, and how the remaining OR genes are kept silent. We have shown that the H region far upstream of the MOR28 OR gene cluster is responsible for the activation of all OR genes in the cluster. How is it that transgenic MOR28 is chosen so frequently when the H region is relocated closer to the promoter? Why does selection of downstream genes in the YAC-H140 construct become down-regulated? We suggest that a transcription-activating complex is formed in the H region that possibly interacts with a promoter site, activating the OR gene (Fig. 6D). It is possible that the smaller the distance between the H region and promoter, the more efficiently the OR gene is activated and expressed. This probably explains why the H region is located far from the OR gene cluster. If the H region were located too close to the cluster, the proximal OR gene would be activated quite frequently and the downstream genes would rarely be chosen for expression. It is important to ask how the expression of one particular OR gene within the activated cluster is ensured. We propose that the activation complex formed in the H region interacts with only one promoter site in the cluster.

Negative feedback regulation by the OR gene product. We have shown that a coding-region deletion of the transgenic MOR28 gene permitted the activation of secondary OR genes in the transgene-expressing OSNs. Similar observations were also made with natural frame-shift mutants. These observations indicate that the OR gene product, most likely the protein, has a regulatory role in the expression of the OR genes. This model explains some curious observations previously made by other investigators. For example, Qasba and Reed (21) reported that the transgenic lacZ gene carrying the OR gene M4 promoter was coexpressed with the endogenous M4 gene in 1% of the lacZ-positive OSNs. The transgene-expressing cells probably allowed the activation of the endogenous M4 promoter because of the absence of the functional receptor. A similar observation was made by Pyrski et al. (22) using the lacZ transgene with the OMP promoter that was integrated into the OR-Z6 cluster. Interestingly, the transgene-expressing OSNs projected their axons to multiple glomeruli in the olfactory bulb, consistent with our observations of the del-MOR28 gene. It is possible that the OSNs expressing truncated ORs allow further activation of other OR genes and project their axons to targets determined by the coexpressed ORs. Furthermore, Wang et al. (23) found that deletions and nonsense mutations in the lacZ-tagged P2 gene resulted in the inability of the OSNs to project their axons to a specific set of glomeruli.

These observations as well as our own results support the idea that the functional expression of an OR gene inhibits the activation of other OR genes. How is this negative regulation achieved? In the immune system, the pre-B cell receptor provides an inhibitory signal via Syk family tyrosine kinases (24) to prohibit the further activation of VH (the variable region of immunoglobulin heavy chain genes), ensuring the allelic exclusion of these genes (2529). We postulate a similar inhibitory role for the functional OR proteins. Stochastic activation of an OR gene and negative feedback regulation by the OR gene product, as shown in Fig. 6D, may ensure the maintenance of the one receptor–one neuron rule in the mammalian olfactory system.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1089122/DC1

Materials and Methods

References

References and Notes

View Abstract

Navigate This Article