Neurons expressing trace amine-associated receptors project to discrete glomeruli and constitute an olfactory subsystem

Abstract

Some chemoreceptors of the trace amine-associated receptor (TAAR) family detect innately aversive odors and are proposed to activate hardwired olfactory circuits. However, the wiring of TAAR neurons, the regulatory mechanisms of Taar gene choice, and the subcellular localization of TAAR proteins remain unknown. Here, we reveal similarities between neurons expressing TAARs and odorant receptors (ORs), but also unexpected differences. Like ORs, TAARs seem to be monoallelically expressed and localized both in cilia, the site of odor detection, and in axons, where they may participate in guidance. TAAR neurons project to discrete glomeruli predominantly localized to a confined bulb region. Taar expression involves different regulatory logic than OR expression, as neurons choosing a Taar5 knockout allele frequently express a second Taar without silencing the deleted allele. Moreover, the epigenetic signature of OR gene choice is absent from Taar genes. The unique molecular and anatomical features of the TAAR neurons suggest that they constitute a distinct olfactory subsystem.

Fig. 1.

TAARs are highly expressed on dendrites. Sections of MOE were stained with antibodies against TAAR4 (A and B), TAAR5 (C and D) or TAAR6 (E and F). Antibodies against TAAR4 and TAAR5 recognized distinct OSN populations in the dorsal MOE (A and C). TAAR6+ OSNs are more ventral in the MOE (E). High levels of staining are observed on dendrites and in perinuclear structures. Scale bars, 50 μm A, C, and E; 20 μm B, D, and F.

Fig. 2.

TAAR neurons project to distinct glomeruli. Sections of OB were stained with antibodies against TAAR4 (A and A’), TAAR5 (B and B’), or TAAR6 (C and C’). Each antibody recognized a distinct group of glomeruli. We show representative glomeruli for each TAAR. (Scale bars, 300 μm in A–C; 50 μm in A’–C’.) (D and E) Costaining with rabbit antisera against TAAR5 (red) and guinea pig antisera against TAAR4 (green) demonstrates that OSNs expressing different TAARs project to adjacent but nonoverlapping glomeruli. (F) Rabbit (red) and guinea pig (green) antisera against TAAR5 label the same glomerulus. (Scale bars, 50 μm.) (G–J) Smoothed density plots for the glomeruli of TAAR4 (G), TAAR5 (H), TAAR6 (I), and M71 (J). The normalized distance from the AOB is presented on the y axis (1 is the anterior tip of the AOB). The normalized distance from the midline is presented on the x axis (0 is the boundary between the two bulbs). (K) Composite distribution of all TAAR4, TAAR5, TAAR6, and M71 glomeruli. Blue, TAAR4; green, TAAR5; red, TAAR6; black, M71.

Fig. 3.

OSNs that choose a deleted Taar5 allele predominantly reselect other Taars. MOE sections from mice that have the Taar5 coding sequence replaced by LacZ (Taar5^LacZ) were analyzed by two-color RNA FISH and by immunostaining. Except where indicated, Taar5^LacZ/+ animals were analyzed in all experiments. (A) A representative two-color RNA FISH experiment with probes for Taar6 (green) and LacZ (red) reveals distinct but overlapping populations of OSNs. The arrow marks a cell labeled by both probes. In a control experiment, two probes for LacZ (green and red) label the same OSNs. (B) Compilation and quantification of two-color RNA FISH experiments with probes for LacZ and for the various Taars or ORs (including a degenerate probe that detects multiple ORs), as indicated. The bars represent the percentage of LacZ+ neurons that were also labeled by the probe for each receptor, with specific cell counts indicated above. The red bar indicates neurons that are positive for Taar5 and for LacZ (indicating switching to the WT allele of Taar5). Note that the probes for Taar7a and Taar8a recognize the entire respective Taar subfamily. (Scale bars, 100 μm.) (C) MOE of Taar5^LacZ/LacZ stained with antibodies against TAAR5 (green) and β-gal (red). As expected, no signal is observed with the antibody against TAAR5. (Scale bar, 100 μm.) (D–F) MOE sections of Taar5^LacZ/+ were costained with antibodies against β-gal (red in all three panels) and TAAR4 (green in D), TAAR5 (green in E), or TAAR6 (green in F). Arrows in the Merge panels mark OSNs coexpressing β-gal and the corresponding receptors, indicating OSNs that reselected a new TAAR. (Scale bars, 50 μm.)

Fig. 4.

Expression of the Taar5-deletion allele persists after selecting another Taar. (A) X-gal staining of whole-mount preparations of OB from 7-mo-old Taar5^LacZ/+ mice. Both dorsal and ventral views are presented. (B–D) Costaining of OB sections from Taar5^LacZ/+ mice with antibodies against β-gal (red in all panels) and TAAR4 (green in B), TAAR5 (green in C), and TAAR6 (green in D). β-gal+ axons are observed in glomeruli for each TAAR. (Scale bars, 50 μm.)

Fig. 5.

TAAR glomeruli express the cell adhesion molecule OCAM. OB sections from WT (A–C) or Taar5^LacZ/+ mice (D) were stained with antibodies against OCAM (red), the presynaptic marker vesicular glutamate transporter VGLUT2 (blue), and TAAR4 (green in A and A’), TAAR5 (green in B and B’), TAAR6 (green in C and C’), or β-gal (green in D and D’). A’–D’ represent high-magnification images of the regions contained in the white boxes in the corresponding low-magnification panels (A–D). In all cases costaining of OCAM and the various TAARs or β-gal is observed. Note that OCAM+ fibers innervate a cluster of dorsal glomeruli that include the TAAR glomeruli. (Scale bars, 300 μm in A–D; 50 μm in A’–D’.)

Fig. 6.

Taar cluster is not covered by repressive heterochromatic marks. (A and B) ChIP-on-chip experiments for H3K9me3 and H4K20me3 in the MOE as previously described (10). (A) Part of an OR cluster on chromosome 10. The blue (H3K9me3) or red (H4K20me3) bars represent significant peaks (false discovery rate ≤5%) identified by model-based analysis of 2-color arrays using parameters: window = 10 kb, minimum number of probes = 20, and maximum gap = 1 kb. The black bars represent the enrichment of signal hybridization intensity over input DNA. (B) The Taar cluster, also on chromosome 10, is devoid of H3K9me3 or H4K20me3. (C–F) ChIP-qPCR analysis using native chromatin preparations from the MOE with various antibodies for repressive chromatin marks. In each case different genes serve as positive controls, and Omp is used as a negative control as indicated. The gene Olfr1320 is shown here as a representative OR. Values are the mean of duplicated qPCR. Error bars indicate SEM. (C) H3K9me3 ChIP-qPCR and (D) H4K20me3 ChIP-qPCR. Both Olfr1320 and major satellite repeats (Maj sat) serve as positive controls (10). (E) H3K9me2 ChIP-qPCR. Wnt2, which is not particularly enriched on ORs from a whole MOE preparation (10), serves as a positive control. (F) H3K27me3 ChIP-qPCR. Neurogenin1, a developmentally regulated gene in the MOE, serves as a positive control. The OR clusters do not exhibit Polycomb-mediated repression (10).