rAAV2/5 as a vector for gene delivery to mammalian OSNs
Given limited prior knowledge of the tropism of different AAV serotypes in the mammalian nasal epithelium, we tested an rAAV5-based vector not reported previously to specifically transduce rodent OSNs but known to display a broad tissue tropism 18. A bicistronic cassette encoding firefly Luciferase and mApple fluorescent protein linked by the furin-cleavage signal and ribosome skipping peptide 2A from a foot-and-mouth virus (Fig. 1A, top), was used to produce the rAAV5 vector 19. A single dose of rAAV5 (10-20 µl, titer ca. 1011 vg/ml) was administered intranasally to anesthetized rats and mice. Robust Luciferase activity, restricted to the nasal cavity only, was detected 7 days later in a mouse (Figure 1A). The same animal used for in vivo Luciferase imaging was subsequently transcardially perfused and analyzed for the mApple expression in the olfactory epithelium (OE) (Figure 1B). rAAV5 successfully induced expression in multiple OSNs along with non-sensory sustentacular cells (Figure 1B). Similarly, en face imaging of the acutely isolated OE from the mouse treated with the rAAV5 revealed a mixed population of transduced OSNs and sustentacular cells (Figure 1C). At higher magnification multiple cilia emanating from the dendritic knob of the OSN could be clearly resolved (Figure 1D).
rAAV5-mediated transduction of the olfactory epithelium and ectopic expression of the OR in rodent OSNs
Following initial characterization of transduction of the OE by rAAV5, we subcloned full-length Olfr73, a Class II OR (other names MOR174-1, mOR-EG) 1 and a genetically-encoded calcium indicator GCaMP3 in to the same vector by substituting the respective ORFs (Figure 2A, top). We hypothesized that OSNs should have all necessary endogenous cellular mechanisms for the correct expression and trafficking of the OR protein to cilia. Hence, no additional modification of the OR gene via fusing the N-terminus tag to improve plasma membrane localization was made, thus ensuring that the expression in OSNs identical to its natural counterparts.
First, functional expression of the OR and GCaMP3 was validated in HEK293T cells by co-transfecting the pTR plasmid encoding the bicistronic cDNA used to produce rAAV5 with the promiscuous G-protein Gα15, conferring coupling to the downstream calcium release signaling cascade. A 5 sec application of eugenol (100 µM) evoked robust elevation of intracellular calcium (Supplemental Figure S6B). Our result confirmed earlier reports of functional expression of untagged full-length Olfr73 in HEK293 cells 20 and proved the functionality of the bicistronic expression cassette.
Next, a single intranasal dose of rAAV5 was used to induce expression of the bicistronic construct in the OE, as evidenced by the fluorescence of GCaMP3 in many cells at a level sufficiently strong to visualize dendritic knobs with their attached cilia (Figure 2, Supplemental Figure S1D). This allowed visualizing functional activity within the transduction compartment in semi-intact tissue with high resolution (Supplemental movie S1, S2). In response to a brief 5-sec pulse of eugenol (1 mM), we observed an odor-evoked global cilia-to-knob response as well as residual activity in the cilia post-stimulation (Figure 2B, arrows). These results clearly suggest that both Olfr73 and GCaMP3 were properly localized to the transduction compartment. Transduced mouse OSNs responded to lower concentrations of eugenol (5 µM) and showed left-shifted concentration-response functions compared to rat OSNs (Figure 2D; 3D). Unexpectedly, methylisoeugenol (MIEG, 1 mM), a previously characterized antagonist of eugenol on Olfr73 15, also evoked a measurable response. The latter finding set the foundation of the current study.
To confirm the general utility of the rAAV5 delivery and expression system we subcloned another mouse OR, a Class I Olfr599 (MOR23-1), using the same bicistronic cassette (Fig. 1A, top). First, we verified that Olfr599 was functionally expressed in our expression system either as a rho-tagged OR or as the untagged OR. Octanoic acid, a ligand of the OR 8 consistently activated a response in a concentration-dependent manner with an EC50 = 70 µM (n=21) (Supplemental Figure S1A,B). Importantly, transfection of untagged Olfr599 using the bicistronic pTR-Olfr599-furin2A-GCaMP3 plasmid also resulted in functional expression of the OR along with the calcium reporter (Supplemental Figure S1C). Similarly, intranasal infusion of rAAV2/5-Olfr599-furin2A-GCaMP3 induced expression as evident by GCaMP3 fluorescence in the knobs of mouse OSNs (Supplemental Figure S1D). Octanoic acid and octanal (each at 100 µM) and a mix of IBMX and forskolin (100/10 µM), a chemical activator of most OSNs, all evoked reliable neuronal responses (Supplemental Figure S1E).
Finally, we asked whether rAAV-mediated potential over-expression of ectopic ORs could have induced an elevated basal activity known to be dependent on the type of the OR 21. As a proxy for basal neuronal activity, we counted Olfr73-GCaMP3-OSNs (Olfr73-OSNs) showing aberrant calcium bursts. In the entire pool of Olfr73-OSNs, we found only 6.44% (13 of 202 cells) spontaneously active Olfr73-OSNs, generating bursts at 0.08 ± 0.03 Hz (n=13). These values were even lower than 12.9% of unidentified OSNs bursting at 0.16 ± 0.01 Hz reported in a recent study 22. This confirms previously published data of low basal activity of native Olfr73-ires-tauGFP OSNs, as well as Olfr73 expressed in HEK293 cells 3,21. This validated the prior understanding that ectopic expression of a different OR in a mature OSN can be used as a reliable model to study ORs in the native cellular environment 13.
Pharmacological profile of OSNs ectopically expressing Olfr73
First, we validated that the OR in question confers its previously characterized molecular receptive range (MRR). Olfr73-OSNs were challenged with two ligands outside the known MRR of Olfr73, amyl acetate and acetophenone (both at 100 µM), followed by increasing concentrations of eugenol. Amyl acetate and acetophenone evoked variable responses in different Olfr73-OSNs, reflecting the expected expression of different endogenous ORs in these cells (Figure 3A), while application of eugenol consistently evoked a response in all cells in a concentration-dependent manner (Figure 3B). In each case, the response profile matched that expected of Olfr73 (Figure 3B). The agonists eugenol and isoeugenol evoked responses in mouse Olfr73-OSNs in a concentration-dependent manner, yielding EC50 values of 4.8 ± 1.1 µM (n=18) and 117 ± 45 µM (n=24), respectively (Figure 3B,C). So did rat Olfr73-OSNs, but they were nearly 10-fold less sensitive to eugenol than mouse OSNs, yielding a right-shifted dose response curve with an EC50 of 41.2 ± 12.4 µM (n=13) (Figure 3D). MIEG was a partial agonist for Olfr73 expressed in native mouse OSNs, yielding an EC50 of 430 ± 130 µM (n=15) (Figure 3D). Isosafrole, a putative antagonist of Olfr73 16, and nootkatone, a putative highly potent agonist of Olfr73 9, even applied at high concentration (1 mM), only activated Olfr73-OSNs at 25% (isosafrole, n=11) and 50% (nootkatone, n=7) of the response to eugenol (1 mM) (Figure 4A, B, D). The potency of nootkatone was consistent for both rat and mouse Olfr73-OSNs (Figure 4B), suggesting that in native OSNs nootkatone acts only as a partial agonist. Lastly, we showed that the inhibitor of adenylyl cyclase, SQ22,536 (200 µM), strongly reduced the response to MIEG, confirming that it was mediated through a canonical cAMP-dependent signaling pathway (Supplemental Figure S2) and that MIEG is a low-potency partial agonist for Olfr73 ectopically expressed in native OSNs.
Ectopically expressed Olfr73 does not mediate ligand-dependent antagonism in native OSNs
Interaction of ligands in binary mixture may show antagonism even though one of the components alone acts as a partial or low-potency agonist 23,24 so we sought to further characterize the interaction between eugenol and several potential antagonists. First, we established the response profile of the Olfr73-OSNs to repeated stimulation with a 5-s pulse of eugenol (100 µM) followed by the pulse of the putative antagonist (1 mM) (Figure 4A). While MIEG evoked responses of similar amplitude relative to eugenol in both rat and mouse Olfr73-OSNs (Figure 4B), we limited these experiments to mouse OSNs to exclude any issues related to the reduced sensitivity of rat Olfr73-OSNs (Figure 3D). In addition to MIEG and isosafrole, we tested the chemical dimer of isoeugenol (di-IEG, 500 µM), identified as an antagonist of eugenol on Olfr73 16. In contrast to MIEG and isosafrole, di-IEG, as well as a dimer of eugenol (di-Eug, 500 µM) failed to activate any response in mouse Olfr73-OSNs (Figure 4B, E). Application of the putative antagonists in binary mixture with eugenol also failed to show any inhibition or additivity relative to the control response evoked by eugenol alone (Figure 4C, F). Furthermore, the response to eugenol at its EC50 concentration (10 µM) mixed with MIEG (1 mM) was not significantly changed from the control response to eugenol alone (Figure 4D, F), suggesting that pre-incubation with a putative antagonist may impose stronger inhibition. Both dimers, di-Eug and di-IEG (500 µM), failed to elicit a response when applied alone for 5-s (Figure 4B, E). However, pre-incubation for 30-s with di-IEG elicited a small truncated elevation of the GCaMP3 signal, with no attenuation of the response to the binary mix with added eugenol (Figure 4E). Co-application of di-IEG resulted in additive increase of the response to the binary mix with eugenol (Figure 4F), suggesting that the dimers of eugenol and isoeugenol did antagonize the response to eugenol. Overall, we failed to identify any antagonistic or inhibitory interaction of eugenol and its putative antagonists in native mouse OSNs ectopically expressing Olfr73.
In vitro-expressed Olfr73 mediates antagonism between ligands independently of the downstream signaling pathway.
Since previous research suggested that antagonism in such binary mixtures of ligands may depend on the signaling pathway downstream of the OR in question 25, we explored the effect of substituting different G proteins in our in vitro assay. First, we confirmed that mammalian ORs can robustly couple to endogenous stimulatory signaling pathways, mediated not only by Gs/olf but also by Gq11, found in ciliary proteome in mammalian OSNs 26. Eugenol (100 µM) consistently activated calcium release in HEK293 cells co-expressing Olfr73 with the complete heterotrimeric G-protein, Gαq11/ß1/γ13 (Figure 5A). Using this expression system, we then measured the concentration-dependence of the response to eugenol, vanillin and isoeugenol (Figure 5B). Co-expression of another mouse OR, Olfr599 with the heterotrimeric G-protein, Gαq11/ß1/γ13, also conferred robust activation by its cognate ligand, octanoic acid (100 µM), as was the case for Gs/olf and Gα15-dependent signaling (Supplemental Figure S3). Thus, mammalian ORs have the capacity of coupling to the heterotrimeric Gq11/b1/γ13 G-protein naturally expressed in olfactory cilia.
We then characterized the interaction between cognate agonists and putative antagonists of rho-tagged Olfr73 co-expressed with either Gαq11/ß1/γ13, Gs/olf , or the promiscuous G protein, Gα15. We validated the earlier finding that Olfr73 transiently expressed in HEK293 cells conferred the antagonism imposed by MIEG on the response to eugenol and another strong agonist, vanillin (Figure 5; Supplemental Figure S4, S5). The response to vanillin (100 µM) was only partially inhibited by co-application of MIEG (1 mM), but was completely blocked following 30-s pre-incubation with MIEG (Figure 5D, E; Supplemental Figure S4, S5). Carvone (1 mM), distantly structurally related to eugenol but a non-agonist for Olfr73 27 failed to antagonize the response to eugenol even following pre-incubation for 30-s (Figure 5E; Supplemental Figure S4, S5). These findings argue that antagonistic interactions between cognate agonists and putative antagonists can occur independently of which signaling pathway is downstream of ORs expressed in vitro.
Interaction of binary mixtures with the in vitro-expressed Olfr73 is dependent on the stimulus paradigm
We used the Cre-SEAP assay, similar to the more widely used Cre-Luciferase assay 5,9 to measure cAMP generated by eugenol applied for a prolonged time to HEK293 cells co-expressing untagged Olfr73 and Gs/olf (Supplemental Figure S6A). MIEG alone also generated cAMP underscoring our finding that in this assay a putative antagonist acts as an agonist (Figure 6A). Surprisingly, we observed no antagonistic interaction between several concentrations of eugenol and MIEG, showing instead an additive effect in net activity (Figure 6A). We used untagged Olfr73 instead of having the receptor rho-tagged on the N-terminus to improve plasma membrane trafficking in assays of in vitro-expressed ORs. In order to address whether the stimulus protocol by itself may have shaped the result, we co-expressed in HEK293 cells the rAAV2/5-targeting plasmid pTR-Olfr73-furin2A-GCaMP3 along with Gα15. In this context eugenol (100 µM) elicited a robust calcium signal that was significantly diminished after addition of MIEG (1 mM). MIEG alone did not elicit any response (Supplemental Figure S6B,C). From this we assumed that modification of the N-terminus of Olfr73 per se did not affect the OR’s ligand binding in vitro nor its coupling to the Gs/olf signaling pathway.
Then, to address the issue of stimulus duration, we measured cAMP-dependent activity in cells co-expressing untagged Olfr73 along with Gs/olf and CNGCmut channel as a cAMP sensor. To ensure that prolonged incubation with ligands did not saturate the cAMP readout, we used isoproterenol (10 µM) as a saturating readout. Isoproterenol applied for 5-min activated endogenous adrenergic receptors coupled to Gs/cAMP signaling and generated a robust sustained response exceeding any odorant-evoked response (Figure 6B). Eugenol (100 µM) applied for 10-min generated a sustained cAMP-dependent calcium influx (Figure 6B; Supplemental movie S3). A 10-min application of MIEG (1 mM) evoked a more slowly developing cAMP-dependent calcium influx resulting of smaller amplitude (Wilcoxon test, p=0.018, n = 3) (Figure 6B, C). Averaged across several independent measurements, however, steady-state elevation of the cAMP-dependent calcium influx evoked by a binary mixture of eugenol (100 µM) and MIEG (1 mM) showed no significant difference from the response evoked by eugenol alone (Wilcoxon test, p>0.9999, n = 4) (Figure 6C; Supplemental movie S4). Thus, two independent methods of detecting Olfr73-activated cAMP-dependent activity confirmed that prolonged stimulation of the OR expressed in vitro results in agonism by the same ligand that with brief stimulation is a potent antagonist.