Optimization of Insect Odorant Receptor Trafficking and Functional Expression Via Transient Transfection in HEK293 Cells

Abstract Insect odorant receptors (ORs) show a limited functional expression in various heterologous expression systems including insect and mammalian cells. This may be in part due to the absence of key components driving the release of these proteins from the endoplasmic reticulum and directing them to the plasma membrane. In order to mitigate this problem, we took advantage of small export signals within the human HCN1 and Rhodopsin that have been shown to promote protein release from the endoplasmic reticulum and the trafficking of post-Golgi vesicles, respectively. Moreover, we designed a new vector based on a bidirectional expression cassette to drive the functional expression of the insect odorant receptor coreceptor (Orco) and an odor-binding OR, simultaneously. We show that this new method can be used to reliably express insect ORs in HEK293 cells via transient transfection and that is highly suitable for downstream applications using automated and high-throughput imaging platforms.

Several methods have been used to express insect ORs in heterologous systems in order to characterize their ligand specificity and to study their functional properties (Fleischer et al. 2018). For example, the so-called "empty neuron system" allows the ectopic expression of ORs in subsets of olfactory sensory neurons of the vinegar fly Drosophila melanogaster lacking the endogenous OrX, but with a fully functional Orco (Dobritsa et al. 2003;Kurtovic et al. 2007;Gonzalez et al. 2016). Moreover, in vitro expression systems allow the heterologous expression of OR proteins in animal cells including-among others-Xenopus oocytes (Nakagawa et al. 2005;Wang et al. 2010;Nakagawa and Touhara 2013), insect cells (Kiely et al. 2007;Tsitoura et al. 2010;German et al. 2013), and mammalian cells (e.g., HEK293 and CHO) using vectors for transient expression (Sato et al. 2008;Wicher et al. 2008) or as stable lines (Grosse-Wilde et al. 2006;Jones et al. 2011;Corcoran et al. 2014).
Each of these methods offers several advantages, but also bears disadvantages: the "empty neuron system" allows the expression of OR proteins in an environment that is very similar to their native olfactory sensillum. However, the generation of transgenic fly lines and the electrophysiological recording tests can be time consuming and are not suitable for high-throughput screening experiments. Xenopus oocytes have been successfully used for extensive screenings of odors and are compatible with automated platforms, but the costs associated with Xenopus rearing or the purchase of oocytes can be prohibitive for very large screenings. Stable and inducible cell lines, on the other hand, originate from cell clones that successfully integrated in their genome an expression cassette containing a tetracycline-dependent promoter driving the expression of Orco and an odor-binding OrX. Such system represents a sort of "gold standard" as cell variability is strongly reduced, compared to transiently transfected cells, by selection of monoclonal populations and non-induced cells constitute an optimal internal negative control. Moreover, the selection of a monoclonal cell population represents-to date-the most effective approach to deal with the limited functional expression of ORs in the plasma membrane in heterologous systems due to an impaired intracellular trafficking of ORs both in insect (German et al. 2013) and mammalian cells (Halty-deLeon et al. 2016).
Although a thorough investigation of the bottlenecks in insect OR intracellular trafficking has not been performed yet, their retention within intracellular membranes (German et al. 2013) may be due to the lack of specialized components of the OR release mechanism from the endoplasmic reticulum (ER) and the Golgi apparatus in heterologous systems. Several membrane proteins have been shown to direct their intracellular trafficking through small peptide regions. Among others (Schülein et al. 1998;Ammon et al. 2002), it has been shown that an N-terminal peptide ( 106 VNKFSL 111 ) from the human HCN1 channel facilitates the exit of HCN1 proteins from the ER (Pan et al. 2015), and the C-terminal portion of the human Rhodopsin ( 344 QVAPA 348 )-containing the "VxPx" motif-is sufficient to promote the formation of post-Golgi vesicles and their trafficking to the plasma membrane via microtubule-mediated transport (Tai et al. 1999;Deretic et al. 2005;Mazelova et al. 2009;Lodowski et al. 2013).
In this work, we investigated whether insect ORs tagged at their N-terminus with the HCN1 and Rhodopsin signal peptides show an enhanced trafficking to the plasma membrane in mammalian cells and reach satisfactory functional expression levels even after transient transfection without clonal selection. Moreover, in order to increase the efficiency of gene cotransfection, we designed a new vector based on a commercially available dual-promoter plasmid for mammalian transient transfection. In this way, we aimed at creating a new fast and inexpensive strategy to reliably express insect ORs in mammalian cells for a broad range of application, including automated and high-throughput imaging systems.
A human codon-optimized version of D. melanogaster Orco (hOrco) tagged at the N-terminus with a myc tag (5′-GAACAGAAA CTGATCTCTGAAGAAGACCTG-3′) was synthesized by Eurofins Genomics GmbH and subcloned into the pCMV-BI vector using the BamHI and HindIII restriction sites. A version of hOrco bearing the β-globin/IgG chimeric intron from the pCMVTNT vector within the hOrco sixth transmembrane domain was constructed by Phusion polymerase amplification using the hOrcoExon1_fwd, hOrcoExon1_ rev, chimeric_intron_fwd, chimeric_intron_rev, hOrcoExon2_fwd and hOrcoExon2_rev primers. A new vector called pDmelOR was then created by cloning this intron-containing version of hOrco into the pCMV-BI vector, after digestion with BamHI and HindIII, using the NEBuilder HiFi DNA Assembly kit.

Targeting of insect ORs to the plasma membrane
We first tested whether the 344 QVAPA 348 (minimal Rho tag, abbreviated as "mRho" or "R" tag) and the 106 VNKFSL 111 (minimal ER export tag, abbreviated as "mER" or "E" tag) peptides from the human Rhodopsin and the HCN1 channel, respectively (Figure 1a) could improve the intracellular trafficking of hOr47a to the plasma membrane in mammalian cells.
For this purpose, we performed functional imaging experiments on HEK293 cells transiently transfected with the constructs shown in Figure 1b and loaded with the calcium dye Fura-2. We stimulated the hOr47a/Orco heteromers with the Or47a agonist pentyl acetate and the Orco synthetic agonist VUAA1 (Figure 1c). We quantified the increase in intracellular free calcium with respect to the base level (∆[Ca 2+ ] i (nM)) ( Figure 1d) and the distribution of Ca 2+ responses (Figure 1e and Supplementary Figure 1) for each tested construct (i.e., treatment), in order to compare the number of excited cells after presentation of the odor pentyl acetate and the Orco agonist VUAA1.
HEK293 cells cotransfected with the R.E.hOr47a and Orco constructs showed significantly higher calcium responses after stimulation with 100 µM pentyl acetate and 100 µM VUAA1, compared to cells transfected with an untagged version of hOr47a ( Figure  1d). Moreover, the R.E.hOr47a construct induced significantly higher calcium responses to pentyl acetate than the E.hOr47a construct bearing the mER tag alone (Figure 1d, left panel). In order to calculate and compare the number of responding cells between the tested treatments, we used the distribution of Ca 2+ responses from HEK293 cells transfected with the empty pcDNA3.1(-) vector to calculate threshold values to classify cells as "responding" or "nonresponding" to an OR agonist. To do so, we calculated the mean response intensity + 2 × SD (Δ[Ca 2+ ] i , in nM) of the top (most responsive) 0.5 percentile of the cumulative distribution of analyzed control cells (transfected with empty vector) after a pentyl acetate and VUAA1 stimulation. The resulting threshold values were 6.75 nM for pentyl acetate and 6.25 nM for VUAA1 (Supplementary Figure 1). Only 2.9 ± 1.05% (mean ± SD) of cells transfected with hOr47a/Orco showed a ∆[Ca 2+ ] i ≥ 6.75 nM after a 100 µM pentyl acetate stimulation, while 21.61 ± 4.16% of cells transfected with E.hOr47a/Orco and 30.51 ± 6.56% of cells transfected with R.E.hOr47a/Orco reached such threshold. When comparing cell responses after a 100 µM VUAA1 stimulation, only 30.24 ± 7.21% of cells transfected with hOr47a/Orco showed an ∆[Ca 2+ ] i ≥ 6.25 nM, while 61.49 ± 5.95% and 71.76 ± 3.29% of cells transfected with E.hOr47a/Orco and R.E.hOr47a/Orco, respectively could be defined as "responding" (Figure 1e and Supplementary Figure 1). These results show that the mRho and mER tags can significantly enhance the intensity of the Ca 2+ response and the number of responding units in functional imaging experiments with insect OR transfected cells. Moreover, they show that the addition of mRho has a significant positive effect compared to the mER tag alone.

Optimization of OR cotransfection in mammalian cells
Although vectors with 2 expression cassettes disposed in series have already been used successfully (Bohbot et al. 2011), we designed a new vector bearing a bidirectional expression cassette to improve the chances of successful cotransfection for a tuning OR and the coreceptor Orco. The rationale for this choice was to minimize the size of the empty vector backbone, by flanking a single CMV enhancer with 2 minimal CMV promoters. The vector consisted of the pCMVTNT plasmid backbone (including the amp resistance gene and a high-copy number ori) and the bidirectional expression cassette of the pBI-CMV1 vector, in order to guide the expression of 2 genes simultaneously (see Methods). We first inserted a human codon-optimized version of D. melanogaster Orco (hOrco) tagged at the N-terminus with a myc tag in correspondence of the multiple cloning site 1 of the vector. As the CMV promoter can initiate expression in Escherichia coli (Lewin et al. 2005) and Orco forms leaky ion channels, the unintended expression of Orco during cloning may reduce the viability of successfully transformed bacterial colonies. To avoid such inconvenience, we inserted a β-globin/IgG chimeric intron within the Orco sixth transmembrane domain. In this way, only mammalian cells can splice out the intron, leading to the production of functional Orco ion channels. The resulting plasmid, named pDmelOR, is intended to serve as a plasmid backbone to insert one of the 61 (including splice variants) tuning ORs of D. melanogaster (Robertson et al. 2003), and the same principle can be adapted to optimize the functional expression of ORs belonging to any insect species.
In order to evaluate the performance of the bidirectional pDmelOR vector against a standard cotransfection protocol, we compared the response profile of HEK293 cells transfected with pDmelOR-R.E.hOr47a to that of cells cotransfected with an Orco and an R.E.hOr47a construct inserted in a pcDNA3.1(-) each (Figure 2a-c). HEK293 cells transfected with pDmelOR-R.E.hOr47a showed a significantly higher ∆[Ca 2+ ] i in response to both pentyl acetate and VUAA1 stimulation, with respect to cells cotransfected with R.E.hOr47a and Orco in pcDNA3.1(-) (Figure 2d). Moreover, a significantly higher number of cells responded to pentyl acetate and after transfection with pDmelOR-R.E.hOr47a compared to cells cotransfected with R.E.hOr47a and Orco in pcDNA3.1(-) ( Figure  2e). These results propose the pDmelOR vector as an efficient transfection tool for the expression of insect OR fusion constructs with higher efficacy than standard plasmid cotransfection procedures.

Optimization of transient transfection for automated imaging platforms
Finally, we took advantage of the high level of functional expression reached combining the increased OR trafficking to the plasma membrane using the mRho and mER tags, together with the optimization of the transfection protocol due to the pDmelOR vector, to validate our method using an automated imaging platform (Figure 3).
A monoclonal stable cell line guarantees a highly homogeneous level of transgene expression within the cell population, which minimizes variability in functional assays. On the other hand, the expected variability in a population of transiently transfected cells is much higher, due to the variation in plasmid copy number between cells that results in different levels of functional expression of the gene of interest. To control for this phenomenon, we stimulated tested cells with VUAA1 100 s after the presentation of the odor stimulus, in order to use the intensity of the response to this synthetic OR agonist as a proxy of the OrX/Orco functional expression level for each cell. In this way, we could remove from downstream analyses those cells that were not expressing ORs at sufficient levels (see Methods), and we could normalize the odor response for each cell to the intensity of the VUAA1 response, thus minimizing the variability induced by the transient transfection protocol (Figure 3a-c).
Using this method, we built dose-response curves for 2 D. melanogaster ORs, namely the broadly tuned receptor Or47a and the narrowly tuned Or56a, stimulated with their main agonists, pentyl acetate and geosmin, respectively, and in both cases, the data could be used to obtain high quality fits (Figure 3d-e). We then tested whether the tuning properties for the Or47a receptor reported in the DoOR database (Münch and Galizia 2016) could be replicated in HEK293 cells. To do so, we selected 9 odors among the known most potent agonists for Or47a (Supplementary Figure 2) and assessed their relative potency using our assay (Figure 3f). Interestingly, when ranked according to their expected potency, compared to the DMSO negative control, the odor agonists did not show a monotonic increasing pattern. Odor properties as solubility in water-based solutions may play a role in explaining such pattern. Moreover in the case of 3-octanol, the odor application significantly affected cell activity independently of Or47a expression, as cells transfected with the negative control plasmid showed a significant reduction in the cell fluorescence base level (Figure 3i). However, we further investigated whether odor purity could have influenced our results. To do so, we tested 2 odors (propyl acetate and 3-methylthio-1-propanol) whose relative potency differed from the expected pattern (compare Figure 3f with Supplementary Figure 2). For each odor, we tested 2 aliquots with different chemical grade. Interestingly, while the responses to propyl acetate were not affected by the odor chemical grade (Figure 3g), the responses to 3-methylthio-1-propanol were significantly affected by this factor, indicating that 3-methylthio-1propanol might not be an actual agonist of Or47a and the source of OR activation might originate from impurities within the extract (Figure 3h).

Discussion
Functional expression in heterologous systems represents a key method to elucidate the function and structure of membrane protein.
When confronted with the choice of which expression system to use, there are several factors to consider: from the codon usage of the gene of interest and necessary post-translational modifications (Gomes et al. 2016), to the type of downstream applications and the level of automation required. HEK293 cells represent a well-understood and very versatile choice: their transcriptome has been extensively profiled (Sultan et al. 2008;Richard et al. 2010) providing fundamental information regarding possible cross-talk between heterologous and native proteins. Furthermore, an extensive set of molecular tools has already been optimized to support protein functional expression (for an overview, see Baser and van den Heuvel 2016), and these tools are amenable to a vast array of downstream applications, from imaging to electrophysiology, that can be performed with automated and high-throughput systems (Mattiazzi Usaj et al. 2016;Obergrussberger et al. 2018).
On such a basis, we decided to implement a fast, inexpensive, and versatile method for the expression of insect ORs in HEK293 cells. In order to achieve this result, we here tackled 2 main problems: the poor surface localization of OR proteins in mammalian cells and the limits imposed by the cotransfection of 2 genes (Orco and an odorbinding receptor) to obtain functional odor-gated ion channels. In order to improve the surface localization of insect ORs, we prepared fusion constructs carrying at the N-terminus small epitopes from the human HCN1 (mER) and Rhodopsin (mRho) that are known to facilitate the release of membrane proteins from the ER and the targeting to the plasma membrane, respectively. Using this method, we obtained a nearly 20-fold increase in the mean response ( Figure  1d) and a 10-fold increase in the number of cells responding ( Figure  1e and Supplementary Figure 1) to odor stimulation. Although we did not quantify how the mRho and mER epitopes affected the abundance of OR proteins in the plasma membrane, a correlation between an increased amount of OR proteins that successfully reach the cell plasma membrane and the increase in the mean Ca 2+ response and the number of responding units is the most parsimonious explanation. Then, by adopting a new high-copy number vector based on a bidirectional expression cassette (Figure 2), we significantly improved the expression efficiency of ORs, showing that three-quarters of cells were stimulated with an overall 100-fold increase in the mean ∆[Ca 2+ ] i in response to an odor stimulation, when compared to a standard cotransfection protocol with ORs lacking the HCN1 and Rhodopsin-derived tags (Figures 1 and 2 and Supplementary Figure  1). Finally, thanks to the high expression level and the possibility to use synthetic Orco agonists as internal stimulus controls to account for intercell variability, we showed that such a system is amenable to be used with automated platforms (Figure 3) and can be consequently used for high-throughput screenings.
Although we proved the effectiveness of such a system for 2 D. melanogaster ORs with different properties-a broadly tuned receptor as Or47a and the narrowly tuned Or56a-it remains to be shown how generalizable such approach is. Mammalian ORs are affected by similar problems regarding an incorrect intracellular trafficking in heterologous expression systems. Although specific classes of proteins have been shown to support the trafficking of mammalian ORs in native olfactory neurons and in heterologous systems (Saito et al. 2004;Mainland and Matsunami 2012), conserved OR residues linked to in silico structural stability were shown to impact their functional expression (Ikegami et al. 2019); we cannot exclude that insect ORs are subject to similar structural constraints.
Taken together, our results show that by optimizing the intracellular trafficking and transfection conditions of insect ORs, it is possible via transient transfection to achieve expression levels in HEK293 cells that are comparable to more time and resourcedemanding methods such as the establishment of mammalian stable cell lines. Hence, we hope that such method can advance the study of insect ORs structure and function even in nonmodel organisms.

Supplementary material
Supplementary data are available at Chemical Senses online.