Drosophila melanogaster with genotype (w, UAS-GCaMP3.0; +; Or22a-Gal4), (w; UAS-GCaMP6f; Orco-Gal4), (w; UAS-GCaMP6f; Or22a-Gal4), and (+; Orco-Gal4/CyO; UAS-Syn21-GFP-p10/TM6B) were reared on conventional cornmeal agar medium under a 12 h light: 12 h dark cycle at 25°C. UAS-GCamP3.0; Orco-Gal4; UAS-GCaMP6f parental line was obtained from Bloomington Stock Center (#32234), (#42747), (#26818), and Or22a-Gal4 line was kindly provided by Dr. Leslie Vosshall, The Rockefeller University.

VUAA1 (N-(4-ethylphenyl)-2-((4-ethyl-5-(3-pyridinyl)-4H-1,2,4-triazol-3-yl)thio)acetamide) was synthesized by the group “Mass Spectrometry/Proteomics” of the Max Planck Institute for Chemical Ecology (Jena, Germany). Ethyl hexanoate (99%, Sigma-Aldrich), L-Cysteine hydrochloride (Cat. Nr. C1276, Sigma-Aldrich), Papain (Cat. Nr. 5125, Calbiochem, San Diego, CA, United States), cOmplete protease inhibitor cocktail (Cat. Nr. 04693116001, Roche, Basel, Switzerland), methanol (≥99.5%, Roth) HCl (≥32%, Sigma-Aldrich), H₂SO₄ (95∼97%, Sigma-Aldrich), sodium metasilicate solution (Cat. Nr. 13729, Sigma-Aldrich), low Ca²⁺-Schneider’s medium (Cat. Nr. S9895, Sigma-Aldrich), Benzaldehyde (Acros Organics, NJ, United States) were used in following experiments.

Flies with the genotype (w, UAS-GCaMP3.0; +; Or22a-Gal4) were anesthetized in ice for 30 min, and then placed into a truncated 1 ml pipette tip, leaving the head out of the tip and fixed using odor-free glue. The truncated tip was fixed using modeling clay on a custom Plexiglas mounting block (Strutz et al., 2012). Next, a custom holder to fix the antenna in vertical position was produced (Figures 1A–C and Supplementary Figure 1). A piece of aluminum foil (5 mm × 5 mm) with a slit (0.5 mm) for inserting the antenna and a small plastic ring (cut from an Eppendorf Microloader™) were glued on a glass coverslip (18 mm × 24 mm, #0 thickness, Menzel-Gläser, Braunschweig, Germany). The arista was then glued on the top of the glass coverslip with odor-free glue and the funiculus was manually cut at around half of its length with a scalpel blade #22 (Fine Science Tools, Heidelberg, Germany). Immediately after cutting the antenna, a small glass coverslip (15 mm × 15 mm, #00 thickness, Menzel-Gläser) moistened with a very thin layer of halocarbon oil 700 (Sigma-Aldrich) was laid on the open funiculus to seal it. Both coverslips were fixed to each other and to the Plexiglas holder with odor-free glue to prevent movement artifacts. Imaging was performed using a BX51W1 wide field fluorescence microscope (Olympus, Hamburg, Germany) equipped with a DCLP490 dichroic mirror and a 60x/0.90 water immersion LUMPFL objective (Olympus). The objective was immersed in a drop of distilled water put on top of the coverslip sealing the funiculus. GCaMP3.0 stimulation with a 475 nm light and an exposition time of 50 ms was performed using a monochromator (Polychrome V, TILL Photonics, Gräfelfing, Germany). Emitted light was filtered with a LP515 filter and acquired at a 4 Hz frequency using a cooled CCD camera (Sensicam, PCO Imaging, Kelheim, Germany) controlled by TILLVision 4.5 software (TILL Photonics). Odor stimuli were sampled from the headspace of a 50 ml volume glass bottle (Schott, Jena, Germany) containing 2 ml of ethyl hexanoate (99% purity, Sigma-Aldrich) diluted in mineral oil to a 10^–2 dilution. A stimulus controller (Syntech, Kirchzarten, Germany) was used to deliver for 2 s the bottle headspace into a charcoal-filtered and humidified constant air flow (0.5 m/s) in a Teflon tube, where inlet was positioned 5–10 mm from the recorded antenna. The stimulus was delivered at 10 s from the start of the recording; the total duration of the experiment was 35 s. The response magnitude was calculated for each frame as the average ΔF/F₀ and expressed in percentage after background subtraction. Regions of interest (ROIs) were selected using the built-in tools of TILLVision 4.5 and F₀ was estimated as the mean fluorescence level calculated for each selected region of interest as the average intensity from 3 to 5.25 s of the recording and paired t-tests were performed using Prism 4 software (GraphPad Software Inc., La Jolla, CA, United States).

Four to eight old female flies of the genotype (w; UAS-GCaMP6f; Orco-Gal4), (w; UAS-GCaMP6f; Or22a-Gal4) were decapitated, then the antennae were excised and fixed on a Sylgard-coated support using the two-component silicon curing medium and immersed in Drosophila Ringer solution (130 mM NaCl, 5 mM KCl, 4 mM MgCl₂ × 6H₂O, 2 mM CaCl₂, 36 mM sucrose, 5 mM Hepes, pH = 7.3, osmolality = 312 mOsm/kg). Samples were immersed in Drosophila Ringer solution and imaged using the imaging setup described above with a LUMPFL 60x/0.90 water immersion objective (Olympus, Hamburg, Germany). Emitted light was separated by a 490 nm dichroic mirror and filtered with a 515 nm long-pass filter. GCaMP6f was excited with a 475 nm light for 50 ms per frame and a temporal resolution of 0.2 Hz. Stimuli consisted of 100 μl of ethyl hexanoate (99%, Sigma-Aldrich) or VUAA1 (N-(4-ethylphenyl)-2-((4-ethyl-5-(3-pyridinyl)-4H- 1,2,4-triazol-3-yl)thio)acetamide, synthesized by the working group “Mass Spectrometry/Proteomics” of the Max Planck Institute for Chemical Ecology, Jena) at the required concentration. Ethyl hexanoate and VUAA1 working solutions were freshly prepared from 100 mM (or 500 mM, for final concentrations >100 μM) stocks in DMSO, kept at −20°C. DMSO at a 1:1000 dilution in Drosophila Ringer was used as control.

Females between 4 and 8 days old were decapitated, the antennae were excised and fixed on a Sylgard-coated support (Sylgard 184, Dow Corning Corp., Midland, MI, United States) using a two-component silicon curing medium (Kwik-Sil, World Precision Instruments, Sarasota, FL, United States). The antennae were immersed in dissecting solution (Au – Sicaeros et al., 2007) (Solution A: 137 mM NaCl, 5.4 mM KCl, 0.17 mM Na₂HPO₄, 0.22 mM KH₂PO₄. Solution B: 9.9 mM HEPES). For 500 ml of dissecting solution: 400 ml ultra-filtered water, 25 ml of Solution A, 14 ml of Solution B, 3.0 g (33.3 mM) D(+)-Glucose, 7.5 g (43.8 mM) Sucrose. Brought to pH 6.7 with 1 N NaOH and to the final volume of 500 ml with ultra-filtered water supplemented with 1 mM EDTA, 5 mM L-Cysteine hydrochloride (Cat. Nr. C1276, Sigma-Aldrich), and after equilibrating the pH to 6.7 with 1 N NaOH, 0.5 mg/ml Papain (Cat. Nr. 5125, Calbiochem, San Diego, CA, United States). Funiculi were cut between one third and one-half of their length and incubated at 27°C for 30 min. After incubation, the dissecting solution was removed and the antennae were rinsed twice for 5 min at 27°C with Ca²⁺-free Ringer (130 mM NaCl, 4 mM MgCl₂ × 6H₂O, 36 mM sucrose, 5 mM HEPES, 5 mM KCl. Osmolality = 312 mOsm/kg, pH = 6.7 adjusted with 1 N NaOH) supplemented with 1:75 cOmplete protease inhibitor cocktail (Cat. Nr. 04693116001, Roche, Basel, Switzerland) dissolved in a 100 mM PBS solution (50.93 mM Na₂HPO₄, 60.22 mM KH₂PO₄, 80.42 mM NaCl; pH = 6.7 adjusted with 1 N NaOH). A stock of Protease inhibitory solution was prepared by dissolving one tablet of cOmplete Protease Inhibitor Cocktail in 2 ml of 100 mM PBS. The solution was aliquoted and stored at −20°C for max 3 months. As an excessive concentration of cOmplete protease may cause cell permeabilization, the protease solution was added to the Ca²⁺-free Ringer in a 1:75 dilution, as suggested by the manufacturer.

Round glass coverslips (12 mm diameter, Cat. Nr. P231.1, Carl Roth, Karlsruhe, Germany) were cleaned by immersion in methanol (≥99.5%, Roth) and HCl (≥32%, Sigma-Aldrich) 1:1 for 30 min and after rinsing in double distilled water, by immersion in H₂SO₄ (95∼97%, Sigma-Aldrich) for 30 min (Cras et al., 1999). Coverslips were then thoroughly rinsed and kept in methanol under N₂; immediately before use, they were thoroughly washed in double distilled water and dried under N₂.

Borosilicate glass capillaries (0.86 mm × 150 mm × 80 mm, Cat. Nr. GB150-8P, Science Products, Hofheim, Germany) were pulled using a P-97 Micropipette puller (Sutter Instrument, Novato, CA, United States) and their tip was cut and fire polished in order to obtain holding micropipettes with an internal diameter ∼0.4 mm. The capillary tip size was found to be critical in order to extract viable neurons. The inner diameter of the tip should be slightly larger than the width of the cut antenna when fixed in a vertical position. The capillaries were subsequently silanized by immersion for 10–15 s in 5% dimethyldichlorosilane in toluene (Cat. Nr. 33065, Supelco, Sigma-Aldrich) and rinsed twice in toluene (≥99.5%, Sigma-Aldrich) and three times in methanol (≥99.5%, Sigma-Aldrich). The capillaries were then dried under N₂ and heated at 200°C for 2 h.

To embed the dissociated antennal tissue, we first used 0.01% sterile filtered poly-L-lysine (Sigma-Aldrich, article no. P4707), Concanavalin A/laminin (Sigma-Aldrich: article no. C2010, L2020) or Collagenase Clostridium (Sigma-Aldrich, article no. C9891) on the coverslips to adhere cells but Drosophila OSNs did not attach to the cover slip. We next used a sodium metasilicate solution (Avnir et al., 2006; Pierre and Rigacci, 2011) to adhere the cells, for that, this solution was prepared immediately before use by mixing 2.71 μl of HCl 0.05 M, 2.43 μl of sodium metasilicate solution (Cat. Nr. 13729, Sigma-Aldrich) diluted 1:10 in double distilled water and 13.86 μl of low Ca²⁺-Schneider’s medium [Cat. Nr. S9895, Sigma-Aldrich, modified with 0.4 g EGTA (1 mM), 22.2 mg CaCl₂ (free Ca²⁺ = 500 nM), 0.4 g NaHCO₃, total volume: 1 L, pH = 6.7 adjusted with 1 N NaOH; sterile filtered and kept at 4°C]. After a 1 μl drop of Ca²⁺-free Ringer was deposited on a cleaned coverslip, the content of the treated antennae was gently sucked using a silanized capillary attached to a 2 μl micropipette. Usually ∼0.5/1 μl of liquid was sucked together with each antenna. Within 10 min from mixing, the complete volume of the silicate gel solution was added to the coverslip and distributed uniformly. The total volume of liquid on a coverslip should be ∼ 25 μl, for a final Na₂SiO₃ concentration equal to 0.972% of the Na₂SiO₃ (≥27% SiO₂ basis) stock solution. Coverslips were incubated for 1 h at 26°C in a high humidity environment, to avoid desiccation.

Imaging data were exported as uncompressed tiff files and analyzed using custom scripts in Fiji-ImageJ2 (Schindelin et al., 2012; Rueden et al., 2017) where the regions of interest were selected using a semi-automatic procedure and the ΔF/F₀ values were calculated after background, flat-field and movement correction. Statistical analysis was performed in R (R Core Team, 2017) using custom scripts including add-on packages (Chang, 2014; Ritz et al., 2015; Wickham, 2016; Arnold, 2017; Auguie, 2017). Parametric statistics for data analysis of the standard deviation of base level ΔF/F₀ values for GCaMP6f (Figure 3I) was used after evaluation of the pooled ΔF/F₀ values distribution of all analyzed cells (ROIs) (Supplementary Figure 5 and see section “Statistical Methods” below). In all preparation types we occasionally observed a reduction in base level fluorescence (e.g., Figure 1 and Supplementary Figure 5). Such artifacts, e.g., due to fluophore bleaching were compensated during data analysis.

The appropriate statistics for data analysis on Supplementary Figures 2D,E and Figure 3I was evaluated after assessment of the data distributions. For GCaMP6f ΔF/F₀ data relative frequency distributions − including all regions of interest (ROIs) selected for each treatment− were evaluated for the basal intracellular free Ca²⁺ concentration ([Ca²⁺]_i) at time t = 0 s (Supplementary Figure 4). Moreover, the standard deviation of ΔF/F₀ values before stimulus application was calculated for each ROI and the relative frequency distribution of values for each treatment was assessed (Supplementary Figure 5). In this case, none of the treatments showed a terminal long tail, meaning that the number of ROIs with skewed high values − and consequently with the chance to skew the mean − was negligible. Consequently, the mean value between all ROIs was selected for each independent measure and the difference between groups was evaluated using two-tailed Welch’s t-tests.

The isolated antenna preparation according to Mukunda et al. (2014) allows to study OR activation in the OSNs. This approach, however, requires dissolved OR activators. To overcome this limitation we developed a whole animal preparation that allows to stimulate OSNs with airborne odorants. To access the antennal olfactory neurons in a living fly, we first immobilized the animal in a custom Plexiglas holder inside a pipette tip and we fixed the antenna in vertical position using a custom holder without disrupting the antennal sensilla (Figures 1A,B and Supplementary Figure 1). After that, we made a transversal cut of the funiculus, stimulated the animal with airborne odorants, and recorded their neuronal activity (Figures 1C–E). The transversal cut did not induce disruption of the antennal tissue sufficient to impair the ability of olfactory neurons to respond to odors. Using a fly line expressing GCaMP3.0 in Or22a olfactory neurons we checked for the viability of these cells by recording their responses first to mineral oil (solvent) and comparing them to the responses elicited by ethyl hexanoate at a 10^–2 dilution in mineral oil. We could show the viability of this preparation as cells were responding to a 10^–2 dilution of the odor, but not to the solvent alone (Figures 1F,G). Moreover, the response was sustained and lasted up to 20 s after stimulation (Figures 1H,I and Supplementary Table 4).

With the whole animal preparation and open antenna preparation, calcium imaging on Drosophila OSNs can be studied under in situ conditions. However, to apply high-resolution techniques such as cryogenic electron microscopy (Cryo-EM) or direct stochastic optical reconstruction microscopy (dSTORM; Heilemann et al., 2008) imaging requires single Drosophila OSNs. We thus established a protocol to mildly digest Drosophila antennal tissue and then isolate OSNs. In detail, after funiculi from excised antennae were transversally cut, the antennal tissue was partially digested with papain − allowing the extraction of the digested content using silanized glass capillary (see schematic diagram Figure 2A). A drop of low Ca²⁺ Schneider’s medium containing Drosophila OSNs from (+; Orco-Gal4/CyO; UAS-Syn21-GFP-p10/TM6B) fly line were kept on a glass slide, covered with a cover slip and then single cell was immediately captured using laser scanning microscopy [Figure 2B, transmission (left), fluorescence (right)]. Next, in order to perform physiological recordings from dissociated tissue, it was necessary to adhere the cells. Initially, we used standard histology procedures to adhere the isolated OSNs on to the coverslip by coating it with Poly-L-lysine, Collagenase Clostridium, or Concanavalin A/Laminin but we failed to attach the cells using this approach.

We thus established a sodium metasilicate hydrogel system to embed biological tissue in hope to successfully apply this embedding system for the isolated Drosophila OSNs. First we started by assessing the effects of the gelation process on the properties of pH-buffered solutions. We found that Na₂SiO₃ had significantly smaller effects on the buffer osmolality than other embedding media as agarose and sodium alginate (Supplementary Figure 2B) and this parameter was not affected by the silicate polymerization process (Supplementary Figures 2C, 3), making sodium silicate hydrogels suitable for functional studies. In particular, sodium metasilicate (Na₂SiO₃)-based aqueous solutions constitute an interesting case as they polymerize through a pH-driven condensation reaction (Pierre and Rigacci, 2011) with salts and water as the only byproducts. Partial neutralization of basic Na₂SiO₃ solutions (pH ∼ 12.5) to physiological (pH ∼ 6.7–7.4) pH levels induce the gelation of colloidal silica particles through a chain of condensation reactions (Rao et al., 2011; Supplementary Figure 2A).

We then explored the effects of Na₂SiO₃ on the dynamics of the morphology and the free intracellular calcium concentration ([Ca²⁺]_i) of adherent HEK293 cells (Supplementary Figures 2D–F). Calcium imaging using the dye Fura2-AM demonstrated that Na₂SiO₃ did not affect the cell basal [Ca²⁺]_i (Supplementary Figure 2D) and did not induce variations of the [Ca²⁺]_i over time (Supplementary Figure 2E). Moreover, we observed changes in the cell morphology due to Na₂SiO₃ only at high concentrations (>1.5% of a ≥27% SiO₂ basis stock solution, see section “Materials and Methods”), where cells tended to aggregate in clusters (Supplementary Figure 2F). We can conclude that Na₂SiO₃ hydrogels are compatible with the embedding of living cells for functional studies. Therefore, we now dissociated Drosophila antennal tissue and then embedded on a clean glass coverslip using a modified Schneider’s medium containing 0.972% Na₂SiO₃ stock solution (see schematic diagram Figure 2C). This concentration of Na₂SiO₃ allowed us to reliably embed the tissue samples while having minimal effects on the cell physiology (Supplementary Figure 3). Following this procedure, we obtained OSNs that retained their morphology, including the ciliated outer dendritic segment (Figures 3A,C). The density of isolated OSNs was about 20 OSNs/100 μm².

We then asked whether the embedded neurons also retained their functional properties. We performed Ca²⁺ imaging from OSNs expressing the calcium indicator GCaMP6f (Chen et al., 2013) under the co-receptor Orco or the odorant receptor Or22a promoters and stimulated OSNs expressing Orco or Or22a with their respective agonists, the synthetic compound VUAA1 (Jones et al., 2011; Figure 3A) and ethyl hexanoate (Münch and Galizia, 2016; Figure 3C and Supplementary Video 1). After extraction of regions of interest (see section “Materials and Methods”), we calculated the changes in GCaMP6f fluorescence intensity with respect to the base level expressed in percent (% ΔF/F₀). In both cases, OR agonists induced calcium responses in a concentration-dependent manner (Figures 3B,D). In order to determine if the OSN response profile was affected by the tissue dissociation process or Na₂SiO₃, we performed calcium imaging from excised antennae immediately after the funiculus cut (Figures 3E–H), and we compared the response profiles of Orco and Or22a-expressing OSNs from dissociated and undigested antennal tissue (Figures 3I,J). The response profiles between the OSNs in dissociated and undigested tissues showed differences in the maximal intensity and time course of Ca²⁺ responses (Supplementary Figure 5). Nevertheless, we did not find significant differences in the fluctuation of the basal fluorescence levels (Figure 3I) and in the EC₅₀ of concentration-response curves for Or22a neurons stimulated with ethyl hexanoate and Orco expressing neurons stimulated with VUAA1 (Figure 3J). In addition to the excitatory ethyl hexanoate, we tested the inhibitory benzaldehyde. At 10^–3 dilution we observed a decrease in GCaMP6f fluorescence intensity ΔF/F₀ by 7 ± 4% (n = 6) (Supplementary Figure 7). This illustrates that after cell isolation and embedding the inhibitory response remained conserved, as observed upon heterologous expression of Or22a (Wicher et al., 2008). OSN isolation removes the direct contact of these neurons to their support cells. Ablation of thecogen cells which directly envelop OSNs did not affect the odor selectivity of tested cells (Prelic et al., 2022).

This suggests that the tissue dissociation and embedding procedures, at the Na₂SiO₃ concentration used, did not significantly affect the viability of OSNs as well as the diffusion of the OR agonists through the embedding medium.