In insects and mammals, olfactory experience in early life alters olfactory behavior and function in later life. In the vinegar fly Drosophila, flies chronically exposed to a high concentration of a monomolecular odor exhibit reduced behavioral aversion to the familiar odor when it is re-encountered. This change in olfactory behavior has been attributed to selective decreases in the sensitivity of second-order olfactory projection neurons (PNs) in the antennal lobe that respond to the overrepresented odor. However, since odorant compounds do not occur at similarly high concentrations in natural sources, the role of odor experience-dependent plasticity in natural environments is unclear. Here, we investigated olfactory plasticity in the antennal lobe of flies chronically exposed to odors at concentrations that are typically encountered in natural odor sources. These stimuli were chosen to each strongly and selectively excite a single class of primary olfactory receptor neuron (ORN), thus facilitating a rigorous assessment of the selectivity of olfactory plasticity for PNs directly excited by overrepresented stimuli. Unexpectedly, we found that chronic exposure to three such odors did not result in decreased PN sensitivity but rather mildly increased responses to weak stimuli in most PN types. Odor-evoked PN activity in response to stronger stimuli was mostly unaffected by odor experience. When present, plasticity was observed broadly in multiple PN types and thus was not selective for PNs receiving direct input from the chronically active ORNs. We further investigated the DL5 olfactory coding channel and found that chronic odor-mediated excitation of its input ORNs did not affect PN intrinsic properties, local inhibitory innervation, ORN responses, or ORN-PN synaptic strength; however, broad-acting lateral excitation evoked by some odors was increased. These results show that PN odor coding is only mildly affected by strong persistent activation of a single olfactory input, highlighting the stability of early stages of insect olfactory processing to significant perturbations in the sensory environment.

Flies

Drosophila melanogaster were raised on a 12:12 light:dark cycle at 25°C and 70% relative humidity on cornmeal/molasses food containing: water (17.8 l), agar (136 g), cornmeal (1335.4 g), yeast (540 g), sucrose (320 g), molasses (1.64 l), CaCl₂ (12.5 g), sodium tartrate (150 g), tegosept (18.45 g), 95% ethanol (153.3 ml) and propionic acid (91.5 ml). All experiments were performed in 2-day-old female flies. The specific genotype of the flies used in each experiment is given in Supplemental Table 1. The transgenes used in this study were acquired from the Bloomington Drosophila Stock Center (BDSC) or the Kyoto Drosophila Stock Center (DGGR), unless otherwise indicated. They have been previously characterized as follows: NP3481-Gal4 (Kyoto:113297) labels DL5, VM7, and DM6 PNs; MZ612-Gal4 (II) (gift of L. Luo) labels VA6 PNs; Or7a-Gal4(KI) (gift of C. Potter) expresses Gal4 from the Or7a locus under the control of its endogenous regulatory elements; UAS-CD8:GFP (X) (RRID:BDSC_5136) and UAS-CD8:GFP (II) (RRID:BDSC_5137) express CD8-tagged GFP, which is targeted to the membrane, under Gal4 control; 20xUAS-IVS-CD8:GFP (attP2) (RRID:BDSC_32194) expressed CD8-tagged GFP under Gal4 control; UAS-brp.S-mStrawberry (II) (gift of S. Sigrist) expresses a red fluorescent protein-tagged short-form of bruchpilot; 20xUAS-IVS-syn21-opGCaMP6f-p10 (su(Hw)attP5) (gift of B. Pfeiffer and D. Anderson) expresses codon-optimized GCaMP6f under Gal4 control; and 13xlexAop2-IVS-CsChrimson.mVenus (attP40) (RRID:BDSC_55138) expresses a Venus-tagged red-shifted channelrhodopsin CsChrimson under lexA control.

Or7a-lexA (III) flies were generated as follows. The Or7a promoter was PCR amplified from a bacterial artificial chromosome (RPCI-98 library, clone 39F18, BACPAC Resources) containing the or7a locus of D. melanogaster using primers 5’-ACCGCATCCCGATCAAGACACAC-3’ and 5’-TGATGGACTTTTGACGCCTGGGAATA-3’. The Or7a promoter was inserted 5’ to nlslexA∷p65 using isothermal assembly in vector pBPnlslexA∷p65Uw, replacing the ccdB cassette. The plasmid pBPnlslexA∷p65Uw was a gift from G. Rubin (Addgene plasmid #26230, RRID:Addgene_26230). The final sequence of the construct was confirmed by Sanger sequencing, and transgenic flies were generated by site-specific integration into the VK00027 landing site (BestGene, Inc., Chino Hills, CA). To verify the selectivity of the driver, Or7a-lexA was crossed to 13xlexAop2-mCD8:GFP (RRID:BDSC_32205), and brains of the resulting progeny flies (2 days old) were dissected and immunostained with antibodies against GFP and nc82. GFP expression was observed selectively in ab4a ORN axons projecting to the DL5 glomerulus; no other signal was observed in the central brain. 

Chronic odor exposure 

With the exception of Figure 4, all data were from flies were chronically exposed to solvent or specific monomolecular odors while reared in standard fly bottles containing cornmeal/molasses food and sealed with modified cotton plugs through which two thin-walled stainless-steel hollow rods (~5 cm length, ~3.2 mm inner diameter) were tightly inserted, serving as an inlet and an outlet for air flow. The bottom of the cotton plug was lined with mesh (McMaster-Carr #9318T45) to prevent flies from entering the rods. The inlet port was fit with a luer connector for easy connection to the carrier stream; the outlet port was vented with loose vacuum suction. 

The odor environment inside the bottle was controlled by delivering to the inlet of the bottle a stream of charcoal-filtered, humidified air (275 ml/min), with a small fraction of the air stream (odor stream, 25 ml/min) diverted into the headspace of a control vial filled with solvent (paraffin oil, J.T. Baker, VWR #JTS894-7) before it was reunited with the carrier stream (250 ml/min). Air flow rates were controlled using variable area valved flow meters (Cole-Parmer). In response to an external 5V command, a three-way solenoid valve redirected the 25 ml/min odor stream from the headspace of the control solvent vial through the headspace of the vial containing diluted odor for 1 second. The diluted odor was continuously stirred using a miniature magnetic stir bar and stir plate (homebrewing.org). Delivery of the 1-second pulse of odor into the carrier stream was repeated every 21 seconds. Tygon tubing (E-3603) was used throughout the odor delivery system, with the exception of a portion of the carrier stream where odor entered and the path from the odor vial to the input to the carrier stream, where PTFE tubing was used. 

The stability of the amplitude of the odor pulse over the course of 24 hours was measured using a photoionization detector (200B miniPID, Aurora Instruments), with the sensor probe mounted at the center of a fresh fly bottle. Based on the observed rundown in the amplitude of the odor pulse (Figure 1B), the odor vial in the odor delivery system was swapped out every 12 hours for a fresh dilution of odor during chronic exposure experiments with flies. Under these conditions, the amplitude of the odor pulse was not expected to decrease more than ~20% at any point during the exposure period.

Flies were seeded in a fresh fly bottle at low density (~7-8 females). The evening prior to expected eclosion, any adult flies were removed from the bottle, and controlled odor delivery was initiated into the bottle.  The next morning (day 0), newly eclosed flies were transferred into a fresh bottle and controlled odor delivery was continued for another ~48 hours. Experiments were typically conducted on day 2, with the exception of those in Figure 4.

In Figure 4, flies were chronically exposed to a sustained high concentration of monomolecular odor from a stationary source placed in the bottle. The evening prior to eclosion, growth bottles were cleared of any adult animals, and a 2-ml glass vial containing 1 mL of geranyl acetate (20% v/v) was placed into the bottle. The vial opening was covered with a layer of porous mesh, held securely in place with a silicone o-ring fitted around the mouth of the vial. The growth bottle was sealed with a standard cotton plug with no inlets/outlets. The next morning, newly eclosed flies were transferred into a fresh bottle containing a fresh odor source, and, thereafter, the odor source was renewed daily until day 5, when experiments were conducted.

Odor concentrations in this study are referred to by the v/v dilution factor of the odor in paraffin oil in the odor vial. For chronic odor exposure in Figures 1-3 and 5-7, flies were exposed to odors at dilution factors of 10^-7 for E2-hexenal, 10^-4 for 2-butanone, and 10^-4 for geranyl acetate. Headspace concentrations were further diluted 1:11 in air prior to delivery to the fly bottle. 

Electrophysiological recordings

PN recordings

Electrophysiological measurements were performed on 2-day-old female flies, except in Figure 4, where recordings were established from 5-day-old female flies. The rate of establishing successful recordings was significantly lower in Figure 4, approximately 1 in 5 attempts compared to the usual rate of approximately 1 in 2 attempts. Flies were briefly cold-anesthetized and immobilized using wax. The composition of the internal pipette solution for current clamp recordings in PNs was (in mM): potassium aspartate 140, HEPES 10, MgATP 4, Na₃GFP 0.5, EGTA 1, KCl 1, biocytin hydrazide 13. The internal solution was adjusted to a pH of 7.3 with KOH or aspartic acid and an osmolarity between 262–268 mOsm. For voltage-clamp recordings, an equal concentration of cesium was substituted for potassium. The external solution was Drosophila saline containing (in mM): NaCl 103, KCl 3, N-tris(hydroxymethyl) methyl-2-aminoethane-sulfonic acid 5, trehalose 8, glucose 10, NaHCO₃ 26, NaH₂PO₄ 1, CaCl₂ 1.5, MgCl₂ 4. The pH of the external solution was adjusted to 7.2 with HCl or NaOH (when bubbled with 95% O₂/5% CO₂), and the osmolarity was adjusted to ~270–275 mOsm. Recording pipettes were fabricated from borosilicate glass and had a resistance of ~6–8 M . Recordings were acquired with a MultiClamp 700B (Axon Instruments) using a CV-7B headstage (500 M). Data were low-pass filtered at 5 kHz, digitized at 10 kHz, and acquired in MATLAB (Mathworks, Natick, MA). Voltages are uncorrected for liquid junction potential. 

Flies were removed from the odor exposure environment at least one hour prior to recordings, briefly cold-anesthetized anesthetized, and recordings were performed at room temperature. Recordings with respect to experimental groups (odor-exposed versus solvent-exposed) were conducted in a quasi-randomized order; the experimenter was not masked to experimental condition. PN somata were visualized with side illumination form an infrared LED (Smartvision) under a 40x water immersion objective on an upright compound microscope equipped with an epifluorescence module and 488 nm blue light source (Sutter Lambda TLED+). Whole-cell patch-clamp recordings were targeted to specific PNs types based on GFP fluorescence directed by genetic drivers. Three PN types were targeted in this study: DL5 and VM7 using NP3481-Gal4 and VA6 using MZ612-Gal4. One neuron was recorded per brain. PN identity was confirmed using diagnostic odors, and the morphology (and thus identity) of all PNs was further verified posthoc by streptavidin staining in fixed brains. Cells with little or no spontaneous activity upon break-in, suggesting antennal nerve damage, were discarded. Input resistance was monitored on every trial in real-time, and recordings were terminated if the input resistance of the cell drifted by more than 20%.

For experiments in Figure 7, the antennal nerves were bilaterally severed immediately prior to recordings using fine forceps.

ORN recordings

Single-sensillum recordings were performed essentially as previously described. Flies were removed from the odor exposure environment at least one hour prior to recordings, briefly cold-anesthetized, and recordings were performed at room temperature. Briefly, flies were immobilized in the end of a trimmed pipette tip using wax, and one antenna or one palp was visualized under a 50x air objective. The antenna was stabilized by tightly sandwiching it between a set of two fine glass hooks, fashioned by gently heating pipettes pulled from glass capillaries (World Precision Instruments, TW150F-3). The palp was stabilized from above with a fine glass pipette pressing it firmly against a glass coverslip provided from below. A reference electrode filled with external saline (see above) was inserted into the eye, and a sharp saline-filled glass recording microelectrode was inserted into the base of the selected sensillum under visual control. Recordings from ab4 and pb1 sensilla were established in the antenna and palp, respectively, based on the characteristic size and morphology of the sensillum, its position on the antenna, and the presence of two distinct spike waveforms, each having a characteristic odor sensitivity and spontaneous firing frequency. Signals were acquired with a MultiClamp 700B amplifier, low-pass filtered at 2 kHz, digitized at 10 kHz, and acquired in MATLAB. Single-sensillum recordings were performed in 2-day old NP3481-Gal4, UAS-CD8:GFPfemales to allow direct comparisons to PN data.

Odor stimuli

Odors used in this study were benzaldehyde, 2-butanone, p-cresol, geosmin, geranyl acetate, 2-heptanone, E2-hexenal, isobutyl acetate, and pentyl acetate. All odors were obtained from MilliporeSigma or Fisher Scientific at the highest purity available (typically >99%).  Odor stimuli are referred to by their v/v dilution factor in paraffin oil. Each 20-ml odor vial contained 2 ml of diluted odor in paraffin oil. Diagnostic stimuli that distinguished targeted PN types from other labeled PN types in the driver line used were as follows: for DL5 PNs, E2-hexenal (10^-8) and benzaldehyde (10^-4); for VM7 PNs, 2-butanone (10^-7) and isobutyl acetate (10^-4); and for VA6 PNs, geranyl acetate (10^-6). Diagnostic stimuli for ORN classes were as follows: for the ab4 sensillum on the antenna, E2-hexenal (10^-7­) for the ab4a “A” spike and geosmin (10^-2) for the ab4b “B” spike; and for the pb1 sensillum on the palp, 2-butanone (10^-5) for the pb1a “A” spike and p-cresol (10^-3) for the pb1b “B” spike.

Fresh odor dilutions were made every 5 days. Each measurement in a fly represents the mean of 5 trials for ORN responses or 6 trials for PN responses, spaced 40 seconds apart. Solvent (paraffin oil) trials were routinely interleaved to assess for contamination. Stimuli were presented in pseudo-randomized order, except for measurements of concentration-response curves where odors were presented from low to high concentrations. 

Odors were presented during recordings from olfactory neurons essentially as previously described. In brief, a constant stream of charcoal-filtered air (2.22 L/min) was directed at the fly, with a small portion of the stream (220 mL/min) passing through the headspace of a control vial filled with paraffin oil (solvent) prior to joining the carrier stream (2.0 L/min). Air flow was controlled using mass flow controllers (Alicat Scientific). When triggered by an external voltage command, a three-way solenoid valve redirected the small portion of the stream (220 mL/min) from the solvent vial through the headspace of a vial containing odor for 500 ms; thus, the concentration of the odor in gas phase was further diluted ~10-fold prior to final delivery to the animal. The solvent vial and the odor vial entered the carrier stream at the same point, ~10 cm from the end of the tube. The tube opening measured ~4 mm in diameter and was positioned ~1 cm away from the fly. We presented 500-ms pulses of odor unless otherwise indicated.

Odor mixtures (Figure 4) were generated by mixing in air. In these experiments, a second solenoid valve was added that diverted another small fraction of the carrier stream (220 ml/min) through either a second solvent vial or a second odor vial before rejoining the carrier stream. When the two solenoids were both triggered, they drew from the carrier stream at the same point, and the two odorized streams also both rejoined the carrier stream at about the same point, ~10 cm from the end of the delivery tube. 

Immunohistochemistry

Intracellular biocytin fills were processed as follows: brains were fixed for 14 min at room temperature in freshly prepared 4% paraformaldehyde, incubated overnight in mouse nc82 primary antibody (1:40, Developmental Studies Hybridoma Bank #AB_2314866), then subsequently incubated overnight in Alexa Fluor 568 streptavidin conjugate (1:1000, Molecular Probes) and Alexa Fluor 633 goat anti-mouse (1:500, Molecular Probes). PN morphologies were reconstructed from serial confocal images through the brain at 40X magnification and 1-µm step size.

LN innervation was quantified in flies of genotype +/UAS-brp.S-mStraw; 20XUAS-IVS-CD8:GFP/NP3056-Gal4; the brp.S-mStraw signal was not measured. Immediately after dissection, brains were fixed for 14 min in freshly prepared 4% paraformaldehyde and incubated overnight in rat anti-CD8 (1:50, Thermo Fisher #MA5-17594) and mouse nc82 (1:40) primary antibodies, then subsequently incubated overnight in Alexa Fluor 488 goat anti-rat (1:500, Abcam #ab150157) and Alexa Fluor 633 goat anti-mouse (1:500, Thermo Fisher #A21050) secondary antibodies. All steps were performed at room temperature, and brains were mounted and imaged in Vectashield mounting medium (Vector labs). In pilot experiments, we compared direct GFP fluorescence in lightly fixed brain with amplified GFP signal using the standard protocol and observed weaker, but qualitatively similar signals. Confocal z-stacks at 1024x1024 resolution spanning the entire volume of the antennal lobe were collected on a Leica SP8 confocal microscope at 1 mm slice intervals using a 63× oil-immersion lens. Identical laser power and imaging settings were used for all experiments.

Calcium imaging

Calcium imaging of ab4a ORN terminals in the DL5 glomerulus was performed on 2-day old female flies. In brief, flies were cold-anesthetized and immobilized with wax. The antennal lobes were exposed by removal of the dorsal flap of head cuticle, and the brain was perfused with Drosophila saline (see above) that was cooled to 21ºC (TC-324C, Warner Instruments) and circulated at a rate of 2-3 ml/min. GCaMP6f signals were measured on a two-photon microscope (Thorlabs, Sterling, VA) using a Ti-Sapphire femtosecond laser (MaiTai eHP DS, Spectra-Physics) at an excitation wavelength of 925 nm, steered by a galvo-galvo scanner. Images were acquired with a 20x water-immersion objective (XLUMPLFLN, Olympus) at 256x96 pixels, a frame rate of 11 Hz, and a dwell time of 2µs/pixel. The same laser power and imaging settings were maintained for all experiments. The microscope and data acquisition were controlled using ThorImage 3.0. The DL5 glomeruli were clearly labeled as bilateral spherical structures ~10 µm from the dorsal surface of the antennal lobes. In each trial, an 8 s period of baseline activity was collected immediately prior to stimulus presentation which was used to establish the level of baseline fluorescence of each pixel. Each odor stimulus was presented for 500 ms, for three trials, with a 45-s interstimulus interval. 

Optogenetic stimulation of ORN axons

A stock solution of all-trans-retinal (Sigma-Aldrich, R2500) was prepared at 35 mM in 95% ethanol and stored at -20º C in the dark. The cross that generated the experimental flies was maintained in the dark.  Newly eclosed experimental flies for optogenetic experiments were transferred to standard cornmeal/molasses food supplemented with 350 mM all-trans-retinal mixed into the food and exposed to odor (or solvent) for two days in the dark.

After exposing the antennal lobes, the antennal nerves were acutely and bilaterally severed at their distal entry point into the first segment of the antennal, eliminating EPSCs derived from spontaneous ORN spiking. Electrophysiology rigs were light-proofed. Whole-cell recordings in voltage-clamp mode were established from DL5 PNs in flies expressing Chrimson in all ab4a ORNs (from or7a-lexA). DL5 PNs were identified based on GFP expression (from NP3481-Gal4) and the presence of light-evoked responses, and their identity was confirmed after the recording by processing the biocytin fill. In pilot experiments, we tested several methods for optical stimulation and were unable to achieve reliable stimulation of only a single ORN axon presynaptic to the recorded PN using excitation with 590 nm light from a fiber optic-coupled LED (Thorlabs M590F3 with Ø200 mm fiber, 0.22 NA). Excitation of ORN terminals was very sensitive to the position and angle of the optrode relative to the antennal nerve, and we found large variability in the amount of light (intensity and pulse duration) required to evoke EPSCs in the PN. 

As an alternative approach, we used wide-field illumination from a 470-nm (blue) LED light source (Sutter Instrument, TLED+). We delivered light pulses of 100 ms duration at 30 s trial intervals, starting at very low levels of light (<0.1 mW/mm²) and gradually increasing the light intensity until an EPSC was observed in an all-or-none manner. At the threshold intensity, trials that fail to evoke an EPSC were infrequently interleaved with successful trials. The ORN-PN synapse is highly reliable, with a single ORN spike evoking robust release of many synaptic vesicles at the ORN terminal; thus, we interpret these failures as a failure to recruit a spike in a presynaptic axon on that trial (as opposed to a failure in synaptic transmission). As the light intensity was further increased, EPSC amplitude remained relatively constant, until it was observed to suddenly double, reflecting the recruitment of a second ORN axon.  The mean uEPSC for each PN was determined by averaging the evoked EPSC in ~8–12 trials at a light intensity approximately halfway between the initial threshold intensity and the doubling intensity. Recordings were discarded if any of the following criteria occurred: 1) a high rate of failures at the light intensity chosen for data collection (>10%); 2) uEPSC amplitude was not stable over a range of at least ~5 mW spanning the chosen light intensity; 2) the shape of the uEPSC was not stable. 

Data Analysis

Unless otherwise stated, all analyses were performed in MATLAB (Mathworks, Natick, MA).

Quantification of electrophysiological responses

Analysis of neural responses was performed masked to the experimental condition (odor- versus solvent-exposed) of the recording.  For each odor stimulus measurement in a fly, a trial block was comprised of 5 stimulus presentations for ORN responses, or 6 presentations for PN responses, at an intertrial interval of 40 s. The first trial for PN responses was not included in the analysis. Spike times were determined from raw ORN and PN voltage traces using custom scripts in MATLAB that identified spikes by thresholding on the first- and second-derivatives of the voltage. All spikes were manually inspected. Spike times were converted into a peristimulus time histogram (PSTH) by counting the number of spikes in 50-ms bins, overlapping by 25 ms. Single-trial PSTHs were averaged to generate a mean PSTH that describes the odor response for each cell. For membrane potential, single-trial voltage traces were averaged to generate a mean depolarization response for each cell. For each DL5 and VM7 PN, the response magnitude for each stimulus was computed as the trial-averaged spike rate (or membrane potential) during the 500-ms odor stimulus period, minus the trial-averaged spontaneous firing rate (or membrane potential) during the preceding 500 ms. For VA6 PNs, the response magnitude was computed over a 1000-ms window that begins at stimulus onset, which better captured the protracted odor response (which extends into the post-stimulus period) observed in this cell type. A mean response magnitude was computed across trials for each experiment, and the overall mean response was plotted as mean ± SEM across all experiments in each condition.

Analysis of LN anatomical innervation

Images of all antennal lobes were collected with identical laser power and imaging settings (magnification, detector gain, offset, pixel size, and dwell time). Brains in which >0.01% of pixels in neuropil regions (excluding cell bodies and primary neurites) were high or low saturated were rejected. Confocal image stacks were imported into ImageJ (NIH) for analysis. Analysis of LN innervation was conducted masked to the experimental condition (odor- versus solvent-exposed). The boundaries of glomeruli of interest were manually traced in every third slice using the nc82 neuropil signal, guided by published atlases, and then interpolated through the stack to obtain the boundaries in adjacent slices. The 3D ROI manager plugin was used to group together sets of ROIs across slices corresponding to each glomerulus to define the volumetric boundaries for each glomerulus and was then used for quantification of glomerular volume (number of pixels) and pixel intensities in each channel. For each 3D ROI corresponding to an individual identified glomerulus, LN neurites or neuropil per volume was computed as the sum of the pixel values in the ROI in the anti-CD8 (LN neurites) or nc82 (neuropil) channels, respectively, divided by the total number of pixels. The ratio of LN neurites to neuropil was computed as the sum of the pixel values in the ROI in the anti-CD8 channel divided by the sum of the pixel values in the ROI in the nc82 channel. To combine measurements across all glomeruli for a given metric (e.g. volume, LN neurites per volume, etc.), the measurement for each glomerulus in an experiment was normalized to the mean value for that glomerulus across all experiments in the solvent-exposed control condition. Plots show mean and standard error across brains. 

Analysis of calcium imaging

The duration of each calcium imaging trial was 15 s, collected at 11 frames per second and 256x96 pixels per frame. Stimulus-evoked calcium signals (∆F/F) were quantified from background-subtracted movies as the change in fluorescence (F-F₀) normalized to the mean fluorescence during the baseline period of each trial (F₀, averaged over 70 frames immediately preceding the odor), computed on a pixel-by-pixel basis in each frame. A Gaussian lowpass filter of size 4×4 pixels was applied to raw ∆F/F heatmaps.

A region-of-interest (ROI) was manually traced around each DL5 glomerulus, which contain the ab4a ORN terminals. ∆F/F signals were averaged across the pixels in the two DL5 ROIs and across three trials for each stimulus in an experiment. The odor response to the 500-ms stimulus presentation was typically captured in ~6 frames. The peak response for each experiment was quantified from the frame containing the maximum mean ∆F/F signal during the stimulus presentation. The overall mean response was plotted as mean ± SEM across all experiments in each condition.

Statistics

A permutation analysis was used to evaluate differences between experimental groups because of its conceptual simplicity and because it does not require assumptions about the underlying distribution of the population. For each measurement (e.g., the response of a PN type to an odor stimulus), experimental observations from flies in odor- and solvent-exposed conditions were combined and randomly reassigned into two groups (maintaining the number of samples in each respective experimental group), and the difference between the means of the groups was computed. This permutation process was repeated 10,000 times without replacement to generate a distribution for the difference between the means of the odor- and solvent-exposed groups (see Figure 6 – figure supplement 1 for an example), under the null assumption that there is no difference between the two populations. We calculated the fraction of the empirically resampled distribution which had an absolute value that equaled or exceeded the absolute observed difference between the means of the odor- and solvent-exposed groups to determine the two-tailed p-value that the observed outcome occurred by chance if the populations are not different. The cut-off for statistical significance of =0.05 was adjusted to account for multiple comparisons in an experiment using a Bonferroni correction. Permutation testing was used in all figures, except in Figure 5 where the Mann-Whitney U-test was used, implemented in MATLAB (Wilcoxon rank sum test).

Sample sizes were not predetermined using a power analysis. We used sample sizes comparable to those used in similar types of studies. The experimenter was not masked to experimental condition or genotype during data collection. For a subset of analyses (analysis of electrophysiological data and quantification of LN innervation), the analyst was masked to the experimental condition, as described above.

CSV and text file readers