Volatile codes: Correlation of olfactory signals and reception in Drosophila-yeast chemical communication

Drosophila have evolved strong mutualistic associations with yeast communities that best support their growth and survival, resulting in the development of novel niches. It has been suggested that flies recognize their cognate yeasts primarily based on the rich repertoire of volatile organic compounds (VOCs) derived from the yeasts. Thus, it remained an exciting avenue to study whether fly spp. detect and discriminate yeast strains based on odor alone, and if so, how such resolution is achieved by the olfactory system in flies. We used two fly species known to exploit different niches and harboring different yeasts, D. suzukii (a pest of fresh fruit) and D. melanogaster (a saprophytic fly and a neurogenetic model organism). We initially established the behavioral preference of both fly species to six Drosophila-associated yeasts; then chemically analyzed the VOC profile of each yeast which revealed quantitative and qualitative differences; and finally isolated and identified the physiologically active constituents from yeast VOCs for each drosophilid that potentially define attraction. By employing chemical, behavioral, and electrophysiological analyses, we provide a comprehensive portrait of the olfactory neuroethological correlates underlying fly-yeast coadaptation in two drosophilids with distinct habitats.


Results
Yeasts derived from field collected fly populations induce distinct preferences. All six yeast species (P. terricola, P. kluyveri, H. uvarum, C. californica, C. zemplinina, and S. cerevisiae), commonly associated with either of the fly species, elicited strong attraction of D. melanogaster and D. suzukii to yeast baited traps in a binary choice assay. Regardless of yeast or fly species, yeast baited traps were consistently more attractive than control traps (p < 0.02). The equal distribution of flies when both the traps were treated with the control bait, potato dextrose broth (PDB), demomsrated that there was no positional bias (p > 0.2) ( Table 1). Further investigation into the relative yeast preference of each fly species under a multi-choice paradigm revealed distinct patterns. While D. melanogaster showed a significant discrimination among the yeasts (F 6,77 = 9.12; p = 1.56 × 10 −7 ), the statistical difference is much more robust in D. suzukii (F 6,77 = 12.85; p = 5.30 × 10 −10 ). Pairwise comparisons of captures for each yeast within D. melanogaster revealed the most significant attraction to H. uvarum and P. terricola (Fig. 1 whereas H. uvarum alone induced the highest attraction within D. suzukii. We also note with interest that a two-way ANOVA, taking yeast and fly species as factors, revealed significant difference in the overall yeast preference by two fly spp. (F 1,132 = 7.78; p = 0.006), despite a relatively conserved preference pattern. Finally, we asked if an unpaired comparison between the two fly species for a given yeast would reveal any differences in the behavioral preference. Difference in preference was non-significant, except for P. terricola, which elicited a significantly higher response in D. melanogaster compared to D. suzukii (p < 0.005).  Tables S1 and S2). To determine if this variation in VOCs can be utilized to resolve yeast species into distinct entities, we subjected the data to Principal Component Analysis (PCA). The area under the ten most abundant constituent odorants extracted from the Total Ion Chromatograms (TICs) from each yeast species sorted the populations into discrete clusters (Fig. 2b). More than 72% of the observed variation in the profiles was explained by the first three principal components (PCs). To identify the most influential constituent odorants (loading factors) separating yeasts, we took a loading value above 0.28 as indicative of significant contribution towards the determination of each PC. PC1, which accounted for 34.1% of the overall variation, was weighed positively by all the ester compounds except one aromatic aldehyde, benzaldehyde, whereas alcohols and ketones contributed negatively. The top positive contributors include isoamyl propionate and isoamyl acetate while 1-pentanol, 2-methyl-1-butanol, and 6-methyl-5-heptene-2-one contributed negatively, in decreasing order of impact. The next highest variation (> 24%) was accounted for by PC2, which was positively impacted by ethyl octanoate, ethyl hexanoate, ethyl decanoate, and 3-(methylthio)-1-propanol, while negatively influenced by ethyl acetate and benzaldehyde. PC3 (explaining 13.6% of the total variation) was influenced positively by an aromatic, phenethyl propionate, whereas negatively contributing compounds were 3-(methylthio)-1-propanol and four esters, namely ethyl octanoate, ethyl hexanoate, ethyl decanoate and ethyl acetate (in decreasing value). Finally, the statistical significance of the differences among yeast profiles was determined by pairwise MANOVA tests of the first three PCs (Fig. 2c). Except for C. californica and S. cerevisiae, all species were significantly different from each other (p = 0.0001), with P. terricola being the most significant.
Peripheral response repertoire resolves yeast species into distinct odor space. Having established the strong preference of vinegar flies to yeast and their ability to discriminate among them, we investigated the sensory physiological basis of this attraction by using gas chromatography link electroantennographic detection (GC-EAD). Many of the yeast headspace odor constituents elicited electrophysiological responses of various intensities (  The area under the 10 most abundant chemical constituents from each yeast species was extracted; since the abundant peaks were not the same in each yeast, extracting the top 10 and retrieving them in the rest yielded a list of 24 constituents overall that were subjected to PCA. The cluster plot was generated using the first three principal components (PC1, PC2, PC3 each explained 34.1%, 24.5% and 13.6% of the total variation, respectively). The insets represent individual mean PC score ± SEM. (C) A heat map based on the hypothesis tests between PCA clusters indicates highly significant differences in the yeast VOC chemistry. Tables S1 and S2). To investigate whether D. suzukii and D. melanogaster can separate yeasts into discrete odor space based on the volatile bouquet, and if this resolution is conserved between the species, we performed principal component analysis (PCA) on the 25 most intense responses compiled from each yeast response profile (Supplementary Table S1) and plotted them three dimensionally. This revealed a distinct sorting of the yeasts by the flies (Fig. 4a,b). The first three PCs together accounted for 56.6% and 58.1% of the total variation observed in D. suzukii and D. melanogaster, respectively. The significant contribution of these three PCs in resolving six yeasts into discrete points are evidenced in the insets of Fig. 4a,b. In order to distinguish which odorants (loading factors) are most influential in fly discrimination among yeasts, we selected the three positive and negative factors of greatest impact on the determination of each PC.
In order to determine the statistical significance of the odor separation represented in the PCA clustering ( Fig. 4a,b), pairwise MANOVA tests were performed between group responses. Responses elicited by all six yeast species were significantly different from each other in a pairwise comparison performed on the two fly species (Fig. 4c). In D. suzukii, the highest difference (p = 2.9 × 10 −8 ) was found between P. kluyveri vs P. terricola whereas the lowest difference was between S. cerevisiae and P. terricola (p = 0.0053). In D. melanogaster the highest difference was between S. cerevisiae and P. kluyveri (p = 1.1 × 10 −8 ) and the lowest (p = 3.4 × 10 −8 ) was between C. zemplinina and C. californica. The resolution pattern and the magnitude of separation among the yeast species were different in D. melanogaster compared to D. suzukii. We note with great interest that D. suzukii could be effectively separated from D. melanogaster (p = 1.96 × 10 −33 ) by subjecting yeast VOC induced responses from both fly species to a single PCA (Fig. 4d). The different chemosensory space of the two fly species suggests that they can resolve complex chemical landscapes as distinct entities.
Finally, to explore if the distinctly resolved sensory spaces for two species can be attributed to a set of compounds, we examined the top ten antennal responses generated by each yeast from D. melanogaster and D. suzukii, which resulted in a subset of constituents eliciting significantly different responses between the two fly species (Fig. 5a,b). Of these, a majority were esters with the exception of two ketones (2-heptanone and 6-methyl-5-heptene-2-one). Responses induced by isobutyl acetate, isoamyl acetate and ethyl hexanoate were significantly different between D. melanogaster and D. suzukii in at least five of the six yeast odor profiles suggesting the salience of these odorants contributing to behavioral discrimination. These compounds were present in varying amounts (ranging from below the detection threshold of MS up to 100 ng) in the natural headspace of each yeast profile, yet were consistently active on the antennae ( Supplementary Fig. S4). We further evaluated the salience by stimulating antennae with known amounts of these compounds to establish relative sensitivity. As a control, we used ethyl isovalerate, which elicited comparable antennal responses from the two fly species for each of the yeasts tested (Fig. 5a). All four compounds elicited electrophysiological responses that were dose-dependent in both species. As expected, the dose-response curves for the test compounds were significantly different between fly species (p < 1.2 × 10 −7 ), whereas ethyl isovalerate (control) induced comparable depolarization (Fig. 5c). Additionally, while the straight chain ester, ethyl hexanoate, elicited stronger responses (lower threshold) in D. melanogaster, the two other test compounds with a branched alkyl group induced responses with lower threshold in D. suzukii. This species-specific sensitivity confirms the critical role of a limited number of compounds potentially contributing to the behavioral differentiation.

Discussion
Yeasts coadapt with each other in their natural habitat to form stable communities, and these communities are further coadapted with Drosophila resulting in mutualistic relationships 21,23 . Drosophila thus appear to have evolved a strong preference for those yeasts which best support their growth and survival, as measured by various fitness traits 15,19,21 . These interactions between yeasts and flies can be directly correlated with the odor induced preference in flies to their coadapted yeasts 17,36 . Furthermore, this mutualistic association is conserved across large taxonomic groups, and the evolution of preference between microorganisms and insects has been shown to be significantly modulated by way of microbial volatile organic compounds (MVOCS) 16 .
Here, we systematically searched for the olfactory correlates of the long described fly-yeast interactions by studying two different fly species, namely D. melanogaster and D. suzukii. While the former species laid the foundation as a model species for neurogenetics, the latter species is emerging as a national threat to food security in the USA 14 . We began our investigation by collecting D. suzukii flies from the field and isolating and identifying the yeast species that were highly enriched in fly alimentary canals. We determined fly preference to those yeasts, both in a binary and multichoice paradigm. Of the six isolated yeasts, H. uvarum induced the highest trap captures under both paradigms (Table 1 and Fig. 1). Not surprisingly, H. uvarum was detected in the highest number of field collected D. suzukii flies 29 , suggesting a strong coadaptation between these two organisms. Of note, a large scale field sampling of Drosophila spp. and their associated microbes (yeasts and bacteria) identified H. uvarum group members as one of the most frequent yeast groups from D. melanogaster flies as well 31 .
Signals and reception rarely, if ever, arise de novo, and this has been elegantly demonstrated to evolve in synchrony in insect chemical communication, for which the term "coevolution" was first coined 37-39 . Fly-yeast interactions, as mediated by the rich yeast volatile signal and fly OR repertoires have been suggested as a powerful tool to study ecological interactions and coadaptation 32 . We used a two pronged approach to characterize the chemosensory correlates defining fly-yeast associations, encompassing yeast chemical analysis (signals) and measuring fly olfactory sensitivity and selectivity (reception). Our chemical analyses revealed species-specific odor signatures that could be separated with high statistical significance between all but C. zemplinina in comparison to S. cerevisiae ( Fig. 2; Supplementary Figs S1 and S2). A recent study involving 14 diverse S. cerevisiae accessions of known genetic background also found that distinct yeast populations could be resolved solely based on volatile constituents 32 . Furthermore, it correlated the major constituents impacting principal components with chemosensory receptor repertoires (ORs and IRs), suggesting them as potential determinants in fly-yeast coevolution. Our meticulous measurements of yeast VOCs are largely in agreement with two other extensive studies on yeasts volatile chemistry of S. cerevisiae, one which analyzed VOCs on polar and non-polar columns 32 , and another which analyzed VOCs produced by yeasts raised on different substrates 40 . In a detailed preliminary analysis we compared and contrasted the VOC profiles from yeast that were either raised on PDB or a synthetic minimal media every 24 hrs. up to 72 hrs. Our results indicated a robust and consistent VOC profile from PDB raised yeast that did not vary significantly between the three time points (data not shown).
We extended our study to identify the olfactory-physiological correlates that potentially encoded the yeast odor separation into chemosensory space, resulting in the behavioral discrimination noted above. Olfactory response patterns from D. melanogaster (Fig. 4b,c) reflected the pattern of separation in the yeast chemistry, demonstrating highly significant differences for C. zemplinina vs. P. terricola as well as H. uvarum vs. P. terricola. In D. suzukii the observations in the chemistry patterns were only partially preserved (Fig. 4a,c). Taken together, these patterns indicate that the chemical differences in the yeast VOC profiles (as described by our PCA) are translated into complex behavioral output, though not comparable between fly species (Table 1 and Fig. 1). This can further be illustrated by the pairwise z-score comparisons (after centered Gaussian normalization) of the top electrophysiological response inducing compounds between two fly species (Fig. 5a,b) wherein esters represented the largest group. Esters are mainly formed via two distinct pathways in yeasts, one resulting in acetate esters (AE) and the other in ethyl esters (EE) 41 . A large screening of 38 yeast strains belonging to 5 genera revealed yeast of the Pichia and Hanseniaspora genera as the best producers of acetate esters 42 . Recent studies additionally demonstrated how certain yeasts strains employ the AE biosynthetic pathway for their active dispersal through flies, and those yeasts are found in highest abundance from field caught flies 43 . Our analyses corroborate those findings in the case of D. melanogaster wherein, Hanseniaspora and Pichia elicited significantly higher trap captures under both the regimes (Table 1 and Fig. 1). Responses from D. suzukii were more intruiging. We note with great interest that two esters of AE class (isobutyl and isoamyl acetate) induced significantly higher olfactory responses from D. suzukii compared to D. melanogaster, whereas ethyl hexanoate (EE class) had an opposite effect suggesting complex interactions ( Supplementary Fig. S4). Overall, our extensive analysis highlights the relative significance of a few shared constituents (Fig. 5b,c) that potentially facilitate D. suzukii and D. melanogaster in discerning their own niche in complex yeast chemical landscapes.
Given that only a handful of compounds, the majority of which transpired to be esters, appear to encode enough information enabling two fly species to detect and discriminate their niches, it poses the question of how this is achieved. In flies, this feat is proposed to be accomplished by multiple means: an overall alteration in the OR repertoires adapted for niche specialization [44][45][46][47][48] ; changes in the amino acid residues of a given OR that renders them differentially sensitive 49 ; changes at the peripheral olfactory apparatus such as an altered number of specialized sensilla/ORNs [50][51][52] ; and finally the modulation of the transduced signal by the interneurons, projection neurons and Kenyon cells 53,54 . Further, though odorants are parsimoniously used in multiple contexts 6,8 , odor constituents of high salience usually have lower sensory thresholds and are detected by a small number of specialized (narrowly tuned) ORs, whereas the majority of chemical signals are detected with higher threshold by a large number of broadly tuned ORs 4,54 . Niche utilization studies in D. sechellia identified methyl hexanoate as a key chemostimulant produced by its specialized host, morinda fruit. Flies responded with extremely low threshold to this ester and D. sechellia were found to have an increased number of sensillae/ORNs detecting hexanoates, also leading to a corresponding increase in the size of DM2 glomerulus which receives the sensory input 50 . Flies thus employ many strategies, alone and/or in concert, to decipher the chemical landscape around them by extracting quantitative and qualitative features of odors which are translated into meaningful percepts. In our study, D. suzukii' s reduced sensitivity to ethyl hexanoate and significantly higher sensitivity to isobutyl acetate and isoamyl acetate as compared to D. melanogaster represent an exciting avenue to study the salience of these ligands at the molecular, cellular, glomerular and perceptual level in this pestiferous fly.
Environment greatly impacts the Drosophila yeast flora 15 . Though the precise origin of D. suzukii remains unknown, its historical range covers much of eastern Asia and now extends widely from East Asia, to Hawaii, North and Central America, as well as Europe. An extensive analysis of 246 individuals from 12 populations did not reveal any population diversity patterns that could contribute to reconstruct its invasion history into the western hemisphere 55 . Based on this, we can assume that D. suzukii have had multiple alternate hosts and habitats before their ongoing major shift to monocultured fruits. An exciting Scientific RepoRts | 5:14059 | DOi: 10.1038/srep14059 avenue for future research is to explore those ancestral and alternate habitats which can potentially offer a rich resource to isolate and identify unique yeasts and/or their derived VOCs for the development of baits towards manipulating the populations of an economically important emerging pest. Such studies would also provide novel insights into the coevolutionary history of the pestiferous and invasive D. suzukii flies and their associated yeasts.

Material and Methods
Fly Husbandry. Oregon R wild-type Drosophila melanogaster and the field-derived Drosophila suzukii Yeast Culturing. Six yeast cultures (Hanseniaspora uvarum, Pichia terricola, Pichia kluyveri, Candida californica, Candida zemplinina, and Saccharomyces cerevisiae) were obtained from the Phaff Yeast Collection at University of California-Davis from Dr. Kyria Boundy-Mills. These yeasts were maintained as stock cultures in the laboratory since 2012 on PDA prepared with 2.5% Potato Dextrose (HiMedia Laboratories) and 2% Agar (Alfa Aesar). Sealed plates were maintained in an incubator at 30 °C. Stocks were re-streaked every two weeks to maintain active cultures. Before experimentation, a single yeast colony from a stock culture plate was picked using a sterile plastic inoculation needle and transferred to a 15 mL test tube containing 5 mL liquid media [2.5% Potato Dextrose Broth (PDB)] to begin high density starter cultures. The tubes were placed on a shaker at 150 rpm at 30 °C for at least 24 hrs. that resulted in an optical density (OD) of ≥ 1.8 which was earlier determined to produce a representative volatile profile (data not shown) in each yeast except C. zemplinina, which needed 48 hrs. ODs were measured, using at least three replicates, at 600 nm absorbance on a spectrophotometer (BioPhotometer plus, Germany) with sterile PDB as the optical blank. Large volume cultures of 50 mL (≥ 1.8 OD) were started in 125 mL glass bottles by using 0.5 mL of inoculum from a mature starter under the same conditions. Larger cultures were aliquoted to 10 mL volumes for parallel odor extraction and physiology experiments. Starter cultures were directly used for behavioral experiments.
Fly Behavior. Attraction to the six yeasts was tested for both the fly species under two regimes. In the first setup, attraction of each yeast species was tested against a control. In the next set of experiments, the relative attraction of all the yeasts was compared with each other.
Fly traps were designed after Syed et al. 57 , with minor modifications: 1.5 mL SeaLRite micro-centrifuge tubes (USA Scientific, Inc.) were cut 3 mm from the tapered end (bottom) into which a 1000 μ L blue plastic pipette tip (Neptune Scientific, CA) was inserted. The pipette tips were cut 0.5 cm from the narrow end and 2.5 cm from the large end. This resulted in an inverted trap design ( Supplementary Fig. S5) wherein the lid of the micro-centrifuge tube could be used to either hold 125 μ L of yeast culture or an equal volume of control broth (PDB). Two to four day old flies that were starved for 20 hours on 1% agar (Alfa Aesar, MA) were used for behavioral assays.
In order to assess the attraction of individual yeasts, a trap was baited with 125 μ L of yeast culture and another trap holding an equal volume of PDB served as the control. Traps were secured to the base of 266.1 mL clear conical plastic cups (4.5 cm base; 7 cm top and 11 cm height from a commercial vendor) approximately 2.5 cm apart using ~0.8 cm squares of double sided adhesive tape (3M Scotch, USA). Next, relative attraction of the six yeast species was compared by exposing flies to all yeasts at once. A 1000 mL glass beaker (10.5 cm base; 16 cm height) served as the arena. Each trap baited with a given yeast species, or control (PDB), was placed 3 cm apart and 3.5 cm from the arena center. Flies were aspirated to the behavioral arenas which were covered with two layers of grade 50 cheese cloth secured with a rubber band. Behavioral assays began approximately at 4:00 pm and ran for a 24 hr. period at 22.5 ± 0.5 °C. Each cup or beaker was considered a replicate. Flies of mixed sex and age were tested. For single yeast experiments, twenty flies were used whereas sixty flies were used for comparative attraction tests.
Yeast Odors Analysis. A 10 mL aliquot of ≥ 1.8 OD yeast culture was transferred into a 20 mL disposable glass scintillation vial (Kimble, IL) and sealed with Teflon ® tape (Sigma-Aldrich). A gray SPME fiber [23 ga StableFlex TM coated with Divinylbenzene/Carboxen/Polydimethylsiloxane (DVB/CAR/PDMS), Supelco Analytical] was cleaned as per the manufacturer's instructions and exposed to the yeast headspace for 3 to 4 hrs. by piercing through the Teflon ® seal. Exposed fibers were then used either for odor profile analysis by Gas Chromatogram-Mass Spectrometry (GC-MS) (GC: Agilent Technologies, CA. 7890A model; MS: Inert XL MSD with a triple-Axis Detector, Agilent Technologies, CA. 5975C) or Gas Chromatogram linked Electro-antennographic detection (GC-EAD) in parallel with Flame Ionization Detection (FID). We used three means by which to maximize the efficiency of our yeast VOC analytical measurements: (1) we employed a SPME for VOC collection which greatly improved the odor profiles obtained relative to previous solvent extractions, (2) we used a general purpose high resolution non-polar column  to resolve yeast at high temperatures, and (3) used two independent means of detection, flame ionization detection and mass-spectrometry to capture more constituents. Additionally, GC-FID Scientific RepoRts | 5:14059 | DOi: 10.1038/srep14059 /EAD and GC-MS analyses were performed simultaneously for a given yeast on SPME fibers that were exposed concurrently.
Chemical analysis by GC-MS. Exposed SPME fibers were desorbed in a spilt-splitless injector of the GC operated under splitless mode for five minutes at 250 °C. The eluted compounds were resolved on Agilent HP-5 capillary column (30 m, 0.25 mm ID, 0.25 μ m phase) using helium (Ultra High Purity 5.0 Grade; Airgas, USA) as the carrier gas at 1 mL/min constant flow. The GC program was: 50 °C for one minute, increased to 280 °C at 5 °C per minute and held for 5 minutes at the final temperature. The MS was operated at 70 eV; data recording and quantification was performed with Agilent MSD ChemStation software (E.02.02.1431). Chemical identity was determined using three methods: NIST 2011 MS library, comparison of Kovat's Retention Indices, and finally confirming biological activity with synthetic standards. Synthetic standards of highest available purity were acquired from Sigma Aldrich.
Isolation and Identification of biologically active yeast odor constituents by GC-EAD. Biologically active constituents from the yeast odors were isolated by SPME-GC-EAD wherein odor exposed SPME fibers were injected onto the GC and the resolved constituents were monitored, in parallel, by a FID and antennae as biological detectors. Exposed SPME fibers were processed as above on an Agilent HP-5 capillary column (30 m, 0.32 mm ID, 0.25 μ m phase thickness) with a flow rate of 3 mL/min. Resolved column effluents were split 2:1 between the antenna and the FID respectively.
Female Drosophila were used throughout all experiments. Flies were restrained as per the original established protocol 58,59 with minor modifications. The restrained fly was mounted upright on a glass slide and the electrodes used were composed of silver wire inserted into drawn-out borosilicate glass capillaries (World Precision Instruments, Inc., FL) filled with 0.1 M KCl (BDH0258-500G, BDH) saline. The reference electrode was placed in the eye of the restrained fly after which the entire preparation was moved to a high magnification microscope (Olympus BX51WI). The recording electrode was maneuvered with a MPM-10 Piezo Translator to make a firm contact on the dorsomedial antennal region. A humidified stream of charcoal filtered air was continuously passed over the fly preparation at ~0.8 cm/s from a glass tube positioned ~5 mm from the fly. Resolved odor constituents from the GC column were added into this flow. Antennal signals were captured using a high-impedance AC/DC pre-amplifier (10x), sent to an IDAC-4 box, and stored on a PC hard disk using GC-EAD 32 (v. 4.6; 2009). Hardware and software were from Syntech, Germany. The antennal signal was band pass filtered between 3 kHz and 0.1 Hz whereas the FID signal was not conditioned; both the signals were fed on to separate channels in the IDAC-4 and the digitized signal was fed onto the PC. At least three flies of each species were tested for each yeast odor with up to three recordings of the same yeast odor per fly. Recordings were performed in the afternoons and typically extended into the evening.
To generate dose-response curves, we used the GC-EAD regime (as above) instead of an offline EAG protocol to ensure comparable delivery of the odorants. The later method differentially affects the dose delivered. GC-EAD recordings were done using a mixture consisting of isobutyl acetate, isoamyl propionate, ethyl hexanoate, and ethyl isovalerate diluted decadicly relative to each compound from 10 pg to 100 ng in double distilled hexane. All the synthetic standards were ≥ 99.0% (Sigma Aldrich) with the exception of isoamyl acetate at ≥ 95% (SAFC).

Data Processing and Statistical Analysis.
Behavior. Percent trap captures from binary choice assay were subjected to a Mann-Whitney test. Data from the multi-choice assay for a given fly species were analyzed by one-way ANOVA and F-test. Subsequently, trap captures from the two fly species to a given yeast were analyzed by unpaired student's t-test. The interaction between the two fly species was assessed using a two-way ANOVA.

GC-MS.
In order to analyze the odor profiles of each yeast, the 10 most abundant peaks (by area in the Total Ion Chromatograms, TIC), were selected for further analysis. Since the top 10 compounds were not the same in all yeast odor profiles, a compilation of the top 10 from each species resulted in 24 compounds. Attempts were made to locate each of the 24 compounds in every yeast profile. Absolute areas under chosen peaks were integrated for each replicate. In order to correct for the possible variation in collection amounts among replicates, the area under all the selected peaks from a given TIC was pooled and the percent contribution of each compound was determined.
GC-EAD. Each yeast species was tested 5 or 6 times on each fly species. The absolute amplitude of the responses were measured (in μ V) from the onset of depolarization (baseline) to the maxima of the deflection. Antennal responses elicited by the biologically active constituents from at least three of five or four of six replicates for a given species were considered reproducible. Of these, only compounds that elicited responses ≥ 250 μ V from either fly species were chosen. The top 25 most active peaks from this data set were selected for further analysis. Since the top 25 responses were not the same to all yeast odor profiles, a compilation of the top 25 from each species resulted in 53 biologically active compounds. Attempts were made to locate each of the 53 responses in every yeast profile. In order to correct for the possible variation in antennal sensitivity between individual D. melanogaster or D. suzukii flies, each individual response was weighed against the maximum. Finally, relative sensitivity of the two fly species to the major biologically active constituents was compared (pairwise) by generating a heat map.
PCA. Yeast volatile profiles and the electrophysiological responses elicited by the biologically active constituents in two fly species were analyzed by PCA using percentile data. Centered Gaussian normalization was applied in order to bring the percentages of all chemicals constituents or the induced response amplitudes to the same scale. The first three principal components (together explaining more than 55% variance) were retained for subsequent statistical analysis and producing a 3-D plot for visualization. Each of the remaining principal components (PCs) accounted for a proportion of variance of single digits, and therefore was not kept. Multivariate analysis of variance (MANOVA) was performed with the first three PCs being the dependent variables. Pairwise MANOVA tests were performed between the yeast VOC profiles to identify quantifiable differences, and between the response profiles of either fly species to the biologically active yeast VOCs. The pairwise comparisons were summarized into p-value matrices. All statistical analyses were conducted using R3.1.1 60 . Python Matplotlib1.3.1 package was used to prepare the 3D PCA plots 61 .
Dose-Response Function. We used a slightly modified sigmoidal function to fit the dose-response data to derive the critical parameters. The fitted model was Response Amp e 1 1

D dilution EC 50
wherein Amp is the maximum amplitude, D reflects the response sensitivity as a function of dilution (steepness of the curve at dilution equaling EC 50 is proportional to D), and EC 50 represents the dilution at half maximum response. Wilcoxon signed-rank test was used to compare the significance between fly species.