Introduction

Many animals use olfactory cues to detect and locate their food, and there is increasing evidence that volatiles produced by microorganisms associated with potential food items can contribute to animal decisions whether to approach and consume the food. For example, the products of microbial fermentation of animal cadavers can be a reliable cue of a protein-rich food source for necrophagous insects and vertebrate scavengers (Verheggen et al. 2017). Similarly, ethanol and other fermentation products of sugar-rich fruits and nectar can be indicative of abundant, accessible calories to frugivorous and nectarivorous animals (Dominy 2004; Dudley 2000; Wiens et al. 2008), while specific microbial products or blends of products, especially at high concentrations, can be repellant, indicating a hazardous level of microbial contamination (Adler 2000; Stensmyr et al. 2012; Tasin et al. 2012).

The basis for this study is the abundant evidence that dipteran insects that feed and oviposit on fleshy fruits are attracted to volatiles produced by microorganisms associated with fruits. These microorganisms are also regularly found in the insect gut, suggesting that they may contribute to insect nutrition. For example, the olive fly Bactrocera oleae is attracted to volatiles, particularly methyl thiolacetate, released from a strain of Pseudomonas putida isolated from its olive host plant (Liscia et al. 2013). Similarly, the apple maggot fly Rhagoletis pomonella is attracted to a strain of the bacterium Enterobacter agglomerans originally isolated from this insect (MacCollum et al. 1992), with 3-hydroxy-2-butanone (acetoin) implicated as a key attractant (Robacker et al. 2004). In addition, the attraction of Drosophila melanogaster to fermenting fruit has been attributed primarily to yeast-derived volatiles, especially ethanol, acetic acid, acetoin, 2-phenyl ethanol and 3-methyl-1-butanol (Becher et al. 2012). These findings illustrate the complexity of ecological and evolutionary processes that shape the trajectory of fruit ripening and rotting (Dudley 2000; Janzen 1977; Ruxton et al. 2014), and they also provide the opportunity for novel behavioral pest control strategies based on microbial volatiles to trap pests for monitoring and suppression of populations (Beck and Vannette 2017; Davis et al. 2013).

The specific focus of this study on the interaction between microbial volatiles and fruit-infesting insects is the invasive “spotted-wing drosophila” Drosophila suzukii (SWD), originating from South East Asia. Unlike most other fruit-feeding drosophilid flies which lay their eggs into overripe fruit, SWD oviposits into ripening soft-skinned fruit such as berries and cherries (Atallah et al. 2014). Combined with its high rate of population increase and lack of natural enemies in introduced regions, SWD is now a major pest in North America and Europe, with chemical insecticide sprays as the dominant control strategy (Asplen et al. 2015). Nevertheless, there is strong demand for alternative control strategies in conventional and especially organic fruit production (Schetelig et al. 2017).

Several lines of evidence suggest that SWD is attracted to microbial volatiles. Under field conditions, apple cider vinegar and baker’s yeast with sugar are widely used as baits for SWD (Iglesias et al. 2014; Knight et al. 2015; Swoboda-Bhattarai et al. 2017), suggesting that SWD is attracted to fermentative yeast odors. Furthermore, laboratory olfactometer assays reveal that SWD are attracted to the volatiles produced by the yeast Hanseniaspora uvarum (Mori et al. 2017) and some acetic acid bacteria (Acetobacteraceae) (Mazzetto et al. 2016), both of which have high prevalence in SWD (Bing et al. 2018; Hamby et al. 2012). Complementary analyses of SWD response to specific volatiles have identified several attractive blends, e.g. a quaternary blend of ethanol, acetoin, acetic acid and methionol (Cha et al. 2018), and a 5-compound blend derived from fermented apple juice (acetoin, ethyl octanoate, acetic acid, phenethyl alcohol and ethyl acetate) (Feng et al. 2018), as well as various commercial formulations, e.g. Scentry (Scentry Inc.), Suzukii Trap bait (BioIberica) and SPLAT SWD (Cloonan et al. 2018). However, the utility of these strategies is constrained by two issues: their relatively low selectivity for SWD, and variability in efficacy, both between laboratory and field trials and across different field trials (Asplen et al. 2015; Cloonan et al. 2018; Schetelig et al. 2017).

Our strategy to study the response of SWD to microbial volatiles comprised three steps: first, to isolate microbial taxa from SWD and fruits infested by SWD; second to quantify the efficacy of these isolates as bait for trapping SWD and other drosophilids from natural populations; and finally to quantify the volatile organic compounds emitted by these microbial taxa. This approach enabled us to correlate specific VOCs and blends with both trapping efficacy and selectivity for SWD, although our exclusive focus on culturable forms precludes global analysis of the microbiota associated with the flies and fruits. Our results are discussed in the context of the ecology of SWD and related fruit-feeding flies and improved design of microbial volatile-based lures for monitoring and control of SWD populations.

Materials and Methods

Cultivation of Microorganisms

The microorganisms were isolated from wild SWD and associated fruits collected from two sites on 6 September 2016: a plot of fruiting primocane raspberry on The Research North Farm as part of Cornell AgriTech, New York State Agricultural Experiment Station in Geneva, NY (42 degrees 52′ 03.50“ N by 77 degrees 02’ 21.09” W); and a bank of fruiting pokeweed Phytolacca americana on a commercial farm near Geneva, NY (42 degrees 47′ 38.03“ N by 76 degrees 59’ 58.97” W). At each site, 40 fruits and 10 adult SWD of each sex were sampled. Individual fruits were transferred using sterile forceps to individual 50 ml sterile plastic tubes (Falcon, Corning, NY). The flies were collected live overnight in traps (Bost et al. 2018; detailed design provided in Supplementary Methods) designed to permit entry but restrict escape and to preclude access to the bait of crushed raspberry fruit, and sorted on return to the laboratory. Indications that the SWD populations were interacting with the fruits came from the presence of Drosophila larvae in many of the collected fruits (Fig. S1), and of fruit-colored material in the crop of many of the collected flies.

To cultivate bacteria and fungi, 20 fruits and 5 flies of each sex from each site were inoculated onto 9 cm diam. plates, using sterile technique in a laminar flow cabinet. In pilot experiments we assessed the microbial growth on four media: nutrient broth (#CM0001, Oxoid Ltd. Hampshire, UK), potato dextrose (#CM013, Oxoid), yeast-peptone-dextrose ((#CM0920, Oxoid) and modified de Man, Rogosa, and Sharpe medium [mMRS 1.25% proteose peptone (Becton Dickinson), 0.75% yeast extract, 2% glucose, 0.5% sodium acetate, 0.2% dipotassium hydrogen phosphate, 0.2% tri-ammonium citrate, 0.02% magnesium sulfate heptahydrate,0.005% manganese sulfate tetrahydrate and 1.2% agar (all constituents from Sigma-Aldrich, St Louis, MO, USA except agar from Apex, San Diego, CA, USA)]. We obtained greatest morphological diversity of colonies with mMRS and nutrient broth, and so the definitive isolations were conducted on these two media. Each fruit was crushed into a pulp on the agar plate and the juices were spread using a sterile glass rod. The gut was dissected from each fly with sterile forceps, and then aseptically transferred to the agar plate, teased open with sterile forceps and spread with 50 μl sterile phosphate buffered saline (PBS) across the plate. Plates were incubated in the dark at 30 °C under aerobic conditions for up to one week.

Microorganisms of diverse colony morphologies grew on all the plates. Each of multiple colonies from each plate was streaked individually onto a fresh plate of the same medium, to obtain pure clonal cultures. Representative colonies of each morphotype were inoculated into 5 ml broth of the same composition and grown at 30 °C to turbidity. Each culture was identified as bacterial or yeast by light microscopy; then, 20% and 10% glycerol stocks, respectively, were prepared and stored at -80 °C.

Identification of Microorganisms

A sample from each glycerol stock (obtained above) was streaked onto an agar plate, and a single colony was grown in 5 ml broth, as above. A 1 ml sample of turbid culture was centrifuged, and the cells were resuspended in 678 μl cell lysis buffer (108 mM Tris-HCl pH 8.0, 1.5 M NaCl, 21.6 mM EDTA) with 30 μl 1 mm diam. Glass beads (Scientific Industries, Vernon Hills, IL, USA) 16 U proteinase K (Qiagen, Hilden, Germany). The homogenate was incubated at 56 °C for 2 h, mixed with 35 U RNaseA (Qiagen), and incubated at 37 °C overnight. DNA was extracted with 750 μl phenol:chloroform:isoamyl alcohol (25:24:1) (Thermo Fisher Scientific, Waltham, MA) with centrifugation at 19,000×g for 15 min at 4 °C. DNA was then precipitated from 550 μl aqueous phase by overnight incubation at −20 °C with 45 μl 3 M sodium acetate (pH 5.2) and 900 μl ethanol. Following centrifugation at 7000×g for 15 min at 4 °C, the pellet was washed in 750 μl cold 75% ethanol, air-dried, resuspended in 50 μl sterile endonuclease-free water and stored at −20 °C.

Microbial samples were identified by Sanger sequencing of PCR amplicons of bacterial 16S rRNA gene using the primers 16SA1 (Forward: 5′-AGAGTTTGATCMTGGCTCAG-3′) and 16SB1 (Reverse: 5′-TACGGYTACCTTGTTACGACTT-3′) of Fukatsu and Nikoh (1998) and the fungal ITS/5.8S rRNA gene regions using the primers ITS1 (forward: 5’-TCCCTACCTGAACCTGCGG-3′) and ITS4 (reverse: 5’-TCCTCCGCITATTGATATGC-3′) of White et al. (1990). The PCR reactions contained 0.2 μM of either bacterial or fungal primers, 1 U OneTaq® 2X Master Mix with Standard Buffer (New England BioLabs, Ipswich, MA) with 94 °C for 30 s, 30 amplification cycles of 94 °C for 30 s, 60 s annealing temperature at 55 °C for bacteria or 55.3 °C for fungi, 68 °C for 60 s, with final extension for 5 min at 68 °C. PCR products were purified with ExoSAP-IT™ PCR Product Clean Up Reagent (Thermo Fisher Scientific) following manufacturer’s protocols, and Sanger sequencing was conducted for both directions on Applied Biosystems 3730xl at the Cornell University Genomics Facility. Sequences were assembled de novo on Geneious Prime® 2019.0.4 and taxonomic identity was assigned by querying against the NCBI nucleotide database using BLAST (https://blast.ncbi.nlm.nih.gov/Blast.cgi). The sequence data are available at GenBank, NCBI (Accession Numbers MN378407-MN378470).

Field Experiments

The attractive characteristics of volatile compounds produced from microbes associated with SWD and its fruit hosts were evaluated in the field using a mixed variety raspberry planting at Cornell AgriTech, Geneva, NY (42°52′5.69“N, 77° 2’19.91”W) during the late summer of 2018. The assessment was based on the number of captures of adult SWD in traps containing a 60 mm diam. Petri dish with 6 ml autoclaved Drosophila food (negative control) comprising 10% Brewer’s yeast (inactive; MP Biomedicals, Santa Ana, CA), 10% glucose (Sigma-Aldrich, St Louis, MO), 1.2% agar (Apex Bio, Houston, TX), each of the six microbial strains (Table 1) and the commercial monitoring lure (Scentry, Billing, MT) as positive control.

Table 1 Microbial strains selected for detailed analysis
Full size table

To prepare the plates for the field trials, a sample of the glycerol stock was streaked onto solid medium and incubated at 30 °C for 24–30 h for colony growth. A single colony was transferred to 5 ml broth culture and grown to mid-exponential phase. A sample (100–500 μl) was transferred to a 1.5 ml microtube, washed with centrifugation in sterile PBS, quantified by optical density (OD), and then diluted to 108 cells ml−1 PBS using OD/cell number calibration curves constructed for each species. Then, a 50 μl sample was transferred to each of 5 replicate 60 mm diam. × 15 mm height tight-fit lid Petri dishes (VWR) containing 6 ml autoclaved Drosophila food. The microbial cell suspensions were spread using 5–6 4 mm sterile glass beads (Merck KGaA, Darmstadt, Germany) that had been cleaned with hexane (Sigma-Aldrich).

The custom-made traps (Fig. 1; see supplementary methods for details of construction) were suspended in the fruiting zone of raspberry canes. Treatments were deployed in a randomized complete block design with 6 replicates, and one replicate of each treatment per block, with order randomly determined. The trial was repeated three times, 27-August, 5-September and 17-September, 2018, with order of treatments re-randomized for each experiment. Trap counts were conducted at 24 h and 48 h by filtering, and then replacing the trap drowning solution. Male and female SWD and non-target drosophilids were identified by reference to Werner et al. (2018) and quantified for each replicate trap.

Fig. 1
figure 1

The SWD trap. The Petri dish bearing the microbial cells is placed in a 120 mL specimen cup covered with insect-proof mesh. Insects gain access to the trap via 3 mm diameter entry holes, and are collected in drowning solution comprising water with a drop of unscented soap. Full details of trap construction are provided in supplementary methods

Full size image

Volatile Analysis

Five replicates of each test microorganism were prepared by the procedure used for the field experiments, but using 50 mm diam. × 9 mm height Petri dishes (Falcon) with 4 ml Drosophila food. The lid was laid gently on top of each Petri dish, which was then sealed in 9 × 9 cm oven bags (Reynolds Kitchens™, Lake Forest, Illinois), transferred into an air-tight Snapware container (World Kitchen LLC, Rosemont, Illinois) and incubated at 24.5 °C for 48 h. Volatile compounds equilibrated within the oven bags were sampled using customized micro-adsorbent traps filled with 5 mg Tenax TA (60/80 mesh size) as described by Arguello et al. (2013). Volatiles were collected for 10 min from each sample replicate (see Supplementary Methods). Three replicate headspace samples were taken from purchased Scentry lures in a similar manner, but sampling time was limited to 10 s. to prevent saturation and/or breakthrough of odorants on the micro-adsorbent traps.

Trapped volatiles were analyzed using direct thermal desorption (TD) coupled with gas chromatography-mass spectrometry (GC-MS; GC2010+, Shimadzu Scientific Instruments, Inc., Kyoto, Japan) using an Optic 3 high performance GC injector, (ATAS GL, International BV, Eindhoven, The Netherlands), as described by Arguello et al. (2013) (see supplemental methods). Control sample chromatograms collected from un-inoculated growth medium (Drosophila food plus PBS) were subtracted from those representing microbial cultures, allowing us to exclude (as background volatiles) any compounds present in microbial cultures that were not at least 3-fold greater in total ion current (TIC) peak area than the corresponding peak in the control samples. Identification of volatile compounds was accomplished by co-injection with authentic standards and by comparison with mass spectral libraries and published chromatographic data available on the NIST Chemistry WebBook (https://webbook.nist.gov/) and websites provided by PubChem (https://pubchem.ncbi.nlm.nih.gov/) and PheroBase (www.pherobase.com/). Peak-integrated chromatographic data have been archived digitally at the eCommons site at Cornell University (https://ecommons.cornell.edu/handle/1813/2162?) using the link https://doi.org/10.7298/4dqz-0k28.

Statistical Analysis

Total adult SWD captured in each 48 h experiment (sum of captures at 24 h and 48 h) were analyzed using generalized linear mixed models (Proc glmmix, SAS 2009) with experiment and block as fixed effects and, using a Poisson distribution with log link, specifying a random effect at the residual level (odor treatment * block * experiment #) to account for overdispersion. Means were compared using the Tukey test. A beta regression was implemented to assess the specificity of SWD captures compared to all Drosophila species identified in the traps using the ‘betareg’ package (Cribari-Neto and Zeileis 2010) in R (R Core Team 2015). The six replicates in each block were summed to give a single estimate per experiment, generating three replicates per treatment. Replicates that resulted in proportions equal to 1 (only SWD captured) or 0 (no SWD captured) were changed to 0.99 or 0.01 to stabilize model results (this occurred for all three replicates of Curtobacerium sp, G. cerinus, and no microbe-free control, and two replicates G. oxydans). Microbial treatments were included as a categorical fixed effect, and replicates were weighted by the total number of flies captured. Residuals were visually assessed for homoscedasticity. A likelihood ratio test was used to identify significance of the fixed effect (compared fixed effect model to intercept-only model), and a post hoc Tukey test was conducted to generate pairwise comparisons between all treatments.

Variation in volatile chemical composition between replicate microbial samples was explored with multivariate statistical methods using the ‘vegan: Community Ecology Package’ (Oksanen et al. 2019) in R. Wisconsin double standardized TIC peak area data for all detected compounds were used to calculate a Bray–Curtis similarity index, which was further analyzed using two alternative statistical tests. First, differences in VOC composition were analyzed using analysis of similarity (ANOSIM (Clarke and Gorley 2006), with the significance of R determined by 999 random permutations to generate an empirical distribution of R under the null model. Second, Permutational Multivariate Analysis of Variance (PERMANOVA) tested for volatile differences within vs. between microbial accessions (species, kingdom, and species nested within kingdom). The calculated F-statistic was compared to 999 random permutations, employing a false discovery rate correction for multiple comparisons (Benjamini and Hochberg 1995). Further exploration of chemical differences used Similarity Percentage Analysis (SIMPER) (Clarke and Gorley 2006), which calculates the average contributions of specific variables (VOCs) to Bray-Curtis dissimilarity in pairwise group comparisons. To visualize the variation in VOC composition among microbial isolates, volatile and microbial species were grouped, and an associated dendrogram constructed, via hierarchical cluster analysis of Euclidean distances (default parameters for the heatmap.2 function in package ‘gplots’ in R).

A probability cutoff of alpha = 0.05 was applied for tests of significance in all statistical analyses.

Results

The Microorganisms Cultured from SWD and Fruits

In total, 217 microorganisms were isolated into culture (Table S1). Of these, 162 (75%) yielded 16S rRNA gene or ITS sequence PCR amplification products (for bacteria and fungi, respectively), comprising 93 isolates of bacteria and 69 isolates of fungi (Table 2). The bacterial isolates could be assigned to 20 taxa in three phyla, Proteobacteria (α-proteobacteria and γ-proteobacteria), Actinobacteria and Firmicutes. Just three taxa, Gluconobacter cerinus (α-proteobacteria), Curtobacterium sp. (Actinobacteria) and Leuconostoc pseudomesenteroides (Firmicutes), accounted for 53 (57%) of the bacterial isolates. The fungal isolates were identified to 10 taxa, comprising 8 taxa of the order Saccharomycetales (Ascomycotina) and just one taxon in each of Dothideales (Ascomycota) and Basidiomycota. Two species of Hanseniaspora (Saccharomycetales) accounted for 52 (75%) of the fungal isolates. Overall, 23 (77%) of the 30 microbial taxa identified were obtained from just one of the four sample types, i.e. from either SWD or fruit (but not both) at one of the two sites. Just three taxa were recovered from both fruits and SWD: the bacterium Curtobacterium sp. and the yeasts Hanseniaspora uvarum and Pichia sp.

Table 2 Microorganisms cultured from swd and fruits
Full size table

Our selection of six microbial strains for detailed analysis comprised taxa that grew well on the Drosophila food used for field traps and were prevalent in our collection (Table 1).

Field Experiments

Over the 48 h test period, up to 35 SWD from natural populations were recovered from each trap set out in a raspberry planting, with 594 SWD trapped in total (Dataset S1). The numbers trapped varied significantly with odor source (F7,119 = 8.79, P < 0.0001) (Fig. 2a), with no significant effect of either experiment (F2,119 = 0.20, P = 0.81) or block nested within experiment (F15,119 = 0.73, P = 0.75). The negative control of autoclaved Drosophila food trapped 0–3 SWD, and significantly higher capture than the negative control was obtained for the commercial lure Scentry and two yeasts, Hanseniaspora uvarum and H. opuntiae, but not for the yeast Pichia sp. or any of the three bacterial species tested (Fig. 2a). The number of male and female SWD recovered from the traps did not differ significantly (Mann Whitney U: W = 20,142, p = 0.315).

Fig. 2
figure 2

Capture of drosophilid flies in traps with different odor sources, in a raspberry planting and the negative control comprising autoclaved Drosophila food. a). Number of SWD captured over 48 h. Mean values (back transformed from LSmeans generated in SAS) are for 18 replicates (6 replicate traps in each of 3 experiments) for each treatment, following nonsignificant effects of block and experiment (see text). *, significantly different from the negative control by Dunnett’s test (p < 0.05). b) Number of SWD captured expressed as proportion of the total drosophilid flies trapped over 48 h. The estimated marginal means and standard errors are plotted from beta regression. Different letters indicate statistical significance as determined by Tukey’s test. n.a., not analyzed: the control and Curtobacterium sp. baits were excluded from Tukey’s test because the total numbers of flies captured was very low

Full size image

Several drosophilid species other than SWD were recovered from the traps: Drosophila melanogaster and species in the D. obscura group, with an occasional capture of D. tripunctata, D. falleni and flies of the genus Chymomyza, closely related to Drosophila (Dataset S1). Visual inspection of the data revealed that, although the total numbers of flies trapped by baits with Gluconobacter was considerably lower than with baits with Hanseniaspora, the specificity of the Gluconobacter baits was high, with SWD accounting for all but one of the 87 drosophilid flies trapped. The proportion of SWD varied significantly with trap bait (χ2 7 = 1810.4, p < 0.001), and Tukey’s post hoc test yielded significantly greater proportion of SWD in traps baited with Gluconobacter species than other treatments (Fig. 2b).

Volatile Analysis

Volatile collections coupled with TD-GC/MS analysis resulted in the detection of 83 VOCs from three yeast and three bacterial accessions (Fig. 3; Table S2). These compounds included 36 aliphatic esters and 9 alcohols (e.g. ethyl acetate, ethanol, isobutyl acetate, 3-methylbutanol), five aliphatic ketones (including acetoin), five aromatic compounds, one alkane and a series of six carboxylic acids ranging from acetic to hexanoic acid. Additional VOCs included nine putative isoprenoids, six compounds whose mass spectra suggested the presence of substituted cyclohexane rings and six compounds not assignable to a structural or biosynthetic class. Yeast accessions were marked by high total emissions (summed TIC peak areas; Table S2), producing 20- to 1000-fold greater TIC peak areas than bacterial accessions and three-fold higher chemical richness (mean + SEM number of VOCs from yeasts 49.3 + 2.6 but only 15.7 + 4.7 compounds from bacteria). Replicate samples taken from Scentry lures were dominated by ethyl acetate, with much smaller amounts (in decreasing order of abundance) of acetic acid, methyl acetate, ethanol, acetoin, vinyl acetate and methionol (Table S2). Three of these compounds (methyl acetate, vinyl acetate and methionol) were undetectable in the volatile headspace of the microbial accessions.

Fig. 3
figure 3

Heat map of volatile data for all microbial samples, organized by species (columns), with numbered replicates and the 83-compound data set (rows; see Table S2). The scaleindicates ln transforms of TIC peak areas for all verified compounds

Full size image

ANOSIM revealed significant variance structure for VOC data collected from yeast and bacterial accessions within the Bray-Curtis dissimilarity matrix, both by kingdom (R = 0.5238, P = 0.001) and species (R = 0.9972, P = 0.001). PERMANOVA performed on the same data set revealed similar outcomes for comparisons between kingdom (R2 = 0.456, P = 0.001, DF = 1) and species nested within kingdom (R2 = 0.506, P = 0.001, DF = 4). Pairwise PERMANOVA between all microbial pairs yielded significant pairwise differences (all P < 0.02) after applying corrections for multiple comparisons. Follow-up analysis of specific volatiles contributing to differences between accessions using SIMPER (Table S3) showed that VOC composition among the three yeast accessions differed significantly primarily through the exclusive detection of ethanol and higher emissions of isoamyl acetate and 2-phenylethyl acetate from Pichia sp. than from H. opuntiae and H. uvarum. Bacterial volatile blends lacked various VOCs, including ethyl acetate, associated with the yeasts (Table S3). Gluconobacter oxydans and G. cerinus had very similar chemical composition dominated by acetoin, acetic acid and other carboxylic acids (heat map, Fig. 3). The headspace of the Curtobacterium accession differed from all other accessions (Fig. 3), consisting of short chain alcohols (including 3-methyl butanol, hexanol) and the ketone nonanone (Table S2).

Discussion

In this study, we demonstrate that different microorganisms isolated from SWD and SWD-infested fruits vary in their attractiveness to wild SWD populations. Because the traps were designed to prevent direct contact between the insects and microbial cultures, these effects can be attributed to insect olfactory responses to microbial volatiles. Our parallel analysis of VOC profiles of the microorganisms in the laboratory has yielded candidate volatiles contributing to this variation. However, these associations should be treated with due caution because we cannot exclude the possibility that the VOC profile of the microorganisms differed between field and laboratory conditions.

Our data confirm and extend current understanding of microbial VOCs as a determinant of insect behavior and ecology, as well as informing strategies for improved control of SWD. In particular, our demonstration that natural populations of SWD are reliably attracted to single-taxon cultures of Hanseniaspora yeasts and Gluconobacter bacteria in the complex field environment of a mixed raspberry planting reinforces the evidence, primarily from laboratory studies, that SWD is responsive to olfactory cues of microbial origin (Cloonan et al. 2018). For example, there is published evidence that laboratory cultures of SWD significantly prefer H. uvarum cultures relative to sterile medium in two-choice assays and to other yeasts in multi-choice assays (Scheidler et al. 2015). Similarly, the two-arm olfactometer trials of Mazzetto et al. (2016) yielded significant attraction of SWD to one of the Gluconobacter species, G. oxydans, used in our study. Mazzetto et al. (2016) did not investigate G. cerinus, the second species used in this study.

In principle, the microbial taxa tested for attractiveness to SWD might be expected to fall into four functional groups: those that attract both SWD and other drosophilid flies, those that attract either exclusively SWD or various drosophilids other than SWD and, finally, those that fail to attract any insects. We obtained representatives of three of these groups, but no microbial taxa that only attracted non-SWD insects. This absence could have arisen by chance, a consequence of the relatively small number of microbial taxa tested, or from bias introduced by our selection of readily-culturable microorganisms isolated from SWD and SWD-infested fruits. Nevertheless, these data are consistent with the conclusion of Kleiber et al. (2014) that SWD is responsive to a wider range of volatiles than other drosophilid flies.

The selective attraction of SWD, and not other drosophilids, to the VOCs of some microorganisms is trait may, in turn, be correlated to the generalist ecology of SWD. Specifically, although both SWD and other drosophilids feed extensively from over-ripe fruits and other fermenting plant material, SWD also feed on macrofungi and oviposit into ripening fruits and also, in non-choice conditions in the laboratory, into mushroom substrate (Atallah et al. 2014; Keesey et al. 2015; Stockton et al. 2019; Werner et al. 2018). Consistent with this evidence that SWD is likely less specialized than some other fruit-feeding drosophilids, GC/EAD studies reveal that SWD is significantly more responsive to leaf odors and various fruit odors compared to D. melanogaster, which utilizes over-ripe fruits (Keesey et al. 2015). However, SWD does display some specificity in its response to microbial odors. Notably, SWD responded significantly in two-arm olfactometer trials to just three of the six species of Acetobacteraceae tested by Mazzetto et al. (2016), who identified ethanol as a likely key component of the VOCs from the three attractive species. As for animal-microbial interactions generally (Douglas 2018), the interactions between SWD and microorganisms are also likely influenced by the life history traits of the microbial taxa.

The evidence that SWD but no other drosophilid flies in the habitat were attracted to Gluconobacter species tested in this study offers a potential route to develop lures of greater specificity for SWD. Future strategies to improve SWD capture can include methods to increase total VOC production by Gluconobacter (which is considerably lower than the emissions from yeasts in our laboratory studies, Table S2) and to test a wider panel of Gluconobacter isolates, following the evidence of among-isolate variation in SWD responses (Mazzetto et al. 2016).

A further feature of this study was the minimal numbers of flies of any species that were attracted to our isolate of the yeast Pichia sp., even though this isolate was more volatile-rich, yielding 30–50% greater TIC peak area than the Hanseniaspora yeasts, which attracted SWD (Table S2). Interestingly, 3-methylbutyl acetate (also known as isoamyl acetate), which is significantly elevated in the unattractive Pichia relative to the attractive Hanseniaspora species (Table S3), has been reported to decrease the attraction of SWD to acetic acid and ethanol under laboratory and field conditions (Cha et al. 2012; Cha et al. 2014). Taken together, the results of this study provide the basis for further research on microbial volatiles and volatile blends that are specific attractants and deterrents under field conditions.

Immediately relevant to these considerations is the evidence that SWD is also highly responsive to visual cues, especially red and black (Basoalto et al. 2013; Kirkpatrick et al. 2016; Rice et al. 2016) and spectral contrasts between fruit and background vegetation (Little et al. 2018). The importance of vision in SWD is reflected in the anatomy of the insect. Compared with D. melanogaster, which is more dependent on olfactory cues, adult SWD have relatively larger compound eyes with more ommatidia, fewer trichoid sensilla on their antennae and, in the brain, larger optic lobes and smaller antennal lobes (Keesey et al. 2019). This difference offers the opportunity to improve the specificity of SWD trapping strategies by combining visual cues, e.g. colored spheres, with specific microbial odors and, for lure-and-kill strategies, insecticidal formulations (Rice et al. 2017), More generally, SWD offers the opportunity for multi-modal investigations of how insects integrate information from volatiles from microbial and other sources, together with visual cues as they navigate the natural environment to identify food and sites for mating and oviposition.