1.In nature, the fruit fly Drosophila melanogaster is attracted to fermenting fruit. Micro-organisms like Saccharomyces yeasts growing on fruit occupy a commonly overlooked trophic level between fruit and insects. Although the dietary quality of yeast is well established for D. melanogaster, the individual contribution of fruit and yeast on host finding and reproductive success has not been established.
2.Here, we show that baker's yeast Saccharomyces cerevisiae on its own is sufficient for fruit fly attraction, oviposition and larval development. In contrast, attraction and oviposition were significantly lower if non-fermented grape juice or growth media were used, and yeast-free grapes did not support larval development either.
3.Despite a strong preference for fermented substrates, moderate attraction to and oviposition on unfermented fruit might be adaptive in view of the fly's capacity to vector yeast.
4.Signals emitted by fruit were only of secondary importance because fermenting yeast without fruit induced the same fly behaviour as yeast fermenting on fruit. We identified a synthetic mimic of yeast odour, comprising ethanol, acetic acid, acetoin, 2-phenyl ethanol and 3-methyl-1-butanol, which was as attractive for the fly as fermenting grape juice or fermenting yeast minimal medium.
5.Yeast odours represent the critical signal to establish the fly–fruit–yeast relationship. The traditional plant–herbivore niche concept needs to be updated, to accommodate for the role of micro-organisms in insect–plant interactions.
Yeasts and other micro-organisms occupy a commonly overlooked trophic level between plants and insect herbivores. Insects from several orders, including beetles, flies, ants and bees interact with yeasts (Ganter 2006; Janson et al. 2008). Many insects feeding on flowers or fruit encounter yeasts (Raguso 2004; Ganter 2006), and yeasts that use insects as hosts and vectors are widespread (Suh et al. 2005; Gibson & Hunter 2010).
Especially in late summer and autumn, the availability of ripe fruit allows wild yeasts to flourish. Niches with freely available fruit sugars are predominantly exploited by Saccharomyces cerevisiae and its close relatives that are able to out-compete other micro-organisms by the accumulation of ethanol (Piškur et al. 2006; Rozpędowska et al. 2011a).
For Drosophila flies, the presence of yeast in larval diet substrates is fundamental for the occupation of niches and separation of species (Begon 1982; Starmer & Phaff 1983), and interactions between yeast and fruit flies have led to mutual coadaptations. Nutritional gains or detoxification of harmful chemicals sustain larval development, while feeding and vectoring by flies mediate dispersal and outbreeding of yeast strains (Starmer & Fogleman 1986; Reuter, Bell & Greig 2007). Attraction and tolerance to ethanol are, in addition, important factors of habitat selection and interspecific competition (McKenzie & Parsons 1972). The model organisms Drosophila melanogaster and S. cerevisiae are particularly suitable experimental models for studying the co-evolution of yeasts and insects. D. melanogaster is attracted to overripe fermenting fruit for feeding, mating and oviposition (Fig. 1; Ganter 2006; Zhu, Park & Baker 2003). Yeasts are known to support larval development and survival (Anagnostou, Dorsch & Rohlfs 2010b; Anagnostou, Le Grand & Rohlfs 2010a). A higher nutritional value of fruit inoculated with yeast over fruit alone is expected to favour sensory and physiological adaptations that facilitate detection and location of fermenting fruit for feeding and egg-laying. However, it is still unclear which compounds constitute the core communication signals in plant–insect–yeast associations.
A fundamental issue towards an understanding of the mechanisms driving insect–fruit–yeast relationships is the individual contribution of the plant substrate and the yeast to habitat quality and habitat choice by the insect. It has so far not been investigated whether attraction and oviposition of D. melanogaster are triggered by the fruit, the yeast or the combination of both. In addition, it is unclear whether volatiles released by fruit and yeast correlate with the nutritional value and suitability of the substrate for larval development.
We used berries and juice of grapes Vitis vinifera and minimal growth medium as substrate for baker's yeast S. cerevisiae and showed that yeast, and not the plant component, was the limiting factor for attraction, oviposition and larval development of D. melanogaster fruit flies. Furthermore, we identified the compounds that are crucial for fly attraction.
Materials and methods
The D. melanogaster wild-type strain Dalby-HL (Ruebenbauer et al. 2008), reared on Bloomington standard cornmeal medium was used for oviposition and attraction experiments. Flies were reared at room temperature (19–22 °C) and under a 12:12 h L/D photoperiod. An Oregon R strain of D. melanogaster free of live yeast (courtesy of L. Søndergaard, University of Copenhagen) was reared on sterilized Bloomington food, containing 10 mL of 10% methyl 4-hydroxybenzoate (MP Biomedicals, Illkirch, France; dissolved in 95% aqueous ethanol; Labscan, Malmö, Sweden) per 1280 g of food. Flies were 2–3 days old for wind tunnel and offspring development experiments and 3–6 days old for oviposition assays. Short CO2 anaesthetization was used for the sexing of flies.
Fungi and yeasts
Three different microfungi, usually associated with grapes, were applied in this study for the inoculation of media or Drosophila feeding; S. cerevisiae (strain S228C) or commercial baker's yeast (Jästbolaget AB, Sollentuna, Sweden), Botrytis cinerea (wild strain isolated from grape, Federal College and Research Institute for Viticulture and Pomology, Klosterneuburg, Austria) and Penicillium sp. (wild strain isolated from decayed Drosophila food and determined by G. Svedelius, SLU Alnarp, Sweden).
Aerobic batch cultivation (n =3) of S. cerevisiae in 0·5 L synthetic minimal medium (Merico et al. 2007) or grape juice was performed in computer-controlled bench-top bioreactors (Mulitfors, INFORS HT, Switzerland) (Rozpędowska et al. 2011b). The reactors were inoculated with a starting culture of 0·4 OD at λ = 600 nm. The cultivation was sustained for 3 days at 30 °C and aerated by an airflow of 0·5 L min−1; dissolved oxygen was measured and concentration >30% of saturation was maintained by the regulation of steering. Samples of 70 mL were taken for headspace collection after 20 h, 30 h and at the end of the fermentation.
Headspace collection and chemical analysis
Headspace from each sample of unfermented and fermented medium and grape juice was collected for 120 min with air filters (Super Q, 80/100 mesh; Alltech, Deerfield, IL, USA) (Becher et al. 2010). The filters were eluted with about 300 μL of methanol (Labscan) every 30 min, and eluates of the same sample were combined. Samples were analysed by gas chromatography–mass spectrometry (GC-MS; 6890 GC and 5975 MS; Agilent Technologies Inc., Santa Clara, CA, USA), on a DB-Wax capillary column (30 m × 0·25 mm × 0·25 μm; J&W, Agilent) using helium as carrier gas at a flow rate of 35 cm s−1. The oven temperature increased from 30 °C (3 min hold) at 8 °C min−1 to 225 °C (5 min hold). The temperature of the injection port was 225 °C. The MS was operated in the EI-mode (70 eV), and mass data were recorded from m/z 29–330 at 2 scans s−1. Compound identities were identified according to retention times (Kovats retention indices), mass spectra (NIST library, Agilent) and by co-injection of authentic standards.
Yeast vectoring and offspring development assays
Experiments on yeast vectoring and offspring development were set-up under sterile conditions in a clean bench. The epidermis of fresh undamaged grape berries was sterilized by immersing them for 15 min in 70% ethanol. After evaporation, berries were cut in quarters and four pieces from different berries were added into individual sterile glass test tubes with metal caps. In a first set of experiments, single female (n =10) or couples of male and female flies (n =10) of the yeast-free Oregon R strain were kept for 1 day on a yeast–cornmeal diet (rearing vials inoculated with additional live S. cerevisiae) and then transferred to tubes with yeast-free grapes to test for their potential to vector yeast and produce offspring on the potentially inoculated fruit. Tubes without flies were used as control. Tubes were closed and kept under the same laboratory conditions as the fly rearing.
In a second set of experiments, half of 14 test tubes with grapes were inoculated with S. cerevisiae. One mated yeast-free female D. melanogaster was then added to each test tube (n =7). Successively emerging flies were counted and removed from the tubes; mother flies were removed together with their first offspring.
On grapes where yeast growth could not directly be determined visually (controls), absence of yeast was double-checked by inoculating a streak from the samples on YPD agar plates (peptone, 2% w/v; yeast extract, 1% w/v; glucose, 2% w/v). In controls, neither flies nor grapes from the S. cerevisiae-free tubes induced growth of yeast on YPD.
Wind tunnel assay
Attraction of Drosophila to headspace odour samples of unfermented and fermented grape juice and synthetic minimal medium was tested in our wind tunnel assay equipped with a piezoelectric sprayer (Becher et al. 2010). Furthermore, a synthetic mimic of fermented minimal medium was sprayed to demonstrate the effect of fermentation products as behavioural most relevant signals. The components of the mimic were selected in a semi-subtractive approach (data not shown) in which we deducted constituents from a full blend corresponding to a headspace sample of 20-h fermented minimal medium, while keeping three previously identified, behaviourally most relevant fermentation compounds co-occurring in vinegar (acetic acid, acetoin and 2-phenyl ethanol; Becher et al. 2010). Our approach allowed subtraction of 8 out of 13 identified headspace components, without losing the behavioural activity of the full yeast headspace blend.
Headspace samples and the five-component synthetic mimic were vaporized at a rate of 10 μL min−1 and flies released in batches of 20 were scored for landing on the odour source during a test period of 15 min (n =5–10).
Oviposition substrates were prepared from grape juice or synthetic minimal medium. Microfungi were grown as batch cultures for 1 or 6 days in the liquid juice or medium before cold thickening locust bean gum (250 mg mL−1; Tørsleffs-Melatin, Haugen-Gruppen AB, Norrköping, Sweden) was added as gelling agent. The substrates were prepared as thin films on glass slides (7·5 × 2·5 cm). For each assay, batches of twenty females (n =12) were starved on humidified cotton 24 h before the experiment. Flies (20) and individual substrates were kept for 24 h in high Petri dishes (6·5 cm high and 12 cm diameter) at 25 °C and 70% rh. Eggs laid on the substrate were counted under a stereomicroscope. To study the effect of S. cerevisiae in more detail, we did additional tests in a slightly modified assay using plastic cups (3 cm high and 4 cm diameter) instead of glass slides. Tested was the effect of Botrytis cinerea or Penicillium sp. in comparison with S. cerevisiae, grown on grape juice substrate (n =10). Furthermore (see Appendix S1), we tested the effect of S. cerevisiae fermentation time (1 or 6 days of fermentation; n =15), pasteurisation (n =5) and yeast feeding prior oviposition (n =15).
Larval development assay
To exclude the factor of egg-laying on offspring development, another set of experiments was carried out with 2-day-old yeast-free larvae. Assays were performed similarly as described above for the offspring development, but larvae instead of gravid flies were transferred to the test tubes with substrate. Groups of five larvae were transferred to a tube containing synthetic minimal medium or pieces of a grape berry, with or without yeast inoculation (n =10–20). The tubes were handled and kept as described for the offspring development assay, and the number of larvae developing to adults was recorded.
Normal distributed data (wind tunnel data and oviposition data from the comparison between fruit- and synthetic minimal media, shown in Figure 3) were square-root transformed to homogenize variances (Bartlett test) and analysed for statistical significances (Tukey's test following an analysis of variance, anova). Nonparametric tests (Mann–Whitney or Dunn multiple comparison following a Kruskal–Wallis test) were calculated when data did not meet the assumption of normal distribution (D'Agostino omnibus K2 test). A Fisher's exact test was used for categorical comparison between larval development with and without yeast.
Fruit is a substrate for D. melanogaster and S. cerevisiae
The human commensal D. melanogaster breeds on fruit, as particularly noticeable in the autumn during the time of harvest and wine fermentation. This is also the season when yeasts like S. cerevisiae proliferate, as they find excellent growth conditions on sugar-rich fruit substrates. To test whether insect vectoring is relevant for yeast proliferation, we tested single D. melanogaster flies for their ability to establish S. cerevisiae growth on grapes.
After the yeast-free strain of D. melanogaster was kept for 1 day on a diet with live S. cerevisiae, single female flies were transferred to sterilized grapes, from where S. cerevisiae subsequently could be isolated in all replicates (n = 10). Yeast was not present on the control grapes and flies that were kept without live yeast. Yeast inoculation of grapes by couples of parent flies (n = 10) led to successful development of the progeny to adult flies.
If inoculated with yeast and flies, the combined effect of fermentation and larval feeding led to decomposition of the grapes. In the absence of flies, S. cerevisiae was developing vigorously on the surface of the cut berries. With flies (yeast-free strain) but without yeast, the visible decay was minor.
D. melanogaster uses yeast to proliferate on fruit
Females produced significantly more offspring on grapes that were inoculated with yeast than on yeast-free grapes (Fig. 2; Mann–Whitney test; n =7; P <0·005). Saccharomyces cerevisiae was a limiting factor for the development of flies (Fisher's exact test; P < 0·05). Larvae hatched from eggs on yeast-free grapes, but only very few flies developed (9 flies on yeast-free grapes vs. 142 in presence of yeast). The first flies eclosed after 28 days from yeast-free grapes, whereas flies from grapes inoculated with yeast emerged already after 19 days.
This experiment did not clarify (i) to which extent substrate-specific food attraction, oviposition rate and larval survival contributed to the overall production of offspring, and it did not allow to (ii) distinguish the respective contribution of grape and yeast to adult behaviour and larval development. Further experiments were carried out using bioreactors for controlled fermentation (Rozpędowska et al. 2011b) and a wind tunnel equipped with a piezoelectric sprayer for the evaporation of grape and yeast volatiles at known and constant rates (Becher et al. 2010) to differentiate the contribution of each partner.
Yeast-produced volatiles mediate the attraction to overripe fruit
Headspace of unfermented grape juice induced significantly stronger attraction (30·5 ± 7·98%) than unfermented synthetic medium (11·5 ± 6·26%). Fermentation significantly increased upwind flight attraction to both grape juice and to synthetic minimal medium (Fig. 3). Fermentation time (20 h, 30 h or 3 days) did not have a significant effect on the attraction response to fermented grape juice (50 ± 11·55% after 20 h; 55 ± 18·4% after 30 h; 51 ± 13·9% after 3 days of fermentation) or fermented synthetic medium (55·5 ± 9·56% after 20 h; 52·5 ± 16·03% after 30 h; 61·5 ± 5·1% after 3 days of fermentation) (anova followed by Tukey's test; n =10, F =24·05, d.f. = 79, P <0·0001). Yeast odours apparently represented the critical signal to induce the fruit fly behaviour but also grape juice alone attracted the flies to some degree.
Thirteen volatile compounds were identified in headspace collections from fermented synthetic minimal medium that attracted flies in the wind tunnel (Table 1, Fig. 3). Attraction to a synthetic five-component fraction of the yeast headspace (selected in a semi-subtractive approach; data not shown) containing acetoin (3·2 ng μL−1), acetic acid (4·7 ng μL−1), 3-methyl-1-butanol (10·0 ng μL−1) and 2-phenyl ethanol (0·2 ng μL−1), dissolved in ethanol was found not to be significantly different (65·0 ± 10·61%) from the average attraction to headspace collected from the fermented substrates (grape juice, 52 ± 4·89%; minimal medium, 56·5 ± 8·97%; Dunn′s multiple comparison test following Kruskal–Wallis statistics; n =5–10, H =5·096, P =0·0783). A blend of these five compounds represents a core signal that attracts the fruit fly.
Table 1. Volatiles from headspace samples produced by fermentation of synthetic minimal medium, 20 h after inoculation with baker's yeast Saccharomyces cerevisiae
Mean ratio ± SD
Ethanol was visible as main component of the headspace, but because of its co-elution with the solvent peak not repetitively recorded for GC–MS quantification. n.d. not determined.
3·0 ± 0·9
2·4 ± 0·4
0·2 ± 0·2
0·3 ± 0·2
4·4 ± 0·6
43·9 ± 9·9
14·1 ± 9·1
6·1 ± 1·2
20·5 ± 4·3
0·3 ± 0·2
4·1 ± 1·9
0·8 ± 0·1
Live S. cerevisiae boosts oviposition of D. melanogaster on fruit substrate
One-day-starved D. melanogaster females were tested for their oviposition preference on jelly substrate based on either grape juice or synthetic minimal medium (Fig. 3). Yeast inoculation increased oviposition on both substrates significantly (anova followed by Tukey's test; n =12, F =37·94, d.f. = 47, P < 0·0001). Oviposition rate was as high on synthetic minimal medium (7·59 ± 3·06 eggs per female) as on substrate made of grape juice (7·56 ± 2·34 eggs per female) when inoculated with S. cerevisiae. Without yeast, females laid more eggs on grape juice substrate (3·91 ± 1·79) than on substrate made of synthetic mineral medium (0·94 ± 0·89; Fig. 3).
The effect of S. cerevisiae on oviposition was further studied (for details, see Appendix S1) showing that only live, fermenting – but not pasteurized – yeast increased oviposition; no difference was found between substrates fermented for 1 day or for 6 days. Flies fed with yeast-based cornmeal diet deposited more eggs than starved females.
In addition to yeast, other microfungi might affect D. melanogaster. Botrytis cinerea and Penicillium sp., microfungi commonly found on grapes (La Guerche et al. 2007), were tested for their effect on egg-laying. Botrytis cinerea (1·03 ± 0·78) and Penicillium sp. (1·35 ± 1·13) did not, in contrast to S. cerevisiae (12·94 ± 4·01), increase oviposition on the grape juice substrate (1·77 ± 1·75), (Kruskal–Wallis test followed by Dunn's multiple comparison test; n =10, H =22·64, P <0·0001).
Yeast supports larval development on fruit
Apart from its effect on fly attraction and oviposition behaviour, S. cerevisiae had a significant effect on larval development (Kruskal–Wallis test followed by Dunn's multiple comparison test; n =10–20, H =30·57, P < 0·0001). In the presence of yeast, 52·67 ± 27·58% of 2-day-old larvae developed to adult flies on minimal medium and 66 ± 31·34% on grapes. In contrast, only 18·0 ± 21·42% of larvae developed on grapes without yeast (Fig. 3). No adults eclosed from larvae on minimal medium free of yeast, although few larvae survived for up to 27 days and 3% of them developed to pupae.
The fruit harvest in our gardens is accompanied by Drosophila fruit flies swarming around fermenting fruits that serve as adult food, rendezvous site for mating and oviposition sites. However, the respective contribution of fruit and yeast to the host-finding behaviour and reproductive success of D. melanogaster, and the chemical background for communication among the three partners, has so far not been clearly established.
We studied the significance of yeast and fruit as individual components involved in fly attraction, oviposition and larval development. We show hereby that yeast is not only required but is also sufficient to support proliferation of D. melanogaster (Figs 2 and 3). Whereas flies reproduced vigorously on yeast-inoculated grapes, only very few and unusually small flies developed from eggs deposited by yeast-free females on yeast-free grapes (Fig. 2).
Fruit provides energy, nutrients and growth factors supporting yeast metabolism and the production of yeast-specific volatile fermentation products. D. melanogaster approaching overripe fruit consequently is exposed to a blend of fruit and yeast-derived odours, which are ligands for odorant receptors in D. melanogaster olfactory sensory neurons (Stensmyr et al. 2003; Hallem & Carlson 2006; Grosjean et al. 2011). Also, oviposition is under neuronal control leading to selective egg-laying on suitable substrates. The choice of the substrate is the result of a sensory evaluation expressed in search-like behaviour, preceding the stereotypical ovipositioning programme (Allemand & Boulétreau-Merle 1989; Yang et al. 2008).
Remarkably, fly attraction to the fermenting synthetic minimal medium was as strong as to fermenting grape juice. Fruit odour did not enhance the attraction of flies to yeast significantly (Fig. 3).
We identified the yeast volatile compounds, which were sufficient to obtain a strong response from D. melanogaster. A synthetic blend of five yeast-produced compounds (acetoin, acetic acid, ethanol, 3-methyl-1-butanol and 2-phenyl ethanol) elicited attraction similar to authentic yeast odour, which confirmed that flies use yeast volatiles for upwind attraction.
Fruit alone was still to some degree able to attract the flies. We suggest that in nature, flies could inoculate fruit with vectored yeast cells. Later on, when the fermentation starts and yeast proliferates, additional volatile signals induce a stronger behavioural response in D. melanogaster.
The inoculum of a single fly derived from a diet containing live baker's yeast induced colony formation and fermentation on sterile grapes. Fly-induced inoculation of ripe grapes was hence sufficient for successful larval development and even enabled colonization of new breeding sites. Despite their strong preference for fermented substrates, moderate attraction to and oviposition on unfermented fruit might be adaptive in view of the fly's capacity to vector yeast.
Yeast vectoring in fruit flies and other insects (Sang 1956; Ganter 2006) has led to mutual coadaptations. Yeasts benefit from targeted dispersal to sugar-rich fruit substrates and from outbreeding by digestion and dissolution of yeast spores (Reuter, Bell & Greig 2007). Insect-derived fitness effects are therefore expected to drive the emission of yeast-produced volatile signals to promote yeast dispersal.
Two other microfungi typically found on grapes, B. cinerea and Penicillium sp., in contrast to S. cerevisiae, did not increase oviposition in our experiments. Interestingly, Penicillium species are known to emit the off-flavour geosmin that diminishes the attraction of D. melanogaster and might serve as a signal for the presence of toxic fungi, such as Penicillium species (Castillo et al. 1999; La Guerche et al. 2007; Becher et al. 2010).
In their pioneering work that has nearly faded into oblivion, (Loeb & Northrop 1916) showed that the development of Drosophila sp. is constrained when reared on sterilized banana without yeast. The dietary quality of yeast has since then been well established. Yeast contains essential food components such as amino acids, lipids and vitamins that fruit alone evidently does not provide. Yeast mediates fly performance by promoting adult attraction, oviposition and larval development, which most likely lead to chemosensory and physiological adaptations towards detection of yeast resources. The behavioural choice of flies, both with respect to enhanced upwind flight response and oviposition rate in response to yeast, matched the enhanced larval performance on yeast diets (Fig. 3) and is in line with the preference–performance concept that predicts reduced investments into unsuitable oviposition substrates (Jaenike 1978; Singer & Thomas 1988; Gripenberg et al. 2010).
The role of yeast volatiles in host location of insects associated with fruit, and possibly even in other insect herbivores, has probably not been sufficiently appreciated. Yeast volatile signatures mediate host finding and discrimination, as shown here for D. melanogaster, and will consequently make an equally important contribution to insect–host plant relationships, including the formation of host races and related speciation events.
It is intriguing that most of the yeast volatiles we found in fermenting minimal medium are also commonly known from flowers and fruit volatiles (Table 1) (Knudsen et al. 2006; Negre-Zakharov, Long & Dudareva 2009; Stökl et al. 2010). Some floral compounds may be produced by yeasts that are associated with flowers, but some volatiles, such as 2-phenyl ethanol, are produced de novo by both plants and yeasts, via different pathways (Tieman et al. 2007).
In the exemplary work on the tephritid fruit fly Rhagoletis pomonella, the physiological sensitivity and attraction behaviour to host odour have been shown to be a key element of sympatric host race formation (Linn et al. 2003; Linn, Nojima & Roelofs 2005; Olsson, Linn & Roelofs 2006). It is eye-catching that the behaviourally active host odours, including 3-methyl-1-butanol are at least in part also typical yeast metabolites (Table 1) (Carrau et al. 2008; Sumby, Grbin & Jiranek 2010). Yeast is a food resource in tephritid fruit flies (Christenson & Foote 1960; Yee 2008) and it is conceivable that yeast volatiles contribute to host location in these flies as well.
Mutualisms with micro-organisms are considered to influence adaptive evolutionary diversification and to shape host specialization in insect herbivores (Janson et al. 2008). A marked response to microbial odours will further facilitate the diversification of phytophagous insects because adaptive, chemically mediated behaviours-like habitat choice are drivers of chemosensory speciation (Smadja & Butlin 2009). S. cerevisiae mediates fly attraction, oviposition and larval feeding and enables the exploitation of fruit as food source by D. melanogaster. This is an essential contribution to our understanding of the behavioural physiology and ecology of D. melanogaster.
This study emphasizes that the traditional bi-trophic, plant–herbivore and tri-trophic, plant–herbivore–carnivore niche concepts (Singer & Stireman 2005) must be updated, to accommodate for the role of micro-organisms in insect–plant interactions.
This study was supported by the INTERREG IVA community initiative ‘Pomerania’ funded by the European Commission (ERDF), the Trygger Foundation, the Swedish Research Council and by the Linnaeus initiative ‘Insect Chemical Ecology, Ethology and Evolution’ IC-E3 (Formas, SLU). Guy Svedelius and Reinhard Eder provided microfungi, Leif Søndergaard flies; Louise Johannson helped setting up the bioreactors.