Chemical signaling and insect attraction is a conserved trait in yeasts

Abstract Yeast volatiles attract insects, which apparently is of mutual benefit, for both yeasts and insects. However, it is unknown whether biosynthesis of metabolites that attract insects is a basic and general trait, or if it is specific for yeasts that live in close association with insects. Our goal was to study chemical insect attractants produced by yeasts that span more than 250 million years of evolutionary history and vastly differ in their metabolism and lifestyle. We bioassayed attraction of the vinegar fly Drosophila melanogaster to odors of phylogenetically and ecologically distinct yeasts grown under controlled conditions. Baker's yeast Saccharomyces cerevisiae, the insect‐associated species Candida californica, Pichia kluyveri and Metschnikowia andauensis, wine yeast Dekkera bruxellensis, milk yeast Kluyveromyces lactis, the vertebrate pathogens Candida albicans and Candida glabrata, and oleophilic Yarrowia lipolytica were screened for fly attraction in a wind tunnel. Yeast headspace was chemically analyzed, and co‐occurrence of insect attractants in yeasts and flowering plants was investigated through a database search. In yeasts with known genomes, we investigated the occurrence of genes involved in the synthesis of key aroma compounds. Flies were attracted to all nine yeasts studied. The behavioral response to baker's yeast was independent of its growth stage. In addition to Drosophila, we tested the basal hexapod Folsomia candida (Collembola) in a Y‐tube assay to the most ancient yeast, Y. lipolytica, which proved that early yeast signals also function on clades older than neopteran insects. Behavioral and chemical data and a search for selected genes of volatile metabolites underline that biosynthesis of chemical signals is found throughout the yeast clade and has been conserved during the evolution of yeast lifestyles. Literature and database reviews corroborate that yeast signals mediate mutualistic interactions between insects and yeasts. Moreover, volatiles emitted by yeasts are commonly found also in flowers and attract many insect species. The collective evidence suggests that the release of volatile signals by yeasts is a widespread and phylogenetically ancient trait, and that insect–yeast communication evolved prior to the emergence of flowering plants. Co‐occurrence of the same attractant signals in yeast and flowers suggests that yeast‐insect communication may have contributed to the evolution of insect‐mediated pollination in flowers.


| INTRODUCTION
Yeasts are microscopic organisms, which communicate through a distinct aroma, rather than by visual signals. Even over distance, fermenting yeasts can be noticed by their characteristic, often sweet fragrance.
The smell of yeast consists of volatile metabolites produced during growth on organic substrates. A sweet smell is not exclusive for yeasts; it is an odor quality that we also attribute to flowers and fruits.
Both flowers and fermenting yeasts attract foraging insects through emission of partially overlapping chemical signals (Stökl et al., 2010). The sweet odor of 2-phenyl-ethanol, for example, is a key constituent of the odor bouquets of yeasts and flowers. Not surprisingly, flower-visiting insects can be lured to fermenting sugar substrates or sweet baits (El-Sayed, Heppelthwaite, Manning, Gibb, & Suckling, 2005;Landolt, Todd, Zack, & Crabo, 2011).
Pre-existing signals that mediated insect attraction to yeasts could have become beneficial for flowers by facilitating pollination. In view of overlapping chemical signals produced by flowers and yeasts, we discuss the concept that yeast-like volatiles emitted by flowers or their associated microbes might have influenced the evolution of pollinator attraction. Clarifying the role of yeasts in insect attraction to flowering plants contributes to our understanding and possibly the management of plant-pollinator interactions, which is an urgent current challenge (Bailes, Ollerton, Pattrick, & Glover, 2015;Brown et al., 2016;Potts et al., 2016).
Yeasts spanning a wide ecological and phylogenetic range were grown under controlled growth conditions in bioreactors. Headspace samples of these yeasts were tested for bioactivity in the vinegar fly Drosophila melanogaster (Figure 1). Moreover, headspace of Yarrowia lipolytica, the most ancient yeast of our selection, was tested for attraction of the springtail Folsomia candida (Figure 1). Following chemical analysis of the headspace samples, a database search was conducted for co-occurrence of yeast volatiles in flowering plants, and the behavioral role of these compounds in insect attraction. For the tested yeast species with known genome, we searched for the occurrence of genes involved in the synthesis of key volatiles. We conclude that insect-yeast chemical communication is a phylogenetically ancient trait and suggest that yeasts may have played a role in the evolution of pollination.
Fresh colonies grown on YPD agar were applied for preparations of precultures in synthetic minimal medium as described by Verduyn, Postma, Scheffers, and Van Dijken (1992). Minimal medium was selected to focus on basic yeast volatiles, and to avoid volatile emissions from more complete or natural media typically emitting a strong smell.
Controlled aerobic batch cultivation was performed in 1 L of medium using bench-top bioreactors (Multifors, INFORS HT, Switzerland). Each reactor was inoculated with an approximately 500-fold diluted starting culture compared to the biomass concentration in the bioreactors, at the end of each experiment (Hagman et al., 2013). The cultivation was maintained at 25°C and aerated with an airflow of 1 L/min, dissolved oxygen was measured, and concentration >30% of saturation was maintained by regulation of steering; a pH of 5 was maintained by automatic pH measuring and buffering with 1M H 2 SO 4 and 2M KOH. All yeasts showed exponential growth and increase in biomass (Table 1).
Cultivations were allowed to grow until reaching maximum cell density in the stationary phase. Yeast growth was followed by continuous measurement of CO 2 and O 2 concentrations in the gas outlet, and regular measurement of OD in the suspension. Samples of 50 ml volume were taken every 3-4 hr during the cultivation of the individual yeasts.
Fermentation was stopped by immediately cooling down the samples to −80°C. For each of the yeasts, the sample representing the highest cell density before reaching the stationary phase was selected for behavioral analysis. Additional samples of earlier growth phases with lower cell densities were selected from the cultivation of S. cerevisiae.

| Yeast taxonomic identification
Ten colonies isolated from three wild D. melanogaster flies were cultivated for analysis of species identity. For genomic DNA extraction, 4 ml of liquid YPD medium was inoculated with single yeast colonies and incubated overnight at 25°C. The cultures were spun down (10,060 g, 2 min) and washed with water. 200 μl of the lysis buffer (2% Triton X-100, 1% SDS, 0.1 mol/L NaCl, 0.001 mol/L EDTA, 0.01 mol/L Tris at pH 8), 200 μl phenol:chloroform:isoamyl alcohol (25:24:1), and 100 μl acid-washed glass beads were added to the pellet. The mix was vortexed for 10 min, and 200 μl of TE buffer (10 mmol/L Tris at pH 7.5 -8, 1 mmol/L EDTA at pH 8) was added. The suspension was centrifuged for 10 min at 10,060 g, and 10 μl RNase A (10 mg/ml) was added to the aqueous phase and incubated for 45 min at 37°C. The DNA was precipitated with 96% ethanol and 3 mol/L sodium acetate.
The mixture was centrifuged for 10 min at 10,060 g at 4°C. The pellet was washed with ice-cold 70% ethanol, air-dried, and re-suspended in 40 μl TE buffer (pH 8).

| Headspace sampling and chemical analysis
We modified the gas outlet system of the bioreactors by inserting air filters (Super Q, 80/100 mesh; Alltech, Deerfield, IL, USA) to collect volatiles emitted by the fermenting yeast cultures. Online sampling from the bioreactors was aiming at collecting odors at specific stages of the fermentation process rather than a blend of odors emitted during changing growth conditions. Adsorbed headspace volatiles were eluted from the air filters with heptane and methanol. Eluents corre- was conditioned for 20 min in a GC injection port at 270°C, passed through a small hole in the foil and exposed above the yeast sample. After 10 min of sampling, the fiber was immediately subjected to GC-MS analysis under similar settings as described above. Headspace compounds were identified according to retention indices and mass spectra, in comparison with a reference library (NIST, Agilent) and authentic reference compounds.
Every few hours, samples of 2 ml were taken from the fermentors for detailed analysis of metabolites produced during the fermentation of S. cerevisiae. Concentrations of glucose, ethanol, acetate, and glycerol were determined with a HPLC 1200 series (Agilent) equipped with a 300*7.7 mm Aminex HPX-87H Column (Bio-Rad). The mobile phase was H 2 SO 4 (5 mmol/L), and flow rate was set to 0.6 ml/min. Column temperature was set to 60°C and RID temperature to 55°C (Hagman et al., 2013). and landing at the odor source. Flies were released from a jar (Becher et al., 2010) at the downwind end of the tunnel and exposed to yeast odor delivered from the upwind end. For odor delivery, yeast samples were brought to room temperature and transferred to a wash bottle prior to testing. Charcoal filtered air (0.5 L/min) was blown through the bottle and, via an attached Pasteur pipette, yeast volatiles were injected into a glass jar in the center of the wind tunnel at the upwind end (Becher et al., 2010). Air blown through minimal medium was used for control. Landing on the Pasteur pipette, the rim of the jar or inside the jar was scored during a test period of 15 min (Becher et al., 2010). Wind tunnels allow discriminative and sensitive testing as flies fly upwind only in response to a behaviorally relevant stimulus. Flies discriminate odor quality as well as quantity, and flies respond in accordance with their internal physiological state; testing flies individually or in batches both allows discriminative testing (Becher et al., 2010;Lebreton et al., 2015). Landing behavior, scored in this study, is the most stringent criterion of measuring odor-mediated upwind flight attraction.

| Y-tube assay
A bioassay using a Y-tube was conducted to test the odor of the most ancient yeast in our study, Y. lipolytica, for attraction of the collembolan F. candida (Terra-Jungle, Germany) as a representative of basal noninsect hexapods. Springtails were 2-3 weeks old and starved on humidified plaster of Paris for 24 hr before the assay.
All tested individuals (parthenogenetic females) were of similar size (ca. 2 mm long). The Y-tube system was based on the olfactometer described by Bengtsson, Hedlund, and Rundgren (1991)

| Blast search
After behavioral and chemical studies of yeast volatiles, we were interested if the ability to produce certain compounds was re-

| Data sources
Data on the occurrence of specific floral volatiles were obtained from the Pherobase (El-Sayed, 2016). Noteworthy, floral emissions of volatiles generally contain a number of metabolites produced by microbes associated with flowers but studies discriminating between plantand microbe-derived metabolites are rare (Lenaerts et al., 2017). The Pherobase was also consulted for behavioral activity of floral odors.
The Saccharomyces Genome Database (SGD) using WU-BLAST2 was consulted for the reciprocal BLAST search.

| Identification of yeasts isolated from Drosophila flies
Yeasts isolated from D. melanogater flies were identified as C. californica (six isolates) and P. kluyveri (four isolates). For each species, we submitted the nucleotide sequences of one isolate to GenBank for assignment of accession numbers (Table 1).

| All yeasts tested attract Drosophila flies
Evolutionarily distant and ecologically distinct yeast species were cultivated to test for their ability to attract flies in a wind tunnel assay.

| Yeasts attracting Drosophila share volatile compounds
Yeasts grown under controlled conditions were used for behavioral and chemical analyses. During short sampling intervals, corresponding to the duration of behavioral tests, we found nine compounds that repeatedly occurred in at least five of the nine yeasts. Two compounds, 2-phenyl-ethanol and 3-methyl-1-butanol, were released by all nine yeasts. Acetoin, ethanol, and ethyl acetate were detected in all but one yeast, Y. lipolytica (Figure 2).
In S. cerevisiae, alcohol acetyl transferase catalyses the synthesis of acetate esters (Verstrepen et al., 2003). In agreement with the absence of ethyl acetate in Y. lipolytica headspace, ATF1 and ATF2 were not found in the published genome of Y. lipolytica (CLIB122) (Appendix S1). Moreover, ATF1 and ATF2 could not be found in the published genomes of D. bruxellensis (AWRI 1499) and C. albicans (WO-1).
Butanediol dehydrogenase catalyzes the synthesis of acetoin in S. cerevisiae (González et al., 2010). Corresponding to the lack of acetoin in Y. lipolytica headspace, BDH1 and BDH2 were absent in Y. lipolytica (CLIB122).
Fusel compounds are fermentation products derived from amino acid catabolism via the Ehrlich pathway (Hazelwood et al., 2008).
All yeasts were producing the fusel alcohols 3-methyl-1-butanol, as a typical catabolite of the branched amino acid leucine, and 2-phenyl-ethanol, derived from the aromatic tryptophane. Each of the yeast genomes contained at least one of the examined orthologs encoding aldehyde dehydrogenases (ALD2-6), aromatic aminotransferases (ARO8-9), and branched-chain aminotransferases (BAT1-2), being enzymes of the Ehrlich pathway (Appendix S1). Overall, for the 13 investigated genes, S. cerevisaie had five orthologs in common with the earliest diverging yeasts Y. lipolytica, and 10 with the more closely related yeast K. lactis (Appendix S1).

| Drosophila attraction is independent of yeast growth stage
Crabtree-positive yeasts like S. cerevisiae aerobically ferment sugar to ethanol and, after depletion of sugar, switch from respiro-fermentative growth to strict respiration and degradation of earlier produced ethanol (Piškur et al., 2006). We tested whether D. melanogaster responds similarly to odors emitted during respiro-fermentative and respiratory metabolism. The sudden drop of CO 2 shows the diauxic shift (23.4 hr after inoculation, Figure 3a). This shift is accompanied by a distinct

| Yeast attracts springtails
Volatiles produced by the phylogenetically most ancient yeast, Y. lipolytica, were tested for attraction of collembolans, early terrestrial noninsect hexapods. A significant number (70%; C.I. = 0.51-0.84) of F I G U R E 2 Upwind flight attraction of Drosophila melanogaster flies followed by landing at the odor source in response to headspace volatiles of nine yeast species. Yeasts were grown as controlled aerobic batch culture on synthetic minimal medium. All yeasts induced significant attraction behavior (ANOVA, F = 25.47, df = 49, p < .0001; different lower case letters indicate significant difference). Predominant volatiles repeatedly identified in yeast headspaces are shown (+)
Attraction of Drosophila to yeasts not associated with fruit or insects was expected to be lower than attraction toward yeasts ecologically linked to host fruit or insects (Palanca et al., 2013). However, our study suggests that production of volatiles that attract insects is a conserved trait, which embraces yeasts of various habitats, including vertebrate pathogens. Furthermore, flies were attracted to yeasts differing in their physiological characteristics of sugar metabolism and, moreover, to S. cerevisiae at different growth phases, suggesting that attraction is a common trait and not limited to a specific type of yeast metabolism.

| Yeast volatiles promote communication with insects
Odorants facilitate recognition of yeasts, fruit, and flowers even from distance (Buser et al., 2014;Palanca et al., 2013;Raguso, 2004;Saveer et al., 2012). Fermenting fruit such as apples or grapes generally have a richer and more intensive odor profile than fresh fruit due to the yeast-derived volatile fraction (Swiegers et al., 2005). For D. melanogaster, yeast-derived volatiles are behaviorally more important than fruit compounds .
Coadaptation between yeast volatile emission and insect olfaction most likely underlies ecological relations between yeasts and Drosophila (Scheidler et al., 2015). Core metabolic processes in S. cerevisiae mediate the production of volatile signals attracting D. melanogaster (Schiabor, Quan, & Eisen, 2014).
Numerous studies report that insects vector yeast internally and externally of their body, for example, D. melanogaster Chandler, Eisen, & Kopp, 2012;Christiaens et al., 2014;Gilbert, 1980;Stamps, Yang, Morales, & Boundy-Mills, 2012;Starmer & Fogleman, 1986). More attractive strains of S. cerevisiae are more likely dispersed by  D. simulans or other insects (Buser et al., 2014). Interestingly, the yeasts we isolated from D. melanogaster flies trapped in an Italian winery split into P. kluyveri and C. californica, representing two of the three species that consistently formed yeast communities with D. melanogaster larvae on banana (Stamps et al., 2012). Together with another described isolation of C. californica from D. melanogaster (Stötefeld, Holighaus, Schütz, & Rohlfs, 2015), these findings support the existence of species-specific adaptations between D. melanogaster and yeasts.
Yeast volatiles mediate attraction of vectors and seem to be less important for functions of cell viability. The ATF1 gene in S. cerevisiae encodes an alcohol acetyl transferase responsible for ester formation and was shown to promote attraction of D. melanogaster (Christiaens et al., 2014). However, ATF1 is not essential for cell survival and plays no known metabolic function apart from the formation of acetyl esters.
A BLAST search revealed the absence of ATF1 and ATF2 in the published genome of Y. lipolytica (CLIB122), which corresponds to a lack of detectable amounts of ethyl acetate and lower fly attraction in the wind tunnel, in comparison with other yeasts tested, except D. bruxellensis ( Figure 2; Appendix S1). Indeed, ATF1 and ATF2 also are absent in D. bruxellensis (AWRI 1499), although ethyl acetate was detected in our headspace analysis of D. bruxellensis (CBS 2499). This compares to reduced, but detectable production of ethyl acetate in S. cerevisiae ATF deletion strains (Christiaens et al., 2014;Verstrepen et al., 2003).
The presence of additional orthologs of genes related to the production of acetoin or fusel compounds in all yeasts included in our BLAST search, together with our behavioral and chemical analysis, supports the view that production of volatile signals is conserved and that other fermentation products in addition to acetate esters contribute to insect attraction. In addition to evolutionary conservatism, also convergent evolution might contribute to similarity in chemical signals (Bohlmann, Meyer-Gauent, & Croteau, 1998;Courtois et al., 2016).
Several compounds found in yeast headspace are by-products of cell metabolic processes like carbohydrate and protein metabolism (Albertazzi, Cardillo, Servi, & Zucchi, 1994;Hazelwood et al., 2008;Lilly et al., 2006;Piškur et al., 2006) and chemical signaling by volatiles might have developed as a secondary function of emitted metabolic products. Additional functions like inhibition of competitive microorganisms by volatiles are likely and could affect compound release and fitness (Hua, Beck, Sarreal, & Gee, 2014;Piškur et al., 2006).

| Yeast coexistence wth insects predates coevolution between insects and flowers
Flies were strongly attracted to yeasts, disregarding their taxonomic position. Five of the examined species contained 2-phenyl-ethanol, acetic acid, acetoin, and 3-methyl-1-butanol, previously shown to induce strong upwind flight attraction in Drosophila .
2-Phenyl-ethanol, present in all yeasts, was the main volatile in headspace of Y. lipolytica, which in our study was the most ancient yeast.
An early origin of yeast chemical signaling and the ability to attract potential vectors was further confirmed by attraction of the basal hexapod F. candida to Y. lipolytica (Regier, Shultz, & Kambic, 2004).
Insect olfactory neurons express three types of chemoreceptors, ionotropic receptors (IRs), gustatory receptors (GRs), and odorant receptors (ORs). Recent work suggests that ORs, being younger than GRs and IRs, evolved in pterygot insects and increased the detection spectrum of compounds but also sensitivity and speed of detection, which is important for odor-sensing during flight (Croset et al., 2010;Getahun, Wicher, Hansson, & Olsson, 2012;Missbach et al., 2014). The evolution of ORs would thus coincide with the evolution of early yeasts, and several yeast volatiles are indeed known as OR ligands (Münch & Galizia, 2016).
Yeast hyperdiversity in insect guts (Boekhout, 2005;Suh, McHugh, Pollock, & Blackwell, 2005) is one aspect reflecting the ecological significance of the diversification of insects for the evolution of yeasts.
Triassic amber samples, 230 my old, show the presence of flies, bacteria, and microfungi . Finally, with eudicot angiosperms being widely distributed in early Cretaceous by about 125 mya ago, the production of fruit provided unparalleled access to sugar (Sun et al., 2011), and the development of new growth strategies in Saccharomyces yeasts (Piškur et al., 2006), leading to fruit-associated yeast-insect interactions.

| Parallels to the pollination concept
Smell the yeasts. Their odors are pleasant and sweet. Yeasts and flowers share volatile signals (Figure 5a) which are attractive to insects ( Figure 5b). The ecological role of such volatiles is well established for flowers but not for yeasts: the metaphoric title "Wake up and smell the roses" (Raguso, 2008) emphasizes the importance of volatiles that had not sufficiently been acknowledged in pollination biology. Most angiosperms require pollinators for reproduction (Schoonhoven, van Loon, & Dicke, 2005) and floral volatiles mediate pollinator attraction ( Figure 5b). There is a clear functional analogy between yeast spores and flower pollen, and insects mediate dispersal as well as outbreeding in both. In return, insects benefit from their visit through a food reward (Knauer & Schiestl, 2015;Yamada, Deshpande, Bruce, Mak, & Ja, 2015).
Pollinator attraction by fungi for the purpose of spore dispersal has been described for ascomycete and basidiomycete fungi (Kaiser, 2006;Roy, 1993;Roy & Raguso, 1997;Schiestl et al., 2006). Furthermore, it was proposed that convergent development has led to the evolution of chemical insect attractants in fungi and plants (Schiestl et al., 2006) and Zygogynum bicolor (Winteraceae) emit a "musty" smell and are pollinated by ancient micropterigid moths of the genus Sabatinca; the association between Zygogynum and Sabatinca is suggested to exist since the early evolution of flowering plants (Pellmyr & Thien, 1986;Thien et al., 1985). Flowers of Z. bicolor and two other species of winteraceae emit ethyl acetate as main volatile and 2-methyl-1-butanol and acetic acid to a minor content (Thien et al., 1985).  (Jürgens et al., 2013;Pellmyr & Thien, 1986;Schiestl et al., 2006).
In summary, we studied a phylogenetically broad range of hemiascomycetous yeasts framed by the alkane-utilizing Y. lipolytica and the sugar degrading Crabtree-positive S. cerevisiae. These species and their volatile signals most likely were present when D. melanogaster and other close related species within the melanogaster subgroup appeared less than 50 mya (Ometto et al., 2013;Wiegmann, Yeates, Thorne, & Kishino, 2003;Wiegmann et al., 2011). As D. melanogaster is attracted to all yeasts, including vertebrate pathogens and other species that do not share habitats with the fly, we conclude that signaling and insect attraction is an ancient trait in yeasts, conserved over millions of years of arthropod and insect coexistence with yeasts, and is vestigial in yeasts that are not primarily associated with insects. Furthermore, coexistence of yeast and insects prior to evolution of angiosperms, overlap of signals attracting insects to yeasts and flowers, as well as functional similarities between insect-yeast interactions and insect pollination suggest to consider yeasts in the evolution of insect-mediated pollination of flowering plants.

ACKNOWLEDGMENTS
This study was supported by the Swedish Research Council Formas, the Swedish Research Council VR, the Sörensen Foundation, the Linnaeus initiative "Insect Chemical Ecology, Ethology and Evolution" IC-E3 (Formas, SLU) and the INTERREG IVA community initiative "Pomerania" funded by the European Commission (ERDF). We thank Tomas Linder for his comments.

CONFLICT OF INTEREST
None declared.

AUTHOR CONTRIBUTIONS
PGB, GF, JP, and PW developed the concept and wrote the manuscript; PGB and AH ran the bioreactors; AC and ER identified the yeasts; PGB and VV performed the assays; PGB, AH, and MB performed chemical analyses; PGB and SL did the database review; AH did the BLAST search.

DATA ACCESSIBILITY
Data used in this manuscript are present in the manuscript and its supporting information.