Microbiota-mediated competition between Drosophila species

Background The influence of microbiota in ecological interactions, and in particular competition, is poorly known. We studied competition between two insect species, the invasive pest Drosophila suzukii and the model Drosophila melanogaster, whose larval ecological niches overlap in ripe, but not rotten, fruit. Results We discovered D. suzukii females prevent costly interspecific larval competition by avoiding oviposition on substrates previously visited by D. melanogaster. More precisely, D. melanogaster association with gut bacteria of the genus Lactobacillus triggered D. suzukii avoidance. However, D. suzukii avoidance behavior is condition-dependent, and D. suzukii females that themselves carry D. melanogaster bacteria stop avoiding sites visited by D. melanogaster. The adaptive significance of avoiding cues from the competitor’s microbiota was revealed by experimentally reproducing in-fruit larval competition: reduced survival of D. suzukii larvae only occurred if the competitor had its normal microbiota. Conclusions This study establishes microbiotas as potent mediators of interspecific competition and reveals a central role for context-dependent behaviors under bacterial influence. Video Abstract Supplementary Information The online version contains supplementary material available at 10.1186/s40168-023-01617-8.


Background
The influence of microbiotas on ecology and evolution is both undebated and poorly understood.It is widely established that most animals and plants harbor complex and usually flexible microbial communities [1].At the level of the individual host, numerous biological functions appear to be affected by microorganisms present in the gut, on the skin, the genital organs, or its flesh.It is even argued that these microorganisms participate in host adaptation, like any other heritable source of phenotypic diversity-such as nuclear genes-would do [2][3][4].When it comes to the ecological consequences of hostmicrobiota associations, the literature is however scarce, in particular regarding effects on interspecific interactions [5][6][7].Although the effects of microbiota on parasite epidemics can be inferred from the well-documented modulation of immunity on individual-level infection dynamics [8], this is not the case for interactions such as competition, facilitation, or predation.Some data do show, nonetheless, that competition and facilitation between invertebrates can be under the influence of microbiotas.For example, the within-host competition between parasitic nematodes is mediated by bacterial symbionts that belong to their microbiota [9,10].Facilitation mediated by the microbiota is exemplified by interaction between Drosophila flies.Microbial growth in berries infested by Drosophila suzukii larvae favors fruit exploitation by the close species Drosophila melanogaster through female behavioral response [11].Given the influence of microbiotas on behavior and brain-function [12], behavior may be an important factor in microbiota-mediated competition and is the focus of the present study.
Over the last 10 years, the Asian fly D. suzukii (Ds) has spread in Europe and the Americas [13] causing major fruit-production losses [14][15][16].As a consequence, considerable research effort has been devoted to the development of strategies to control this species and protect crops.It was soon observed that co-culturing of Ds with D. melanogaster (Dmel) led to the rapid competitive exclusion of Ds [17].This phenomenon can be partly explained by the observation that Ds females avoid laying eggs in resource sites that already contain Dmel eggs [18].The prevention of larval crowding does however not explain this behavior as Ds females did not avoid oviposition on sites with conspecific Ds eggs in the first study on competition [18] and in the conditions of our experiments (Fig. S1).The literature on Ds however reports both oviposition preference and avoidance of sites with cues from conspecifics, possibly because of context dependency of this response [19,20].We hypothesized that Dmel eggs might carry specific cues that deter Ds females from depositing eggs.We investigated the mechanisms and variability of Dmel repellency on Ds oviposition.We determined that the oviposition deterrence is mediated by Dmel symbiotic bacteria and that the repellency is plastic and conditional on the Ds carrying a microbiota distinct from that of Dmel.We infer that the inhibition of Ds oviposition is a microbiota-mediated adaptive response to reduce larval competition between the two species.

Variable response of D. suzukii females to D. melanogaster cues
In an initial experiment, we offered groups of Ds females the choice to oviposit either on substrates previously exposed for 24 h to Dmel females or on control substrates (Fig. 1).We followed Ds egg-laying preferences over 4 days with the oviposition substrates replaced daily.During the first 2 days, Ds females laid more than 75% of their eggs on sites that had not been exposed to Dmel (p = 0.008 and p = 0.006 on days 1 and 2, respectively; Fig. 2a) (Table S2 aggregates experimental details and analyses of the data presented in Fig. 2).However, Ds females did not avoid substrate contaminated by Dmel during the final 2 days of the assay (p > 0.1 on days 3 and 5, Fig. 2a).Avoidance of conspecific cues as reported by [20] was temporary too, but disappeared much faster, after 4 h in choice conditions.The presented experiment showed that Ds have a strong preference to oviposit on sites that have not been visited by Dmel, but that the  S2 of the Materials and methods.Note a male D. suzukii is depicted in the plastic cage as only males harbor the spotted wings that give the species its vernacular name, the Spotted-Wing Drosophila avoidance behavior is plastic and can wear off with time or female experience.
In order to determine how universal the Ds avoidance behavior is, we tested Ds females from different inbred laboratory populations founded with insects captured in France (our reference population used throughout this study), the USA, China, and Japan (see the "Materials and methods" section).Because in these behavioral investigations individual females were the essential unit of replication, and to consider potential inter-individual differences, this and all following experiments were carried out with single females, rather than groups of flies, and over 24 h.Females from all populations except the Japanese exhibited significant avoidance of oviposition on substrates that were visited by Dmel (China p = 0.019, France p < 0.0001, Japan p = 0.53, USA p = 0.014; Fig. 2b).Ds originates from mainland China, invaded Japan at the beginning of the twentieth century, and invaded Europe and North-America in the past 10-15 years [21].These results show that avoidance behavior is restricted neither to invasive populations nor to those from the area of origin.
Our initial experiments demonstrated that Ds females actively avoid oviposition on substrates that had been previously visited by Dmel females (Fig. 2a and b).But these experiments do not distinguish whether the aversion is due to the presence of Dmel flies or Dmel eggs.To test this, we repeated the repellency assay using substrate conditioned by Dmel males.The experiment showed that both Dmel males and females induced oviposition avoidance (p = 0.0007 and p = 0.0013, respectively; Fig. 2c).This rules out Dmel eggs or oviposition-associated cues as driving Ds oviposition avoidance and contrasts with Tephritid fruit flies that use host-marking pheromones to limit oviposition and avoid larval crowding [22].
Because our initial experiments were performed using a laboratory population of Dmel, we wanted to determine whether Ds oviposition avoidance could also be triggered by wild Dmel and by the Dmel sister species, D. simulans (Dsim), whose ecology is very close to that of Dmel [23].We tested the repellency of wild Dmel trap-captured in Southern France, lab-reared F1 offspring of the same wild Southern France Dmel population, the Oregon-R lab strain of Dmel used for all previous experiments.Similar to the experiments performed with laboratory Dmel, substrate conditioned by the wild Dmel flies was repellent to Ds females (p = 0.0032, Fig. 2d).Surprisingly, however, the F1 offspring of the wild-caught Dmel, which had spent one generation in the laboratory, did not induce oviposition avoidance (p = 0.4, Fig. 2d).The Dsim population we tested was also not repellent (p = 0.4).Similarly, exposure of fruit to Ds did not elicit Ds oviposition avoidance (Fig. S1).Repellency is therefore a feature of wild and laboratory Dmel populations that may nonetheless be sensitive to rearing conditions.
Finally, we tested whether wild Ds also avoid substrates that have been visited by Dmel.We trapped wild Ds adults from the Montpellier region, Southern France, using classical vinegar traps modified to prevent the drowning of captured flies (see the "Materials and methods" section).These traps attracted various species of Drosophilid flies, including both Ds and Dmel.To our surprise, wild Ds females did not exhibit avoidance behavior to Dmel-exposed substrates (p = 0. 28 and p = 0.98 for the two groups tested, Fig. 2e).We can envision three alternative explanations for this variation among groups of Ds: (1) avoidance behavior is a laboratory artifact; (2) uncontrolled fly age or pre-capture history affects female selectivity; or (3) exposure to other Drosophilid flies, including Dmel, during the time spent in traps eliminates the avoidance behavior, similar to the third and fourth days our first experiment (Fig. 2a).
Our results show that Ds oviposition avoidance of sites with Dmel cues varies among populations (i.e., the Japanese population did not avoid) and with individual experience or physiological condition.This contrasts with the sustained and hard-wired oviposition avoidance that Dmel females display in response to geosmin [24], a molecule produced by microorganisms responsible for late-stage fruit rot that is detrimental to Dmel larvae.Unveiling the nature of the Dmel cues perceived by Ds females may shed light on how Ds females lose their aversive response.

Bacterial symbionts of D. melanogaster are involved in repellency and D. suzukii avoidance loss
Our observation that exposure to males or females of Dmel was sufficient to reduce Ds oviposition (Fig. 2c) but that Ds avoidance behavior was lost after 2 days of exposure to Dmel (Fig. 2a) led us to hypothesize that the repellent agent was something shed by all adult Dm.To test whether the hypothesized agent was volatile or stationary, we conducted an additional experiment testing whether repellency was restricted to substrates directly contacted by Dmel or whether the adjacent substrate also became repellent to Ds.We did not observe Ds avoidance to substrates neighboring Dmel-exposed medium (Fig. S3), so we concluded that repellent agent could not diffuse through the air.A logical alternative was that Dmel might condition the substrate with bacteria they shed, and that the bacteria were aversive to Ds. Drosophilids possess the sensory and neuronal circuitry to perceive specific bacteria and compounds produced by them, and the presence of microbiota on substrate has previously been shown to affect behaviors in Dmel such as adult foraging preferences [25,26].Furthermore, the effect of substrate microbes on the behavior of Dmel depends on their endogenous microbiota [13,14].We thus hypothesized that microbial symbionts of Dmel excreted on the substrate could perhaps be perceived by Ds females and that oviposition avoidance, or its lack, could be a function of the symbiont community carried by Ds.
As a first test of this hypothesis, we experimentally removed the microbiota from Dmel and tested whether these axenic flies remained repellent to Ds.Because we suspected the Ds microbiota might also influence oviposition avoidance, we performed this test with both axenic and conventionally reared Ds females (Table S4 aggregates experimental details and analyses of the data presented in Fig. 3).Axenic Dmel flies did not elicit oviposition avoidance in Ds (p = 0.11 and p = 0.39 for conventional and axenic Ds, respectively), and both axenic and conventional Ds were significantly repelled by conventionally reared Dmel (p = 0.0001 and p = 0.002 for conventional and axenic Ds, respectively, Fig. 3a).Thus, we conclude that some component of the Dmel microbiota is directly or indirectly required to repel Ds but that Ds does not require microbiota to perceive repellent cues.
We investigated the capacity of symbiotic bacteria to generate repellency in Dmel by inoculating axenic flies with candidate bacteria (i.e., creating gnotobiotic flies).The bacterial microbiota of Dmel has been extensively described over the last 10 years, showing it largely varies among populations and environmental conditions but almost always includes species of the genera Lactobacillus and Acetobacter [27][28][29][30].We therefore chose to associate Dmel flies with a strain of Lactobacillus brevis, or with one of Acetobacter pomorum, both of which had been isolated from a laboratory population of Dmel and are frequently used for microbiota studies [31,32].In order to investigate whether any generic bacterium could restore repellency in axenic Dmel, we also associated Dmel flies with a strain of Escherichia coli previously shown as non-pathogenic to flies [33].Dmel inoculation with L. brevis made Dmel repellent to Ds (p = 0.0043) while association with A. pomorum produced a marginally non-significant effect (p = 0.0745) (Fig. 3b).Inoculation with E. coli did not elicit repellency (p = 0.11).This experiment proves repellency can be restored in axenic Dmel adults following the association with bacteria that belong to fly microbiota.We identified L. brevis as a bacterium able to induce Dmel repellency.This bacterium has beneficial effects on Dmel larval development, but that depends on the identity of the other microorganisms that constitute its microbiota [34,35].This bacterium is also beneficial to the larval development of Ds when nutrients are scarce [36].Interestingly, L. brevis was shown to repel Dmel oviposition, a phenomenon probably based on females' avoidance of lactic acid, a metabolite produced by Lactobacillus bacteria in anaerobia [37].Our data show all bacteria do not have the same effect.Because we only tested 3 bacterial strains, it is not possible to predict which species or strain will or will not elicit repellency in general.It is probable other microorganisms, beyond L. brevis, can recapitulate repellency in axenic Dmel adults.Indeed it is very unlikely we selected by chance the sole bacterium with this property.In addition, L. brevis is a common but not universally reported species in Drosophila microbiota [30,38].Studies of bacterial microbiota by 16 s rRNA sequencing show extensive variations among populations, and even among individuals exposed to identical inoculas [39,40].Likewise, extensive variations exist among strains of the same species, even in the context of microbiota [41].The variations in repellency intensity that occurred among our replicate experiments (Figs. 2 and 3) may be originated from temporal and among-vial variations in microbial community composition.This would explain why the Dmel F1 population we tested did not elicit repellency when the previous F0 generation did (Fig. 2d).We could, and should, have followed bacterial communities by 16 s rRNA sequencing in the present study, but unfortunately did not.Our candidate approach nonetheless showed some bacteria can underpin Dmel repellency.Determining the full range of microorganisms able to render Dmel repellent, and the natural variability of this phenomenon, will require comprehensive assays with wild strains of microbes and insects in the field and field-like conditions.
In our initial experiments, we observed that Ds females lose avoidance behavior after 2 days of exposure to Dmel cues (Fig. 2a).How to explain this change?We hypothesized that the decrease in oviposition avoidance was due to the colonization of Ds by the microorganisms deposited by Dmel on oviposition sites.In order to test this hypothesis, we mono-associated adult Ds for 5 days with the strain of L. brevis that elicited strong repellency by Dmel (Fig. 3b).As expected, Ds females associated with L. brevis did not avoid oviposition on substrate that had been exposed to Dmel adults bearing the same bacterium (p = 0.11; Fig. 3c).Ds females hence avoided sites with cues indicative of the presence of Dmel unless they carried similar bacteria.This result could also explain why trap-captured wild Ds females did not avoid Dmel cues (Fig. 2e).In the traps, wild Ds were in close contact with other Drosophilids from which they may have acquired microbiota.An influence of Ds microbiota composition on avoidance behavior could also explain why the Japanese population did not respond to Dmel cues, they may have harbored a different bacterial community.The oviposition behavior of Ds females in response to cues of conspecifics, Dmel, or microbial agents has now been studied several times independently [18][19][20]42].Some phenomena, such as Ds oviposition avoidance of Dmel, did repeat even though they exhibited substantial variations among experiments, while other behaviors even contradicted among studies.Our discovery that Ds symbiotic status affects its oviposition behavior sheds light onto these discrepancies and establishes clearly that oviposition decision is a complex phenomenon that not only depends on external cues but also on internal ones.
To investigate the possibility of transferring our results to application in pest management, we investigated whether bacteria deposited by Dmel were sufficient to repel Ds oviposition even in absence of Dmel individuals, or if Ds flies perceive cues produced by the interaction between Dmel and its symbionts.A recent study indeed shows Ds females respond to bacterial contamination and avoid oviposition in sites inoculated with bacteria-rich wash water from Dmel-exposed media [42].To investigate the effect of L. brevis inoculation, we carried out two experiments.In the first, we tested the repellency of medium inoculated with 1,000,000 L. brevis bacterial cells.In the second, we inoculated the medium with only 5000 cells, which corresponds to the approximate number of live bacteria retrieved from substrates exposed to Dmel under our experimental conditions (personal observation).Ds females did avoid oviposition on media inoculated with the larger number of bacterial cells, though only weakly with less than 60% of avoiding individuals (p = 0.011; Fig. 3d left).With lower, more realistic, L. brevis numbers females did not avoid inoculated sites (p = 0.58; Fig. 3d right).Together, these results suggest that when Dmel adults are associated with bacteria, the interaction produces compounds that are shed and perceived by Ds females, but that neither the Dmel fly nor her associated bacteria are sufficient for full repellency on their own.A recent study reported that bacteria deposited during oviposition by the oriental fruit fly, Bactrocera dorsalis, induce the host fruit to produce a molecule, b-caryophyllene, that is perceived by female flies and repels them from ovipositing [43].In the case of Ds ovipositional avoidance, prospects for crop protection will necessitate identifying the compounds produced by the interaction between Dmel and its bacteria and testing them as pure molecules.

D. suzukii larvae suffer from competition with symbiont-associated D. melanogaster larvae
Avoidance behavior by Ds females could be an adaptation that ensures offspring do not develop in poor-quality sites.In order to test whether Ds larvae suffer from competition with Dmel larvae, we reproduced in-fruit competition between the two species.Surface-sterilized grape berries were pierced with a fine-needle and a single Ds egg was deposited in each hole (6 holes per berry), mimicking Ds oviposition (Fig. S4a).Berries further received 0, 1, or 5 Dmel eggs per hole.Our goal was to compare the effects of microbiota-free or conventional Dmel larvae on Ds larval development.In half of the replicates (i.e., berries), we therefore used axenic Dmel eggs instead of conventional ones.In berries without Dmel eggs, Dmel microbiota was inoculated by exposing pierced berries to conventional Dmel males prior to Ds egg deposition.Ds developmental success (i.e., proportion of eggs that reached adulthood) was impaired by competition with Dmel larvae that were associated with their microbiota, but not with axenic Dmel larvae (Fig. 4).In the wild, Dmel eggs are never axenic, so the normal outcome of larval competition should therefore be poor Ds development.These results support our hypothesis that Ds oviposition behavior prevents costly larval competition with Dmel.They are in line with the results of Bing [36] who observed that the effects of Dmel bacteria, such as L. brevis, on Ds larval development depend on the environmental context, in their case nutrient availability.Similarly, here the presence of Dmel bacteria had no visible effect in the absence of Dmel larvae but reduced Ds larval survival in their presence (Fig. 4, left).Our data show unambiguously that the combination of Dmel larvae and their microbiota is detrimental to Ds development.Whether Ds larvae suffered directly from bacterial presence, from direct interactions with microbiota-associated Dmel larvae, or from metabolic byproducts of the Dmel-microbiota association is unknown.Each of these mechanisms is plausible.Microbiota effects on Drosophila larvae like antagonistic interactions among Drosophila larvae are both reported as being context-dependent (e.g., [32,36,[44][45][46]).
The effect of Dmel microbiota on larval competition with Ds suggests an adaptive explanation to why Ds females did not respond to cues produced by axenic Dmel.Ds females should only avoid oviposition in environmental contexts that are detrimental to their offspring, which was the case when Dmel larvae associated with their microbiota.Furthermore, the lack of oviposition avoidance exhibited by Ds females that bore the repellency-inducing bacterium L. brevis (Fig. 3c) may be another adaptation.Drosophila females transmit bacteria and yeasts from their microbiota to their offspring [47].When Ds females associate with a detrimental bacterial species there would be no point in avoiding sites contaminated with the same microorganism.The plastic decision by Ds to oviposit, or not, as a function of microbiological presence may enable the use of all suitable oviposition sites, with the avoidance of sites only necessary when they are contaminated with costly competitors.Alternatively, the lack of avoidance may be driven by the microorganisms themselves in order to promote their dispersal and transmission [12,48].In Dmel, it is established that adults associated with specific bacteria, including Lactobacillus species, are attracted to feeding sites inoculated with the same bacteria [25].

Ecological significance and prospects for crop protection
Our study shows that commensal microbiota can mediate the competition between insect species with overlapping ecological niches.In our particular example, Ds females rely on combined cues from the competitor Dmel and its symbiont L. brevis to avoid oviposition sites that are likely to incur competition costs.Microorganisms can impact the outcome of competitive interactions between hosts [49].Often, parasitic microorganisms shed by tolerant species have detrimental effects on less-tolerant competitors (e.g., [50]); the spill-over hypothesis that facilitates the spread of some invasive species is based on this very mechanism [49].Symbiotic microorganisms can also elicit beneficial effects for heterospecific neighbors.For example, mycorrhizal fungi can mediate mutualism between plants species [51].In the present case, a frequent bacterium of Dmel microbiota was detrimental to Ds larvae, eliciting oviposition avoidance by Ds females.A remarkable feature of our study is the implication of behavior (i.e., oviposition avoidance) in the mediation of interspecific competition.Our study hence connects ecological dynamics with the wide literature on the effects of microbiotas on behavior and brain function [12].The interplay between microbiota, behavior, and competition may also be related to the recent realization that fear of predation, a form of behavioral avoidance, can have a greater effect on predator-prey dynamics, another important type of ecological interaction, than mere prey consumption [52].We conclude that the microbiota can drive competitive interactions between species through direct and indirect effects, in the present case through decreased larval survival and behavioral adaptation to avoid these situations.
Few species in the Drosophila genus oviposit in undamaged, ripening fruit.A phylogenetic perspective indicates that the ability to exploit ripening fruits is a derived character that evolved in Ds ancestors and presumably alleviates competition with other Drosophilids [53].Dmel arrived in Asia less than 60,000 years ago, long after the species origin of Ds [54].The larval niche of Ds, and possibly female oviposition preferences, hence probably evolved in response to other species of competitors.Several studies have reported that Ds larvae share their fruit with species such as Dmel, Drosophila subobscura, and Zaprionus indianus in a variety of crop and wild plant species [11,55,56].In our experiments, Ds did not avoid D. simulans cues.It is nonetheless plausible Ds females avoid cues produced by other Drosophilid species or populations, in particular those from the region it originates and possibly including other strains of D. simulans, and this avoidance may depend on the symbiotic status of those flies.
Ds is responsible for heavy crop losses throughout the globe due to the development of larvae in farmed fruit.It is tempting to exploit Ds oviposition avoidance to shelter fruit from Ds damage.Field tests of repellents based on 1-octen-3-ol, a molecule produced by fungi that compete with Drosophila larvae, gave encouraging results [57,58].In the present case, the microbiota associated with Dmel clearly cannot be sprayed directly in orchards because of the plastic avoidance loss exhibited by Ds females if they acquire those symbionts (Figs.2a and 3c).A better solution may be to identify and use as a repellent the compounds produced by bacteria-inoculated Dmel (Fig. 3d).Future experiments should test whether Ds can become habituated to the aversive compound [59,60] and whether management strategies such as refugia or alternating application need to be deployed.Characterizing D. suzukii's chemosensory receptors and circuitry involved in the recognition of Dmel cues and its consequential behavioral response may enable the design of an optimized repellent.

General experimental design
The study is based on a simple assay where female D. suzukii (Ds) are given the choice to lay eggs on two substrates: either a blank control or a substrate that had previously been exposed to D. melanogaster (Dmel) adults (Fig. 1).By changing the nature of the Ds and Dmel flies employed, we were able to reveal the factors that govern Dmel's repellency and Ds's corresponding avoidance.
In most cases, a single Ds female was placed in a 9-cm diameter plastic cylindrical box for 24 h.Boxes contained two 2 cm × 2 cm × 2 cm plastic receptacles each half-filled with oviposition substrates, generally an agar-jellified strawberry puree or a piece of strawberry inserted in blank agar.These two substrates were prepared the day before, one of them was exposed to 3 adult Dmel flies overnight.Because these experiments were conducted over 5 years with variable objectives, some experimental details varied among assays.In all experiments, a variable fraction of assayed females (usually around 50%) did not oviposit during the 24 h they spent with the tested substrates.These females were excluded from further analyses.Table S2 describes the experimental details, sample sizes, and statistical analyses of each of the results reported in the article.
All flies were reared, and experiments conducted, in climatic chamber with a 13 h:23 °C/11 h:19 °C day/night cycle, an artificial dawn and dusk of 45 min.Humidity was maintained constant at 75% relative humidity.

Biological material
Most experiments were carried out with our standard Ds population that was founded by the authors in 2013 from a few dozen individuals that emerged from blackberries harvested in Gaujac, Southern France (44.0794, 4.5781), and the classical Dmel population Oregon R, founded in 1927 and shared among laboratories since then.These fly colonies were maintained in standard drosophila vials with banana artificial medium (see below) or 30-cm cubic cages when we needed larger numbers of flies.
Additional laboratory populations of Ds were as follows.The Japanese population was founded from individuals captured in Matsuyama, Japan (33.8389, 132.7917)Wild Ds were captured during summer 2016 in two localities 10 km apart near Montpellier, Southern France (43.6816, 3.8776), and tested about a week after capture, once they started laying eggs in the laboratory.Wild Dmel were also captured near Montpellier.For the experiment reported in Fig. 2d, Dmel flies were captured in several instances.Flies from a first group were reared in the laboratory and their offspring (i.e., F1) tested along with freshly-captured flies (i.e., F0).All wild flies were captured using custom-designed traps based on c.300-mL plastic cups, covered with cling-film, pierced on the sides for fly entry, and containing an attractant (a mix water, vinegar, wine, and sugar) separated from the flies by netting.The netting prevented fly drowning but allowed occasional access to the attractant as cups were readily shaken by wind or operators, which caused the netting to become soaked with the liquid bait.Traps were checked daily and usually contained various fly species, including Dmel and Ds.

Axenics, mono-associated flies, and microbiological work
Axenic flies were produced following a protocol derived from [61].Briefly, Drosophila eggs were collected on grape-juice medium (see previous recipes section) before being bleached and rinsed twice (1.2% sodium hypochlorite).Eggs were then transferred to 50-mL centrifugation vials with 10-mL autoclaved banana medium (see recipes section) which lids were either incompletely screwed or harbored breathing membranes.All manipulations were conducted under a laminar flow hood.With care, it is possible to transfer freshly emerged adults to new vials aseptically and therefore maintain the population microbe-free for several generations.The axenic nature of the flies was regularly confirmed by the absence of cultivable microbes.
To produce mono-associated (i.e., gnotobiotic) adult flies, axenic flies were added to vials that had been surface-inoculated with suspensions (i.e., c. > 10 5 cells) of the relevant bacterium at least 4 days before experiment onset.The presence of inoculated microbes in adults was verified by culturing the bacteria retrieved from homogenized insects several days after their nutritive medium was inoculated.

Larval competition between D. suzukii and D. melanogaster in fruit
This assay aimed at testing whether the development of Ds larvae was affected by the presence of Dmel larvae and their associated microbiota.We took great care of reproducing field-like conditions (i.e., in-fruit interactions) as competition costs notoriously depend on ecological conditions [62] and the effects of Drosophila bacterial symbionts on larval development change with medium richness [63].A key parameter was to choose a fruit species in which both Ds and Dmel had been reported to develop simultaneously in the field, and we chose grape [11].Given the large effect of grape variety on Ds oviposition, we first confirmed that Ds would oviposit on the batch of grapes we used (fruit of an unknown cultivar bought in April 2018 in a food retail store) and that this behavior was reduced by exposure to Dmel (data not shown).In order to mimic realistic competition conditions, we manually pierced the skin of grape berries with fine needles, making a hole close in size as those Ds females do with their ovipositor [53].Each hole was first inoculated with a wild strain of the yeast Hanseniaspora meyeri isolated from wild Ds adults and received a single Ds egg (Fig. S5a).There were 6 holes per berry.Each berry was allocated to one of three treatments: addition of no Dmel eggs, addition of one Dmel egg per hole, and addition of 5 Dmel eggs per hole.The larger ratio of Dmel to Ds eggs was chosen as it reflects relative infestation intensity observed in grapes collected in the field [11].Half of the berries with Dmel eggs received conventional eggs (i.e., with microbiota); the other half received Dmel eggs that had been made axenic by bleaching (see previous section on the production of axenic flies).Note that an important design choice was to either use conventional Dmel eggs, or axenic Dmel eggs artificially inoculated with microbes harvested from conventional flies.We rejected the second option because it would have been challenging to ensure eggs artificially associated with cocktails of microorganisms bore all the relevant strains.Bleaching eggs impose additional mortality compared to conventional (i.e., non-bleached) eggs; however, this effect was easily controlled for statistically (see the "Statistical analyses" section; Fig. S5b).In the treatments without Dmel eggs, two-thirds of the berries served as Ds-only controls; the other third received Dmel microbiota alone.To this end, pierced berries were exposed to 10 conventional Dmel males for 24 h prior to Ds egg deposition.All grape berries were incubated in individual plastic vials until adult flies emerged.This assay comprised 25-30 individual berries per treatment (50 replicates for the control treatment with Ds eggs and no Dmel microbiota) spread over 8 temporal blocks.

Statistical analyses
In all reported experiments except the one on larval competition (Fig. 4), Ds females deposited their eggs on either treated or untreated oviposition substrates.Egg counts on each type of medium were therefore not independent because these were produced by the same females.Additionally, total number of eggs varied among females and experiments and largely followed a Poisson distribution, which prevented the use of traditional linear models that assume normal distributions of the residuals.We therefore used a simple, robust statistical approach to analyzing the proportion of eggs deposited on treated and untreated sites: a nonparametric, one-tailed Wilcoxon signed rank test that took into account data pairing was compatible with the data distribution and has often been used in comparable studies (e.g., [25]).We noticed that paired t-tests, which assume data follow a normal distribution, provided similar results.The aim of our experiments was to investigate female behavior determinants rather than infestation intensities, so the units of replication were the females and their individual preferences towards different types of substrates.For this reason, the statistical methods we employed were not affected by variation in the fecundity of individual females, and the most fertile females could not skew the results towards their specific preferences.With this in mind, it appeared preferable to include all females that oviposited, even if those that deposited only a single egg.Because of the plasticity of the avoidance behavior, all experiments included a positive control-usually the response of standard Ds to laboratory Dmel flies.This ensured that lack of avoidance in an experiment was not due to unidentified factors or inappropriate conditions.Note that several of our most important results were repeatedly observed in distinct experiments.Compare, for example, loss of avoidance in Figs.2a and  3c, effect of axenic Dmel in Fig. 3a and b, and restoration of Dmel repellency by Lactobacillus brevis inoculation in Fig. 3b and c.
Results from the larval competition assay were analyzed using a linear mixed model with the REML method.Numbers of Ds adult that emerged from each fruit were Log(x + 1)-transformed and complied with tests assumptions.This model contained discreet, fixed terms describing the number of Dmel eggs deposited, whether Dmel microbiota was present, and their interaction.It was also very important that the model included the (logtransformed) number of Dmel adults that emerged from the fruit as a fixed, continuous factor.Indeed, this term was necessary to control for the additional mortality of Dmel larvae caused by bleaching eggs in the axenic treatment (Fig. S3b).The presence of this term in the analysis ensures the significant effect of axeny was not an artifact due to reduced Dmel larvae numbers.The model also included a block term (treated as random).Differences among treatments were tested with independent contrasts and pairwise Student's tests.

Fig. 1
Fig. 1 Schematic drawing of the experimental procedure for testing D. suzukii oviposition avoidance of egg-laying sites previously exposed to D. melanogaster.Details of each experiment, among which origin, sex, and numbers of D. suzukii and D. melanogaster flies, cage size, and oviposition substrate, are described in TableS2of the Materials and methods.Note a male D. suzukii is depicted in the plastic cage as only males harbor the spotted wings that give the species its vernacular name, the Spotted-Wing Drosophila

Fig. 2 Fig. 2 (
Fig. 2 Oviposition avoidance of D. suzukii females for egg-laying site previously exposed to D. melanogaster.Values significantly below 0.5 indicate D. suzukii preference for sites unexposed to D. melanogaster.Repeated tests of the same females (a) showed plastic avoidance loss.D. suzukii populations from different geographical origins (n = 9 D. suzukii cages) and (b) exhibited variable avoidance (n = 14, 57, 27, and 16 D. suzukii individuals, from left to right).(c) D. melanogaster males, like females, induce repellency (n = 16 and 21 D. suzukii individuals to test the effects of D. melanogaster females and males, respectively).(d) Trap-captured, wild D. melanogaster flies (F0 in d) induced repellency; however, this property was not induced by laboratory-reared offspring from wild-caught flies (F1 in d) nor by D. simulans (n = 17, 16, 23, and 19 D. suzukii individuals, from left to right).(e) Trap-captured, wild D. suzukii females did not avoid oviposition on D. melanogaster exposed substrates (n = 19, 12, and 8 D. suzukii individuals, from left to right).Symbols indicate means and error-bars standard errors.Significant deviation from equal number of eggs on sites exposed to D. melanogaster, or control sites, were produced by one-tailed Wilcoxon signed rank tests; * for p < 0.05; ** for p < 0.01; *** for p < 0.001 (See figure on next page.)

Fig. 3
Fig. 3 Investigation of the role of extracellular symbionts on D. melanogaster repellency and D. suzukii oviposition avoidance.(a) Axeny, the removal of extra-cellular microorganisms, had different effects on D. melanogaster and D. suzukii (n = 16, 16, 17, and 27 D. suzukii individuals, from left to right).Oviposition sites exposed to axenic D. melanogaster were not avoided by D. suzukii, showing the importance of symbionts in D. melanogaster for repellency.By contrast, axenic D. suzukii behaved like conventionally reared flies; D. suzukii microorganisms were therefore not required for perceiving the repellent.(b) Tests of candidate bacteria in association with D. melanogaster (n = 14, 13, 11, and 17 D. suzukii individuals, from left to right) revealed the bacterium Lactobacillus brevis can restore repellency in formerly axenic flies (note the axenic and Acetobacter pomorum treatments were marginally non-significant, p = 0.071 and p = 0.075, respectively).We hypothesized D. suzukii avoidance loss was due to their colonization with D. melanogaster symbionts.(c) As expected, D. suzukii females experimentally associated with the bacterium L. brevis did not avoid oviposition on sites exposed to L. brevis-associated D. melanogaster (n = 29 and n = 28 D. suzukii individuals for axenics and gnotobiotics, respectively).(d) Direct inoculation of medium with L. brevis cells in large numbers or at a dose similar to that naturally shed by D. melanogaster (i.e., 1,000,000 vs 5000) produced different results (n = 29, 29, 22, and 11 D. suzukii individuals, from left to right).The low, natural dose of deposited bacteria failed to elicit avoidance, suggesting D. melanogaster repellency is largely due to the production of unidentified molecules when in symbiosis.Symbols indicate means and error bars standard errors.Statistical tests produced by Wilcoxon signed rank tests; * for p < 0.05; ** for p < 0.01; *** for p < 0.001

Fig. 4
Fig. 4 Effect of D. melanogaster larvae and their associated microbiota on the development of D. suzukii eggs until adult emergence.Eggs were individually deposited in grape berries where we mimicked natural oviposition by Drosophila females and field-like conditions.The greater ratio of D. melanogaster to D. suzukii egg follows relative infestation intensities observed in the field [11].The statistical interaction between number of D. melanogaster eggs and the presence or absence of their microbe was significant (F 2, 172 = 6.46; p = 0.002).Independent contrasts indicate a significant difference between the treatments with and without D. melanogaster microbes at high D. melanogaster density (F 1, 174 = 15.6;p = 0.0001).Overall REML model results: Number Dmel eggs per Ds egg; F 2, 165 = 4.83; p = 0.009; Dmel axenic or not; F 1, 162 = 0.41; p = 0.52; Number of Dmel eggs * axenic or not; F 2, 172 = 6.46; p = 0.002; Number of emerging Dmel adults; F 1, 174 = 7.74; p = 0.006.Symbols indicate means; error bars indicate standard errors; *** for p < 0.001 , in 2015 (courtesy A. Fraimout and V. Debat), the US population in Watsonville, CA, USA (36.9144, -121.7577) in 2014 (individuals captured by S. F.), and the Chinese population in Shiping, China (23.7048, 102.5004) in 2015 (courtesy P. Girod and M. Kenis).The D. simulans population tested was founded from individuals captured in 2015 in Lyon, France (45.7835, 4.8791) (individuals captured by P. G.).All populations were initially composed of a few individuals and experienced repeated population bottlenecks during maintenance.They were thus largely inbred at the time of testing in 2017.