Natural biocide disrupts nestmate recognition in honeybees

Honeybee colonies are under the threat of many stressors, biotic and abiotic factors that strongly affect their survival. Recently, great attention has been directed at chemical pesticides, including their effects at sub-lethal doses on bee behaviour and colony success; whereas the potential side effects of natural biocides largely used in agriculture, such as entomopathogenic fungi, have received only marginal attention. Here, we report the impact of the fungus Beauveria bassiana on honeybee nestmate recognition ability, a crucial feature at the basis of colony integrity. We performed both behavioural assays by recording bee guards’ response towards foragers (nestmate or non-nestmate) either exposed to B. bassiana or unexposed presented at the hive entrance, and GC-MS analyses of the cuticular hydrocarbons (CHCs) of fungus-exposed versus unexposed bees. Our results demonstrated that exposed bees have altered cuticular hydrocarbons and are more easily accepted into foreign colonies than controls. Since CHCs are the main recognition cues in social insects, changes in their composition appear to affect nestmate recognition ability at the colony level. The acceptance of chemically unrecognizable fungus-exposed foragers could therefore favour forager drift and disease spread across colonies.

Social insects have evolved a number of behavioural adaptations to restrict the diffusion of diseases inside the colony, which go under the name of social immunity [41][42][43] . Among these adaptive responses to pathogens, social insects recognize individuals from other colonies as well as sick colony members and avoid, exclude, isolate or even kill them 41 . Cuticular hydrocarbons (CHCs), which are the cues at the basis of the recognition processes in social insects 44 , are also responsible for the discrimination of alien and sick or parasitized individuals [45][46][47][48][49][50] . The activation of the immune system, following an infection, can alter insect CHC profiles [45][46][47]50 . Consequently, biocontrol agents, including entomopathogenic fungi, can interfere with the colony recognition system in a subtle way by altering the individual chemical signature, as recently demonstrated in Myrmica scabrinodi naturally infected by ectoparasitic fungus Rickia wasmannii 51,52 and in Lasius neglectus pupae artificially infected with Metarhizium brunneum 50 . If B. bassiana infection would cause similar alterations in bees, the allegedly safe biocontrol agents might have considerable effects on colony integrity and survival by impairing its recognition systems, therefore favouring the entering and the spread of parasites and pathogens among colonies vectored by drifting foragers.
In the present study, we investigated the effect of B. bassiana on the epicuticular hydrocarbon profile of fungus-exposed foragers and if these individuals are differently treated by guards at the hive entrance compared to unexposed conspecific individuals.

Results
Behavioural assays at the hive entrance. Guard bees at the hive entrance responded differently to the presented stimuli in the bioassays with freeze-killed lures and live bees exposed to B. bassiana spores (see Methods, Fig. 1). As expected, nestmates were attacked less than non-nestmates but exposed bees were in general less attacked than non-exposed individuals (significant effects for colony membership and exposure in Table 1). An important difference also emerged between experiments with a lower reaction toward lures compared to live bees and with the occurrence of a significant interaction between colony membership and exposure in the case of lure experiment (see below). The different results in the two experiments are likely due to the fact that in our first experiment with lures we tested only guards' response to the stimuli chemical components involved in recognition, while in the second bioassay, with live freely-moving bees, the behavioural interactions between guards and the presented stimuli were also maintained. However, although the overall aggressive response was lower in the bee-lure experiment (Fig. 1a), in both cases with freeze-killed lures and live bees, exposure to B. bassiana spores had a significant effect on recognition and a decrease in the guards' aggressive response towards fungus-exposed non-nestmate foragers was observed.
Presentation of lures experiment. A mixed model GLM showed that guard bees had a lower aggressive reaction toward nestmates (Table 1) and that the fungus exposure lowered the aggression (Table 1). Furthermore, fungus exposure interacted with nestmate recognition and fungus-exposed bees (both nestmates and non-nestmates) were subjected to a similar and intermediate level of aggression between unexposed nestmates and non-nestmates (Table 1, Fig. 1a).
Live bees experiment. A mixed GLM revealed that also in this experiment, guard bees had a lower aggressive reaction toward nestmates (Table 1, Fig. 1b) and the fungus exposure lowered the aggression (Table 1). However, in the case of live bee exposure and colony membership (nestmate, non-nestmate) did not show a significant Mean number of agonistic acts (±standard error) received by unexposed nestmates, fungus-exposed nestmates, unexposed non-nestmates and fungus-exposed non-nestmates. (a) freeze-killed lures, (b) freely moving live bees.

Chemical analysis.
To test whether bees showed any change in their epicuticular CHC profile after exposure to Beauveria bassiana, the CHC profile of fungus-exposed foragers was analysed with GC-MS. Analyses of the chromatograms for all the bees allowed the identification of 49 peaks corresponding to different compounds. The data set was reduced (see Methods) by cutting off 11 hydrocarbons; statistical analyses were thus performed on 38 compounds ( Table 2). The identified compounds belong to four different classes: alkanes (N = 12), alkenes (N = 16), methyl-branched hydrocarbons (N = 1), esters between fatty acids and fatty alcohol (N = 8), plus one unidentified compound (Table 2).
When plotted separately, the scores obtained by individuals of different colonies in a Partial Least Squares discriminant analyses, showed that unexposed individuals were characterized by a good separation based on the composition of their epicuticular lipids (Fig. 2a). Conversely, the scores obtained by fungus-exposed individuals revealed a much weaker diversification (Fig. 2b). Accordingly, a jackknife attribution of the specimens resulted in 63.4% of unexposed individuals correctly assigned to their colony against a 27.2% scored by fungus-exposed individuals. It must be noted that a random attribution of specimens to three colonies is expected to return a 33.3% of individuals correctly assigned by chance.

Discussion
Our study demonstrated that the entomopathogenic fungus Beauveria bassiana, used as natural biocide in pest crop control, affects nestmate recognition in honeybees. As expected, guard bees at the hive entrance discriminated nestmate from non-nestmate foragers, being less aggressive toward unexposed nestmates. Interestingly, a reduction of the aggressive response was recorded when guards faced non-nestmate foragers experimentally exposed to the fungus B. bassiana. The decreased aggressive response towards exposed non-nestmates was observed when testing guards both with freeze-killed forager lures and with live foragers. With nestmate bees, instead, the guards' reaction towards exposed individuals differed between the two experiments. In the experiment with freeze-killed lures, although the aggressive response was much lower than with live bees, exposed nestmates received a higher number of agonistic contacts when compared with unexposed ones at a rate that was comparable with exposed non-nestmates, Conversely, exposed live nestmates presented at the hive entrance received the lowest aggressive reactions by guards. Such a discordant result is most likely due to the different experimental protocols. In facts, it is plausible that in our first experiment with freeze-killed lures we tested exclusively the chemical cues involved in recognition. Guard bees might perceive the alteration in the chemical profile of their nestmates following fungus exposure which could prompt a relatively higher reaction towards them. In the case of live bees, fungus exposed individuals both nestmates and non-nestmates were less attacked compared to their unexposed controls and exposed nestmates received the lowest number of agonistic acts, while exposed non-nestmates were attacked at a rate that was comparable with unexposed nestmates. The lowered aggressive reaction of guards toward exposed bees regardless of colony membership is most likely due to the fact that live foragers presented at the hive entrance behaviourally interacted with the guards. If exposure to the fungus also altered their behaviour alongside their chemical profile, for example making exposed bees less reactive or aggressive, they could more easily sneak inside the colony. In facts, previous studies 53,54 demonstrated that context or type of stimulus can have major effects on recognition. In the case of bees, guard are significantly more likely to make acceptance errors when tested in an unnatural context 53 , while the level of aggressive response can vary depending on the characteristics of presented stimulus as for example with Polistes wasps which show a lower www.nature.com/scientificreports www.nature.com/scientificreports/ degree of aggression when presented with only the cuticular fraction of a conspecific compared to a whole body wasp lure 54 . Overall, exposure to the fungus alters the CHCs of foragers bearing the mycosis after exposure to B. bassiana spores and this appears to be correlated to a significant change in the acceptance behaviour by guards towards fungus-exposed bees, especially increasing acceptance of exposed non-nestmates regardless of the experimental context, and such alteration.
Previous work on the CHCs dynamics on the tegument of insect hosts (Ostrinia nubilalis and Melolontha melolontha larvae) exposed to B. bassiana showed a high reduction (variable from 50% to 80% depending on the host) of the amount of such compounds within 96 h from infection 55 . Entomopathogenic fungi catabolize the long-chain hydrocarbons and penetrate inside the waxy cuticle of their insect hosts 56,57 and such metabolic interactions may be responsible for the CHC alterations of mycosis-affected insects 50,56,57 .
By analysing the chemical profiles of fungus-exposed and unexposed foragers we found that differences between groups are mainly due to the different amount of six alkenes and two alkanes, which were consistently www.nature.com/scientificreports www.nature.com/scientificreports/ less abundant in fungus-exposed individuals, and one alkane and one ester of palmitic acid that was more abundant in the fungus-exposed bees rather than in the unexposed bees. The differential decrease in the alkene fraction in exposed individuals resulted in a homogenization of chemical profiles among colonies (Fig. 2) and it is likely at the basis of the observed reduction in aggressivity against exposed non-nestmates. This alteration of the hosts' chemical profile may arise during the first phases of the infection process from the breakdown of hydrocarbons by B. bassiana 58 or after infection due to an immune response affecting the synthesis of specific hydrocarbons 45,46 .
The selective reduction of specific alkenes is of particular interest, since it has been demonstrated that bees supplemented either with a synthetic mixture of alkenes or with the colony-specific alkenes fraction, are treated more aggressively by nestmate guards than control bees or nestmates applied either with a synthetic mixture of alkanes or with the colony-specific alkanes fraction 59 . Such findings indicate that alkenes play a crucial role in nestmate recognition in honeybees and that the relative increase of such compounds on the cuticle of incoming foragers elicits aggressive behaviour in guards, preventing them from recognizing nestmates 59 . Indeed, alkenes appear to be responsible for differences in the average chemical profile of different honeybee colonies as well as other social hymenoptera 60 . Moreover, alkenes have been found to be more easily learned and discriminated in PER (proboscis extension reflex) trained bees with respect to alkanes 60,61 . Thus, the lower amount of the six specific alkenes found on the cuticular profile of B. bassiana-exposed bees, with respect to unexposed ones, are most likely responsible for the higher degree of acceptance of guard bees towards fungus-exposed non-nestmate foragers with respect to unexposed non-nestmate ones.
Recent work has demonstrated that the selective increase of specific alkenes in pupae of the ant Lasius neglectus exposed to the enthomopathogenic fungus Metarhizium brunneum induce the killing of infected brood during the non-transmissible incubation period of the pathogen 50 . Our bees exposed to Beauveria showing a selective reduction in alkenes might instead be depleted of those compounds relevant for colony discrimination 60,61 . Such a modification interferes with the nestmate recognition system and guard bees at the hive entrance are unable to correctly assess foragers affected by the mycosis.
Several bee parasites have evolved adaptive strategies based on chemical mimicry to their host: Varroa mites, for instance, show an almost perfect colony specific mimicry with their host to avoid detection [62][63][64][65] . Chemical mimicry, replicating the colony specific pattern of CHCs, has also been reported for the death's head hawkmoth Acherontia atropos and the parasitic fly Braula coeca 66,67 , and appears to be widespread in social insect parasites mimicking the host chemical profile to favour their spread and survival inside colonies 67 . Since B. bassiana is an unspecific insect pathogen, the modification of cuticular hydrocarbon profile observed in our fungus-exposed bees may be a mere by-product of the pathogen's metabolism 56-58 rather than an adaptive strategy evolved to infect honeybee colonies; nevertheless the selective decrease of cuticular olefins in our fungus-exposed bees could represent a serious threat for colony integrity.
Artificial infection with the non-specific pathogen E. coli also alters the cuticular profile of bee workers, but such changes did not consist of a selective reduction of the alkene fraction and the resulting alterations in cuticular profiles elicited an increase in the nestmates' aggressive response towards bacteria-injected individuals 46 . A significant change in the CHC profile of honeybees has also been reported for workers infected with deformed wing virus (DWV) which are detected and removed from the colony 47 , however, also in this case, the chemical alteration involves different fractions of the cuticular compounds rather than the specific alkene fraction important for nestmate recognition 59 .
Overall, our results showed a clear disrupting effect of B. bassiana on the nestmate recognition system of honey bees, potentially favouring the drift of foragers to alien colonies and in turn the potential spread of parasites and pathogens. Indeed, an efficient nestmate recognition system is crucial for honey bee colonies as it limits drifting and robbing behaviour 68 , which are the main causes for the spread of infectious diseases, such as www.nature.com/scientificreports www.nature.com/scientificreports/ varroatosis, foulbrood and nosemosis [69][70][71] . Foragers carrying the fungal pathogen Nosema ceranae are even more prone to drift toward nearby hives than uninfected ones 71 . Given the use of B. bassiana as a natural biocide in organic agriculture and the possible use of pollinators as vectors for microorganisms in biocontrol 19,21,24,25 , future research should aim at investigating whether foragers exposed to the fungus spores show an altered behaviour (i.e. higher degree of drifting) with respect to unexposed bees.
Many recent efforts to limit the use of chemical pesticides and to adopt more sustainable agricultural models have led to a considerable increase in the use of bio-insecticides for pest management 13,14,17 . Although the adoption of natural biocides is desirable, we showed that treatments with the allegedly safe B. bassiana interfere with nestmate recognition in honeybees, possibly fostering inter-colony transmission of diseases. We thus warn that natural biocides used for pest management should not only be tested for their direct toxicity at the individual level on target and non-target species. The same attention should be dedicated to carefully testing for side effects on non-target ecologically relevant social insect species to understand the potential effects on those behavioural features linked to complex social organization determining the colony health status and survival. Behavioural assays. Behavioural assays were performed, by using two different procedures, to test whether guard bees at the hive entrance are able to recognize fungus-exposed nestmate/non-nestmate foragers. The first bioassay consisted of presentation of lures at the hive entrance, where the importance of the mere chemical cues for recognition was tested. The second bioassay consisted of a more natural condition in which guard bees had to recognize fungus-exposed or unexposed live foragers induced to approach the hive entrance 72 ; in this experiment, the behavioural components are maintained alongside the chemical ones. All the experiments were carried out in accordance with Italian laws and regulations.

Presentation of lures experiment.
i) Lure preparation: forager bees were collected at the hive entrance while departing for their foraging flights (20 foragers per hive; 28 hives from apiary A; four hives from apiary C). Foragers were then brought to the laboratory, about half of the collected bees for each colony were exposed to the fungus by applying 1 μl of a 0.01% Triton X-100-water solution, containing about 10 6 conidia of B. bassiana, on their thorax by using a 2 μl micropipette. We chose to use this concentration, corresponding to 10 9 conidia/ml, since it appears similar or lower to those of the solution commonly sprayed in the field for pest biocontrol 19 or pollinator biocontrol vector technology 24,25,27 . The other half of the foragers was applied with 1 μl of triton-water solution, containing no fungus conidia using the same procedure. Experimental and control bees were separated based on treatment and colony of origin in 7 cm Petri dishes, supplied with honey and water ad libitum for three days. A period of three days is sufficient for the fungus to reach the hemolymph and trigger an immune response 21,73 , potentially inducing detectable changes in epicuticular CHCs, which could in turn affect the bee guards' response [46][47][48][49]72,74,75 . Foragers were kept in the dark, under controlled laboratory conditions (28-29 °C T, 55-65% RH); bees that died during treatment were daily removed from each dish. After three days, bees were killed by freezing and brought to the apiary A for field bioassays. ii) Lure presentation: to perform the behavioural assay we followed a procedure similar to that used by Cappa et al. 76 : each of the selected colonies (N = 28) in apiary A was presented with four stimuli in a random sequence: two nestmate and two non-nestmate foragers from apiary C either unexposed or exposed to the fungus three days before the behavioural assays. Each lure was fixed to a small clip at the top of a 40 cm thin metal rod, slowly brought near to the hive entrance, and gently moved alongside the flight board contacting as many bees as possible in order to record their reactions 76 . Each presentation lasted 1 min starting from the first interaction of a bee with the lure. To avoid habituation effects, assays were performed on a same colony with at least a 15-min interval between each other. All the assays were carried out on a single day, between 10:00 and 16:00. All the assays were videotaped. Each lure was used only once and we carried out a total of 112 assays.
Videotapes were blind-watched in slow motion (0.25 s) by a viewer who noted the total number of bees interacting with each lure and the total number of agonistic contacts (biting, stinging) of resident bees towards the lure. As the lure was actively moved on the flight board to record the reactions of the highest possible number of www.nature.com/scientificreports www.nature.com/scientificreports/ bees, which could vary during the 1-minute assay, the average number of bees at the entrance was calculated by counting the number of bees on the flight board at the beginning and at the end of presentation 76 .
We used the "glmmadmb" function in the "glmmADMB" package (http://glmmadmb.r-forge.r-project.org/) to fit a Generalized Linear Mixed Model using a Poisson distribution. As predictor variables we used i) average number of bees at the entrance, ii) colony membership (alien vs nestmate), iii) exposure to B. bassiana as fixed factors together with the iv) interaction between colony merandom factor.
Live bee experiment. The following year, a different behavioural assay was performed by presenting live foragers to the colonies. These presented foragers were: a) unexposed nestmates, b) unexposed non-nestmates, c) fungus-exposed nestmates, d) fungus-exposed non-nestmates. Nestmate foragers (apiary B) and alien foragers (apiary C) were collected and treated following the same procedure described in the lure presentation experiment. Three days after the exposure with B. bassiana, both exposed and unexposed bees were individually placed into eppendorf tubes and carried to apiary A for the bioassay. Tubes were kept for several minutes in a ice-cooled styrofoam box before presentations, in order to calm down the experimental bees, therefore preventing them from fidgeting or flying away once freed from the tubes 57,72,74 . When the assay started, each focal bee was gently released on the landing board and was free to move, to interact with other bees, to enter the hive or fly away. Bee guards' reactions were recorded for 3 minutes after the first interaction. When the focal bee entered the hive, or flew away without interacting with guards the assay was discarded. The videotaped assays were blind-watched to record the number and the duration of each agonistic interaction (biting, stinging) among focal bees and guards on the landing board. A total of 259 assays (108, 65, 86 respectively on hives a, b and c in apiary B) were performed (Table S1). We modelled the number of agonistic acts with the same design of the same Generalized Linear Mixed Model and contrast analysis as described above.
Chemical analyses. Data collection. We collected 100 foragers from the three tested colonies a, b and c (at least 30 bees per colony) in the experimental apiary B. Following exactly the same procedure used for bioassays, we exposed half of these bees to B. bassiana conidia while the remaining half were used as control and received only the 0.01% Triton X-100-water solution. Experimental and control bees were separated by treatment and colony in 7 cm Petri dishes, supplied with honey and water ad libitum for three days. All the bee groups were maintained in the dark, under controlled laboratory conditions. Three days after the bees were killed by freezing for subsequent chemical analyses. Before the analyses, each bee was thawed, then the apolar fraction of the cuticular compounds of each worker (N = 85, 44 fungus-exposed, 41 unexposed controls) was extracted by washing the entire body for 10 min in 1 ml of pentane. The extracts were dried under a stream of nitrogen and then re-suspended in 10 µl of heptane with 70 ng/μl of heptadecane (n-C 17 ) as internal standard. One μl of extract was injected in a Hewlett Packard (Palo Alto,CA, U.S.A.) 5890 A gas chromatograph (GC) coupled to an HP 5970 mass selective detector (using 70 eV electronic ionization source). A fused ZB-WAX-PLUS (Zebron) silica capillary column (60 m x 0.25 mm × 0.25 mm) was installed in the GC. The injector port and transfer line temperatures were set at 200 °C and the carrier gas was helium (at 20 PSI head pressure). The temperature protocol was: from 50 °C to 320 °C at a rate of 10 °C/min and the final temperature was kept for 5 minutes. Injections were performed in splitless mode (1 min purge valve off). Data acquisition and analysis were done using the Chem Station G1701 BA (version B.01.00) -Copyright © Hewlett-Packard 1989-1998.
CHCs were identified on the basis of their mass spectra, and equivalent chain length. For the preparation of the dataset used in statistical analysis we only included compounds quantified in at least 25% of specimens with the exception of compounds only identified in one of the treatment groups (exposed/unexposed). We used Mixed-effects models for repeated-measures ANOVA (glmmPQL function of the MASS R package) to evaluate differences both in the total amount of cuticular hydrocarbons present in the extracts and in the amount of single compounds between fungus-exposed and control bees; colony membership was included as a random factor. Multivariate analyses have been applied to the dataset to verify the possibility of attributing fungus-exposed and unexposed specimens to their colony based on chemical composition. With this aim we first performed a Partial Least Square Discriminant analysis using colony membership as a priori grouping variable. This allowed us to visualize the general pattern of similarity among individuals. The possibility to attribute specimens to their colonies has been tested by a jackknife procedure where a Partial Least Square Discriminant analysis has been performed on all the uninfected specimens but one. Then the colony membership of the excluded specimen and of one fungus-exposed individual (in random order) has been predicted on the basis of their CHCs composition. We used the percentage of correctly attributed cases as a measure of the possibility to blindly attribute fungus-exposed and unexposed individuals to their colonies.

Data Availability
Data files and R script used are uploaded alongside supplementary material.