Effects of sublethal doses of clothianidin and/or V. destructor on honey bee (Apis mellifera) self-grooming behavior and associated gene expression

Little is known about the combined effects of stressors on social immunity of honey bees (Apis mellifera) and related gene expression. The interaction between sublethal doses of a neurotoxin, clothianidin, and the ectoparasite, Varroa destructor, was examined by measuring differentially expressed genes (DEGs) in brains, deformed wing virus (DWV) and the proportion and intensity of self-grooming. Evidence for an interaction was observed between the stressors in a reduction in the proportion of intense groomers. Only the lowest dose of clothianidin alone reduced the proportion of self-groomers and increased DWV levels. V. destructor shared a higher proportion of DEGs with the combined stressors compared to clothianidin, indicating that the effects of V. destructor were more pervasive than those of clothianidin when they were combined. The number of up-regulated DEGs were reduced with the combined stressors compared to clothianidin alone, suggesting an interference with the impacts of clothianidin. Clothianidin and V. destructor affected DEGs from different biological pathways but shared impacts on pathways related to neurodegenerative disorders, like Alzheimer’s, which could be related to neurological dysfunction and may explain their negative impacts on grooming. This study shows that the combination of clothianidin and V. destructor resulted in a complex and non-additive interaction.

causing acetylcholine receptors (nAChRs) to open ion channels leading to exhaustion and death 18 . However, sub-lethal doses of neurotoxic insecticides can also be damaging 19,20 . For example, for neonicotinoids, sub-lethal doses of imidacloprid reduced memory retention and altered motor function in honey bees 21,22 and thiamethoxam impaired learning and memory in bumblebees 23 . Honey bees can be exposed to multiple sub-lethal doses of pesticides by gathering nectar and/or pollen when foraging (reviewed in refences 24 and 25 ). While there is no controversy on the effect of acute exposures to lethal doses of insecticides, like neonicotinoids, to non-target insects like honey bees, there is controversy about the effects of chronic exposure to sublethal doses (reviewed in reference 25 ).
There are few studies on the importance of neural activity on grooming behavior in honey bees 13,26,27 . Also, there is little information about how neural activity can be affected by compounds and ectoparasites and their impact on grooming behavior. Thus, a study on the effect of a neonicotinoid, clothianidin, and V. destructor on neural mechanisms was undertaken using self-grooming behavior as the marker of the neural response. In this study, three sublethal doses of clothianidin with or without V. destructor were examined for the frequency and intensity of self-grooming behavior, DWV levels and honey bee gene expression, in order to better understand how a ACh agonist can act alone or interact with a honey bee parasite to affect a mechanism of social immunity.

Results
Self-grooming behavior. Significant effects of treatments were found for the proportion of bees that selfgroomed in any manner versus bees that did not groom (Chi 2 (7) = 36.019, p < 0.003) ( Table 1). A post-hoc analysis revealed that the only significant difference was with the exposure to 9 × 10 −4 ng/µl clothianidin alone which had the lowest proportion of bees that self-groomed (p < 0.003; Table 1). Also, the post-hoc analysis showed that there were no effects on the proportion of bees that self-groomed in the groups parasitized by V. destructor, with or without clothianidin. These results indicate that the only factor linked to a decrease in the proportion of bees that groomed in any manner was the exposure to the lowest dose of clothianidin without V. destructor and no interaction between the two stressors for that parameter was found.
Significant effects of treatments were also found for the proportion of bees that self-groomed intensively versus not groomed or self-groomed lightly (Chi 2 (7) = 48.85, p < 0.0002) ( Table 2). Post-hoc analysis showed that at 4.2 × 10 −3 ng/µl clothianidin there was significantly less intense groomers with V. destructor than without V. destructor. There was also a significant reduction in the proportion of intense groomers for bees treated with  Table 1. Contingency table showing the number of bees that self-groomed in any manner or that did not selfgroom, and the proportion of bees that self-groomed in any manner within 3 min after placing 20 mg of flour on their thoraces that had previously been treated with clothianidin (0 ng/µl, 9 × 10 −4 ng/µl, 4.2 × 10 −3 ng/µl and 1 × 10 −2 ng/µl) and/or V. destructor. The asterisk indicates a significant reduction in the proportion of bees that self-groomed based on Chi 2 analysis and adjusted residuals and p value of 0.0031.  Table 2. Contingency table showing the number of bees that self-groomed intensively or that self-groomed lightly, and the proportion of bees that self-groomed intensively within 3 min after placing 20 mg of flour on their thoraces that had previously been treated with clothianidin (0 ng/µl, 9 × 10 −4 ng/µl, 4.2 × 10 −3 ng/µl and 1 × 10 −2 ng/µl) and/or V. destructor. The asterisks indicate a significant reduction in the proportion of bees that self-groomed intensively based on Chi 2 analysis and adjusted residuals and p value of 0.00019.
1 × 10 −2 ng/µl clothianidin both with or without V. destructor. The significant reduction in the proportion of intense groomers in bees treated with 4.2 × 10 −3 ng/µl clothianidin only in combination with V. destructor indicates a possible interaction between the two stressors for that parameter.
DWV quantification. Bees exposed to V. destructor had 1.8 log 10 more DWV GCs per µg RNA than control bees (no clothianidin and no V. destructor) (F (1,64) = 197.85, p =<0.0001) (Fig. 1). By comparison, bees exposed to 1 × 10 −4 ng/µl clothianidin had 1.4 log 10 more DWV GCs per µg RNA relative to the control (F (3,64) = 3.84, p = 0.014). Thus, V. destructor parasitism has a much greater effect on DWV levels compared to the control than the lowest dose of clothianidin only. With treatments with higher doses of clothianidin, lower levels of DWV were observed in bees exposed to clothianidin alone compared to the lowest dose, whereas there was no significant change in DWV levels in bees exposed to both stressors compared to bees only parasitized with V. destructor. Therefore, the main factor contributing to DWV loads in the bees was V. destructor parasitism and the lowest dose of clothianidin without V. destructor. Up-regulated DEG pairwise comparisons. A Venn diagram of the up-regulated DEGs showed that 92.1% of the DEGs up-regulated by 1 × 10 −2 ng/µl of clothianidin alone were unique to that treatment, none were shared solely with V. destructor alone, 3.4% were shared solely with the combined stressors, and 4.5% were shared among all the treatments ( Fig. 2A and Supplementary Table S7). Thus, almost all those DEGs were unique to the effects of clothianidin. For the V. destructor alone treatment, 64.8% of the DEGs were not shared with the other treatments, none were shared only with clothianidin alone, 21.6% were shared only with the combined stressors and 13.6% were shared among all the pairwise comparisons. The majority of those DEGs were unique to V. destructor but many were shared, particularly with the combined stressor treatment. The combined stressors had ) of adult bees that were exposed to 0, 9 × 10 −4 , 4.2 × 10 −3 and 1 × 10 −2 ng of clothianidin per µl and/or V. destructor. Data points with different letters above them represent significant differences using Tukey's HSD tests after a two-way ANOVA showed a significant effect.

RNAseq.
www.nature.com/scientificreports www.nature.com/scientificreports/ just 35.5% of the DEGs being unique to that treatment, 14.5% shared only with clothianidin alone, 30.6% shared only with V. destructor alone and 19.4% shared among all the treatments. Thus, the combined stressors had the lowest number of DEGs unique to that treatment and was the only treatment where more DEGs were shared with other treatments (40) than were unique to that treatment (22).
Based on the number of up-regulated DEGs, it appears that the significantly up-regulated DEGs with the combination of V. destructor and clothianidin was more similar to that of V. destructor alone than to clothianidin alone. In contrast, there was relatively little in common among up or down-regulated DEGs between clothianidin alone and V. destructor alone showing that each stressor alone had different effects. The much smaller number of DEGs up-regulated by the combined stressors (62) compared to clothianidin alone (267) suggests that there was a possible interaction between clothianidin and V. destructor with the parasite reducing the number of DEGs when combined with clothianidin.

Down-regulated DEG pairwise comparisons.
A Venn diagram of the down-regulated DEGs showed that 87.1% of the DEGs down-regulated by 1 × 10 −2 ng/µl of clothianidin alone were unique to that treatment, 6.5% were shared solely with V. destructor alone, none were shared solely with the combined stressors, and 6.5% were shared among all the libraries (Fig. 2B and Supplementary Table S7). Thus, like the up-regulated DEGs, almost all those DEGs were unique to the effects of clothianidin. For the V. destructor alone treatment, 55.1% of the DEGs were not shared with the other treatments, 2.6% were shared only with clothianidin alone, 39.7% were shared with the combined stressors and 2.6% were shared among all the treatments. Thus, a very similar number of those DEGs were shared between V. destructor and the combined stressor treatments. The combined stressors had just 41.07% of the DEGs being unique to that treatment, none were shared with clothianidin alone, 55.4% were shared only with V. destructor and 3.6% were shared among all the treatments. Like up-regulated DEGs, the combined stressors had the lowest number of DEGs unique to that treatment and was the only treatment where more DEGs were shared with other treatments (33) than were unique to that treatment (23).
Based on the number of down-regulated DEGs, similar conclusions as for up-regulated DEGs can be made about the greater similarity between the down-regulated DEGs of the combined stressors to that of V. destructor alone than to clothianidin alone. However, the number of DEGs down-regulated by the combined stressors (56) was not very different compared to clothianidin alone (31) as that observed for up-regulated DEGs, suggesting less possible interaction between clothianidin and V. destructor. KEGG analysis. KEGG pathway analysis was used to examine the link between DEGs and biological pathways and to compare which biological pathways were shared between treatments or unique to a particular treatment. KEGG analysis was able to assign approximately one quarter of all the up or down-regulated DEGs in this study to biological pathways (Supplementary Tables S8-S13). More than twice the number of KEGG pathways (256) were identified for up-regulated DEGs compared to down-regulated DEGs (106).
There were a number of noteworthy KEGG terms among the up-regulated DEGs (Supplementary Tables S8, S10 and S12). Some of the pathways among the 144 KEGG terms assigned to DEGs up-regulated by clothianidin that were unique to that treatment were associated with Parkinson's disease. Among the 74 KEGG terms for up-regulated DEGs by V. destructor, viral myocarditis was notable among the pathways that were unique to V. destructor, and for the 27 KEGG terms for up-regulated DEGs with the combined stressors, notable KEGG terms www.nature.com/scientificreports www.nature.com/scientificreports/ unique to that treatment included insulin signaling pathway. For up-regulated KEGG pathways common between clothianidin and other treatments, noteworthy KEGG terms were Alzheimer's disease, dopaminergic synapse and glutamatergic synapse shared with V. destructor, cellular senescence shared with the combined stressors, and calcium signalling pathway shared between all three treatments. For KEGG pathways common between V. destructor and the combined stressors, a noteworthy term the hippo signalling pathway-fly.
There were also a number of noteworthy KEGG terms among the down-regulated DEGs (Supplementary Tables S9, S11 and S13). For the 31 KEGG terms assigned to down-regulated DEGs by clothianidin, a notable term was sphingolipid metabolism. Among the 45 KEGG terms linked to down-regulated DEGs only by V. destructor were neuroactive ligand-receptor and among the 30 KEGG terms associated with the combined stressors was NOD-like receptor signaling pathway. For down-regulated KEGG pathways common between clothianidin and V. destructor, a possibly significant KEGG term was peroxisome, while longevity regulating pathway-multiple species was notable among the down-regulated DEGs shared by clothianidin, V. destructor and the combined stressors.

Discussion
This study examined the effect of two stressors, sublethal doses of clothianidin and V. destructor on gene expression in the brain and increases in a mite-transmitted virus, DWV, to better understand how these two stressors act separately and together to affect honey bees on aspects of neural activity related to self-grooming behavior. Among the two stressors, clothianidin alone had a greater effect than V. destructor alone to reduce self-grooming behavior, based on the proportion of bees that self-groomed and the proportion of bees that self-groomed intensively, as well as the number of up-regulated DEGs. However, the results with the combination of V. destructor with clothianidin showed that combining each stressor did not result in a simple additive effect in the number of DEGs or the KEGG pathways associated with them.
One unexpected result of this study was that both DWV levels and the proportion of bees that self-groomed were affected only by the lowest dose of clothianidin without V. destructor. This may indicate that effects of DWV and grooming could be linked in some manner. Unexpectedly, higher doses of clothianidin in absence of mites, or the same dose of clothianidin combined with V. destructor, did not have the same impacts, indicating that there was a relatively specific impact from the lowest dose of clothianidin tested. Hormesis occurs when there is a beneficial biological response to a low exposure to a stressor 28 . In this case, a beneficial response did not occur at the lowest dose, but the effect could be similar to hormesis. At the lowest clothianidin dose, it would still bind to nicotinic acetylcholine receptors of the neural cells resulting in neural stimulation but may not overstimulate it to the level of toxic doses that block receptors impeding the action of the neurotransmitter ACh 29 . At very low doses, clothianidin stimulation of the central nervous system may just be sufficient to somehow interfere with self-grooming and resistance to a virus. While these results imply hormesis, future research should investigate more sublethal doses of clothianidin to confirm this by determining the range of doses that results in these effects.
While DWV levels were increased by the lowest dose of clothianidin without V. destructor, they were still much less than with V. destructor alone, which was not surprising as DWV can replicate in V. destructor, which also acts as a vector of DWV to bees 30 . An increase in DWV with the lowest dose of clothianidin without V. destructor indicates that the bees were latently infected with DWV, and the treatment permitted DWV to multiply in the bee. Latent infections of DWV in bees have been widely reported, and the levels of DWV in a bee may reflect the degree of resistance to the virus 31,32 . One explanation for increased DWV loads could be that the lowest dose of clothianidin suppressed humoral immunity in bees, which has been reported for a dose of clothianidin similar to the lowest dose in this study 33 . The mechanism proposed by Di Prisco et al. 33 for this was that clothianidin negatively affected the NF-ƙβ signaling pathway, which regulates gene expression related to antiviral defence mechanisms 34 . However, another explanation could be that the neonicotinoid reduced hemocyte density, which occurred following sublethal doses of thiacloprid and imidacloprid in adult honey bees comparable to the lowest dose of clothianidin in this study 35 . Hemocytes are important for cellular defence mechanisms against viruses in insects through the phagocytosis of infected cells after induced apoptosis 36 . Higher levels of DWV infection could have contributed to impairment of the proportion of bees that groomed due to the virus affecting neural processes. Iqbal and Mueller 6 showed that bees infected with DWV showed impairment of associative learning using the proboscis extension response (PER) assay. This could be due to DWV multiplication in the nervous system, including the brain 37 . Perhaps this is also true for DWV and grooming. Viral infections of insect brain tissues have been shown to affect behaviors, such as increased feeding activity in Aedes aegypti infected with dengue virus 38 .
Another explanation for the effects of the lowest dose of clothianidin on the expression of grooming behavior could be directly related to neural stimulation by clothianidin 29 . To perform self-grooming behavior, bees first have to perceive the stimulus of a substance, particle or pathogen on their bodies 39 . They then have to process that signal in the central nervous system and send the appropriate response through peripheral nerve conduction to the muscles 40 . Williamson et al. 22 provided honey bees with sugar syrup containing even lower sublethal doses of clothianidin and thiamethoxam than in this study. After exposure, bees were not evaluated for the proportion that groomed, but for the time spent grooming, which was increased with the thiamethoxam but not with the clothianidin treatment. While the results are not directly comparable to this study, it does show that very low levels of a neonicotinoid can affect neural activity and grooming behavior.
Another parameter of grooming behavior, the intensity with which bees groom themselves, is important as it correlates well with the number of mites that they can remove from their bodies 10 . Thus far, there have been no studies of a neonicotinoid on this aspect of bee behavior. This study showed that at the medium dose, a significant reduction in the proportion of bees that groomed intensively occurred when exposed to both clothianidin and V. destructor, but not clothianidin alone, suggesting an interaction between the two stressors. This was (2019) 9:5196 | https://doi.org/10.1038/s41598-019-41365-0 www.nature.com/scientificreports www.nature.com/scientificreports/ not observed at the lowest and highest clothianidin doses. At the highest clothianidin dose, neural stimulation appears to be sufficient to affect neural activity associated to perceiving irritants on the bee's body and reacting to them; including V. destructor did not increase that. In contrast, the lowest clothianidin dose apparently provided insufficient neural stimulation to affect intense grooming, even if V. destructor was included. Treatment with V. destructor alone was also insufficient to affect grooming intensity, despite the expected effect on energy stress by feeding on hemolymph 4 and/or possible neurological damage from viruses 6 . Thus, the dose of clothianidin is critical for demonstrating an interaction between V. destructor and clothianidin on grooming intensity. It was also notable that the proportion of intense groomers was much more decreased by the stressors compared to the proportion of bees that groomed in any manner, possibly due to a greater effect on neural processes required for intense grooming, such as neurotransmitters like acetylcholine. Thus, future work is warranted to shed light on the neural processes involved in grooming and on differences in neurological activity between light and intense groomers. This appears to be first report of possible interactions between biotic and abiotic factors affecting grooming intensity in insects. Based on these results, it would be expected that colonies of honey bees exposed to V. destructor would be more susceptible to negative effects of clothianidin, possibly resulting in less resistance to V. destructor and thus larger populations of the parasite in the colonies, due to the impaired ability of the bees to groom themselves intensively.
Gene expression analysis using RNAseq with insect brains can be a powerful way to reveal impacts on biological pathways in the brain. Differential gene expression has been linked to pathogens, such as Black queen cell virus and Nosema ceranae in honey bees and nucleopolyhedrovirus in silkworms 41,42 or behaviors, such as hygienic behavior in honey bees 43 .
One surprising result in this study from the RNAseq analysis was that the total number of DEGs significantly up-regulated uniquely by clothianidin was more than eleven times higher than with the combined stressors. Thus, the addition of V. destructor parasitism appeared to reduce the impacts of clothianidin on gene expression in the brain. One might have expected the reverse, where even more significantly up-regulated DEGs would be detected with the combined stressors due to effects of V. destructor parasitism making the bee more sensitive to clothianidin damage. The results could be due to V. destructor or viruses causing such strong damage to brain tissue that it limits the effects of clothianidin. This may also explain why the number of up-regulated DEGs was only slightly lower with V. destructor parasitism than by the combined stressors. This implies that the effects of combining stressors is unpredictable, at least for affecting gene expression in honey bee brains.
KEGG analysis provided some insights into the effects of clothianidin, V. destructor and clothianidin plus V. destructor on the brains of the treated bees. There were a number of DEGs that had KEGG terms that were unique to each stressor. KEGG terms solely observed with clothianidin exposure included Parkinson's disease pathway, which is associated with degeneration of the central nervous system 44 , and sphingolipid metabolism, which is involved in signalling for diseases like the neurodegenerative Alzheimer's disease 45 . Those terms indicate that sublethal effects of clothianidin on acetylcholine receptors may also be causing degeneration of the brain. KEGG terms solely found with V. destructor parasitism, included viral myocarditis that is involved in heart muscle inflammation commonly caused by viral infections 46 and neuroactive ligand-receptor, which has been associated with the early stage symptoms of Parkinson's disease in the Drosophila melanogaster disease model 47 . While D. melanogaster does not show the symptoms of Parkinson's disease like humans, Whitworth 48 noted that it is a good model as there is extensive conservation of neuronal function and development between D. melanogaster and vertebrates when examined at a cellular level. The association of those biological pathways with V. destructor parasitism indicates that certain impacts of the parasite or viruses transmitted by the parasite are detrimental to neural function. KEGG terms unique to the combined stressors included insulin signaling, whose disruption in brain tissue has been associated with neurodegenerative disorders, like Alzheimer's disease in rats 49 . Insulin signaling has also been linked to cast differentiation 50 , longevity 51 and division of labour 52 in honey bees. Another KEGG term unique to the combined stressors was the NOD-like receptor signaling pathway, which regulates innate immune receptors whose disfunction is also associated with neurodegenerative disorders, like Alzheimer's and Parkinson's disease 53 . The association of those biological terms with the combined stressors suggest that the combined stressor treatment was resulting in DEGs related to neurodegeneration that were not detected with each stressor alone.
The fact that the three stressors shared common KEGG terms indicates some common impacts of clothianidin and V. destructor alone or in combination. KEGG terms shared between clothianidin and V. destructor included Alzheimer's disease, which is a characterized by pathological alterations in neuronal receptors 54 , dopaminergic and glutamatergic synapse, which control functions like locomotor activity, learning and memory 55 , and peroxisome which is important in detoxification of free radicals that could damage neurological functions in brains 56 . KEGG terms found both with clothianidin and the combined stressors included cellular senescence related to an irreversible cellular arrest associated with brain pathology 57 . Among the KEGG terms associated with both V. destructor and the combined stressors were hippo signalling pathway-fly related to cell proliferation and whose dysregulation is linked to pathologies like cancer 58 . The KEGG term longevity regulating pathway-multiple species was found with all the three treatments. That biological pathway promotes cellular fitness through autophagy and stress defence 59 . All of the above indicates that a parasite and sub-lethal doses of a neonicotinoid have some shared as well as unique negative impacts on the honey bee brain, like cell damage, proliferation and senescence, that are consistent with aspects of neurodegeneration.
In conclusion, this study showed that a sublethal, chronic exposure to clothianidin, similar to that expected under field conditions, can more negatively impact the self-grooming behavior of honey bees when combined with V. destructor parasitism than applied alone. Additionally, RNAseq analysis of the brains of treated bees revealed different impacts on gene expression by each stressor, which was also observed when the stressors were combined based on the number of DEGs. Surprisingly, the interaction between the stressors decreased rather than increased the number of up and down-regulated DEGs found with clothianidin. Although a variety of www.nature.com/scientificreports www.nature.com/scientificreports/ biological pathways were associated with the DEGs, it was notable that many terms were associated with neurodegeneration and cell damage implying that each stressor alone or in combination may negatively affect neural activity, which could help explain their impact on grooming behavior potentially reducing the bees' survival.

Methods
Sources of honey bees and varroa mites. Honey bees were obtained from colonies of the Buckfast strain at the University of Guelph's Honey Bee Research Centre in Ontario, Canada. The queens of the colonies for the brood and workers in this study were mated under controlled conditions in isolation at Thorah Island, Simcoe, ON, to guarantee the purity of the Buckfast strain and uniformity of its genotype. The colonies that were used as source of bees were not previously exposed to treatments, such as pesticides. The newly emerged bees used for the four different biological repetitions did not come from the same colony, but the bees shared the same genotype. Female Varroa mites were collected from highly infested colonies as per Arechavaleta and Guzman-Novoa 11 and placed in a Petri dish for immediate use in experiments.
Exposure to clothianidin and/or V. destructor. Newly emerged bees (<24 h) were obtained from frames after incubation in screened emerging cages (50.3 × 7.3 × 25.2 cm) at 35 °C and 60% RH overnight. Forty newly-emerged bees were randomly assigned to each of the eight treatments. The treatments were 0 ng/µl (control), 9 × 10 −4 ng/µl, 4.2 × 10 −3 ng/µl, 1 × 10 −2 ng/µl, 0 ng/µl + V. destructor, 9 × 10 −4 ng/µl + V. destructor, 4.2 × 10 −3 ng/µl + V. destructor, and 1 × 10 −2 ng/µl + V. destructor. For the four treatments with V. destructor, one female mite was taken from a Petri dish using a fine paintbrush, placed on the abdomen or thorax of a bee and observed to verify that the mite was attached to the bee's body. The bees were incubated in a sterilized hoarding cage (12.7 × 8.5 × 14.5 cm) at 35 °C and 60% RH for seven consecutive days and were fed 50% sugar syrup ad libitum containing the desired dose of clothianidin, using a 20 ml gravity feeder. The sugar syrup contained in the feeders was weighed on the 3 rd and 7 th day of treatment to determine the amount of syrup that had been consumed by the 40 bees in each hoarding cage, which was within the amounts reported in the literature. The mortality of the bees was also recorded at the end of the 7th day of treatment, but it was negligible, as it was less than 2% in all the treatment groups. The experiment was repeated four times. The treatments that corresponded to the same biological repetition were done at the same time.
Grooming behavior assays. Seven days after treatment, 1,056 bees from the eight treatments were assessed for self-grooming behavior as per Aumeier 64 with some modifications 65 . Briefly, a treated bee was placed in a Petri dish (100 mm × 15 mm, FisherScientific ® Mississauga, ON, Canada) that was lined with Whatman ™ white filter paper (Fisher Scientific ® ) covered with perforated plastic foil. After being introduced into the Petri dish, the bee remained for 1 min to become accustomed to the testing arena. Then, approximately 20 mg of wheat flour were put on the dorsal surface of the bee's thorax using a fine brush, and for 3 min, self-grooming instances exhibited were recorded and classified as per Guzman-Novoa et al. 10 A study conducted by Espinosa-Montaño 65 showed that using 20 mg of flour was as reliable as using a varroa mite as an irritant for the grooming behavior assay. Class "light grooming" occurred if slow swipes were observed and the bee used one leg or two legs at most, class "intense grooming" occurred if the bee performed vigorous wiping and shaking using more than two legs, and class "no grooming" occurred if the bee did not show any kind of grooming activity. After the assessments, all the bees used in the trials were immediately frozen at −70 °C for RNA analysis.
RNA extraction and RNAseq analysis. Total RNA was extracted from 15 to 25 brains obtained from randomly selected bees per treatment for three replications, using TRIzol ® reagent following the manufacturer's instructions with modifications as per Boutin et al. 43 15 µl of the RNA from each of the three biological replicates per treatment were pooled to obtain the equivalent of RNA from 45 brains, which was used for RNAseq analysis. RNA samples were sent to McGill University (Génome Québec Innovation Centre, Montreal, QC, Canada) to perform a high throughput sequencing analysis, using a HiSeq. 2500 v4 (Illumina, San Diego, CA, USA). DWV quantification. 15 µl of the RNA from three biological replicates per treatment was also used for DWV quantification. To calculate the number of DWV genome copies per sample, primers (5′-3′ F: GGGTAACGTGCGACGTTTTA; R: GACGTAAAGGCGGTAGTTGC) specific for the DWV helicase were used 33 . PCR conditions were one cycle at 48 °C for 15 min, one at 95 °C for 10 min, 40 at 95 °C for 15 s and 60 °C for 60 s, followed by one cycle at 68 °C for 7 min. The reaction volume was 25 µl containing 2 µl template, 3 µl 200 nM primers,12.5 µl Maxima SYBR Green/ROX qRT-PCR Master Mix (2×) and 9.5 µl nuclease free H 2 O per sample. As a negative control, nuclease free H 2 O was included instead of cDNA, and a positive control from previously identified DWV positive bee samples by qRT-PCR were included in each qRT-PCR run. Calibration curves to convert Ct values to DWV genome copies were done using 300 bp gBlocks ® (Integrated DNA Technologies, IA, USA) that included the sequence of the forward primer, amplicon and reverse primer. The lyophilized gBlock ® was diluted with 20 µl of ds H 2 0 to obtain an initial concentration of 10 ng/µl that was used to make serial dilutions from 10 9 to 10 1 copies. Using a plot of Ct values versus DWV copy number (log 10 ), a linear equation was used to calculate the DWV genome copy numbers for each of the samples of interest 66 . Statistical analyses. To compare the proportions of the grooming classes, contingency tables using Chi 2 tests of independence with α of 0.05 were used, and adjusted residuals were calculated for post hoc analysis.
www.nature.com/scientificreports www.nature.com/scientificreports/ DWV quantities were tested with the Shapiro Wilk test and transformed to a base 10 logarithm (due to lack of normality) before being subjected to two-way ANOVAs and Tukey HSD tests with α of 0.05. The above statistical analyses were performed using R, version 3.4.3© (The R Foundation for Statistical Computing, 2017).
RNAseq was performed at Génome Québec Innovation (Montreal, QC, Canada). Library preparation was done using the NEB kit Illumina (San Diego, CA, USA) for poly(A)+ enriched RNA prepared with a KAPA kit (Roche, Mississauga, ON, Canada) according to the manufacturer's instructions. Sequencing was performed as 125 bp, paired-end reads using a HiSeq2500 v4 (Illumina, San Diego, CA, USA).
Bioinformatic analysis was performed at the Canadian Centre for Computational Genomics (C3G) (Montreal, QC, Canada). Sequence trimming was done with Trimmomatic software 67 . Read sets were aligned to a reference genome of the honey bee, Apis mellifera (ftp://ftp.ncbi.nlm.nih.gov/genomes/Apis_mellifera) (ver Amel_4.5) using STAR 68 . Aligned RNAseq reads were assembled into transcripts, and fragments per kilobase of exon per million fragments mapped (FPKM) was determined with Cufflinks 69 . Differential gene analysis (DGA) to identify differentially expressed genes (DEGs) was done using the DESeq R Bioconductor package 70 and edgeR Bioconductor package 71 . Transcript expression levels and test for significant differences (P < 0.05) was calculated with Cuffdiff 69 using the FPKM values.

Data Availability
The datasets generated from grooming behavior during the current study are available from the corresponding author on reasonable request. All other data generated or analysed during this study are included in this published article (and its Supplementary Information files).