Ma et al BMC Genomic (2019)20592 Page 10 of 15 [36,52].Whitfield et al.[52]compared nurses and for- ontogeny of foraging behavior [60].Pankiw et al. ers,cont or8eand exposure to increased poller on ge and found more ging to nollen forag which the authors found than 200 DEGs betweer age-matcl ed bees that were ex explanation may posed to usly for multiple days that short exposur an se tha t were notx are ake the ora tudies we comr ed 1)the DEGs betw e which may be a way to buffer aga inst ephemeral swing oragers in our sti ly with those identified by Whit t al emands of dev loping larvae 52.an the I A mo ated to b We e of Whitfield et al (P<001)and of Alau et al (P. but not epige netic pathways.which sug sts that metabol )The een our st and the sign may play that for s time scales (see 59)),and sup tore and changes in insulin signaling [61].Thes ports the idea hat ogical changes during the tion from in-hIv lar of fo references for nectar s genes foragin spec bu olism lipid signa ing path study tha 41 bel es been shown to elicit astrated the imp rtance of brain oli ollen agers.For exam ple, exp on influe individual variation in be old 42 ony-l enr and the individual e ollen for [44]How ng nath way in our ever,prior to this study,there were no docu nted im orts the role of sulin signaling pathways in mediating pacts of exposure to brood phe mo on ne inct and or in insects 64,65].For ex ommo t of DEC necialization (Table 3)which was driven primarily nd is related to FOXO signalin DEGs in nectar foragers but not pollen forag ers (Tabl Module3 was enriched for FOXO signaling and significantiy )Hierarchical clust ring analysis showe that,for the EBO treatment,so its hul genes may serve P s were nto po nt s nt he im on phe ectar foragers exp sed to EBO had expre ofiles ion and fora that were more like those of pollen foragers (Table 4) Although the results of this study are consistent with PC that ne rage exp erpre that ph tion may other n tar fora 3)The on which this in vely small revealed that two modules were associated with The consistency of the expression differences between our pheromon treat ent and foraging, one of which a tudy and previous studies 36 52],the patterns obtained ane compe ent rgy m he upe A g.hi stering ism hy which larval nheromone modulate colon indicate that our data ma reveal biologically meaningfu pollen foraging behavior could be by downregulating patterns despite the small sample size.However,future metabolic path role that energy metabolism plays in variation among foragers[36, 52]. Whitfield et al. [52] compared nurses and for￾agers, controlling for age, and found over 1000 DEGs. Alaux et al. [36] were the first to study the effects of brood pheromone on gene expression, and found more than 200 DEGs between age-matched bees that were ex￾posed to BP continuously for multiple days (i.e., five or 15 days) and those that were not exposed. To test the de￾gree of overlap between our results and those from previous studies, we compared 1) the DEGs between nectar and pollen foragers in our study with those identified by Whitfield et al. [52], and 2) the DEGs between pheromone treatments in our study with those identified by Alaux et al. [36] . We found sig￾nificant overlaps between DEGs identified in our results and those of Whitfield et al. (P < 0.001) and of Alaux et al. (P < 0.001). The significant overlap between our study and the two microarray studies, which validate the expression patterns re￾lated to foraging specialization and brood pheromone expos￾ure, suggests that foraging-related gene expression shows a degree of consistency across time scales (see [59]), and sup￾ports the idea that pheromones regulate the transcriptional pathways underlying foraging specialization. Our data supported the hypothesis that exposure to larval pheromones alters expression of foraging related genes depending on foraging task specialization, but contrary to our prediction, the pheromones had more pronounced effects on gene expression in nectar for￾agers than pollen foragers (prediction 4). Larval phero￾mones have been shown to elicit specific responses in pollen foragers. For example, exposure to brood phero￾mone (BP) increased colony-level pollen foraging 2.5 fold [42], the ratio of pollen to non-pollen foragers [44], and the individual effort of pollen foragers [44]. How￾ever, prior to this study, there were no documented im￾pacts of exposure to brood pheromones on nectar foraging. There was a common set of DEGs that were associated with both pheromone treatment and foraging specialization (Table 3), which was driven primarily by DEGs in nectar foragers but not pollen foragers (Table 4). Hierarchical clustering analysis showed that, for the most part, samples were clustered into pollen and nectar foraging “branches,” with the intriguing exception that nectar foragers exposed to EBO had expression profiles that were more like those of pollen foragers (Table 4). Similarly, PCA showed that nectar foragers exposed to EBO clustered more closely with pollen foragers than other nectar foragers (Fig. 3). The gene network analysis revealed that two modules were associated with both pheromone treatment and foraging, one of which was enriched for membrane components and energy metab￾olism (Table 8). These results suggest that one mechan￾ism by which larval pheromones modulate colony-level pollen foraging behavior could be by downregulating metabolic pathways in the nectar forager brain, which is consistent with the role that energy metabolism plays in the ontogeny of foraging behavior [60]. Pankiw et al. [44] found that short exposure to BP increased pollen foraging, but did not observe task-switching of nectar foragers to pollen foraging, which the authors found puzzling. Our results indicate that one explanation may be that even after short exposures to larval pheromones, nectar foragers are primed to switch to pollen foraging even before they actually make the behavioral transition, which may be a way to buffer against ephemeral swings in the nutritional demands of developing larvae. DEGs and WGCNA modules related to both pheromone treatment and foraging specialization were enriched for sev￾eral metabolic pathways, including fatty-acid metabolism, but not epigenetic pathways, which suggests that metabolic processes and lipid signaling in integration centers of the honey bee brain may play a role in behavioral plasticity. The transition from nursing to foraging involves large-scale changes in metabolic pathways, including reductions in lipid stores and changes in insulin signaling [61]. These physio￾logical changes during the transition from in-hive tasks to foraging are associated with changes in energy metabolism (including insulin signaling), gustatory response, and foraging preferences for nectar vs pollen [62, 63]. Therefore, the prominence of energy metabolism, lipid signaling pathways, and related metabolic pathways in our study’s brain tran￾scriptome data supports the idea that these pathways in the brain play a role in insect behavior [64, 65]. Other studies have demonstrated the importance of brain metabolic pro￾cesses on influencing individual variation in behavior, par￾ticularly aggression [65–67]. The enrichment of metabolic pathways in DEGs and the prominence of the FOXO signal￾ing pathway in our gene co-expression networks further sup￾ports the role of insulin signaling pathways in mediating neuronal function and behavior in insects [64, 65]. For ex￾ample, an insulin binding protein, Queen brain-selective protein-1 (Qbp-1), was differentially expressed in response to pheromone treatment and is related to FOXO signaling. Module 3 was enriched for FOXO signaling and significantly correlated with EBO treatment, so its hub genes may serve as useful candidate genes for subsequent studies investigating the impact of insulin signaling on pheromone communica￾tion and foraging. Although the results of this study are consistent with the interpretation that pheromone communication may possibly regulate foraging task specialization, the sample sizes on which this interpretation rests are relatively small. The consistency of the expression differences between our study and previous studies [36, 52], the patterns obtained in the PCA, and the results of the unsupervised clustering strategies (WGCNA & hierarchical clustering) all serve to indicate that our data may reveal biologically meaningful patterns despite the small sample size. However, future studies will be required to assess whether any confounding factors—such as individual variation among foragers, Ma et al. BMC Genomics (2019) 20:592 Page 10 of 15