Effects of immunostimulation on social behavior, chemical communication and genome-wide gene expression in honey bee workers (Apis mellifera)

Honey bees are an outstanding model system for studying the molecular, physiological and social basis of disease transmission and resistance. Honey bees are plagued by a number of parasites and pathogens (reviewed in[1]), and the social colony environment (with up to 50,000 densely packed worker bees[2]) provides excellent conditions for disease transmission. While the innate immune systems of invertebrates and vertebrates are surprisingly conserved[3], there can be large differences in the numbers of genes involved in the different molecular arms of the immune response system across species. As is the case with many insects with recently sequenced genomes[4–8], bees have a much smaller number of known canonical immune response genes relative to Drosophila[9], which is one of the best characterized models of insect immunity[10]. Thus, bees (and other insects) may utilize alternative genetic and physiological mechanisms to respond to infections. Furthermore, honey bees can resist pathogens and parasites by employing sophisticated behavioral defense mechanisms, termed “social immunity” (reviewed in[11]). Here, we examine the effects of a panel of general immune elicitors (injection with saline, Sephadex beads or bacteria) on worker-worker social interactions and chemical communication, and use whole-genome microarrays to characterize global gene expression responses to these immune elicitors.

Honey bee populations have been in decline worldwide, with beekeepers recently reporting massive annual losses (>30%) (reviewed in[1]). This decline is undoubtedly due in part to the multitude of parasites and pathogens that target honey bees, several of which have only recently been identified. Honey bees are host to over 20 viruses[12–14], as well as a number of bacterial and fungal pathogens[15–18] including the gut microspordian parasites Nosema apis and Nosema ceranae. Honey bees are also severely impacted by Varroa mites (Varroa destructor), tracheal mites (Acarapis woodi) and other ectoparasites[19]).

Sequencing of the honey bee genome identified 177 genes associated with the canonical immune response pathways in insects[9]. Innate immune pathways in insects, mostly obtained from studies in Drosophila, consists of both cellular and humoral responses, which can be systemic or local (reviewed in[10, 20–24]). Cellular immune responses involve a number of differentiated hemocytes. Pathogens activate phenoloxidase and associated immune cascades, resulting in phagocytosis, encapsulation, and/or melanization of invading organisms or wounds. Humoral responses include cytotoxic molecules (such as reactive oxygen and nitrogen species), lysozymes, cytokines, and antimicrobial peptides (AMPs). AMPs are primarily produced by the fat bodies in insects. A number of signal transduction pathways are involved in moderating immune responses, including the JAK/STAT (which seems to respond primarily to tissue damage), JNK (which mediates wound repair), Imd (which primarily regulates responses to Gram-negative bacteria), and Spaetzle/Toll (which generally regulates responses to Gram-positive bacteria and fungi) pathways. Activation of one of these signal transduction pathways may lead to up- or down-regulation of the others depending on the host-parasite system. For example, studies in Drosophila have shown that the relationships between immune pathways and effectors can be highly pathogen specific and much still remains to be characterized[25]. Furthermore, canonical pathways derived from Drosophila experiments, may not be fully generalizable to other systems.

Honey bees also employ a number of behavioral mechanisms – termed “social immunity” – to reduce the impacts of parasites and diseases (reviewed in[11, 26]). Grooming can remove ectoparasites, and hygienic behavior can reduce levels of brood diseases and depress Varroa mite populations (mites feed on developing pupae). Workers also coat the interior of their colonies with propolis (derived from plant sap), which has antimicrobial properties and results in lowered levels of immune response genes in individual bees[27]. Furthermore, diseased or otherwise stressed bees undergo accelerated behavioral maturation, moving from in-hive tasks to foraging, which removes them from the brood nest and colony, thereby potentially reducing transmission[28, 29]. Finally, studies in termites suggest that healthy nestmates can become “socially vaccinated” by exposure to infected nestmates, thereby improving their resistance to a pathogen[30]. Future investigations may uncover similar defense mechanisms in honey bees.

Honey bees can also use cuticular hydrocarbons as chemical cues to distinguish between healthy and immuno-stimulated adult nestmates, and respond differently to them[31]. Cuticular hydrocarbons are synthesized in oenocytes, which are embedded in the fat body tissue under the epithelia, and deposited on the cuticular surface[32]. Cuticular hydrocarbon patterns can be modified by genotype, physiological state, and environmental context (reviewed in[33, 34]), including social status in honey bees[35]. Immunostimulation of honey bee workers with lipopolysaccharides derived from bacterial cell walls caused significant changes in cuticular hydrocarbon profiles in worker bees after four hours, and resulted in altered social interactions[31]. Healthy bees were more aggressive towards their nestmates that were coated with extracts of immunostimulated nestmates than to nestmates that were coated with extracts of healthy nestmates, indicating that these changes in chemical profile could alter worker-worker interactions. Other studies have demonstrated that Varroa parasitized pupae and adults have modified cuticular hydrocarbon profiles[36], while infection with a virus can elicit aggressive reactions from healthy nestmates in a colony[37].

Here, we examined the behavioral, chemical and molecular responses in honey bee workers to injection with saline, Sephadex beads, and Gram-negative bacteria (freeze-killed E. coli cells), six hours after injection. We determined if these different immune elicitors stimulate unique responses at these three levels; for example, theoretically only E. coli injection should activate the Imd pathway, which could lead to distinct chemical profiles and behavioral responses. We also determined if immunostimulation resulted in significantly altered expression of previously annotated honey bee immune genes[9], if there was overlap with other studies of the effects of immunochallenge or parasitization in honey bees and Drosophila[38–41], and if there were changes in expression of genes associated with cuticular hydrocarbon biosynthesis pathways.

We have demonstrated that immunostimulation of adult honey bee workers results in altered social interactions only in the case of bacteria-injected bees. However, all the treatment groups have significantly different cuticular hydrocarbon chemical profiles, and all treatments caused large scale changes in gene expression patterns in abdominal fat body and epithelial tissue, which includes oenocytes. It is important to note that these dramatic changes occurred in a relatively short timeframe, only six hours after immunostimulation. Importantly, this study demonstrates that expression of a large number of genes, not just those canonically associated with immune response pathways, are modulated by immunostimulation, as has been demonstrated by several other studies of genome-wide responses to immune challenges (for example,[38–41]). However, comparisons with studies in Drosophila found limited overlap of gene expression patterns, and the handful of commonly regulated genes were key members of canonical immune response pathways.

It is somewhat perplexing that only bacteria-injected bees were subjected to significantly increased allo-grooming and aggression. All treatments caused changes in chemical profiles, and the changes did not appear to be substantially larger in bacteria-injected bees, though there was a trend for lower total quantity of branched alkanes in this group. However, previous studies have suggested that alkenes, rather than alkanes, play a significant role in nestmate recognition[55]. Based on the hierarchical clustering analysis, the gene expression patterns of bacteria-injected bees were clearly distinct from saline- and bead-injected bees. However, overall it did not appear that bacterial injection altered expression of an entirely unique set of genes (out of 302 transcripts, expression of 68 transcripts were significantly regulated only by bacterial injection, relative to controls, and expression of many of these genes changed non-significantly in the other treatments as well). Thus, perhaps bacteria injection did not cause change in expression of a large, unique subset genes, but rather caused more substantial changes in the magnitude of expression levels of some genes, as indicated by hierarchical clustering. Alternatively, it may be that expression of a few key genes and/or chemical profile components in bacteria-treated bees provoked behavioral changes in nestmates.

It remains to be determined if infection with honey bee specific parasites and pathogens causes similar changes in gene expression and behavioral responses, though there is some indication that infection with Nosema microsporidia does stimulate expression changes in a significantly overlapping set of genes (Holt, Aronstein and Grozinger, unpublished data). It also remains to be determined if these behavioral changes are adaptive: it is possible that bees have evolved to “match” the strength of the signal to the virulence of the pathogen, and thus infection with bacteria would elicit a greater change in the behavioral responses of nestmates than simple cuticular wounding. However, colony-level assays would need to be performed in order to determine if these changes in social interactions actually result in differences in the spread of pathogens through the colony and their impacts on infected individuals. For example, the increased grooming behavior we observed towards bacteria-injected individuals could facilitate the spread of pathogens through the colony (as observed in carpenter ants,[56]), slow the spread by causing isolation or removal of infected individuals[56, 57] or reduce the impact on the infected individual[58, 59].

Our studies also demonstrate that immunostimulation elicits complex gene expression changes in the epithelial tissues in honey bee workers. Immunostimulation resulted in significant expression changes of 302 common transcripts in worker bees from the two colonies examined. Only 14 of these genes corresponded to previously annotated immune genes identified from the honey bee genome[9]. Several other biological processes were modified, including cell growth and proliferation, cytoskelatal structure, metabolism and components of the Notch signaling pathway. Cell growth, proliferation, and migration, particularly involving actin-mediated cytoskeletal changes, are required for repairing epithelial wounds[24]. The Notch signaling pathway has not yet been linked to immune response or wound repair, but wound repair uses many of the same developmental pathways that function during dorsal closure in Drosophila development, which does involve Notch signaling[60]. Alternatively, since the insect fat body is involved in regulating many key processes, including metabolism (which was commonly regulated in our study and by Varroa parasitization[38]), these changes may reflect general physiological changes after immunostimulation or stress[61]. Expression changes in three genes of the Notch signaling pathway (apterous, groucho, and pebbled) were confirmed using quantitative real-time PCR. Interestingly, expression was significantly higher in saline-injected bees relative to controls for all three genes, but only expression of apterous was affected in bacteria-injected bees, suggesting that changes in Notch signaling may be modulated temporally or by other signaling pathways.

We found significant expression changes of a number of key immune response genes (for a review of the function of these genes, see[10, 23, 24]). The JAK/STAT pathway is regulated by the Domeless receptor; domeless expression was significantly regulated by immune stimulation in both genotypes of bees in our study. Activation of the IMD pathway requires cleavage of Relish by the caspase DREDD; both Dredd and Relish were significantly regulated in our study. We also observed significant regulation of PGRP-SC2 which suppresses activation the IMD pathway, with bacteria-injected bees showing the highest levels of PGRP-SC2 expression (data not shown). In fruit flies, PGRP-SC1 and PGR-SC2 may function in preventing over-activation of the IMD pathway[62]. The IMD pathway triggers the JNK pathway, which activates the transcriptional regulator AP-1 (which contains Kayak/D-fos), and AP-1 in turn negatively regulates Relish-dependent transcription. We found significant regulation of kayak in our study. Interestingly, we also found significant changes in expression of cabut, which is regulated by the JNK pathway but has not yet been linked to immune function[45]. The Toll pathway operates through transduction factors including spirit, which acts extracellularly and upstream of Spaetzle[46] and the NF-κB protein Dorsal (which is negatively regulated by cactus). We found significant regulation of spirit and cactus in our study. Pale, which plays an important role in melanization and wound repair, and draper, which functions in phagocytosis, were also significantly regulated in our study. As a Gram-negative bacteria, it would be expected that E. coli would primarily stimulate activation of the IMD pathway. However, we observed changes in gene expression of members of the Toll pathway, including cactus, spirit and defensin-1, which exhibits Gram-positive antimicrobial activity and is not regulated by the IMD pathway[63]. Thus, there is likely considerable cross-talk between the pathways.

Our studies also identified several genes which may play a role in altering cuticular hydrocarbon patterns. Cuticular hydrocarbons are synthesized primarily in the oenocytes (reviewed in[47]), which are embedded in the fat body of adult honey bees[64]. Cuticular hydrocarbon biosynthesis involves activation of fatty acids by an acyl-CoA synthetase, chain elongations of fatty-acyl-CoAs to produce very long chain fatty acids, and subsequent conversion to a hydrocarbon, likely by a p450 enzyme[47]. Fatty acids are stored in lipid droplets in the adipoctye cells of the insect fat body[61]. Fatty acids can be released from droplets in the adipocytes and accumulate in the oenocytes; this occurs under starvation conditions in particular[65]. This process is mediated in part by lipid storage droplet-2 (Lsd-2): increased Lsd-2 expression in the fat bodies decreases lipid movement to the oenocytes. We found increased expression of Lsd-2 in immunostimulated bees (see Figure6), suggesting reduced movement of lipids to oenocytes, and perhaps reduced levels of cuticular hydrocarbons. We did observe a decrease in the total relative quantity of all branched alkanes in bacteria-injected workers in both colonies but this difference was not significant.

Bubblegum (bgm) activates long chain fatty acids to form acyl-CoAs (reviewed in[66]), a key step in cuticular hydrocarbon biosynthesis. Bgm was originally described as a Drosophila mutant that resulted in elevated levels of very long chain fatty acids and neurodegeneration[48]. Bgm homologs have been identified in numerous species, including humans and mice, and have been demonstrated to activate long chain (C16) and very long chain (C24) fatty acids[67]. In our study, bgm expression was significantly decreased relative to controls (see Figure6).

Despite large differences in study designs and analysis methods, we found some overlap in gene expression with previous studies examining the effects of Varroa mite parasitization on honey bees[38, 40]. Varroa-responsive genes were significantly associated with basic cellular processes, including cell organization, biogenesis and metabolism. We also found significant changes in functional categories associated with basic cellular processes, such as cell growth, proliferation and cytoskeletal structure. Varroa parasitization also caused changes in expression of pale. As discussed above, this may represent cellular mechanisms for wound-healing. Expression of potential hydrocarbon synthesis genes, namely bgm and Lsd-2, were also regulated by Varroa parasitization. Indeed, Varroa parasitized pupae and adults have modified cuticular hydrocarbon profiles[36]. These differences are likely responsible for stimulating hygienic behavior, in which diseased larvae are removed by adult worker bees, a key component of Varroa resistance[11].

Comparison with two previous studies[39, 41] examining the effects of immunostimulation on Drosophila global gene expression patterns revealed conserved changes in expression of key immune genes in Drosophila and honey bees (including relish, cactus, defensin-1, spirit, PGRP-SC2, and pale), but otherwise limited overlap in the significantly regulated genes. The lack of similarity could represent species-specific immune responses or simply technical differences – for example, Roxstrom-Lindquist[41] orally infected young male flies with bacteria, fungi and microsporidia, and measured whole-body gene expression changes in only two replicates using Affymetrix microarrays.

Our results suggest that immunostimulation of honey bee workers causes significant changes in gene expression patterns, cuticular hydrocarbon chemical profiles, and, in the case of bacteria-injection, social behavior. While bacteria-injection did not cause expression changes in an entirely unique set of genes or production of unique cuticular hydrocarbons compared to other treatment groups, it does appear that the magnitude of the expression and production changes was greater, and this may have resulted in the altered behavioral responses. As demonstrated by other studies examining genome-wide expression changes associated with immune challenges[38–41], we found that even a short-term immune challenge can result in dramatic changes in gene expression that encompass far more genes than those represented by canonical immune response pathways. For example, we found several genes associated with the Notch signaling pathway, which suggests this pathway may also play a role in mediating immune responses. Furthermore, our study has highlighted potentially new candidate genes for regulating cuticular hydrocarbon synthesis in insects. These chemicals serve many functions in insects, including operating as sex and caste-specific pheromones, and most research on their biosynthesis has focused on desaturase enzymes and p450s[34]. These studies lay the groundwork for future work examining the molecular pathways that mediate immune responses to acute stimulators and chronic infections in bees and other insects, and has paved the way for research into the genes that regulate social immunity.