Behavioral defenses are “strikingly effective” mechanisms for combating parasites but are often overlooked by studies addressing the effects of a parasite on the host immune system and related physiological processes 1 . For a behavior to be considered as an anti-parasite defense the parasite must impose a cost on the host and the behavior should limit or eliminate the parasite 2 3 . Parasite avoidance behavior is found across animal taxa, where it manifests as behaviors such as cleaning and avoiding feces or avoiding diseased conspecifics 4 . For example, normally gregarious spiny Caribbean lobsters shun lobsters that carry a lethal virus, even before the diseased lobster is visibly infected 5 . However, not all behaviors are anti-parasite, as parasites can also alter the behavior of their hosts in ways that are ultimately beneficial to the parasite or its offspring 6 7 .

Living in groups can further complicate host-parasite interactions since group members can also play a role in the regulation of anti-parasite behavior. Collective defenses against parasites within a social group, called social immunity, are physiological, behavioral or organizational adaptations that prevent the transmission of the parasite 8 9 . A common social defense is the removal of the parasitized individuals from the social group. Indeed, in social insects, parasitized individuals can remove themselves from the group, which corresponds to an altruistic self-removal 10 , or nestmates can modify their interaction with those individuals to prevent parasite transmission 6 9 . For example, in honey bees, very different types of stress, including exposure to the mite Varroa 11 12 , the microsporidia Nosema ceranae 13 14 , or immune challenge 15 have been shown to induce precocious foraging or forager-like physiological and behavioral characteristics. Leaving the colony to perform outside activities, like foraging, limits contact in the hive, especially with castes of great importance (e.g. queen, brood), and thus, the spread of parasites into the colony 9 . However, despite parasite modification of host behavior being widespread in animals, the underlying physiological and neuronal mechanisms are poorly understood, especially in social insects.

In order to better understand this phenomenon, we analyzed how two different parasites, an ectoparasite (Varroa destructor) and endoparasite (Nosema ceranae), affect the physiology and the brain neurogenomic state of honey bees. The Varroa destructor mite infects the larval cell immediately before capping where it feeds on the developing pupae and completes its reproductive cycle. It weakens the honey bee by feeding on its hemolymph and transmitting viruses, such as deformed wing virus (DWV), which are correlated with its effect on honey bee survivorship 16 17 18 . Nosema species are obligate, intracellular spore-forming fungal parasites that infect honey bee adults from emergence by spreading through the hive, most likely through the activities of cleaning and trophallaxis 19 . Once a worker has ingested Nosema spores, the spores develop in the intestine of the bee, where the germinated microsporidian infects the epithelial cell layer of the midgut and consumes the energy of the cell 20 21 . However, despite Varroa and Nosema infections varying greatly in their pathologies, they affect honey bee behavior and learning abilities in similar ways (see Table 1). For example, both Varroa-infested and Nosema-infected bees showed impaired orientation abilities at the hive entrance 22 23 .

Since there is a robust association between brain gene expression in the honey bee and its behavioral state 29 32 33 , we measured the brain transcriptional changes induced by V. destructor and N. ceranae and determined whether they induce similar brain host responses. We also characterized the cuticular hydrocarbon (CHC) profiles of honey bees parasitized by V. destructor or N. ceranae and recorded nestmate interactions in observation hives in order to detect nestmate aggression toward parasitized bees. Indeed, challenging the immune system of bees with lipopolysaccharides or other non-living immune stimulants changed the CHC profiles of the bees, involved in social recognition, which lead to modified and aggressive conspecific contacts in a laboratory-based nestmate recognition assay 34 35 . Bees infected with the virus DWV, that showed changes in their CHC profiles, were also ejected from the hive at higher rates than healthy bees, notably from healthy hives 36 . Finally, we determined whether parasitism can affect the level of production of 10-HDA that contributes to social immunity of the colony. This compound, produced in the mandibular glands of bees, displays antiseptic properties in the royal jelly 37 .

In this study, we demonstrated that two parasites, Varroa destructor and Nosema ceranae, with distinct, differing pathologies both modified the physiology and transcriptomic profiles in the brain of their honey bee host. Parasitized honey bees exhibited changes in their CHC profiles but showed no differences in behaviors between parasitized and healthy bees. In addition we observed no significant aggressive behavior towards parasitized bees, nor any change in social interactions.

A previous study found that bees parasitized by Varroa exhibit a modified CHC at their emergence 42 . Our data shows that this change is lasting. However, our behavioral results differ from recent studies that observed increased general social interactions or aggressive behaviors towards immune-challenged bees, but differences in the experimental design of our study may account for those differences. In a series of studies, increased social and aggressive behaviors towards immune-suppressed workers were observed a few hours after treatment in assays that were conducted in the laboratory 34 35 . Our study employed natural conditions of a four-frame observation hive using bees that had been parasitized in the days prior to the experiment, in order to understand if nestmates respond to parasitized bees within the context of general activity of the hive. Moreover we chose to focus on bees that were Varroa-parasitized but asymptomatic for DWV. In contrast, a second study found that DWV-infected bees that exhibited deformed wing symptoms were detected and removed from the hive by nestmates 36 . Thus our results do not contradict previous studies but reflect the subtle nature of parasitism by Varroa or Nosema that, when resulting in precocious departure from the hive, is more likely due to altruistic self-removal, acting as a mechanism of social immunity. Indeed, it may be less costly for the colony that sick or parasitized bees leave the colony of their own accord, rather than recruiting nestmates to exclude those bees via aggressive behaviors. In that case, bees might distinguish sick bees based on different CHC profiles but not discriminate them, except in the case of extremely sick bees that cannot leave on their own, such as bees exhibiting deformed wing symptoms. We also found a change of the chemical profile with age which is consistent with previous studies (nurse vs forager, see 43 ). In honey bee hives, older bees segregate themselves from young bees by olfactory discrimination of cuticular hydrocarbons as they correspond to different age groups 44 . Indeed, older bees both emit and respond to a more complex bouquet of cuticular hydrocarbons than younger bees 43 44 . Since Nosema and Varroa-parasitized bees age faster, it is possible that they exhibited a CHC profile of old bees. In addition, the CHC profile is shaped by the genotype, nutrition, environment and physiological state 45 46 . Therefore, it is possible that their nestmates did not respond to the parasitized bees because their chemical profiles could not be distinguished from the chemical profile of others bees of different ages, physiological status and genotypes. These results highlight the importance of testing for biological effects within the hive when trying to draw conclusions about honey bee behavior.

If parasitism by Varroa or Nosema induces precocious foraging, one would expect the parasitized bees to show physiological changes similar to the transition to a forager bee. Levels of 10-HDA increased with the age of the bee confirming the study of Plettner et al. 47 , but did not change in response to Nosema or Varroa parasitism. Thus the production of antiseptic compounds, like 10-HDA, in the food is age- or task-dependent but not regulated by the presence of parasites. However, further investigation of different type of pathogens or parasites, would give more insight as to whether antiseptic production can vary according to the infection level of the hive.

Our results also demonstrate that parasites alter the brain of the honey bee host, whether they were parasitized at the pupal (Varroa) or the adult stage (Nosema). In addition, twenty genes, nearly one half of those detected in Nosema-infested bee brains, show a shared expression pattern between Varroa and Nosema-infested bees. Their functions are diverse but several genes stand out as interesting for their possible roles in oxidative stress, neural function and foraging behavior. Flavin-containing monooxygenase FMO GS-OX-like 3-like (FMO3) and torsin-like protein (torp4a) are overexpressed and replication factor C subunit 5-like (RfC5) is downregulated in Varroa-infested and Nosema-infected bees. FMO3 is part of the FMO family known to react to xenobiotic stress in other organisms 48 . In Drosophila, the Torp4a ortholog (dtorsin) is involved in dopamine metabolism and locomotion 49 and the RfC5 ortholog (RfC3) plays a role in neurogenesis 50 , indicating that both parasites can modify brain function. The honey bee gene, Pheromone biosynthesis-activating neuropeptide (PBAN) is also over-expressed in both Varroa-infested and Nosema-infected bees compared to controls. In honey bees, PBAN neuropeptide levels are significantly higher in nectar foragers than pollen foragers 51 . In Lepidoptera, PBAN is linked to regulation of sex pheromone production 52 , where pheromone production is JH-dependent, and JH primes the pheromone gland in adult females to respond to PBAN 53 . JH is also responsible for “priming” the foraging behavior in honey bees, though no link has been established between PBAN and JH in honey bees.

The presence of viruses may also account for similarities in gene expression between Nosema and Varroa-parasitized bees. We found increased levels of DWV in the brains of both types of parasitized bees and therefore cannot exclude that the observed changes are actually caused by an increase in DWV titer. DWV is a positive-strand RNA picorna-like virus that has been detected and can actively replicate in the heads 54 and brains of honey bees, specifically the mushroom bodies, visual and antennal lobe neuropils 55 . The virus is closely associated with Varroa infestation 56 and thus it was not unusual that it occurred in higher levels in Varroa-infested brains. On the other hand, we did not expect to find an increase in DWV levels in N. ceranae-infected bees, as a negative correlation between these two pathogens in the midgut of the bee was recently reported 57 , which suggests that N. ceranae and DWV may compete for resources in the degenerated midgut, but not in the brain, where Nosema is not found.

Despite the statistically significant number of shared genes that change, Varroa and Nosema-infested brains demonstrate different patterns of expression that may reflect the different pathologies of the two parasites. Bees that were parasitized by Varroa as developing pupae exhibit more gene changes compared to controls than bees that were inoculated with Nosema ceranae as one-day old adults. This apparent disparity in gene expression changes may be due to long-lasting brain developmental changes induced during pupal development that persist in adult bees. The genes affected by N. ceranae infection could not be sorted by functional group analysis but several immune-related and antioxidant genes, including defensin-1, peroxidase, esterase A2, glucose oxidase, were upregulated indicating that the blood–brain barrier in honey bees, although not well studied, may be compromised by a parasitic attack. Genes involved in the oxidative response to stress were also upregulated in the guts of N. ceranae-infected honey bees 58 , suggesting a systemic response throughout the honey bee in response to the microsporidian.

The impact of Varroa on the brain transcriptome suggests a decrease in learning and memory that may result from parasitism during development. This brain response would explain the actual learning impairment and losses of foragers induced by Varroa 23 59 . Based on functional group analysis, Varroa-infested bees show decreased expression of glutamate and GABA receptor-related genes, the dopamine receptor, Amdop1, and overexpression of ascorbate/aldarate metabolism genes. Inhibition or suppression of glutamate receptors disrupts memory formation in honey bees 26 30 60 . GABA receptors are present throughout the mushroom bodies 61 , a region important for learning and memory, especially in foragers 62 . GABAnergic interneurons also form part of the olfactory conditioned learning pathway 63 . Yet the simultaneous increase in the expression of glutamate decarboxylase and GABA neurotransmitter transporter with a decrease in GABA receptor targets signals either compensatory mechanisms at work or a disruption in GABAnergic network. Analysis of the neuroanatomical changes in the Varroa-parasitized brain could resolve whether decreased expression of GABA and glutamate receptors leads to a reduction in their numbers. The dopamine receptor Amdop1 is higher in newly born cells in the mushroom body than older cells 64 and in Drosophila, it is required for aversive and appetitive learning 65 . Finally, the cAMP pathway and its targets in the mushroom bodies are important mechanisms for learning in bees 66 . Several genes linked to the cAMP and calcium signaling cascades are downregulated in Varroa brains: Adenylate cyclase type 10-like, Ryanodine receptor 67 and voltage-dependent calcium channel subunit (GB10696).

Compared to the transcriptomic changes in the honey bee brain that accompany the switch from nurse to forager, we found relatively few genes that changed in response to Nosema ceranae or Varroa destructor infestation. Brain expression profiles of Varroa and Nosema- parasitized bees bear a greater resemblance to each other than to the reported profiles of typical foragers or nurses. Thus, their early departures from the hive may not be induced by mechanisms of normal behavioral development, but perhaps by an alternative mechanism that results in altruistic self-removal. Indeed, certain genes that are typically upregulated in foraging bee brains (Inos, Kr-h1) are downregulated in Varroa-infested honey bee brains 29 41 68 69 . Foraging activity is an especially demanding activity for learning and memory in the honey bee 63 , but parasitized bees seems to have deficiencies at this level (see above). Therefore, altogether these results suggest that Varroa and Nosema-parasitized bees seem to not be true foragers, much like CO₂-treated bees that left the hive, but also disappeared, at higher rates than control bees 10 . However, this will require confirmation in a more natural context (colony level).

Neither of the parasites, Varroa destructor nor Nosema ceranae, attacks the honey bee brain directly, yet we observed transcriptional changes in the brains of honey bees in response to parasitism. Thus, these changes, that are most likely triggered by a reaction in another tissue (e.g. midgut, fat bodies, hemolymph), highlight a link between the immune system, the brain, and perhaps, behavior in the honey bee. While parasitized bees are reported to behave like foragers, by leaving the hive, their brain transcription profiles suggest that their behavior is not driven by the same molecular pathways that induce foraging behavior. Whether the transcriptional changes observed are due to host immune response, parasite protein release or viruses that propagate in the brain is not known. LPS-challenged bees also behave more like foragers than same-aged bees 15 , even without parasitic or viral challenges, but proteomic analysis of parasitized insects, grasshopper (by a nematode) and tsetse fly (by Trypansoma brunei) detected changes in the host brain 70 71 and proteins released by the parasite that may affect host behavior 70 .

Stress response in the honey bee to parasitism by Varroa destructor or Nosema ceranae shows similarities in their features; both parasites induce changes in CHC profiles and similar transcriptional profiles in the brain. That these parasitized bees are not attacked by their nestmates suggests that they leave the hive voluntarily, perhaps propelled by gene expression changes in the brain, showing altruistic behavior as predicted by Rueppel et al. 10 . This social removal may be a general and conserved response to parasitism, given that it was observed with extremely different types of parasites: a mite (ectoparasite) 11 12 and a single cell microsporidian (endoparasite) 13 14 . As to what these bees do once they have left the hive still needs to be examined. Are they normal foragers but with shorter life spans, less efficient foragers due to learning and memory deficiencies or do they leave the hive and wander aimlessly in the landscape? Emerging tracking technology will allow us to answer these questions and determine the role of the parasitized honey bee within the colony. In addition, such studies that incorporate behavioral, genomic and physiological components will help us to better understand current worldwide declines of honey bee populations that are often characterized by an unusual loss of adult bees from the colony 72 .

The authors have declared no competing interests.

YLC, CMM and CA conceived of the study and its experimental design. CMM, DC and CA conducted the experimental protocol with the honey bees, ED and DB performed the chemical analysis, ED and CA performed the behavioral observations, CMM performed brain dissections and RNA isolation, HP and JPD carried out the sequencing and statistical analysis of sequencing data, CMM and CA analyzed and interpreted the data, and CMM, CA and YLC drafted manuscript. All authors read and approved the final manuscript.