The Pennsylvania State University. The Graduate School. Entomology Department HONEY BEES AND INTESTINAL DISEASE:

Transcription

1 The Pennsylvania State University The Graduate School Entomology Department HONEY BEES AND INTESTINAL DISEASE: MOLECULAR, PHYSIOLOGICAL AND BEHAVIORAL RESPONSES OF HONEY BEES (APIS MELLIFERA) TO INFECTION WITH MICROSPORIDIAN PARASITES A Dissertation in Entomology by Holly L. Holt 2015 Holly L. Holt Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy August 2015

2 The dissertation of Holly L. Holt was reviewed and approved* by the following: Christina Grozinger Professor of Entomology, Director of The Center for Pollinator Research Dissertation Advisor Chair of Committee Diana Cox-Foster Professor of Entomology Kelli Hoover Professor of Entomology James H. Marden Professor of Biology Gary W. Felton Professor of Entomology Head of the Department of Entomology *Signatures are on file in the Graduate School

3 iii ABSTRACT Pollinators are integral to modern agricultural productivity and the continued survival and vitality of natural ecosystems. However, recent declines in pollinator populations and species diversity threaten both food security and the architecture of natural habitats. Due to their vital role in agriculture, honey bees (Apis mellifera) have served as a model organism for investigating the alarming and widespread diminution of pollinator populations. Indeed, surveys from both North America and Europe report large annual colony losses. Parasites along with chemical exposure, poor nutrition, climate change and habitat destruction are frequently cited as leading causes of colony loss. Honey bee colonies are assaulted by a battery of bacterial, fungal and viral pathogens in addition to other parasitic arthropods including mites and beetles. Novel, costeffective disease management practices are desperately needed to preserve colony health. Basic studies investigating honey bee immunity and disease pathology lay the groundwork for developing efficacious diagnostic tools and treatments. Here, we present a series of studies characterizing honey bee immunity and the molecular, physiological and behavioral responses of honey bees to two important fungal pathogens, Nosema apis and Nosema ceranae. Chapter 1 reviews the current state of research on these prevalent and destructive disease agents and highlights future studies that are needed to develop effective management practices. Chapter 2 investigates worker honey bee genomic responses to general immune stimulation. Findings from these experiments provide a contextual framework for Chapter 3 s studies which characterize worker honey bee genomic responses to infection with Nosema parasites and offer a molecular model for explaining previously documented disease symptoms. Chapter 4 investigates drone (male) honey bee molecular, physiological and behavioral responses to Nosema infection and underscores potential caste-specific responses to infection that have larger implications for colony fitness. Finally, Chapter 5 summarizes novel findings from this dissertation, integrates results with current scientific literature and discusses the future of Nosema management.

4 iv TABLE OF CONTENTS List of Figures... viii List of Tables... x Acknowledgements... xi Chapter 1 Towards an integrated pest management (IPM) approach for Nosema parasites in honey bee (Apis mellifera) colonies... 1 Abstract Introduction Nosema parasites and noseomosis Nosema apis and Nosema ceranae are microsporidian parasites of honey bees Nosema spp. global distribution and factors affecting virulence in managed honey bee colonies Nosema spp. global distribution in native pollinator populations: Nosema spp. pathology: consequences for individual workers and honey bee colonies Nosema spp. pathology: consequences for adult queens and drones and immature castes Nosema virulence factors and host defense mechanisms Methods for diagnosing Nosema infections in colonies Light microscopy Molecular techniques ELISA (Enzyme-Linked Immunosorbent Assay) test Measuring acoustic or odor signatures of colony stress Challenges with establishing treatment thresholds Sampling colonies for Nosema infection Setting EILs (Economic Injury Levels) Currently available and potential future Nosema treatments Chemical treatments Hive sterilization methods Colony management practices... 18

5 v 6.0 Summary and future directions Box 1: Integrated Pest Management Box 2: Microsporidia Box 3: Honeybee colony demographics and division of labor (see [167] for a review) Acknowledgements Chapter 2 Effects of immunostimulation on genome-wide gene expression in honey bee workers (Apis mellifera) Abstract Introduction Methods Honey bee stocks Honey bee rearing Experimental treatment Microarrays Analysis of Gene Expression Validation of candidate gene expression patterns using quantitative realtime PCR Results Global gene expression responses to immunostimulation Effects of individual immune elicitors on gene expression Functional analysis of regulated genes Comparisons of gene expression patterns with previous studies Quantitative real-time PCR validation of expression of candidate genes Discussion Conclusions Acknowledgements Chapter 3 Chronic parasitization by Nosema microsporidia causes global expression changes in core nutritional, metabolic and behavioral pathways in honey bee workers (Apis mellifera) Chapter 4 Molecular, physiological and behavioral responses of honey bee (Apis mellifera) drones to infection with microsporidian parasites... 43

6 vi Abstract Introduction Methods Colony management and drone samples Nosema spore isolates Drone infection Impact of Nosema on drone sperm count Impact of Nosema infection on drone electroantennogram (EAG) response to 9-ODA Impact of Nosema infection on drone metabolic activity during flight Impact of Nosema infection on drone starvation rate Impact of Nosema infection on drone flight behavior Impact of Nosema infection on drone gene expression Infection confirmation Statistical analyses Results Impact of Nosema on drone sperm count Impact of Nosema infection on drone EAG response to 9-ODA Impact of Nosema infection on drone metabolic activity during flight Impact of Nosema infection on drone starvation rate Impact of Nosema infection on drone flight behavior Impact of Nosema infection on drone gene expression: Nosema infection confirmation: Discussion Physiological assays Flight assays Molecular assays Infection type and conclusions Box 1: Drone biology Box 2: Adult drone sexual maturation and flight behavior Acknowledgements Chapter 5 Conclusions and future directions... 81

7 vii Appendix A Reprint of Chapter Appendix B Supplementary Tables and Figures for Chapter Appendix C Supplementary data for Chapter References

8 viii LIST OF FIGURES Figure 1-1 Nosema spp. spores in whole abdomen homogenates Figure 2-1 Hierarchical clustering of significantly regulated genes Figure 2-2 Effects of specific immunostimulants on gene expression Figure 2-3 Quantitative real-time PCR validation of expression patterns of candidate genes Figure 4-1 Modified runway for drone observation experiments Figure 4-2 Drone electroantennogram (EAG) response to the queen sex attractant 9- ODA Figure 4-3 Maximum (A) and average metabolic (B) rates produced by drones over 4 minutes of consecutive flight Figure 4-4 Cumulative drone mortality during starvation assays Figure 4-5 Average spore counts for all surviving infected drones collected at the end of each observation experiment Figure 4-6 Average flight duration for all completed drone flights Figure 4-7 Average flight duration for all long ( 12 minutes) drone flights Figure 4-8 Average number of long flights ( 12 minutes) taken by drones in each treatment group that took at least 1 long flight during each experimental trial Figure 4-9 Daily cumulative percent of drones taking their first flight out of all drones that completed at least one flight Figure 4-10 Interaction between drone infection status, flight experience and flight length Figure 4-11 Effects of Nosema infection on drone gene expression patterns Figure B-1 Average flight duration for all short (<12 minutes) drone flights Figure B-2 Average inter-flight duration between all consecutive drone flights Figure B-3 Average inter-flight duration between all consecutive long ( 12 minutes) drone flights Figure C-1 PCR confirming N. apis and N. ceranae infection status in a subset of drones (Chapter 4) and workers (Chapter 5) from molecular experiments

9 Figure C-2 Effects of Nosema infection on worker gene expression patterns ix

10 x LIST OF TABLES Table 2-1 Primer sequences used to quantify expression of candidate genes using quantitative real-time PCR Table 2-2 Analysis of overlap among treatment groups Table 2-3 Overlap of gene lists between studies Table 4-1 Primers used for drone fat body gene expression analysis Table 4-2 Number of returned drift events (when a drone left the observation colony and returned on a later date) and the percent of returned drift flights taken by infected drones per experimental trial Table 4-3 Number of returned flights and the number of days each experiment lasted per trial Table 4-4 The percent of drones that completed at least 1 flight (any duration) and at least 1 long flight ( 12 minutes) per treatment group and experimental trial Table 4-5 Percentages of drones surviving to the end of each experimental trial Table B-1 Average flight duration and flight number across observation trials Table B-2 Model estimates and (standard errors) for transformed flight lengths Table B-3 Model estimates and (standard errors) for transformed inter-flight duration Table B-4 Model estimates and (standard errors) for transformed flight rate

11 xi ACKNOWLEDGEMENTS I am extremely grateful to my committee and many, many colleagues, friends and family members for their encouragement and support. Christina, Diana, Kelli and Jim, thank you so much for your mentorship and guidance! To all members (past and present) of the Grozinger Laboratory, I already do and will continue to miss your fabulous company and all those summer afternoons with you and the bees. I am also deeply appreciative of the academically challenging and supportive environment fostered by the excellent faculty, staff and students of the Penn State Entomology Department. Sally, Stephanie, Sarah, Tugba, Deniz, Erin, Scott and Susan, thanks for cheering me along when the goings got tough, and Becky, thank YOU for fixing Microsoft Word s formatting. And, Mom, Dad, Brendan and Emily, thanks for always being there. This dissertation is dedicated to pollinators, without which I would not have had the chocolate and coffee needed to finish.

12 1 Chapter 1 Towards an integrated pest management (IPM) approach for Nosema parasites in honey bee (Apis mellifera) colonies Holly L. Holt and Christina M. Grozinger Department of Entomology, Center for Pollinator Research, Huck Institutes of the Life Sciences, Pennsylvania State University, University Park, PA, USA Abstract Nosema apis and Nosema ceranae are common intestinal parasites in honey bee (Apis mellifera) colonies. Though globally prevalent, there are mixed reports of Nosema infection costs, with some regions reporting high parasite virulence and colony losses. To effectively manage these parasites, beekeepers may implement Integrated Pest Management (IPM) strategies, which employ diverse, complementary and cost-effective control tactics. Before an IPM plan can be successfully deployed against Nosema spp., however, basic and applied studies are urgently needed. Beekeepers need novel and affordable technology that facilitates disease diagnosis and science-backed guidelines for when to treat. Furthermore, new treatment methods are needed, as there are several problems associated with the chemical use of fumagillin (the only currently extensively studied, but not globally available treatment) to control Nosema parasites. Though selective breeding of Nosema-resilient bees may offer a long-term, sustainable solution to Nosema management, other treatments are needed in the interim. Furthermore, the validation of alternative treatment efficacy in field settings is needed in addition to toxicology assays to ensure that treatments do not have unintended, adverse effects on honey bees and/or humans. Finally, given variation in Nosema virulence, development of regional IPM guidelines, rather than universal IPM guidelines may provide optimal and cost-effective Nosema management. 1.0 Introduction Despite heroic efforts, beekeepers lose a large percentage of their colonies every year, with average overwintering losses hovering near 30% in the US and first reports of total summer losses (2014) nearing 25% [1]. Several factors are contributing to this high mortality rate, including pesticides, parasites, pathogens and poor nutrition. Many of these stressors act synergistically to undermine bee health [2, 3]. However, because these stressors are widespread, difficult to diagnose and can have long-term sublethal effects, it is frequently hard for beekeepers to know when and why colonies are experiencing stress. Furthermore, the thresholds at which many of these stressors seriously damage colonies are unknown. Management strategies to effectively mitigate these stressors are also not well developed, and in some cases, the

13 management approach can also unintentionally create stress for the colony. Thus, it is difficult to develop Integrated Pest Management (IPM) approaches for honey bee husbandry, though IPM has been widely and successfully employed to control pest damage in agricultural crop systems (Box 1). Here we outline the diagnostics and treatments for Noseomosis, a common honey bee disease, discuss what is known about the costs and benefits of applying treatments and highlight gaps in our knowledge that must be addressed in order to develop an effective IPM approach for this disease Nosema parasites and noseomosis 2.1 Nosema apis and Nosema ceranae are microsporidian parasites of honey bees Nosema apis and Nosema ceranae are two globally prevalent parasites of European honey bees (Apis mellifera), with the latter parasite species representing an emerging disease agent (see reviews [4-7]). Both Nosema spp. belong to a larger group of fungal parasites called microsporidia which generally cause progressive, chronic infections (Box 2). In honey bees, Nosema infection is transmitted via fecal/oral exposure when bees clean comb or consume food and/or water tainted with the durable, thick-walled environmental Nosema spores. Infection may also spread through nestmate grooming since molecular studies detect high levels of Nosema DNA in bee washes [8]. Additionally, N. ceranae DNA was isolated from brood food, suggesting that workers may transmit infections to their larval siblings, though it is unclear if parasites were secreted from infected workers glands with food, or if spores originated from external worker or comb surface contamination [9]. There is limited evidence for vertical Nosema transmission: molecular studies (incorporating wash steps prior to dissection) detected N. ceranae DNA in queen spermatheca and ovaries [9]. Histological studies of infected queens are needed to confirm intracellular parasite presence in queen reproductive tissues and to determine if parasites are directly passed to offspring. In horizontally transmitted Nosema infections, consumed spores are carried through the honey bee digestive tract until currently unidentified cues cause spore germination and intracellular invasion of midgut cells (Box 2). Once inside a host midgut cell, Nosema parasites multiply. These intracellular parasitic stages are termed "vegetative states." During this reproductive phase, the Nosema vegetative states procure energy molecules (ATP) from host cells [10, 11] and likely undergo sexual reproduction [12, 13]. Ultimately, new spores are formed which either infect adjacent midgut cells, or are evacuated by the host. If eaten by a new bee, these spores propagate the infection. Until recently, it was thought that N. ceranae might escape from the host midgut to infest other tissues while N. apis remains confined to the gut [14], potentially explaining N. ceranae s alleged greater virulence. However, follow-up studies suggest that both Nosema species are limited to the midgut in workers [15]. Thus, findings from previous studies that indicate broad parasite tissue distribution based on molecular evidence alone may need to be re-evaluated.

14 2.2 Nosema spp. global distribution and factors affecting virulence in managed honey bee colonies 3 N. apis was discovered in 1909 and has been subject to epidemiological scrutiny in European honey bees (Apis mellifera) since the early 20 th century [16]. In contrast, N. ceranae is an emerging parasite of Apis mellifera. N. ceranae was thought to only naturally infect Asian honey bees (Apis cerana) when it was first described in 1996 [17]. Less than a decade later, N. ceranae was discovered in European honey bees in Spain [18]. Shortly thereafter, mixed Nosema infections were correlated with colony collapse disorder (CCD) in the US [19]. Studies from Spain found that N. ceranae was highly virulent in caged bees [11] and could cause colonies to collapse in the field [20]. These and other studies suggested that N. ceranae was potentially more virulent to A. mellifera than N. apis and that N. ceranae was replacing N. apis [21]. N. ceranae s recent host shift from A. cerana to A. mellifera was a hypothesized driver behind its alleged greater virulence. Indeed, pathogens that successfully traverse host species barriers are frequently more damaging to their new hosts than the hosts that they co-evolved with [22]. Furthermore, complete genome sequencing of both parasite species [23, 24] points to some differences in parasite virulence factors (see Section 2.6) that may offer N. ceranae a competitive advantage over N. apis. Given potential links between N. ceranae and world-wide declines in honey bee health, scientific literature concentrating on one or both Nosema species has rapidly expanded and continues to grow. However, the evolving picture of N. ceranae and N. apis virulence and culpability in colony losses is complex. Currently, both species of Nosema have a global distribution, but there are temporal and regional variations in prevalence in addition to rates of singly infected and co-infected colonies (see recent publications for examples [25-28]). To our knowledge, only one survey from Thailand has reported complete absence of both Nosema parasites in several species of bees (Apis spp.) [29]. While numerous studies have documented Nosema spp. prevalence, fewer longitudinal studies, summarized here, have recorded costs of infection in the field. In Spain, historical samples show increasing N. ceranae prevalence [30] and today, N. ceranae appears to be both highly prevalent and in most studies, highly virulent [20, 31-34] (but see [35]). Other European and North American countries have reported dissimilar findings. In Serbia for example, N. ceranae is nearly ubiquitous and the only microsporidian species that infects A. mellifera, but it is not associated with colony loss [36]. Longitudinal studies in Germany and Sweden found that N. apis was more prevalent than N. ceranae and that N. ceranae prevalence was not increasing in Sweden [37, 38]. Furthermore, Nosema spp. infection was not associated with colony overwintering mortality in Germany, Switzerland, and Canada (Ontario), but was associated with smaller spring adult populations in Ontario and overwintering mortality in Belgium [38-41]. In the US, N. ceranae is generally more prevalent than N. apis but respective species prevalence varies regionally [28, 42] and differs between managed and feral populations [27]. Co-infections are associated with US collapsed colonies, but longitudinal studies are still needed to characterize costs independently from those associated with the specific colony collapse disorder (CCD) syndrome [19, 43]. In South America, retrospective molecular analyses have detected the earliest documented N. ceranae infections in A. mellifera populations from Brazil [44], with similar early detections in Uruguay [45]. Both studies note that N. ceranae s prior undocumented presence seems to have caused limited damage, but, they acknowledge that longitudinal studies are needed to confirm this speculation. Finally, cage studies conducted across Europe and in North America also report heterogeneous findings on Nosema speciesspecific virulence [11, 46-50], where N. ceranae in turns has been found to be more virulent, no

15 more virulent, or less virulent than N. apis. Several factors are thought to explain the global variation in reports of Nosema spp. distribution and virulence: (1) Climate is likely a strong driver of Nosema spp. distribution and virulence. N. ceranae spores are more tolerant of heat and desiccation than N. apis spores. Also, N. ceranae infections proliferate faster in bees incubated at warmer temperatures [6, 51, 52]. Field studies in Taiwan, where N. ceranae levels tightly correlate with temperature, corroborate laboratory findings [53]. These temperature studies may partially explain why N. ceranae is both more damaging and prevalent in the warmer, drier climate of Spain than in cooler, temperate regions such as Germany, Switzerland and Sweden [6, 37, 38, 40]. However, even within Spain, distribution of both Nosema spp. varies, indicating that local climate and/or other conditions shape Nosema spp. establishment [54]. Comparative analysis of Nosema spp. genomes also support observed differences in each parasite s capacity to survive under different climatic conditions. For example, functional gene categories related to response to abiotic stimuli were more represented in N. apis than in N. ceranae, possibly providing N. apis with superior ability to respond to lower temperatures or changes in temperature [23, 24]. (2) Differences in virulence between N. ceranae strains may contribute to global trends, but additional studies are needed. In accordance with its recent invasion of A. mellifera populations, molecular analyses of N. ceranae isolates from distinct geographical regions do not segregate based on origin, suggesting that variation in host populations may be a more important factor in explaining global virulence trends than variation among Nosema strains [12, 55]. Indeed, a high degree of Nosema sequence variability has been noted within honey bee colonies [12, 56] and even within the same bee [57], indicating that multiple strains of N. ceranae may have simultaneously infiltrated A. mellifera and spread together. Widely distributed, raw genetic variability in N. ceranae global isolates, along with strong indirect evidence for recombination could mean that local adaptation of strains will occur in the future, giving rise to more heterogeneity across regions [12, 56-58]. Indeed, N. ceranae strains in A. cerana populations distributed throughout China group based on geographic origin and cluster separately from N. ceranae strains derived from A. mellifera populations [59]. (3) Variation in host bee population susceptibility likely contributes to global parasite virulence trends. Though it is early yet to conclude that variation in A. mellifera populations rather than N. ceranae strains predominantly governs global virulence trends, new research offers support for this hypothesis. For example, when N. ceranae isolates from France and Spain were compared, genetic variation amongst isolates could not be correlated with parasite regional origin [55]. Furthermore, honey bees from the A. m. iberiensis subspecies found in Spain were equally susceptible to infection from both isolates across all measured parameters (mortality, spore production, midgut lesions). In a similar study, A. m. iberiensis was equally susceptible to infection with N. ceranae isolates from the Netherlands and Spain (though there was a nonsignificant trend for greater survival in cohorts infected with N. ceranae from the Netherlands) [58]. Again, molecular analyses did not separate N. ceranae isolates based on geographic origin. Interestingly, when 3 A. mellifera taxa from 2 different regions (North Mediterranean, Near and Middle East) were infected with the same N. ceranae isolate, source colony rather than geographic origin emerged as the most important factor modulating host performance [60]. Additional studies have shown that some strains of bees are more resilient to infection, giving rise to selective breeding programs (see Section 5.3). Further studies are needed to determine the respective roles of Nosema isolates and host strains in global virulence patterns. 4

16 (4) Parasite-specific developmental trajectories and Nosema interspecies competition likely contribute to global patterns. In single-species infections, N. ceranae tends to produce greater numbers of mature spores than N. apis, hinting at a competitive advantage for N. ceranae [15, 46, 49, 50] (but see [61]). However, the minimum infective dose for N. ceranae may require more spores than N. apis [50]. In mixed infections, the order of microsporidian species exposure matters. In simultaneously acquired mixed infections, there is limited evidence that N. apis may produce more spores than N. ceranae [46] (but see [49, 61]). Alternatively, prior infection with one Nosema species dampens reproduction of a subsequently acquired infection of the other species [61]. This mutual repression determined by infection order of Nosema spp., however, is not symmetric: primary infection with N. ceranae hampers N. apis reproduction far more than primary infection with N. apis inhibits later N. ceranae reproduction. Indeed, the authors hypothesize that the greater competitive advantage of N. ceranae in initial infections may explain N. ceranae s predominance in some regions. Unbalanced effects of infection primacy have also been found for N. ceranae and deformed wing virus (DWV), where N. ceranae shows a strong competitive advantage if administered as the primary infection in cage studies [62]. Potential differences in virulence factors between N. apis and N. ceranae that may contribute to differential pathogen success are discussed in Section 2.6. (5) The presence of other stressors, including pesticides [63-66], other pathogens and parasites [67, 68], and potentially a common Nosema treatment (see Section 5.1) [69], can impact Nosema success and virulence. (6) Finally, differences in experimental design affect study findings [70]. For example, adult worker susceptibility to infection changes with age. Thus, timing of inoculation and bee age may influence experimental outcomes [71] Nosema spp. global distribution in native pollinator populations: Nosema spp. research has frequently focused on characterizing pathology and virulence in managed honey bee colonies. However, there is growing awareness that N. ceranae can infect other insect pollinator species. This pathogen spillover from honey bee colonies to wild bee populations endangers the health of pollinators and the ecosystem services they provide [72]. N. ceranae has been found in solitary bee populations in Belgium [73] and in wild bumble bee species (Bombus spp.) across several continents, demonstrating that N. ceranae has a broader host range than A. mellifera [74-76]. In addition, N. ceranae may be more lethal to Bombus spp. than Apis spp. [74]. Additional studies are needed to assess the magnitude of risk and damage imposed by N. ceranae in other pollinator species. 2.4 Nosema spp. pathology: consequences for individual workers and honey bee colonies Infected workers are energetically deprived, exhibit precocious foraging and are more likely to die prematurely [5]. Nosema replication appears to be a key proximate driver of the energetic costs of infection. While reproducing within host midgut cells, Nosema parasites cause tissue damage and expropriate host ATP energy molecules [10, 77] (see Box 2). Thus, host

17 digestion is likely hindered while host resources are redirected to support parasite replication, indirectly and directly depriving the host of sustenance and molecular fuel. Indeed, if permitted, N. ceranae infected workers will consume extra food, and if food access is restricted, workers will starve faster than uninfected siblings [78]. Additional studies have documented nutritional and metabolic deficits and changes in feeding behavior in both N. apis and N. ceranae infected workers [79-83]. Furthermore, both molecular and metabolomics studies indicate changes in infected workers lipid, carbohydrate and amino acid metabolic profiles [84, 85]. The energetic costs associated with Nosema infection contribute to workers behavioral symptoms of infection, including accelerated maturation from nursing/brood care to foraging behavior (Box 3) [86-88]. Normally behavioral maturation rates are governed by workers internal nutritional and hormonal status which in turn are sensitive to diverse colony cues [89]. Comparative molecular studies of gene expression in worker fat body tissue suggest that energetic costs of infection may starve workers, preventing them from either reaching or maintaining the nutrient-rich physiology and attendant molecular profiles associated with nursing [85]. Interlinked changes in worker nutritional and hormonal status (likely involving the insulin signaling pathway and the vitellogenin/juvenile hormone axis) as a result of energy deprivation may subsequently promote foraging behavior and physiology. Since many stressors are known to cause precocious foraging, premature foraging in infected individuals may represent a conserved stress response [90]. Supporting the hypothesis of a convergent stress response, common genes are regulated in worker brain tissue following exposure to ecto- (Varroa) and endoparasites (N. ceranae) [91]. However, some elements of worker foraging response to Nosema appear diseasespecific since foraging patterns differ between N. ceranae infected workers and workers given a sterile wound [92]. For Nosema-infected workers, premature foraging results in premature death, which undermines colony stability [86]. Foraging is the terminal vocation for workers and individuals with shorter life expectancies due to infection or another stressor will undertake foraging tasks prematurely [88, 93]. Since foraging is energetically intensive and dangerous, precocious foraging may offer colonies a way to minimize resource losses by assigning hazardous tasks to short-lived individuals. Infected foragers may also suffer greater extrinsic mortality than healthy foragers since infection is associated with disorientation and other metabolic and behavioral abnormalities [92, 94-96]. However, when a colony does succumb to infection, its failure in part likely arises from imbalances in worker division of labor leading to population declines. Since infected workers forage precociously and therefore die prematurely, younger workers are compelled by colony cues to fill the foraging void, perpetuating a vicious cycle of early adult death. For example, simulation models suggest that precocious foraging and early forager death are strong predictors of colony failure [97]. Thus, colonies unable to compensate for resources invested in workers that die early (and therefore also contribute less to their colony s fitness), may dwindle until the weakened colony succumbs to Noseomosis or another stressor (as summarized in [5]). Interestingly, infected workers produce higher levels of a pheromone (ethyl oleate) that slows worker behavioral maturation [98]. Excessive EO production could cautiously be interpreted as a colony attempt to slow behavioral maturation of healthy workers to help infected colonies maintain a balanced nurse:forager ratio. In addition to premature death of infected forager bees, Nosema infections can also lead to other individual behavioral changes that negatively impact colony function. Not surprisingly, healthy foragers appear more efficient at gathering resources than infected foragers [94]. Moreover, harmonic radar tracking studies show that infected foragers take longer rests and are less likely to return to the colony during homing experiments [96]. Thus, colonies that are not 6

18 killed by infection with either Nosema spp. still suffer costs: they are slower to grow, have smaller adult populations relative to brood area, and produce less honey [4, 32] Nosema spp. pathology: consequences for adult queens and drones and immature castes Few studies have characterized Noseomosis in queens, drones and immature host stages. Briefly, infection in adult queens and drones, as in workers, results in aberrant physiology and metabolism [99, 100]. N. apis infected queens are more likely to be superceded [101] while N. ceranae changes queen pheromone profile [99]. Additional studies are needed to determine if N. ceranae also precipitates queen replacement. Caged drones exhibit increased mortality while infected drones in the field are more likely to drift and/or have greater mortality rates (Chapter 4) [100]. Additional laboratory and field studies are needed to fully characterize molecular, physiological and behavioral symptoms of Noseomosis in these castes. Only a handful of molecular studies have investigated Nosema infections in pupal drone stages [102], in larval and pupal workers [103] and in larval queens [9] suggesting that immature stages across castes can become infected, but Nosema incidence and prevalence in immature stages as well as disease etiology and ramifications for colony health remain uncharacterized. 2.6 Nosema virulence factors and host defense mechanisms Overall, virulence factors that enable Nosema spp. to successfully invade host midgut tissue are poorly understood. The insect midgut is lined with the peritrophic membranes (PM), a protective mucosal film secreted by intestinal cells [104]. How Nosema pervades and/or subverts the honey bee PM is poorly understood. Recent studies, however, suggest that N. ceranae does impair local host immune defenses in midgut tissue, including ROS (reactive oxygen species) production and apoptosis [77, 105, 106]. ROS production is a conserved, non-specific immune response and ROS molecules can be highly toxic to both parasite and host cells. Apoptosis, or programmed cell death, is also a general immune strategy where host cells systematically selfdestruct to undermine parasite reproduction. Whether N. apis employs similar host manipulation strategies remains to be confirmed. Once successfully established within a host cell, Nosema parasites employ an arsenal of molecular virulence factors to purloin host nutrients (see Box 2). Indeed, comparative genome analysis of N. ceranae and N. apis provides a better understanding of how these parasites have metabolically adapted to import essential nutrients and energy from host cells [23, 24]. Both N. ceranae and N. apis contain genes that code for proteins including glycerophosphate shuttle enzymes, the ATP-binding cassette (ABC) transporter complex, sulfate transporter family protein, NADPH oxidoreductase, iron-sulfur cluster (ISC) assembly enzyme, hydrolase, heat shock protein 70, cation efflux protein zinc transporter, exportin, and pyruvate dehydrogenase E1 (PDH) enzymes that are involved in metabolite transport, ISC assembly and export, and antioxidative stress. Although the core genes and their associated functions have been retained in both Nosema species, the key enzymes involved in energy transporter activity and metabolic processes are more represented in N. ceranae than N. apis, suggesting that N. ceranae may have a greater capacity for using host ATP. Thus, N. ceranae s relatively stronger ability to obtain

19 energy from its host may confer a survival advantage to N. ceranae and promote successful competition with N. apis [23]. Nosema spp. also appear to modify systemic expression of canonical worker immune genes, though changes are often transient and particular to the Nosema species of infection, incubation period and other factors incorporated in experiment design [ ]. Indeed, comparative analysis of N. apis and N. ceranae genomes shows that proteins involved in responding to stress and biotic stimuli were significantly more represented in N ceranae than N. apis, suggesting that N. ceranae may have a better ability to survive under stress of host immune defense potentially contributing to N. ceranae s dominance in some geographic regions [23]. However, expression of members of the canonical Toll signaling pathway is altered in infected workers [85] and some Toll pathway genes are upregulated in drones from a Nosema tolerant honey bee strain [110], pointing to the Toll signaling pathway s likely involvement in host defense against microsporidia. Additional molecular studies have identified other genome regions and non-canonical immune genes in fat body tissue that modulate worker response to infection [85, 111]. Part of the challenge of dissecting host immune response is that worker immune, hormonal, metabolic and nutritional statuses are interlinked. Thus some changes in immune function may be a byproduct of disease costs and/or of generalized stress [85]. Furthermore, host age and timing of exposure interact. For example, older bees survive incipient infection with N. ceranae better than younger bees even though older workers also produce higher spore loads [71]. Changes in individuals behavior following infection influences parasite growth trajectories and likelihood of transmission. For example, recent choice tests found that infected workers prefer honey with greater antimicrobial activity, and that consumption of favored honey could reduce N. ceranae pathogen loads in cage trials [112]. Thus, self-medication through selective diet may be one way that individuals repress infection but field studies are needed. N. ceranae infected bees also prefer warmer temperatures and are more likely to be found in the center of the colony [113]. Workers suffering from Noseomosis may inherently prefer warmer temperatures because their ability to thermoregulate is potentially restricted by infection costs. Alternatively, since N. ceranae develops better at warmer temperatures, this thermotactic predilection could cautiously be interpreted as a host-parasite manipulation to enhance parasite reproduction [51]. Regardless, congregation of infected workers in certain hive regions likely influences parasite transmission. For example, cage experiments investigating permutations in diseased and susceptible host density, with workers or drones serving as the initial source of N. ceranae infection, found not only that N. ceranae transmission exhibited density dependent properties, but also that drones transmitted N. ceranae at higher rates than workers [114]. The number of spores produced by individual workers and drones also varied with initial infection density and caste. Together, these finding suggest that multiple factors regulate Nosema transmission within the complex and dynamic context of colonies. Changes in social interactions (or lack thereof) also contribute to disease dynamics. Interestingly, workers may perceive if nestmates have been exposed to an immune challenge and treat infected nestmates more aggressively [115]. However, healthy workers do not treat N. ceranae infested workers more aggressively than controls in observation hive studies [91]. Therefore, workers harboring infections escape social persecution which may have consequences for disease transmission. Finally, when examined in a providential light, precocious foraging serves as a general social immune response. As previously discussed, infected individuals have shorter life expectancies, and thus optimize their contribution to colony fitness by performing the riskiest task 8

20 of foraging, sparing their healthy siblings with longer life expectancies. By leaving the colony, infected workers may further reduce chances of in-hive transmission. But, accelerated behavioral maturation may indirectly benefit the parasite since infected foragers are disoriented [95] and may drift into neighboring colonies (but see [96]) or contaminate flowers with spores [116] Methods for diagnosing Nosema infections in colonies Without laboratory assistance, the majority of beekeepers do not have the ability to determine if their colonies are infected with Nosema, let alone determine infection species or severity [117]. Accurately diagnosing Nosema infection is difficult because there are few obvious clinical symptoms, and those that may be present (diarrhea for N. apis only (but see [118]) or swollen, milky digestive tract for both Nosema spp.) are not unique to microsporidian infection [4]. Thus, while beekeepers may speculate that their colonies are infected with Nosema, national self-reports of Nosema prevalence and corresponding colony loss likely underestimate the true disease distribution and losses attributable to microsporidia (e.g. [119, 120]). Currently available diagnostic techniques comprise microscopy and molecular tools. These tools come with the obvious limitation that they can be expensive and molecular techniques in particular are not accessible to beekeepers. Depending on the diagnostic method, beekeepers may only be able to confirm presence or absence of infection and may not be able to determine infection intensity, and/or parasite species. Currently, only molecular techniques can determine both infection intensity and species of infection. However, several alternative and more accessible tools can be developed which would greatly improve the efficiency and cost of Nosema diagnosis. 3.1 Light microscopy Nosema spore presence and numbers can be easily determined using a light microscope and standard practices are detailed in [70]. In this method, the abdomens or dissected midguts of one or a pool of bees (see Section 4.1 for further discussion about the populations and numbers of bees to use) are ground up, excess tissue is removed by filtering or centrifugation, and the solution containing the spores is placed on a microscope slide and visualized at a magnification of 400x (Figure 1). The numbers of spores in a particular volume can be determined using a hematocytometer, and, after back-calculating (using the volume each bee was homogenized in), it is possible to obtain the average number of spores/bee [121]. Aside from the start-up costs of purchasing the microscope and hematocytometer, the overall costs of using light microscopy in Nosema diagnosis are low. Also, US and Canadian beekeepers may access microscopy services for free (aside from shipping costs) by sending samples to the USDA [117] for spore detection and quantification.

21 10 Figure 1-1 Nosema spp. spores in whole abdomen homogenates. Light-refracting, Nosema spp. spores are circled in the image (400x). The bar is approximately 6 um in length. Other artifacts (pollen, bacteria) are visible.however, there are several disadvantages to using light microscopy for diagnosis. First, the methodology cannot distinguish between N. apis and N. ceranae infections (though N. ceranae spores tend to be smaller than N. apis spores) [70], which may be important for effective disease treatment. Second, this methodology requires instruction and practice to accurately identify spores versus other artifacts. Third, preparing, viewing and counting spores in samples can be time consuming. Fourth, light microscopes only allow users to view mature spores. Vegetative states, because they lack thick spore walls and refractive properties, are not easily detected. Therefore, while spore counts may be indicative of pathogen load, they do not provide an absolute measure. However, spore counts using microscopy do correlate well with molecular techniques, which capture total pathogen load (both vegetative and spore states) [122]. 3.2 Molecular techniques Molecular approaches can be used to diagnose infections, distinguish between Nosema species, and determine parasite titers based on DNA copy number. Standard molecular methodology has been outlined in [70]. In all cases, DNA is extracted from the abdomens or midguts of sampled bees and processed with Polymerase Chain Reaction (PCR) using primers that are specific for genetic sequences found in Nosema. These reactions amplify the Nosema DNA sequence between the primers, and the amplified DNA fragments can be visualized either by staining with a fluorescent dye and viewing the fragments on an agarose gel or using special instrumentation to measure the amount of fluorescence in the reaction solution (for quantitative real-time PCR) [70]. Using two sets of primers that match specific sequences in the genomes of either N. ceranae or N. apis, it is possible to determine which Nosema species are present. Alternatively, primers that match both Nosema species can be used in the PCR reaction, and the resulting fragment can be cleaved by a restriction enzyme that targets a sequence that is only found in either N. apis or N. ceranae. Using this "Restriction Fragment Length Polymorphism" or RFLP approach results in fragments of different sizes for each parasite species [70]. Finally, the titers of Nosema can be determined from the amount/number of copies of the DNA fragment

22 produced, either by examining the amount of fluorescence on an agarose gel (this method is considered to be less accurate and therefore "semi-quantitative") or using quantitative real-time PCR. For quantification, it is possible to create a standard curve using samples with different numbers of spores, in order to calculate the number of spores/bee. However, this method is not entirely accurate since molecular approaches can quantify both vegetative and non-vegetative spores, and spore counts using microscopy can only identify non-vegetative spores [122]. Molecular techniques offer many advantages. First, these techniques can distinguish between N. ceranae and N apis infections. Secondly, these techniques incorporate DNA from both vegetative and non-vegetative spores, and thus can capture total infection levels. Third, these approaches are very sensitive, and can identify low levels of infection in individual bees and low prevalence infections where only a small percentage of the bees in a pooled sample may be infected. Finally, molecular approaches readily lend themselves to automation and can be scaled up so that hundreds of samples can be processed at the same time. However, the preparation of each sample may still be time-consuming depending on methods used. The primary limitations of molecular approaches are the cost of the instruments and materials and the amount of technical skill needed to perform these studies. In addition, due to variation in Nosema spp. genome sequences, developed molecular techniques must be carefully vetted to prevent misdiagnosis [123] ELISA (Enzyme-Linked Immunosorbent Assay) test Recently, Aronstein and collaborators adapted the Enzyme-Linked Immunosorbent Assay (ELISA) technique to detect Nosema infections [124, 125]. In an ELISA assay, an antibody that is specific for a protein or hormone of interest is coupled to a dye, and thus it is possible to determine if that protein or hormone is present and semi-quantitatively determine its levels visually. Using specialized equipment to measure the concentration of the dye (a spectrophotometer) and a standard curve with known quantities of material, it is possible to quantify the amount of the target molecule (or spores). Alternatively, a simple color change can be used to document presence or absence of the target molecule, as in the case of pregnancy tests, which measure levels of the human chorionic gonadotropin hormone in urine. Aronstein and collaborators developed antibodies that specifically bind to the spore wall protein (SWP32) of Nosema ceranae, and incorporated these in an ELISA assay. This method was validated with qpcr and not surprisingly, both ELISA and qpcr methods were shown to be more sensitive than microscopy. The published ELISA method requires expensive equipment and is time consuming. However, this methodology could readily be adapted to produce dipstick tests that could be easily used by beekeepers in the field. To test a colony, sample workers would be collected and macerated. A drop of the sample would then be applied to the dipstick. If N. ceranae spores are present, the dipstick would change color. Infection intensity could be estimated by comparing the color results to a reference card provided by the manufacturer. If produced and commercialized, this test could offer beekeepers a cheap, fast and easy way not only to detect but also to quantify N. ceranae spores in colonies, as well as other internal parasites (such as Nosema apis) if specific antibodies were developed.

23 3.4 Measuring acoustic or odor signatures of colony stress 12 New technology may offer novel means of diagnosing Nosema infection as well as other colony disorders. For example, Bee Alert Technology (Missoula, MT) is currently testing a prototype of the Honey Bee Acoustic System (HAS) [126]. The HAS operates on the proposition that honey bee colonies, when exposed to different stressors, produce unique sound waves which can be diagnostic of specific problems. Peer-reviewed studies however, are needed to validate this system. An analogous diagnostic method would allow beekeepers to identify colony stressors based on chemical odorants [127]. To our knowledge, beekeeping commercial technology of this nature does not yet exist. However, there is potential for its development. Indeed, electronic noses have already been developed for diverse applications ranging from contagious disease detection to food quality or environmental monitoring (see [128] for review). If infected bees produce unique volatile signatures (as do Varroa mite parasitized-pupae [129] or chalkbrood infected- larvae [130]), future technology might be able to capture chemical evidence of Nosema parasitization. Though acoustic and odorant technologies are either in development or remain hypothetical, if proven through independent tests such systems would greatly advance beekeepers ability to diagnose diseases. Such technology might in theory be minimally invasive and far less-time consuming since it would obdurate the need to collect and process bee samples. Second these systems could be harnessed to diagnose many different problems. Third and finally, if manufacturing costs are not prohibitive, this technology could offer better access to beekeepers as well as long term time- and cost-savings. 4.0 Challenges with establishing treatment thresholds At this time, there is no consensus on the treatment thresholds for Nosema. Older analyses suggested treatment for N. apis when spore loads matched or exceeded one million spores/bee [131]. However, this is an arbitrary threshold. Further research is desperately needed to determine the best practices for sampling colonies and quantifying Nosema intensity. Furthermore, studies are needed to link Nosema levels with colony damage so that Economic Threshold (ET) and Economic Injury Levels (EILs) can be established and serve as guidelines for when to treat (Box 1). 4.1 Sampling colonies for Nosema infection Since Nosema spp. disease progression is chronic and infected workers forage precociously, it is clear that foragers are the best population of bees to sample to maximize detection sensitivity [70, 132]. However, within the same colony, infection levels can vary dramatically between foragers, so larger sample sizes are needed [133]. If samples are to be screened via light microscope, at least 60 bees should be included in a pooled sample to increase detection sensitivity (95% confidence of identifying a 5% infection prevalence in the colony) [134]. Alternatively, some studies suggest that determining the percent of infected bees within a

24 sample is a better gauge of infection intensity. This latter method is time-intensive and impractical. Since the percent of infected bees in a sample correlates with the overall infection intensity [132], the former method is preferred. If samples are to be screened with molecular techniques, additional research is needed to determine appropriate sample sizes as molecular techniques offer higher levels of sensitivity than microscopy. Collecting the requisite number of foragers from the colony entrance can be timeconsuming especially if a large number of colonies must be sampled. If foragers cannot be collected, it is possible to sample workers from outer frames of top supers where older bees are more likely to reside, but this increases sample heterogeneity [135]. Unfortunately, sampling dead bees may underestimate the actual prevalence and intensity of infection [49]. However, collecting worker fecal matter from the bottom of colonies may allow diagnosis of both infection intensity and Nosema spp. [36, 136]. If further developed and validated, these methods may circumvent time-intensive bee collections in addition to eliminating the need to kill colony members. Establishing when to sample colonies also presents a challenge. Forager spore loads can fluctuate dramatically from one week to the next and even within the same day [30, 133, 137]. There have also been reports of seasonal oscillations in Nosema colony levels that may inform sampling guidelines. Historically, N. apis levels have been reported to maximally peak in the spring, with a small peak in the fall [4]. In Spain, however, N. ceranae prevalence is seasonally stable [5], but studies elsewhere have reported that N. ceranae colony levels cycle [36, 38, 138] hinting that temperature/climate [53] and other regional differences may govern seasonal parasite levels Setting EILs (Economic Injury Levels) Setting thresholds for Nosema economic injury levels (EILs) is especially challenging since there is global variation in reports of Nosema virulence, and damage incurred by infestation may vary with climate, bee subspecies, Nosema strain and the presence of other hive stressors (see Section 2.2). Studies to date have used spore counts or parasite DNA copy number to estimate parasite burden in individuals and colonies. However, parasite burden does not always directly correlate with parasite virulence. For example, protein-rich diets may enhance both worker longevity and parasite reproduction (see Section 5.3). Furthermore, different populations of honey bees may be more tolerant of or resistant to infection (see Sections 2.2 and 5.3). Tolerance traits allow hosts to carry parasite burdens without suffering the same infection costs as less tolerant individuals with commensurate infection levels. Resistance traits allow hosts to actively suppress infections, effectively reducing the number of parasites they carry [139]. Studies are desperately needed to determine at what point infection levels threaten colony survival, and, how local conditions and host population traits may regionally affect thresholds for EILs. Given variation in Nosema virulence (see Section 2.2), regional IPM guidelines, rather than universal IPM guidelines may need to be developed for cost-effective Nosema management.

25 5.0 Currently available and potential future Nosema treatments 14 Numerous Nosema treatments exist, but none alone are ideal for parasite management. 5.1 Chemical treatments Chemical treatments against Noseomosis include fumagillin, application of bacteria- (surfactins, organic acids) or plant- (essential oils) derived compounds with fungicidal activity, and potentially the future use of genetic products (RNAi). These compounds may relieve heavy infestations in the short-term and some treatments may cross taxonomic boundaries to counter other honey bee pathogens including bacteria and viruses. However, chemical treatments also have several drawbacks. First, these compounds only inhibit active infections within bee midgut cells, but will not kill spores contaminating colonies. After the treatment wears off, future applications may be necessary to prevent re-infection of the same colonies from residual spores. Repeated application of the same treatment may select for resistant Nosema strains and be costly to beekeepers. Second, these compounds may have unintended, negative off-target effects. In order to inhibit vegetative parasite growth within bee midgut cells, these compounds must be fed to honey bees in liquid or patty form. Alternatively, honey bee combs may be drenched or sprayed with these products in sugar solution, promoting consumption by bees. Unfortunately, some of these compounds may negatively affect honey bee health and/or human health if chemically contaminated hive products (e.g. honey) are eaten. Beekeepers using these products may also be at risk through repeated exposure. Finally, only the effectiveness of fumagillin has been formally tested in multiple cage and field trials, and as will be discussed below, there are still several unresolved issues regarding use of fumagillin products in hives. The effectiveness of plant- or bacteria-derived products and RNAi has primarily been studied in cage trials with limited or no field studies to date. Clinical use of these chemicals must be thoroughly evaluated to ensure safety to both bees and humans in addition to efficacy in the field. Fumagillin: A recent publication has reviewed the pros and cons of fumagillin use against Nosema parasites in honey bees [140]. Briefly, in 1949 fumagillin was isolated from the fungus Aspergillus fumigatus and discovered to have far-reaching antimicrobial properties. It has historically been deployed against N. apis in the commercial dicyclohexylamine (DCH) salt formulation which is dissolved in sugar water. Both cage and field assays demonstrate that fumagillin application can control Nosema spp. infection. However, N. ceranae infections lamentably reemerge in colonies within 6 months of treatment, presumably due to lingering spores contaminating colonies or new infections introduced by drifting workers [141]. The rate at which Nosema spp. infections reoccur is partly governed by how rapidly the fumagillin treatment is consumed by the colony and how quickly the chemical breaks down. Fumagillin is degraded by both heat and UV light exposure. Thus, chemical storage conditions and climate may affect

26 the duration of Nosema control achieved. Furthermore, timing of fumagillin application (relative to honey removal) is restricted because fumagillin is toxic to humans. Fumagillin targets a conserved protein (methionine aminopeptidase type 2) that is present in Nosema spp., bees and humans. Due to fumagillin s non-specificity and thus potential for human toxicity, its use is banned in the EU barring exceptional circumstances and its handling must be supervised by a veterinarian. Where its employment is legal, fumagillin cannot be fed to colonies before a nectar flow since some of the product may be sequestered in honey stores that will be taken for human consumption. Instead, the label recommends fumagillin application in the spring or fall, well before or after honey is removed. Recent research has highlighted potential problems associated with the expected degradation of fumagillin within colonies in addition to negative, off-target effects [69]. Cage studies showed that at decreasing concentrations of fumagillin (approximating fumagillin degradation over time in colonies), N. apis and N. ceranae are eventually able to begin reproducing. However, at low fumagallin concentrations, N. ceranae levels actually surpass those achieved in control workers that are never exposed to fumagillin. The authors speculate that excess rebound in N. ceranae but not N. apis populations at low fumagillin concentration may in part explain why N. ceranae appears to be supplanting N. apis in some regions. Worryingly, these findings also suggest that fumagillin may relieve N. ceranae infection in the short-term but ultimately intensify infestation. Furthermore, fumagillin alters protein production in worker midguts at treatment concentrations that do not repress either Nosema spp., suggesting that workers accrue off-target effects without gaining protection against Nosema infection as fumagillin degrades. Another important (and until recently) overlooked consideration of commercially prepared fumagillin is that individual components of the formulation may degrade at different rates and exhibit different toxicities [142]. In the marketed salt formulation, fumagillin is the negative ion while dicyclohexaylamine (DCH) is the positive ion. DCH alone is toxic to rats and can cause chromosomal changes in human cell cultures (summarized in [140]). Moreover, DCH is far more temperature stable and degrades more slowly than fumagillin, which has implications for DCH s persistence in sequestered colony honey and potential accumulation in bee cuticles [142]. Given fumagillin s ubiquitous usage where legal, field studies are needed to determine if current recommendations for fumagillin use against Nosema spp. must be reassessed. 15 Bacterial metabolites: Antimicrobial molecules called surfactins produced by bacteria may also be used to treat Nosema. Surfactins have unique properties that allow them create pores in cell membranes and the resulting perforations are typically lethal for targeted cells [143]. Biomedical research has identified many potential anti-fungal, anti-viral, anti-tumor, and anti-bacterial therapies for surfactin use in humans [143]. Likewise, surfactin treatment alone or in conjunction with essential oil application (see next section) inhibits growth of honey bee pathogens, including Paenibacillus larvae, the destructive bacterial agent of American Foulbrood disease and Ascosphaera apis, the fungal pathogen that causes chalkbrood disease [144, 145]. However, these early studies only assayed surfactin efficacy against dish-cultured pathogens, and in vivo

27 cage and field studies are needed. However, feeding caged bees surfactins produced by Bacillus subtilis bacteria isolated from honey samples reduced N. ceranae spore counts in inoculated workers [146]. Also, a separate study tested whether organic (lactic) acids isolated from Lactobacillus johnsonii bacteria could reduce spore loads in the field [147]. These bacteriaderived organic acids did not cause acute mortality over 72 hours, and workers in treated colonies generally had lower spore loads than control colonies. In both these experiments, however, infected workers still carried thousands to several million spores, and in the case of colonies, the prevalence of infected workers did not vary amongst treatment groups. Thus, while some aspects of these results are promising, spore counts remain high even with treatment and bees are likely to still suffer negative effects of Nosema. Additional development and testing is required to make sure these treatments alleviate disease costs without harming bees and that they are labor and cost-efficient (e.g. one study fed bees organic acids 5x over a 25 day period [147]). 16 Essential oils: Essential oils (EOs) may represent another class of anti-microsporidian substances. EOs are aromatic blends of compounds isolated from plants, with different plant species and even tissues yielding different oils (see [148] for a review). EOs aid in plant defense against bacterial, viral and/or fungal infection. Historically, humans have distilled EOs from plant tissues and exploited their antimicrobial and fragrant properties and EOs continue to be economically important components of products ranging from disinfectants to perfumes. Beekeepers routinely use EOs, especially menthol and thymol (Apilife var, Apiguard), to control Varroa mites as an alternative to manufactured chemicals such as tau-fluvalinate (Apistan), coumaphos (Checkmite) and amitraz (Apivar). These EO formulations are placed in colonies and released volatiles create a toxic environment for mites. EOs have also been shown to be effective at reducing Nosema spore loads in studies with caged bees. Feeding cages bees with thymol suppresses Nosema reproduction while leaving worker life span unaffected or even extended [149, 150]. However, as in the case of bacterial metabolites, thymol application reduced spore counts but did not eliminate Nosema infection. After 25 days of feeding workers thymol-laced syrup or control syrup, thymol-fed workers had an average of 60.2 ± 9.2 million spores while control workers had an average of ± 15.8 million spores [149]. Thus, it remains to be determined if thymol treatment actually reduces Nosema loads sufficiently to prevent the negative effects of Nosema infection. Several other plant extracts have also been tested for anti-microsporidian activity in bees (e.g. wormwood (Artemisia absinthium), garlic (Allium sativum), sweet bay (Laurus nobilis) yerba mate (Ilex paraguariensis), beet root (Beta vulgaris), oak bark (marketed as Nosevit) and lemongrass and spearmint oil (marketed as Honey-B-Healthy) [ ] with variable results. Translating the results of these cage studies to the field is challenging. First, some extracts are not palatable to bees, and therefore there may be poor consumption of EOsupplemented food in the field where alternative food sources are available. Second, ensuring that bees have high, oral doses of these extracts may be challenging, expensive, and laborintensive in colonies. Third, it is essential to determine if EO treatment actually improves worker longevity and colony survival, rather than simply assessing spore counts.

28 Importantly, additional research is needed to determine if EOs inflict unintended, sublethal costs in honey bees. For example, EOs can have cytotoxic effects on invertebrates including mollusks and insect larvae [148]. In bees, volatile exposure in a laboratory setting caused changes in expression of genes related to detoxification, immunity and behavioral maturation [154]. Furthermore 24 hours of volatile exposure to thymol in a colony setting resulted in detectable levels of thymol in worker brains and altered worker phototactic behavior [155, 156]. Since the chemical properties of EOs vary depending on plant source, and oral or volatile exposure could result in different toxicity effects, additional studies are needed to characterize EO mechanisms in honey bees under field-treatment conditions. There are several possible mechanisms by which EO compounds may reduce Nosema spore loads. EOs penetrate mitochondrial membranes, causing a break-down of mitochondria function and release of toxic reactive-oxygen species (ROS), ultimately causing cell death (summarized in [148]). As discussed earlier, microsporidia siphon energy stores (ATP) from their honey bee hosts mitochondria (Box 2), and thus EO disruption of the mitochondria may limit microsporidian growth and reproduction. Additionally, if EOs cause host cell death via apoptosis, any associated, immature Nosema vegetative states will also die. Finally, ROS release serves as a basic invertebrate immune defense. Therefore, if EOs promote ROS release, honey bee defenses may be enhanced. Indeed, molecular studies suggest that enzymes involved in ROS production are upregulated in gut tissue of workers infected with N. ceranae [77]. However, all of these mechanisms damage the Nosema parasites by first damaging the host cells, and thus there may be sublethal effects of EO treatment that could be mitigated if an IPM approach is used, and EOs are only employed when necessary. Despite relatively little scientific investigation, there is enthusiasm in the beekeeper community for use of EO formulations such as Honey-B-Healthy [157] and other plant extracts. Plant-based treatments may generally appeal to beekeepers since organic farming practices are gaining mainstream interest and organic products may command a premium on the market. Also, since Fumagillin is banned in the EU, plant-derived products may serve as alternative therapies. As these formulations gain popularity, it is imperative that scientific research validate treatment efficacy and investigate potential negative effects on colonies. 17 RNA interference (RNAi) The comparative genome analysis of N. ceranae and N. apis led to the identification of parasite-specific genetic elements that are potentially related to virulence, which could be harnessed for developing RNA interference (RNAi)-based therapeutics against Nosema diseases [23]. Such RNAi technology would exploit antiviral defense mechanisms found in honey bees. By synthesizing and subsequently feeding bees double-stranded RNA (dsrna) for a target gene sequence, it is possible to dramatically reduce expression of the gene target. A previous study evaluating the feasibility of RNAi for controlling N. ceranae, showed that ingestion of dsrna homologous specific for Nosema ADP/ATP transporter gene could lead to reduction of the parasite load as well as the Nosema disease in the infected host [10]. Similarly, RNAi has been successfully deployed in field trials against viral pathogens [158]. These results provide evidence that RNAi holds therapeutic potential for the treatment of Nosema parasites and other diseases in honey bees. RNAi offers the advantage of target specificity since dsrna sequences would be unique to bee parasites though follow-up studies would be needed to ensure that dsrna exposure

29 does not negatively affect molecular processes in all honey bee castes and life stages. However, dsrna presumably would be administered orally, and studies are needed to determine how frequently treatments would need to be applied. Routinely feeding colonies large quantities of dsrna might be costly and time consuming Hive sterilization methods An alternative or complementary strategy to chemically treating vegetative growth of Nosema parasites is to inactivate environmental spores by sterilizing hive equipment. Previous studies have found that N. apis spores are deactivated after being incubated at 60 C but N. ceranae spores are both more heat and desiccation tolerant than N. apis and would require higher temperatures for spore deactivation [52]. UV exposure, however, can kill spores of both Nosema spp. and gamma radiation has been successfully shown to deactivate N. apis spores in liquid suspension [159]. Naturally, all these sterilization methods pose some logistical challenges. Heating colony equipment, especially frames, may be impractical since it would cause wax to melt. For example, beekeepers have previously heated hive bodies and dry comb to 120 F for 24 hours and significantly reduced N. apis infections [160], but such methodology would not work with wax comb. Gamma radiation, however, has been previously employed at large scale in Australia to sterilize hive bodies and frames contaminated with Bacillus larvae, the highly virulent agent of American Foulbrood (AFB) [159, 161]. Since gamma radiation does not damage hive equipment (materials are only heated ~3 C) it is perhaps the only viable, current method for hive sterilization. In sufficient doses, gamma radiation has the added and large benefit of eradicating other bee pathogens including fungi, viruses, bacteria. On the other hand, use of gamma radiation requires that bees are either first killed (in the case of AFB) or are removed from contaminated equipment before colonies can be sterilized. Also, materials must be transported to a radiation facility which can incur additional costs. Since hive bodies and frames are expensive, gamma radiation can be cost-effective [161]. For example, the Australian government has sponsored gamma radiation treatment for AFB eradication. Whether such programs would be effective in other countries depends on facility availability. 5.3 Colony management practices In addition to chemical treatments and hive sterilization techniques, beekeepers may employ colony management practices to mitigate Noseomosis including selective breeding, queen replacement and potentially nutrient supplementation.

30 Selective breeding and/or queen replacement 19 Human-mediated selection may produce robust honey bee breeds. Encouragingly, Danish beekeepers have selected for a N. ceranae tolerant strain of honey bees and genetic mapping has been used to identify chromosomal regions that underpin resilience to Nosema [110, 111, 162]. Other studies provide further evidence of genetic resistance in sampled Russian subspecies (but not Italian) honey bees [163]. In Uruguay, there are also reports that Africanized honey bees have more natural resistance to N. ceranae infections than Italian honey bees [164]. These studies suggest that there is enough raw genetic variability in host bee populations to choose bee strains that are either tolerant of or resistant to Noseomosis. Selecting for Nosema resilient bees offers many long-term benefits. Resilient bees would require fewer chemical applications to control Nosema infestations, saving beekeepers both time and money and reducing the chances of Nosema parasites developing resistance to available treatments. However, selective breeding programs require intense time and resource commitments [165]. For example, the Danish selection program was conducted over 20 years, started with 500 colonies and required annual screening for Nosema and replacing the queens of susceptible colonies [162]. Another important consideration when breeding Nosema resilient bees is whether the selected strains can withstand other colony stressors. For example, if selected colonies are always treated for Varroa, the resulting bees might be Nosema resilient but susceptible to mite infestation. An ideal goal of a long-term breeding program might be to combine traits of bees that are resistant to unique stressors. For example, such a program might aim to combine traits from Varroa-resistant bees [166] with Nosema resilient bees. Selective breeding represents a long-term strategy for Nosema management and would be a large task for an individual beekeeper to undertake. However, inducing queen replacement may serve as a short-term measure that beekeepers can employ with other tactics to control Nosema. Honey bee queens generally live 1-3 years but their fecundity and vigor decline over their lifetime [167]. Since Nosema causes premature worker death, queen fertility is important as lost workers must be replaced. Researchers tested whether forcing Nosema infested colonies to rear new and potentially more fertile queens could ultimately reduce N. ceranae and N. apis infestation [168]. Queen replacement did reduce the percentage of parasitized workers in colonies. However, there were a number of short-term, negative effects, including a reduction in the number of adult bees and therefore lower food stores from reduced foraging rates. Breaking the brood cycle with forced queen replacement might be beneficial in controlling brood parasites, including Varroa mites, but the reduced food stores may render the colony more sensitive to other stressors (see below). Thus, beekeepers must employ this strategy carefully, during peak blooming periods with abundant nutritional resources, which would allow the colony to rapidly recover. Nutrient supplementation Improving overall colony nutrition may help bees cope with multiple abiotic and biotic stressors, including Nosema spp. Honey bees derive all requisite nutrients from consumption of nectar and pollen. Nectar serves as an important source of carbohydrates and other micronutrients while pollen provides protein, fats (including essential sterols), vitamins and

31 minerals. Since honey bees are generalists, diverse (multifloral) nectar and pollen promotes colony health [169]. Multiple studies have underscored the role of pollen (or protein supplements) in adult worker health and longevity (as reviewed in [170]). For example, caged workers fed on polyfloral pollen diets exhibit higher constitutive levels of immune biomarkers and other physiological parameters related to immunity compared with workers fed on monofloral diets [171]. In addition, caged pollen-fed workers had lower titers of naturally acquired DWV infections than workers fed on sugar water alone suggesting that higher protein diet augments worker ability to suppress viral infection [172]. Moreover, pollen ingestion activates detoxification molecular pathways in caged adult workers [173]. These pathways also help workers process ingested pesticides and thus diets incorporating pollen can improve worker survival after pesticide exposure. Several cage experiments have highlighted a complex relationship between Nosema parasitism, worker nutrition and severity of Noseomosis symptoms. In short, high protein diets increase both the longevity of caged workers and the spore loads that they carry [ ]. In these experiments, newly emerged workers were fed high quality protein diets (bee bread or pollen) as opposed to protein supplements (in the form of amino acid mixtures) and/or sugar water before infection with either N. apis or N. ceranae (0-7 days post-emergence). After an incubation period, pollen/bee bread fed workers had higher levels of protein in their blood and lived significantly longer than bees fed on poorer quality diets. Furthermore, pollen diversity improved survival of workers infected with N. ceranae [178]. However, across studies, bees fed on high quality protein also had greater spore loads. This suggests that good nutrition improves workers ability to tolerate rather than resist (actively suppress) infection [139]. Also, several of these studies indicated that feeding workers high quality pollen extended worker life span more than Nosema reduced worker survival, accentuating the importance of nutrition in worker longevity. These cage experiments have interesting implications for colony health. Results suggest that supplementing colonies with pollen might improve worker longevity even in the presence of Nosema. However, pollen supplementation might also raise spore counts, potentially increasing parasite transmission within and between hives. Unfortunately, few studies have examined the relationship between pollen availability and outcomes of Nosema infection in the field. One study supplemented colonies with pollen or protein during worker larval development and part of adulthood and inoculated newly emerged adult workers with N. apis (controls were fed sugar water) [179]. As expected, N. apis infection decreased worker life expectancy across feeding regimens, but contrary to cage experiments, pollen and/or protein supplementation did not improve infected worker survival. Spore counts were not compared across diet regimens so it cannot be determined if superior diet in the field amplified parasite reproduction as in caged studies. In subsequent field analyses, pollen/protein availability during larval development positively impacted worker longevity while N. apis had no effect when workers reared in supplemented or deprived colonies were housed in the same observation colony as adults [179]. Collectively, these field experiments show that the outcomes of worker nutrition and infection status can be strongly affected by colony context and do not always reflect predictions based on cage studies. Adequate nutrition is essential to colony health and productivity, but additional field studies are needed to determine if and when nutritional supplementation in the field can ameliorate Nosema infections. For example, Spanish investigators gave asymptomatic, N. ceranae infested colonies one of four treatments: 1) pollen supplementation and fumagillin treatment; 2) pollen supplementation only; 3) fumagillin treatment only; 4) no treatment (control). Researchers found 20

32 no difference in overwintering survivorship across experimental groups [180]. However, the authors noted that the winter was mild and workers had access to flowering plants. They concluded that if colonies could forage on high quality natural food sources and Varroa populations were controlled, treating for Nosema through supplementation or fumagillin application was an unnecessary expense Summary and future directions The collision of apiculture with modern stressors induced by human modifications to earth s biomes has created many challenges for the sustainable maintenance of healthy honey bee colonies. Colonies are subject to dynamic permutations of stressors including exposure to pathogens and parasites, pesticides, poor nutrition, habitat destruction and global warming. Beekeepers must track moving management targets as these stressors change and interact over time within and between colonies. In this context, both longstanding European honey bee (Apis mellifera) parasites such as Nosema apis and emerging parasites including Nosema ceranae can acquire emergent properties as they intermingle with other colony stressors and co-evolve with their honey bee hosts and the interventions applied by beekeepers. Thus, constructing Integrated Pest Management (IPM) plans that counter costs of Nosema infection and foster colony health and productivity is challenging (Box 1). Here, we discussed the current state of research on Noseomosis etiology and summarized steps to creating an IPM strategy for managing Nosema spp in honey bee colonies. Before such an IPM plan can be tactically deployed, however, additional research and commercial production of affordably priced tools are needed. First, beekeepers need science-backed recommendations for when and how to sample for Nosema infection and accessible and time-saving methods for detecting parasites. Molecular technology such as dipstick tests that yield both spore loads and identification of Nosema spp. would provide an affordable diagnostic method to beekeepers. Next, based on sample results, beekeepers need science-based directives for when to treat Nosema infections. Given the global variability in reports of damage inflicted by Nosema parasites and the inconstant nature of spore loads in colony samples, establishing treatment thresholds will require intensive field studies and likely the development of novel sampling methodology. Furthermore, due to worldwide heterogeneity in disease virulence, future treatment recommendations may vary regionally. Finally, novel interventions for Noseomosis management are needed. Fumagillin is currently the only well-studied treatment that reduces spore loads in colonies, but new research indicates that it may have problematic long-term effects on parasite populations and unintended health consequences for bees. Furthermore, fumagillin use is banned throughout most of Europe. New molecular based therapies such as RNAi or microbe- or plantderived compounds may suppress Nosema reproduction while limiting the risks to both bees and humans. However, these new treatment options are in the early stages of development and require rigorous tests of efficacy and investigation of non-target effects in the field before recommendations can be made. A challenge posed by all chemical interventions (existing or hypothetical) to-date is that only parasite vegetative states are targeted, meaning that Nosema diseases may re-emerge from reservoir spores contaminating colony equipment after treatments wear off. Thus, the number of times treatments must be applied to effectively suppress infection must be factored into cost-effectiveness analyses and management plans to reduce the chances of selection for treatment resistant Nosema strains. Gamma radiation could be employed to

33 deactivate spores contaminating colony equipment, however, this treatment option presents logistical challenges. Preventative measures to manage Nosema infection include honey bee breeding programs. If successfully implemented at large scale, selective breeding could help bees to cope with multiple stressors and would reduce the overall need to treat colonies, providing a long-term cost-savings. However, comprehensive selective breeding would require a large initial input of resources and continued government and scientific oversight to maintain funding and ensure efficacy. Other management efforts to improve honey bee nutrition through landscape diversification will likely yield additional returns by benefiting both managed and native pollinator populations which face many of the same or similar stressors imposed on honey bees [181]. Knowledge gained through characterizing Nosema infection in honey bees and successful development and implementation of IPM practices may serve as a template for best management practices for other honey bee diseases, including those caused by other emerging intestinal parasites. Indeed, recent reports indicate that the trypanosomal parasite Lotmaria passim is globally distributed in European honey bees and may be linked to colony loss [41, 182]. Improved Nosema control may also benefit wild pollinators by reducing pathogen spillover and/or providing means to help control microsporidia in other pollinator populations. Finally, honey bees and their microsporidian parasites may serve as a model disease system for human microsporidiosis and knowledge gained through honey bee studies may be translated to human medical practices [ ]. 22 Box 1: Integrated Pest Management In agricultural systems, Integrated Pest Management or IPM refers to a pest management paradigm where multiple, complementary strategies are implemented to control disease agents or pest populations (see [186] for a review). Farmers practicing IPM monitor crop damage incurred by pests and intervene only when pest levels reach a predetermined economic threshold (ET), where intervention is necessary to prevent additional crop damages which would be more costly than taking action to control pest activity (EIL: Economic Injury Level). Depending on the crop, farmers may have multiple means of controlling pests at their disposal, including: chemical pesticides, biocontrol agents (i.e. pathogens or parasites of the pest insect), introduction of natural enemies of the pest insect (e.g. introducing lady bugs), growing genetically modified crops or resistant cultivars that inhibit insect feeding, baiting pest populations with pheromones to disrupt mating, and landscape management practices that foster pest control and/or provide refuge for natural enemies of pests. Treating crops only when necessary is cost-effective and reduces the chance of contaminating adjoining natural habitats or residential communities connected to the agricultural system with chemicals. In addition, limiting pesticide applications reduces damage to off-target, beneficial insects such as honey bees. Finally, judicious application of pesticides and taking a cocktail approach to pest management reduces the chances that pest populations will develop resistance to a particular treatment. This extends the useful lifetime of available control technologies. Thus IPM strategies seek to balance the need for sustainable farming approaches from both economic and environmental perspectives [187].

34 Box 2: Microsporidia 23 Microsporidia comprise a group of fungal parasites. Recent phylogenetic studies place microsporidia either as a basal clade to all fungi or a sister group to the zygomycetes [188, 189]. Microsporidian species exhibit narrow to broad host spectrums and collectively parasitize diverse hosts ranging from single-celled protists to vertebrates, including humans (reviewed in [189]). Microsporidia parasitize several economically important, beneficial and pest animals. For example, microsporidia can infect salmon, reducing their market value. Likewise, microsporidia jeopardize honey bee, bumble bee and silk worm colony productivity and survival [189, 190]. However, microsporidia may be seen as beneficial in some agricultural systems where they infect pests [189]. Depending on the species, microsporidia can be transmitted horizontally (fecal/oral) and/or vertically (mother to offspring). Naïve hosts are infected after exposure to tough, thickwalled microsporida spores (see for [191] a review). Once having gained access to a host (e.g. after being eaten), host chemical cues signal to the spores that they are in the proper environment for germination. When a spore germinates, it extrudes an internally coiled tube called the polar filament which penetrates a host cell membrane. Next, the spore contents (sporoplasm) are pushed through the polar filament and delivered directly to the host cell cytoplasm where the sporoplasm begins to reproduce. The resulting parasite progeny, termed vegetative states, undergo a number of reproductive phases before ultimately forming new spores. As obligate parasites, microsporidia are completely dependent on their hosts for furnishing the proper environment and energy for reproduction, which is reflected by their compact and simplified genomes and loss of some internal structures [192]. For example, microsporidia lack mitochondria, cell structures that efficiently produce large quantities of energy molecules (ATP) via oxidative phosphorylation. Instead, microsporidia marshal host cell mitochondria from which they commandeer ATP [193]. Thus, while chronic microsporidian infections may be slow to build, infections are frequently energetically costly for hosts. Box 3: Honeybee colony demographics and division of labor (see [167] for a review) Honey bees are social insects that inhabit densely populated colonies consisting of approximately 10,000-50,000 individuals. As is common in social insect systems, honey bees divide tasks amongst colony members. Each colony has one mated queen who is the sole reproductive individual. She lays 1,000-2,000 eggs per day during active summer months. The majority of the remaining individuals that comprise a colony are worker bees, the queen s daughters. Adult workers perform all colony tasks other than reproduction, with worker age serving as a general template for labor responsibilities. Newly emerged adult workers (nurses) tend to the needs of their larval siblings. As they age, nurses ultimately transition to foraging behavior (a process termed behavioral maturation), leaving their nest to collect nectar, pollen, water or propolis (tree sap). While transitioning from nursing to foraging behavior, workers may perform intermediate tasks such as tending the queen, storing food, building comb, removing dead workers from the colony and/or guarding the nest entrance, though different individuals may perform different tasks. Though roughly governed by age, shifts in worker behavior are sensitive to colony needs and can be elastic depending on context. Male honey bees (drones) are produced

35 during summer months. They contribute to their colony s fitness only if they successfully mate with a virgin queen from a different colony Acknowledgements We would like to thank Dr. Judy Chen (USDA-ARS, Beltsville, MD) for contributing expert knowledge on Nosema spp. genomics and members of the Grozinger lab for critical reading of this manuscript. This material is based upon work supported by the National Science Foundation under Grant No. DGE to HLH. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. Additional funding was provided by USDA-AFRI (PI J Chen, copi: CMG).

36 Chapter 2 25 Effects of immunostimulation on genome-wide gene expression in honey bee workers (Apis mellifera) Holly L. Holt 1, Freddie-Jeanne Richard 2,3 and Christina M. Grozinger 1,3 1 Department of Entomology, Center for Pollinator Research, Center for Chemical Ecology, Huck Institutes of the Life Sciences, Pennsylvania State University, University Park, USA 2 Laboratoire Ecologie Evolution Symbiose, UMR CNRS 6556, University of Poitiers, 40 avenue du Recteur Pineau, Cedex, F-86022, Poitiers, France 3 Previous address: Department of Entomology, North Carolina State University, Raleigh, NC 27695, USA This chapter has been modified from a previously published manuscript (Richard, F.J., Holt, H.L. and C.M. Grozinger (2012). "Effects of immunostimulation on social behavior, chemical communication and genome-wide gene expression in honey bee workers (Apis mellifera)." BMC Genomics 13(1): 558.) available online. In this original peer-reviewed article, FJR performed the treatments, collections, behavioral analyses and chemical analyses. CMG performed the microarray analysis and HLH analyzed the microarray data and identified candidate genes. FJR and CMG designed the study. FJR, HLH, and CMG wrote the manuscript. All authors read and approved the final manuscript. Here, findings from molecular studies performed by HLH and CMG are summarized. Abstract Social insects, such as honey bees (Apis mellifera), use molecular, physiological and behavioral responses to combat pathogens and parasites. The honey bee genome contains all of the canonical insect immune response pathways, and several studies have demonstrated that pathogens can activate expression of immune effectors. Here, we test the specificity of honey bee genomic responses to immunostimulation by challenging workers with a panel of different immune stimulants (saline, Sephadex beads and Gram-negative bacteria E. coli). Immunostimulation caused significant changes in expression of hundreds of genes, the majority of which have not been identified as members of the canonical immune response pathways. Furthermore, several new candidate genes that may play a role in cuticular hydrocarbon biosynthesis were identified. Effects of immune challenge on expression of several candidate genes involved in immune response, cuticular hydrocarbon biosynthesis, and the Notch signaling pathway were confirmed using quantitative real-time PCR. Finally, we identified common genes

37 regulated by pathogen challenge in honey bees and other insects. These results demonstrate that honey bee genomic responses to immunostimulation are substantially broader than the previously identified canonical immune response pathways, and may mediate the behavioral changes associated with social immunity by orchestrating changes in chemical signaling. These studies lay the groundwork for future research into the genomic responses of host honey bees to native honey bee parasites and pathogens Introduction 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 [194]), and the social colony environment (with up to 50,000 densely packed worker bees [195]) provides excellent conditions for disease transmission. While the innate immune systems of invertebrates and vertebrates are surprisingly conserved [196], 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 [ ], bees have a much smaller number of immune response genes relative to Drosophila [202], which is one of the best characterized models of insect immunity [203]. 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 [204]). 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 [194]). 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 [ ], as well as a number of bacterial and fungal pathogens [ ] including the gut microspordian parasites Nosema apis and Nosema ceranae. Honey bees are also severely impacted by Varroa mites (Varroa destructor) and tracheal mites (Acarapis woodi), and reviews of these different parasites have been published, see [212]). Sequencing of the honey bee genome identified 177 genes associated with the canonical immune response pathways in insects [202]. 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 [203, ]). 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 gramnegative 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

38 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 [218]. Furthermore, canonical pathways derived from Drosophila experiments, may not be fully generalizable to other systems. Changes in individuals gene expression may also affect defense mechanisms employed by honey bees at a social level. For example, honey bees display a range of behavioral mechanisms termed social immunity to reduce the impacts of parasites and diseases (reviewed in [204, 219]). Honey bees can also use cuticular hydrocarbons as chemical cues to distinguish between healthy and immune-stimulated adult nestmates, and respond differently to them [220]. Cuticular hydrocarbons are synthesized in oenocytes, which are embedded in the fat body tissue under the epithelia, and deposited on the cuticular surface [221]. Cuticular hydrocarbon patterns can be modified by genotype, physiological state, and environmental context (reviewed in [222, 223]), including social status in honey bees [224]. For example, 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 [220]. Here, we examined the 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; for example, theoretically only E. coli injection should activate the Imd pathway. We also determined if immunostimulation resulted in significantly altered expression of previously annotated honey bee immune genes [202], if there was significant overlap with other studies of the effects of immunochallenge or parasitization in honey bees and Drosophila [ ], and if there were changes in expression of genes associated with cuticular hydrocarbon biosynthesis pathways Methods 2.1 Honey bee stocks Honey bee colonies were maintained according to standard practices at the Lake Wheeler Honey Bee Research Facility at North Carolina State University in Workers for these studies were obtained from three source colonies, headed by queens each instrumentally inseminated with semen from a single, different, male (Glenn Apiaries, Fallbrook, CA). The specific source colonies used for the different experiments were Colonies 1, 2 and 3, and are listed in the experimental details below. Since honey bees are haplodiploid, the coefficient of relatedness among the workers in each colony was 0.75, thereby reducing variation in gene expression responses, chemical profiles, and presumably behavioral responses among the nestmates. Honeycomb frames of late-stage pupae were removed from the colonies and placed in incubators overnight at 33 C, 50% relative humidity (RH).

39 2.2 Honey bee rearing 28 Newly emerged bees (<12 hours old) were brushed from the frames and placed into modified 10 cm Petri dishes in groups of 15. Dishes were maintained under red light in a temperature and humidity-controlled environmental room (33 C, 50% RH, Phytotron Facility, NCSU). Bees were fed 50% sucrose/water solutions and 45% honey/45% pollen/10% water paste ad libitum. Food was replaced every two days. Bees were also exposed to 0.1 queen bee equivalents of queen mandibular pheromone (QMP) (Pherotech, Vancouver, Canada). Every day, 10 μl of QMP (0.01 queen equivalents/µl in an isopropanol/1% water solution) was placed on a microscope slide and allowed to evaporate before being placed in the cage. This amount of QMP mimics a live queen in assays of worker behavior and physiology [229, 230] and thus should help simulate normal rearing conditions. 2.3 Experimental treatment When the bees were 10 days old, five individual bees were removed from each cage (leaving 10 in the cage). Three of these were then subjected to one of four treatments. The first group of bees was handled and anesthetized with CO 2 for 1 minute (control treatment). A second group of bees was anesthetized and injected with 8 μl of sterile bee saline (130 mm NaCl, 6 mm KCl, 4 mm MgCl 2, 5 mm CaCl 2, 160 mm sucrose, 25 mm glucose, 10 mm 4-(2-hydroxyethyl)- 1-piperazineethanesulfonic acid in distilled water, ph 6.7, 500 mosmol, as in [220]). A third group of bees was anesthetized and injected with 8 μl of a CM-25 Sephadex beads (Sigma- Aldrich, Steinheim, Germany) solution (0.01 g of beads mixed with 500 μl sterile bee saline and vortexed, resulting in approximately 110 beads/8 μl). The fourth group of bees was anesthetized and injected with 8 µl of an E. coli (JM101, Sigma, St Louis, MO) solution suspended in sterile bee saline (3.8*10 5 cells/bee, following a protocol modified from [231]). The bacteria were grown in LB media (Sigma, St Louis, MO), collected and resuspended in the sterile bee saline at the appropriate concentration (4.75*10 4 cells/μl), and then stored at -80 C until it was thawed for the injections. Injections were performed into the abdominal cavity through tergites with a nanoinjector equipped with a glass needle (Schley Compact Model II Instrument; Honey Bee Insemination Services, Davis, CA, US), using a binocular microscope (Leica MZ6 stereomicroscope, Leica Microsystems, Buffalo Grove, IL). Treated bees were marked with a dot of Testor s paint on their thorax and maintained individually in 10x10x7 cm 3 Plexigas cages in an incubator under red light at 33 C, 50%RH for 6 hours, with 50% sucrose. After 6 hours, one treated bees per cage was collected onto dry ice and stored at -80 C for microarray analysis. Microarray analysis was performed on 6 individuals/treatment in Colony 1, and 4 individuals/treatment in Colony Microarrays Individual bees were thawed and dissected under cold RNAlater (Qiagen, Valencia, CA). Abdomens were eviscerated and the cuticles, with the associated fat bodies and oenocytes, were collected. RNA was extracted using an RNeasy kit (Qiagen). RNA was quantified using a Nanodrop 1000 spectrophotometer (Thermo Fischer Scientific, Wilmington, DE) and sample

40 integrity and quality was monitored using agarose gel electrophoresis to confirm the presence of ribosomal RNA bands. 500 ng of RNA/individual was amplified using the Ambion MessageAmp II arna Amplification kit (AM1751, Life Technologies, Grand Island, NY). 5 µg of amplified RNA from each sample were labeled independently with Cy3 and Cy5 dyes using the ULS arna fluorescent labeling kit (EA-006, Kreatech, Amesterdam, Netherlands). Samples were hybridized to microarrays (two samples/array) in a loop design with dye swaps incorporated, using 24 microarrays for Colony 1 (6 biological replicates, 2 technical replicates per sample) and 16 arrays for Colony 2 (4 biological replicates, 2 technical replicates per sample). Whole genome microarrays were purchased from the W.M. Keck Center for Functional Genomics at the University of Illinois, Urbana-Champaign. Arrays were scanned using the Axon Genepix 4000B scanner (Molecular Devices, Sunnyvale, CA) using GENEPIX software (Agilent Technologies, Santa Clara, CA) Analysis of Gene Expression Any spots with an intensity of less than 100 (the background level on the arrays) were removed from the analysis. Also, spots present on less than 12 out of 24 arrays for the first colony and 8 out of the 16 arrays for the second colony were excluded from the data set as well. Expression data was log-transformed and normalized using a mixed-model ANOVA (proc MIXED, SAS, Cary, NC) with the following model: Y = μ + dye + block + array + array*dye + array*block + є where Y is expression, dye and block are fixed effects, and array, array*dye and array*block are random effects. Transcripts with significant expression differences between groups were detected by using a mixed-model ANOVA with the model: Y = μ + treatment + spot + dye + array + є where Y represents the residual from the previous model. Treatment, spot and dye are fixed effects and array is a random effect. p-values were corrected for multiple testing using a false discovery rate < 0.01 (proc MULTTEST, SAS). Hierarchical clustering, using the ward method was performed in JMP (SAS, Cary, NC). Gene ontology analysis was performed using DAVID version 6.7 [232, 233] with a cutoff of p <0.05. For all gene ontology (GO) analyses, array transcripts were matched to their Drosophila orthologs in Flybase. All of the array transcripts with Drosophila orthologs were used as a background list. Analysis of overlap between the significantly regulated transcripts in our study or between significantly regulated genes in our study and other gene expression studies was performed using Fisher s Exact Tests (Dr. Oyvind Langsrud, Statistics Norway, < Common, differentially expressed transcripts or genes between studies were identified with Venny [234]. For overlap comparisons between our study and other studies examining gene expression in honey bees ([202, 225, 227]), we converted microarray transcript identifiers (AM numbers) to GB identifiers from the honey bee genome annotation (see BeeBase

41 For comparisons between our study and other studies examining gene expression in fruit flies ([226, 228]), we converted all genes to Flybase identifiers to find common overlap. The array data has been deposited on the ArrayExpress website (E-MEXP-3708) Validation of candidate gene expression patterns using quantitative real-time PCR. Worker bees from a third source colony were reared as before. Control, saline-injected, and bacteria-injected treatment groups were generated (with 9, 8 and 10 individual bees per treatment group, respectively) as before, but the injected E. coli bacteria were collected from an actively growing culture just prior to injection. Bees were collected onto dry ice four hours after treatment and stored at -80 C. RNA was extracted from eviscerated abdomens of individual bees as above. cdna synthesis and quantitative real-time PCR was performed as in [235]. Expression levels of candidate genes were normalized to actin. Significant differences in expression levels among treatment groups were determined using an ANOVA followed by posthoc pairwise comparisons with Tukey HSD tests (JMP 9.0.2, SAS, Cary NC). Primer sequences are given in Table 1.

42 Table 2-1 Primer sequences used to quantify expression of candidate genes using quantitative real-time PCR. 31 Gene Name Primer Direction Sequence Accession number/ reference Lipid Storage Droplet 2 Forward Reverse AATCGTGCTCGCAAAATACC TTCGCAATAGGTAATGATTTTTCA XM_ Bubblegum Forward Reverse TTTTGCAAATCTTGGTGCAA TTCTTCTGGTTTGCCTTCGT XM_ Pale Forward Reverse TACTTCGTCGCGGATAGCTT TACGCGAACGATGTCTTCAG NM_ Groucho Forward Reverse ACTGTGGACAGGAGGATTGG ATGAAGGACCTCGACGTTTG XM_ Draper Forward Reverse ATCCTGCCACTGGTTATTGC AGTCAACTCCCTCCCAACCT XM_ Apterous Forward Reverse AGGAAGAGGAAACCCAAGGA TCTGCGAGAGTTGCTTCAGA XM_ Pebbled Forward Reverse CGCTGATACGTCACCTCAGA GCGAACTTCTCTCCACGTTC XM_ Defensin 1 Forward TGCGCTGCTAACTGTCTCAG (Evans et al, 2009) Reverse CGAAACGTTTGTCCCAGAG NM_ Actin Forward Reverse CCTAGCACCATCCACCATGAA GAAGCAAGAATTGACCCACCAA (Huising & Flik, 2005)

43 3.0 Results Global gene expression responses to immunostimulation Microarrays were used to monitor global gene expression patterns in the eviscerated abdomens (containing epithelial tissue, fat bodies, and oenocytes) of bees from the four treatment groups. We examined all transcript expression levels across all pairs of treatment groups for significant differences (control x saline, control x bead, control x bacteria, saline x bead, saline x bacteria, bead x bacteria) and found that 670 unique transcripts were significantly regulated among treatment groups in Colony 1, and 1610 unique transcripts were significantly regulated among treatment groups in Colony 2 (FDR <0.01). Thus, expression levels for these transcripts were significantly different between at least two of the treatment groups. Of these, 302 transcripts were significantly regulated in both colonies. This overlap in transcript expression was greater than expected by chance (Fisher s Exact Test; p<0.001). The higher numbers of transcripts in Colony 2 may be due to the lower number of biological replicates (4 vs 6 replicates) or the effect of genetic background, which can strongly affect responses to immunostimulation [236, 237]. Despite some differences in relative expression patterns for individual transcripts across colonies, hierarchical clustering of the 302 common transcripts revealed the same overall clustering of treatment groups, (Figure 1). Treated groups clustered separately from control groups, and saline- and bead-injected groups clustered separately from the bacteria- injected group.

44 Figure 2-1 Hierarchical clustering of significantly regulated genes. We performed hierarchical clustering analysis on the 302 significantly, differentially expressed transcripts (FDR<0.01) similarly regulated in Colony 1 (A) and Colony 2 (B). Both colonies demonstrated the same overall grouping for experimental treatments: saline (S) and bead (B) injected bees formed a sister group and were closer in transcript expression to bacteria (Bac) injected bees than controls (C). Control and bacteria-injected bees had the most disparate transcript expression. Colors denote differences in log2 expression relative to the mean expression across the four treatment groups, according to the scale shown. 33

45 3.2 Effects of individual immune elicitors on gene expression Pairwise comparisons identified sets of transcripts differentially regulated between saline-, bead-, bacteria-injected bees and the control bees (FDR < 0.01, Table 2). Overlap between colonies for each treatment-control comparison was significantly greater than expected by chance (Table 2). 111, 70, and 117 transcripts, respectively, were significantly regulated in these pairwise comparisons in both colonies, though these transcripts did not necessarily show the same directional patterns of expression between colonies. While each treatment resulted in significant expression changes in a unique set of transcripts relative to controls, there was considerable overlap across treatment groups (Figure 2). Indeed, 22 transcripts were significantly regulated by all three treatments. Table 2-2 Analysis of overlap among treatment groups 34 Treatment group # of significantly, differentially regulated transcripts Colony 1 Colony 2 In both p-value (Fisher's Exact Test) Saline x Control p < Bead x Control p < Bacteria x Control p < Total *557 *1453 *206 *duplicate transcripts between groups removed Figure 2-2 Effects of specific immunostimulants on gene expression. Pairwise comparisons identified sets of transcripts differentially regulated between saline-, bead-, bacteria-injected bees and control bees in both colonies. A Venn diagram demonstrates that there is considerable overlap in the effects of each treatment, but also treatmentspecific effects on gene expression. The numbers represent the number of genes in each category.

46 3.3 Functional analysis of regulated genesgene ontology (GO) analysis of the significantly regulated 302 transcripts found in both colonies (207 of which had unique Drosophila orthologs with Flybase annotations and thus were used in the analysis) revealed an overrepresentation of genes involved in immune response, spermatogenesis, wing disc dorsal/ventral pattern formation, tissue development, post-transcriptional regulation of gene expression and protein polymerization (p<0.05). However, none of these categories survived the Benjamini correction. Immune genes corresponding to the major immune pathways were significantly regulated (defensin-1, relish, domeless, cactus, melanization protease 1, death related ced-3/dredd, PGRP-SC2, kayak, spirit among others). Pale (ple), a tyrosine hydroxylase involved in melanization and wound repair, was also significantly regulated by bacteria, bead and saline injection in Colony 1 and bacteria and bead injection in Colony 2 [238]. Other genes included those involved in cell growth and proliferation (insulin-like receptor), cytoskeleton structure (basigin, chickadee, twinstar, annexin ix, and isoforms of tubulin), extracellular matrix components (pericardin and laminin A), Notch signaling (apterous, pebbled, groucho), phagocytosis (draper), and cabut, which encodes a transcription factor that is regulated by the JNK cascade [239]. Of the 22 transcripts (corresponding to 17 unique Flybase genes) significantly regulated by injection with saline, beads, and bacteria relative to controls, two categories were significantly overrepresented: organ morphogenesis, p<0.005, and developmental process, p< Again, neither of these categories survived the Benjamini correction. Several of the previously discussed genes (domeless, insulin-like receptor, basigin, twinstar, and groucho) are part of this group, as well a serine protease immune response integrator (spirit) which functions in Toll pathway activation [240]. Notably, two genes involved in lipid metabolism were also found in this group, which is of particular interest since fatty acids are the precursors of cuticular hydrocarbon synthesis in insects (reviewed in [241]). Bubblegum (bgm) encodes a very long chain fatty acid CoA ligase, which plays a role in fatty acid metabolism [242]. Lipid-storage droplet 2 (Lsd-2) is involved in lipid storage and accumulation [243]. Lipid storage droplet 1 (Lsd-1), a component of lipid droplets in fat bodies [244], was found to be significantly regulated in both colonies, but not by all treatments. GO analysis of the 68 transcripts whose expression was specifically altered in bacteriainjected workers (corresponding to 50 unique Flybase genes) yielded only one significantly overrepresented cluster (immune response; p=0.001). These genes included cactus, kayak, draper, defensin-1, relish, PGRP-SC2. Other regulated genes included those that may be involved in metabolism, such as Gr28b, a gustatory receptor that seems to mediate diet-related changes in immune response [245], sorbital dehydrogenase-2 (Sodh-2), an alcohol dehydrogenase of sugars [246], and trehalose-6-phospate synthetase 1 (Tps1) which may play a role in protection from hypoxia and anoxic injury [247]. 35

47 3.4 Comparisons of gene expression patterns with previous studieswe examined overlap in gene expression patterns between our study and previous studies in honey bees and Drosophila. It is important to note that there were large differences in study design (including pathogen challenge, timecourse and tissue) as well as gene expression platform and statistical analyses, which makes direct comparisons challenging. However we sought to identify common regulated genes that could indicate conserved immune response mechanisms across different host organisms and pathogen challenges. Fourteen genes were found in common between the significantly regulated 302 transcripts (239 of which were annotated honey bee genes with GB identifiers) and the canonical immune response genes annotated from the honey bee genome by Evans and colleagues [202]. This overlap was significantly greater than expected by chance (Fisher s Exact Test, p<0.0001; Table 3). However, clearly the majority of significantly regulated genes are not members of these canonical immune response pathways. Table 2-3 Overlap of gene lists between studies 36 Gene List # of significant transcripts # of transcripts in common between studies p-value (Fisher's Exact Test) Evans et al. (2006) p < Navajas et al. (2008) 22 2 p > 0.09 Alaux et al. (2011) p < The 302 genes that were significantly regulated by immunostimulation in both colonies in our study were matched to 239 GB numbers and compared with the immune genes identified during annotation of the genome (Evans et al., 2006), and genes significantly regulated by Varroa parasitization of honey bees (Navajas et al., 2008 and Alaux et al., 2011). A Fischer s Exact Test was used to determine if this overlap was significant, assuming that all genes present in the background lists for the arrays were in common between studies. In order to ascertain overlap, only genes in previous studies that were present on our array platform were included. Navajas et al [227] performed a microarray analysis of control and Varroa miteparasitized honey bee pupae from two strains (resistant and sensitive to Varroa mite infestation). Expression levels of 32 genes (with 22 matching unique GB identifiers) were significantly altered by Varroa parasitization. There was no significant overlap with the 302 transcripts (matched to 239 GB identifiers) in the current study (Fisher s Exact Test, p>0.09; Table 3). However, two genes, including ple, were regulated in both studies, though directionality was not necessarily conserved between studies. In a separate study, Alaux and colleagues examined the effects of Varroa parasitization and nutrition on honey bee worker gene expression using an RNA-seq approach [225]. We compared our 302 regulated transcripts (matched to 239 GB identifiers) with the list of genes that were significantly, differentially regulated by mite parasitization in pollen-fed bees and found 117 overlapping genes, including genes involved in immune and metabolic processes. Though directionality was not necessarily conserved between studies, the overlap was greater than expected by chance (Fisher s Exact Test, p<0.0001; Table 3). Bgm, Lsd-2, and domeless were

48 among the genes significantly regulated in both studies. A GO analysis of the overlapping transcripts (corresponding to 102 unique Flybase genes) found one significantly overrepresented category which did not survive Benjamini correction (metabolic process, p<0.04). Finally, we compared the 302 significantly regulated transcripts in our study (matched to 207 unique Flybase identifiers) with significantly regulated genes in two separate studies examining immune response in fruit flies. De Gregario and colleagues [226] identified 400 significantly regulated genes after septic injury of male Drosophila with bacteria contaminated needles or feeding with fungal spores (corresponding to 497 unique Flybase genes). Roxstrom- Linquist and colleagues [228] identified approximately 390 upregulated genes (corresponding to 464 unique Flybase genes) in Drosophila melanogaster orally infected with protozoa, viruses, bacteria or fungi. Eight genes were found in all three studies, including a number of genes with immune related functions: ple, cactus, defensin-1, relish and PGRP-SC2. Serpin 28D (CG7219) was also regulated in all three studies, and is involved in melanization [248] Quantitative real-time PCR validation of expression of candidate genes We examined expression levels of several candidate genes identified in the microarray study in a third biological replicate (Figure 3). Three treatment groups were used: control, salineinjected, and bacteria-injected worker bees. As in the microarray study, defensin 1 levels increased with treatment, and were significantly higher in bacteria-injected bees relative to controls, and intermediate in saline-injected bees (F(2,24)=6.49, p=0.0056, see Figure 3 for results of Tukey HSD post-hoc pairwise comparisons). As in the microarray study, levels of bubblegum decreased significantly with treatment (F(2,24)=32.56, p<0.0001), while levels of Lsd-2 (F(2,24)=392.1, p<0.0001) and pale increased (F(2,24)= 27.98, p<0.0001). Similar to the results of the microarray analysis, the effects of treatment on expression of three genes involved in Notch signaling were less dramatic, but nonetheless expression of apterous (F(2,24)=6.76, p=0.0047), groucho (F(2,24)=8.44, p=0.0017) and pebbled (F(2,24)=4.41, p=0.023) were significantly affected by treatment. Expression levels were higher in saline-injected bees than controls for all three genes, and higher in bacteria-injected bees relative to control bees for apterous.

49 38 Relative RNA levels (mean +/- se) Control Saline Bacteria 0 Figure 2-3 Quantitative real-time PCR validation of expression patterns of candidate genes. Expression levels of seven candidate genes (relative to actin) were analysed in a third biological replicate using quantitative real-time PCR. Mean expression levels for each treatment group are normalized to expression in the control treatment group, for graphical representation. Significant differences in expression levels across the three treatment groups were determined using an ANOVA with treatment as a variable, followed by post-hoc Tukey HSD pairwise comparisons. Different letters denote significant differences in expression (p<0.05). Nine, eight, and ten individual bees were used for the control, salineinjected, and bacteria-injected treatment groups, respectively. 4.0 Discussion Our studies demonstrate that general immunostimulation elicits complex gene expression changes in the epithelial tissues in honey bee workers. It is important to note that these dramatic changes occurred in a relatively short timeframe; only six hours after immune challenge. General immunostimulation resulted in significant expression changes of 302 common transcripts in worker bees from the two colonies examined. Importantly, only 14 of these genes corresponded to previously annotated immune genes identified from the honey bee genome [202]. Thus, 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 in this and other model systems (for example, [ ]). Futhermore, injection with saline, beads or bacteria resulted overlapping and unique changes in worker gene expression, suggesting that infection with natural pathogens may result in pathogen-specific changes in worker gene expression.

50 Several biological processes were modified by immunostimulation, 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 [217]. 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 [249]. 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 [225]), these changes may reflect general physiological changes after immunostimulation or stress [250]. 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 [203, 216, 217]). 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 overactivation of the IMD pathway [251]. 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 [239]. The Toll pathway operates through transduction factors including spirit, which acts extracellularly and upstream of Spaetzle [240] 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 [252]. 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 [241]), which are embedded in the fat body of adult honey bees [253]. 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 [241]. Fatty acids are stored in lipid droplets in the adipoctye cells of the insect fat body [250]. Fatty acids can be released from droplets in the adipocytes and accumulate in the oenocytes; this occurs under starvation conditions in particular [254]. 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 Figure 6), suggesting reduced movement of 39

51 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 [255]), 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 [242]. 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 [256]. In our study, bgm expression was significantly decreased relative to controls (see Figure 6). 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 [225, 227]. 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 [257]. 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 [204]. Comparison with two previous studies [226, 228] 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 [228] orally infected young male flies with bacteria, fungi and microsporidia, and measured whole-body gene expression changes in only two replicates using Affymetrix microarrays. Companion chemical and behavioral analyses to these molecular studies have been previously published as outlined at the beginning of this chapter. However, detailed mechanisms behind how changes in gene expression may lead to changes in worker cuticular hydrocarbon profiles and elicit subsequent changes in behavior by infected workers and/or nestmates require future studies Conclusions Our results suggest that immunostimulation of honey bee workers causes significant changes in gene expression patterns. 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

52 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 [223]. 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 Acknowledgements We would like to thank Davis Murphy for assistance with the cage experiments, Joe Flowers for expert beekeeping assistance, Coby Schal for use of his GC instruments, Mike Roe for the use of his GC-MS, Sarah Kocher for assistance with statistical analysis of the data, Elina Lastro Niño and Tracey Baumgarten for valuable assistance with the qrt-pcr validation, and members of the Grozinger lab for helpful discussions and critical reading of the manuscript. This research was funded by a USDA-NIFA-AFRI grant to CMG and FJR and an NSF Predoctoral Fellowship to HLH.

53 Chapter 3 42 Chronic parasitization by Nosema microsporidia causes global expression changes in core nutritional, metabolic and behavioral pathways in honey bee workers (Apis mellifera) Holly L. Holt 1,3, Katherine A. Aronstein 2 and Christina M. Grozinger 1 1 Department of Entomology, Center for Pollinator Research, Center for Chemical Ecology, Huck Institutes of the Life Sciences, Pennsylvania State University, University Park, USA 2 Honey Bee Breeding, Genetics & Physiology Lab, USDA, Baton Rouge, USA Chapter 3, a previously published original manuscript (Holt, H.L., K.A. Aronstein, and C.M. Grozinger. (2013). "Chronic parasitization by Nosema microsporidia causes global expression changes in core nutritional, metabolic and behavioral pathways in honey bee workers (Apis mellifera)." BMC Genomics 14(1): 799.), is included as Appendix A. HLH, KAA and CMG designed the study and interpreted the results. KAA conducted infection experiments, collected samples and conducted qrt-pcr. HLH performed the microarray experiments and analyzed data. All authors wrote, read and approved the final manuscript.

54 Chapter 4 43 Molecular, physiological and behavioral responses of honey bee (Apis mellifera) drones to infection with microsporidian parasites Holly L. Holt, Gabriel Villar and Christina M. Grozinger Department of Entomology, Center for Pollinator Research, Center for Chemical Ecology, Huck Institutes of the Life Sciences, Pennsylvania State University, University Park, USA HLH, GV and CMG designed assorted studies and interpreted the results. HLH conducted all experiments and GV conducted EAG assays. Abstract Nosema apis and Nosema ceranae are damaging intestinal pathogens of European honey bees (Apis mellifera). Nosema pathology has primarily been characterized in workers where infection is energetically costly and accelerates worker behavioral maturation. Few studies, however, have examined infection costs in drones to determine if Nosema similarly affects male energetic status and sexual maturation. We conducted a series of molecular, physiological and behavioral assays to characterize Nosema etiology in drones. Newly eclosed drones were infected with Nosema spores, placed in surrogate colonies, and assessed days postinoculation. Nosema infection is chronic and symptoms are commonly catalogued in workers starting at 10+ days post-infection. Nosema infection had no impact on drone sperm numbers or antennal responsiveness to a major queen sex pheromone component (9-ODA). Infection also did not alter drone metabolic rates over 4-minute flights. However, infected drones starved faster than controls and exhibited altered patterns of flight activity in the field, consistent with either energetic distress and/or altered rates of sexual maturation. Finally, the expression of candidate genes with metabolic, hormonal and/or immune functions, including members of the insulin signaling pathway, were regulated by parasitization. Interestingly, while drone molecular responses generally tracked predictions based on worker studies, several aspects of infected drone flight behavior contrasted with the flight behavior of infected workers. We compare Nosema pathology in drones with previous studies describing symptoms in workers and discuss ramifications for drone and colony fitness. 1.0 Introduction The production of healthy, virile drones (male honey bees) that successfully mate is a key component of honey bee (Apis mellifera) colony fitness [167, ]. However, drones may also sexually transmit or vector diseases between colonies if they drift from their natal nests [261]. Thus, drone health can have a large impact on the success of existing and future honey bee

55 generations. Despite their fundamental role in colony reproduction, drones remain an understudied colony demographic (see Box 1 for a summary of drone biology). Nosema apis and Nosema ceranae are damaging parasites of honey bees (Apis mellifera) (see Chapter 1 and [6] for review) that cause chronic intestinal disease. Both Nosema spp. belong to a class of fungal parasites called the microsporidia. These obligate parasites persist in the environment as dormant, durable spores. Once consumed by a honey bee host from a contaminated food or water source, the spores are carried through the bee s digestive system until they reach the midgut, a specialized tissue for digestion and absorption. Unidentified midgut cues cause spores to activate and penetrate host cells. Once inside a midgut cell, Nosema parasites begin replicating. New spores are produced which either invade neighboring midgut cells or are eliminated by the host, perpetuating the disease cycle [6]. Molecular, physiological and behavioral costs of Nosema infection have been investigated in worker honey bees, but few studies have characterized disease in drones. Pathology studies in workers show that infection damages midgut tissue [77], is energetically costly [78, 79] and accelerates the pace of adult worker behavioral development [87, 88] (see Chapters 1and 3 for a review). Early studies in drones suggest some parallels in disease etiology. For example, pre-adult and adult drones may naturally acquire N. apis or N. ceranae in a colony setting [102, 262] and N. ceranae-infected, caged drones are lighter and die more rapidly than controls [100], indicating infection costs. However, due to differences in life history, these castes may exhibit unique responses to the same pathogens. Not surprisingly, biomarkers of immune function differ between adult drones and workers [263]. Furthermore, since drones are haploid while workers are diploid, studies have postulated that drones may be more susceptible to infection in general. Indeed, caged drones inoculated with the same number of spores carry lower N. ceranae titers than workers, yet die sooner [100]. Also, infected drones disseminate disease in a density dependent manner to uninfected worker cagemates more easily than infected workers transmit disease to other workers held in the same cage [114]. Finally, to our knowledge, only one study has investigated whether Nosema infection (specifically N. apis) alters flight behavior of adult drones since adult worker flight behavior is modified [264]. Researchers found no effect of parasitization on first day of drone flight initiation or average flight duration. However study results must be interpreted with caution due to limitations in experiment design: This study only included one trial, and drones were only followed for a period of 10 days. Infection costs may have been missed due to variability in colony resilience to infection and/or the shorter follow-up period employed for this chronic disease. Taken together, these studies lay the groundwork for exploring caste-specific responses to Nosema infection. However, additional studies, measuring drone responses to infection in natural as opposed to cage settings are clearly needed to ascertain pre- and postcopulatory costs in male honey bees (see Box 1). Here, we characterize molecular, physiological, and behavioral responses of drones to infection with Nosema parasites. We hypothesized that if infection is energetically costly in drones as in workers, infected drones fitness may be reduced by several physiological and behavioral changes. Specifically, we predicted that infected drones would: 1) have lower sperm counts, 2) have a lower electroantennal response to a major queen pheromone component (9- ODA) that serves a sex attractant during mating flights 3) starve more rapidly, 4) have lower metabolic outputs during flight and 5) initiate flights later and fly for shorter periods of time than controls. Additionally, 6) we predicted that candidate molecular pathways that govern worker metabolic, behavioral and immune responses to infection would be modulated by Nosema infection in drones. Specifically, we tested changes in expression of seven genes (vitellogenin; ultraspiracle; insulin-like peptide 1; insulin-like peptide 2; forkhead box-containing protein O 44

56 subfamily; peptidoglycan recognition protein S3; and evolutionarily conserved signaling intermediate in Toll pathways ) that were previously shown to be regulated by Nosema infection in workers [85] Methods Experiments were conducted during the summer months (May-August) of Colony management and drone samples Honey bee colonies were maintained in Penn State apiaries according to standard beekeeping practices. To obtain drones for experiments, honey bee queens were caged onto honeycomb frames containing drone-sized cells for hours until solid egg patterns were observed on both sides of each frame. After laying eggs, queens were released and comb was retained in the host colony until developing drone larvae were capped. Afterwards, frames were transplanted to a host colony near field laboratory facilities (University Park, PA). Shortly before drones emerged (24 days after eggs were laid), the frames were again caged and replaced in the host colony. This allowed drones to emerge in a natural colony environment with the assistance of workers. Nascent adult drones were brushed from frames, separated from workers, and temporarily incubated in Plexiglass cages (10 x 10 x 7 cm) in groups of ~100 before inoculation. 2.2 Nosema spore isolates In May (2012), adult foragers were collected from the hive entrances of colonies and screened using light microscopy for Nosema spores [70]. Once heavily infected colonies were identified, additional foragers were collected and spores were isolated from whole abdomen extracts using methods adapted from [70] (see next paragraph). These spores were mixed with 50% sucrose solution and fed to a selected colony (hereafter referred to as the IC, infected colony) maintained in a separate research apiary. Infection of the IC was confirmed 2 weeks later by checking collected foragers for spores. To prevent the IC from clearing the infection, foragers were collected approximately every 2 weeks from the colony, spores were isolated and fed to the IC as before. One new IC was established each year in 2013 and To isolate spores for experiments, foragers were collected at the IC hive entrance and anesthetized at 4 C. Worker abdomens were homogenized (~ 1 ml of water/abdomen) using a mortar and pestle and poured through screen mesh to filter out cuticle pieces. Coarse filtrate was then poured through polypropylene mesh with 105 μm openings (Spectrum Laboratories Inc, Rancho Dominquez, CA). Next, the filtrate was centrifuged in 50 ml falcon tubes at 3500 rpm for 5 minutes to collect particulate matter. The supernatant was removed and the pellet was resuspended in water. The centrifugation and rinse steps were repeated for a total of 3 washes. After the third wash, particulate matter was again pelleted and suspended in a small volume of

57 water. Spore concentration was determined according to standard practices using a hemocytometer [121]. Concentrated spore isolates were stored at 4 C to forestall bacterial growth and were generally used for infection experiments within hours but no later than 8 days after isolation. Before feeding to drones, spore samples were suspended to the appropriate concentration in 50% sugar water Drone infection Pilot studies showed that newly emerged drones would not consume 50% sugar water (presumably because they had recently been fed by workers). Thus, newly emerged drones were incubated (34 C, 50% RH) in Plexiglass cages (10 x 10 x 7 cm) without workers for ~ 3 hours. After this time, drones were individually restrained and fed 200,000 Nosema spores suspended in 2 μl of 50% sugar water, while controls were fed 2 μl of sugar water. Drones that did not consume the entire dose of spores or control treatment were not used in subsequent experiments. After feeding, drones were paintmarked and housed in unrelated surrogate colonies that had shown no or low levels of Nosema infections in pre-screened forager samples. Drones were sprayed with sugar water to enhance acceptance before introduction. Surrogate colonies were expanded as needed and supplemented with sugar water and MegaBee patties (Tucson, AZ) on a weekly basis. Queen cells were removed each week. 2.4 Impact of Nosema on drone sperm count In 2012, control and infected drones were marked with different colors of paint on their thoraxes and released into a surrogate colony. The entrance of the host colony was blocked with queen excluder until drones were 6 days old to maximize later retrieval. At day 7 postincoluation (pi), colony entrances were opened and drones were allowed to undertake flights until collection. When drones were 14 days old (13 days pi), drones were collected at the host colony entrance in the afternoon while attempting mating flights. Collected drones were anesthetized on ice. The right seminal vesicle of each drone was dissected from the abdomen and placed in 250 μl of modified Kiev buffer [265]. Once in the buffer, the seminal vesicle was torn using fine forceps to release sperm. If the right seminal vesicle was accidentally ruptured before being placed in the buffer, the left seminal vesicle was used. The remainder of each drone s abdomen was frozen at 20 C for later verification of infection status. Dissection equipment was cleaned between samples. Sperm was allowed to diffuse for ~12 hours. Each seminal vesicle was lightly pressed with a pestle to encourage expulsion of any remaining contents. Before quantification with a hemocytometer, each sample was inverted 10 times to evenly suspend sperm. This experiment was repeated twice using drones from different source colonies and unique surrogate colonies. The total sperm count distributions for both trials were log-transformed and the effects of trial, treatment and treatment x trial on sperm count were assessed with a general linearized model (normal distribution, identity link function).

58 2.5 Impact of Nosema infection on drone electroantennogram (EAG) response to 9-ODA 47 In 2013, infected and control drones were individually paint-marked and released into a surrogate colony. The colony entrance was blocked with queen excluder for 48 hours to facilitate drone acceptance. After 48 hours, the queen excluder was removed. Drones were collected at 14 days pi (13 days old) during the afternoon from the colony entrance and temporarily incubated (34 C, 50% RH) with workers from the surrogate colony and ad libitum access to sugar water. Drone EAG responses to 9-ODA, a major component of queen sex pheromone were determined as before [266]. Briefly, drone right antennae were removed and antennal responses to a series of 9-ODA concentrations (1 ng, 10 ng, 100 ng, 1 ug, 10 ug, 100 ug, 1 mg) were recorded and normalized to the average of two concentrations of a floral odorant 2-PE (2-phenylethanl;10 ug, 100 ug). Drone bodies were frozen for later confirmation of spore counts. This experiment was repeated twice using drones from two independent source colonies housed in two unique surrogate colonies. Drone EAG responses were log-transformed and select outliers were removed. After removing outliers, 8 control and 10 Nosema infected drones were included in analyses for the first trial, and 9 control and 10 Nosema infected drones were included in the second trial. All data met criteria for normality (Shapiro-Wilk W Test, p > 0.05) except for 1 mg concentrations for the first trial (p < ). The effects of trial, treatment and concentration on EAG response to 9-ODA were analyzed with repeated measures ANOVA. 2.6 Impact of Nosema infection on drone metabolic activity during flight In 2012, control and infected drones were marked with different colors of paint on their abdomens and released into two separate surrogate colonies. The hive entrance was blocked with queen excluder to prevent drones from leaving. Drones were collected days post-infection and housed with workers as in the EAG study. To assess in-flight metabolic activity, a drone was removed from the cage using tweezers and restrained, dorsal side up. Beeswax was melted and a small portion was collected at the tip of a 50 μl capillary tube. The capillary tube was then gently held perpendicular to the drone s thorax until the wax hardened so that wax created a seal between the drone and the capillary tube. The drone, attached to the capillary tube, was then suspended from the lid of a clear air-tight container, connected to a respirometer airflow system, described as before [267]. The container was inverted (so that drones were upside-down and would not fly) and immediately flushed with CO 2 -free air. When flushing was complete, the container (and drone) were returned to an upright position in front of a window. Drones were given a short acclimation period (~2 minutes total, including container flushing) and then were encouraged to fly by lightly shaking the flight chamber and briefly inverting it when necessary. Analogue system output (CO 2 flow, airflow rate, ambient temperature, system temperature) was recorded for the next 4 minutes. Studies were replicated using drones from 3 source colonies, hosted in 2 unique surrogate colonies (one designated as control and one as Nosema ). Maximum and average metabolic rates (expressed as ml of CO 2 produced per hour) over the four-minute period were calculated for each drone as before [267]. After checking data for

59 normality, the effects of trial, treatment and trial x treatment on the average and maximum metabolic rates over the 4-minute flight period were determined by ANOVA Impact of Nosema infection on drone starvation rate In 2013, control and infected drones were marked with different colors of paint on their thoraxes and placed in separate cages made of queen excluder mesh. A wooden spacer was fit between the two supers of the selected surrogate colony and the cages were laid between the two supers in the created space. Both treatment groups were housed in the same surrogate colony. Workers were able to enter the cages to tend the drones but drones could not leave the cages. Using cages allowed quick retrieval of large numbers of drones. At 13 days post-infection, cages were retrieved, and drones were individually restrained in an upright position in 1.7 ml Eppendorf tubes. The tips of the tubes were previously cut to allow drone heads to fit through, and cotton was placed at the other open end to prevent drones from backing out of the tubes. This allowed movement of drone heads and their front pairs of legs, but their other legs, thoraxes and abdomens were restrained. Drones were held in an incubator at 34 C and 50% relative humidity. Their survival was monitored every hour for the next 14 hours. At the end of the experiment, drones were frozen at -20 C for later confirmation of infection status. This experiment was repeated twice using drones from separate source colonies and housed in two unique surrogate colonies. Drone survival was compared across treatment groups for each trial using Kaplan Meir Survival Analysis. 2.8 Impact of Nosema infection on drone flight behavior In the summer of 2014, infected and control drones received a unique paint-mark on their thoraxes and abdomens resulting in a total of 49 drones/treatment group. (Pilot studies had previously indicated that using larger sample sizes made monitoring drones too difficult as described in [264]). Drones were released into an unrelated 5-frame nucleus colony with an extended runway. This runway was covered with Plexiglas and had gates at both ends, allowing an observer to control drone entrance and egress from the colony (Figure 1). Workers were able to leave and enter at will through queen excluder mesh. After drone introduction, an additional empty super with a hole drilled in the front (covered with tape) was added. For the first two days after drone introduction, the gates were closed to encourage drone acceptance. Starting on the third day of the experiment, drone flight activity was monitored each day starting at ~1 pm until drones ceased flying in the evening (usually from 5-7 pm) or when experimental progress was halted by inclement weather. Returning drones that did not use the runway entrance but that landed on the colony were returned to the colony interior using the hole drilled into the empty upper super (previous pilot studies had showed that returning drones through the hole gave more accurate flight time records, since returning drones that could not find the entrance might take 30+ minutes to find it, or leave and not return). The times that individual drones left and returned to the colony were recorded by an observer. Flights were recorded to the minute level, thus drones that flew for less than 1

60 minute were recorded as flying for one minute. At the end of each experimental trial, a census of remaining drones was taken and drones were collected and stored at -20 C for later confirmation of infection status. This study was repeated using drones from 3 different colonies housed in 3 unique surrogate colonies. 49 Figure 4-1 Modified runway for drone observation experiments. Drone movement in and out of the observation colony was controlled via 2 gates (marked G). Drones traversed the runway through a corridor (C). Worker traffic proceeded through queen excluder mesh (marked QE).Data from handwritten forms was entered into Excel and screened for accuracy. Incorrect records were excluded from the analysis. (e.g. if a drone departed twice in a row or returned twice in a row, one data point in each case was conservatively eliminated to produce the shortest flight possible. If a return was observed without a departure, the return was excluded from analysis). Records were also checked for returned drifts (e.g. if a drone departed from the colony one day and returned on a future date) and these data points were not included in flight duration analyses. Also, flights that were longer than 50 minutes were not incorporated in analyses as it was possible that drones had drifted to another colony and then later returned within the same day (also, see Box 2). Finally, records indicating that drones had departed but not returned over the remainder of the experiment were excluded from flight duration analyses. Completed flights were compiled by day across treatment groups and trials and divided into short (<12 min) and long ( 12 min) flights. Short flights were considered to be best representative of orientation flights while long flights were considered to be representative of mating flights (see Box 2). Optimal Box-Cox transformations were applied to data in SAS and we used linear mixed models (proc mixed function with the covtest function) to: 1) compare average flight duration for all, long and short flights across treatment groups by experimental trial (infection status was the fixed effect and date, bee and bee*date were random variables), 2) compare length of inter-flight duration across treatment groups by experimental trial (infection

61 status was a fixed effect with prior flight length, sequential long flight, infection status x prior flight length and infection status x sequential long flight included as covariates. Bee and date were included as random effects) and 3) compare the rates at which infected and control drones begin taking longer flights by experimental trial (infection status was included as a fixed effect and flight number and infection status x flight number were included as covariates. Bee was included as a random effect). As appropriate, terms were dropped from these models for parsiomony. Also, we compared the average number of long flights taken amongst all drones that took at least one flight by treatment group and trial using Kruskal-Wallis Rank Sum tests or Welch ANOVA where appropriate (Welch ANOVA was used if standard deviations differed significantly across treatment groups). Finally, using Kaplan Meir Survival Analysis, we compared the daily cumulative percentages of drones that completed their first successful flight (any duration) out of all drones that flew across treatments and trials Impact of Nosema infection on drone gene expression In the summer of 2013, control and infected drones were paint marked and introduced into a surrogate colony. A queen excluder mesh was placed across the colony entrance for the first 24 hours of the experiment to encourage drone acceptance. Drones were allowed to exit the colony after this initial period. Drones were subsequently collected between 10 AM-12 PM on dry ice at 14 days post-infection at stored at -80 C. Individual drones were dissected with RNALater (Applied Biosystems Life Technologies, Grand Island, NY) on wet ice. Whole drone midguts were set aside to confirm infection status. Drone fat bodies (eviscerated abdominal tissue) were removed and immediately frozen on dry ice for later RNA extraction. Dissection equipment was cleaned between samples. RNA was extracted from drone fat body tissue using the RNeasy Plus Universal Mini Kit (Qiagen, Valencia, CA). Drone abdomens were homogenized at 6.0 m/s 2 for 45 seconds in UV sterilized tubes (6000 kj delivered over 2 minutes) containing 3 beads. Samples were stored on wet ice for 10 minutes before a second homogenization. Total RNA was then extracted according to manufacturer s directions and concentration and purity were checked with a Nanodrop (Thermo Scientific, Wilmington, DE). Relative expression of 7 genes with metabolic, hormonal and/or immune functions was assessed. Primers sequences and sources are given in Table 1. Primers specially designed for this study (ECSIT, Accession Number: XM_ ; PGRP-S3 Accession Number: NM_ ) were selected using the NCBI Primer Blast service [268]. Briefly, for qrt- PCR, 225 ng of RNA per sample were converted to cdna using Cloned RNase Inhibitor, ArrayScript Reverse Transcriptase, Random Hexamers and dntps (Applied Biosystems Life Technologies, Grand Island NY). We used SYBR green (Applied Biosystems Life Technologies, Grand Island, NY) to detect target RNA levels using the ABI Prism 5700 Sequencer (Applied Biosystems Life Technologies, Grand Island, NY). Expression levels were calculated based on elapsed PCR cycles (Ct values) until samples crossed a fluorescence threshold using the 2 - Ct method as before [269]. Levels of each target gene were assayed in duplicate and processed as the average across technical replicates. Next, expression levels were normalized to the geometric mean of two housekeeping genes (eif-s8 and GADPH1, Table 1) or 1 housekeeping gene if analyses showed that expression of the other housekeeping gene significantly differed between treatment groups. Finally, relative expression of all candidate genes was compared across

62 treatment groups and trials using a generalized linear model (GLM) incorporating trial, treatment and trial x treatment as model effects (normal distribution, identity link function). No significant trial x treatment interaction factors emerged (p > 0.05) except for PGRP-S3, so data was combined across both trials for all genes except PGRP-S3. Goodness of fit tests indicated that it was inappropriate to analyze PGRP-S3 with GLM. Thus, PGRP-S3 expression was compared using non-parametric Wilcoxin Rank Sum ChiSquare Approximation. A total of 20 control and 20 Nosema infected drones (10 drones per treatment and experimental trial) were included in each treatment group for each gene. Table 4-1 Primers used for drone fat body gene expression analysis. 51 Target Gene eif-s8 GAPDH1 vg foxo ilp1 ilp2 usp ECSIT PGRP-S3 Primer sequence 5 3 F: TGAGTGTCTGCTATGGATTGCAA R: TCGCGGCTCGTGGTAAA F: GCTGGTTTCATCGATGGTTT R: ACGATTTCGACCACCGTAAC F: AGTTCCGACCGACGACG R: TTCCCTCCCACGGAGTCC F: TTCGCAGAACAACGTGATAGGT R: GCATTGGTGCTCACGTAAACA F: CGATAGTCCTGGTCGGTTTG R: CAAGCTGAGCATAGCTGCAC F: TTCCAGAAATGGAGATGGATG R: TAGGAGCGCAACTCCTCTGT F: GCGAAGAGAAATCCTGCATC R: TCCCTTTCCTTGGTACGTTG F: ACCTGATTTTGAGATGCAAGAACT R: CTGTTGGTATCGGTCTGGGT F: ACGAAGGTTGTGGCTGGAAT R: ACCTCCGATTACCCGAACATC Source (Grozinger et al, 2003) [269] (Huang et al, 2012) [110] (Ament et al, 2011) [270] (Grozinger et al, 2003) [269] (Ihle et al, 2014) [271] (Ihle et al, 2014) [271] (Wang et al, 2012) [272] Designed by authors [268] Designed by authors [268]

63 2.10 Infection confirmation 52 For all experiments, collected control and infected drone midguts were dissected, homogenized and spores were quantified using a hemocytometer [121]. A subset of 8-10 drones from each treatment group and trial were assessed for spores from sperm, EAG, metabolic rate and starvation experiments. All drones were checked for infection (n=10) for molecular experiments and all surviving infected drones (n = 6-8) and a subset of 8 control drones were checked for observation trials. Furthermore, DNA was extracted from midgut homogenates using the protocol outlined in [85] from a subset of 2-3 control and 2-3 infected drones from each treatment group in each trial. DNA samples were screened for N. apis and N. ceranae infection using published methods [123] Statistical analyses All statistical analyses were conducted in JMP Pro (v , SAS Institute Inc., Cary, NC) except for linear mixed model analyses which were completed in SAS (v. 9.4, SAS Institute Inc., Cary, NC). 3.0 Results 3.1 Impact of Nosema on drone sperm count No significant model effects (trial, treatment, treatment x trial) on drone sperm counts were detected (GLM, X 2 (3) = 0.88, p > 0.83). The average sperm count for control drones was 4.4 x 10 6 ± 1.7 x 10 5 (mean ± sterr, n = 38) and 4.3 x 10 6 ± 1.7 x 10 5 for infected drones (n= 35) from both trials. 3.2 Impact of Nosema infection on drone EAG response to 9-ODA Repeated measures ANOVA indicated that trial, treatment and treatment x trial had no significant effect on drone EAG response to 9-ODA concentrations (F 3,33 = 0.10, p = 0.36). However, as expected, drone EAG responses significantly increase with increasing 9-ODA concentration (F 6,28 = 57.5, p < ). Also, there was a significant concentration x trial effect (F 6,28 = 0.59, p = 0.03). Data is presented in Figure 2.

64 53 A EAG response (mean ± se) Control Nosema B EAG response (mean ± se) ng 10 ng 100 ng 1 ug 10 ug 100 ug 1 mg 9-ODA concentration Control Nosema 0 1 ng 10 ng 100 ng 1 ug 10 ug 100 ug 1 mg 9-ODA concentration Figure 4-2 Drone electroantennogram (EAG) response to the queen sex attractant 9-ODA. Data is provided for Trial 1 (A) and Trial 2 (B). The electroantennal response of drone antennae to 9-ODA was normalized to the floral odor 2-phenyl ethanol. Drone were 15 days old (14 days post-inoculation). In Trial 1, control n = 8 and Nosema n = 10. In Trial 2, control n = 9 and Nosema n = 10. Repeated measures ANOVA indicated that trial, treatment and treatment x trial had no significant effect on drone EAG response to 9-ODA concentrations (F 3,33 = 0.10, p = 0.36). However, as expected, drone EAG responses significantly increase with increasing 9-ODA concentration (F 6,28 = 57.5, p < ). Also, there was a significant concentration x trial effect (F 6,28 = 0.59, p = 0.03).

65 3.3 Impact of Nosema infection on drone metabolic activity during flight 54 Maximum and average metabolic rates (expressed as ml of CO 2 produced per hour) over fourminute flights were calculated for each drone. Data were analyzed by ANOVA (Trial 1, control n = 10, Nosema n = 10; Trial 2, control n = 13, Nosema n = 13, Trial 3, control n = 14, Nosema n = 14) and no significant effects of trial, treatment or trial x treatment were found (maximum metabolic rate, F 5,66 = 1.33, p = 0.26; average metabolic rate, F 5,66 = 1.22, p = 0.31) (Figure 3).

66 55 A ml CO2/hour (mean ± se) Maximum metabolic rate Control Nosema B ml CO2/hour (mean ± se) Average metabolic rate Control Nosema Figure 4-3 Maximum (A) and average metabolic (B) rates produced by drones over 4 minutes of consecutive flight. Data was combined for three trials. Metabolic rates are expressed as ml of CO 2 per hour. Assayed drones were days old (12-14 days post-infection). No significant effects of trial, treatment or trial x treatment were found (maximum metabolic rate, F 5,66 = 1.33, p = 0.26; average metabolic rate, F 5,66 = 1.22, p = 0.31). Treatment sample sizes are given in the base of each column.

67 3.4 Impact of Nosema infection on drone starvation rate 56 Almost all drones starved by 14 hours post-setup in the first trial and by 9 hours post-setup in the second trial. In both trials, infected drones starved significantly faster than controls. In the first trial, infected drones starved faster than controls over the duration of the observation period (Kaplan Meir Survival Analysis: control n = 39, Nosema n = 40, Wilcoxin X 2 (1) = 8.3, p = 0.004; Log-Rank X 2 (1) = 10.1, p = 0.002) while in the second trial, infected drones starved significantly faster than controls early in the observation period (Kaplan Meir Survival Analysis: control n = 59, Nosema n = 56, Wilcoxin X 2 (1) = 5.7, p = 0.02) (Figure 4).

68 57 A 100% Cumulative mortality (%) 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% B 100% Cumulative mortality (%) 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Starvation rate (Trial 1) * Control Nosema Hours elapsed Starvation rate (Trial 2) * Control Nosema Hours elapsed Figure 4-4 Cumulative drone mortality during starvation assays. Data provided for Trial 1 (A) and Trial 2 (B). In both trials, infected drones were 14 days old (13 days post-infection). Infected drones in the first trail starved faster than controls over the duration of the observation period (Kaplan Meir Survival Analysis: control n = 39, Nosema n = 40, Wilcoxin X 2 (1) = 8.3, p = 0.004; Log-Rank X 2 (1) = 10.1, p = 0.002) while in the second trial, infected drones starved significantly faster than controls early in the observation period (Kaplan Meir Survival Analysis: control n = 59,

69 Nosema n = 56, Wilcoxin X 2 (1) = 5.7, p = 0.02).3.5 Impact of Nosema infection on drone flight behavior 58 Spore counts revealed that control drones were not infected or were infected at low levels (data not shown) while treated drones had high levels of infection (Figure 5). However, levels of infection in Nosema treated differed significantly across the trials (Welch ANOVA, F(2) = 12.0, p = ), with infected drones in Trial 3 carrying significantly greater average spore loads than drones in Trial 1 (Steel-Dwass Multiple Comparisons, Trial 3 vs 1: Z = 3.0, p = 0.007; Trial 3 vs 2, Z = 2.0, p = 0.1, Trial 2 vs 1, Z = -2.0, p = 0.1). Across all 3 experimental trials, drones began flying at 6 days of age (5 days post-inoculation). Weather had a strong impact on drone flight * p = Average number of spores/bee 5.0E E E E E E+00 NS, p < 0.05 NS, p > Trial 1 Trial 2 Trial 3 Figure 4-5 Average spore counts for all surviving infected drones collected at the end of each observation experiment. Levels of infection differed significantly across the trials (Welch ANOVA, F(2) = 12.0, p = 0.003), with drones in Trial 3 carrying significantly greater average spore loads than drones in Trial 1 (Steel-Dwass Multiple Comparisons, Trial 3 vs 1: Z = 3.0, p = 0.007; Trial 3 vs 2, Z = 2.0, p = 0.1, Trial 2 vs 1, Z = -2.0, p = 0.1). behavior; drones did not fly during overcast, rainy and/or windy periods as previously observed (reviewed in [273]). Few drift events where departing drones returned to the colony on a later date were recorded (Table 2). Across trials, there did not appear to be a pattern in likelihood of drones returning from a drift to belong a particular treatment group. These flights were not included in later analyses. Also, few completed flights lasted longer than 50 minutes (Trial 1: 7; Trial 2: 10, Trial 3: 1). Since drones might drift to a different colony in the apiary and later return to the surrogate colony within the same day, flights longer than 50 minutes were excluded from the analysis (see Box 2). Overall, approximately completed drone flights were recorded per experimental trial, with each trial lasting between days (Table 3).

70 Table 4-2 Number of returned drift events (when a drone left the observation colony and returned on a later date) and the percent of returned drift flights taken by infected drones per experimental trial. 59 Trial Total # returned drift flights % infected Table 4-3 Number of returned flights and the number of days each experiment lasted per trial. Trial Total # flights Length (days) For additional analyses, we divided completed flights into two groups: short (<12 min) and long flights ( 12 min). Table 4 summarizes the percent of drones in each trial and treatment that completed at least 1 flight and at least 1 long flight. The data suggest that infection reduced the likelihood of taking a flight (any length) in replicate 3 and the likelihood of taking a long flight in replicates 2 and 3. Data also suggested that Nosema infection had limited impact on drone survival in the first trial, while infection decreased survival in trials 2 and 3 (Table 5). (Survival values are based on the numbers of drones remaining in the colony at the end of each trial compared with the numbers of introduced drones). Given that infected drones starve faster than healthy drones (see Section 3.3) we predicted that Nosema infection would be energetically costly during flight. However, we suspected that energetic costs would manifest over longer periods of flight exertion since metabolic rates did not differ between infected and control drones during 4-minute flights (Section 3.2). Therefore, we predicted that infected drones would fly for shorter periods of time on average and take more time to recover after taking a flight than controls. Also we predicted that amongst drones that succeeded in taking longer flights, infected drones would take fewer flights on average than control drones.

71 Table 4-4 The percent of drones that completed at least 1 flight (any duration) and at least 1 long flight ( 12 minutes) per treatment group and experimental trial. 60 Trial Treatment % completed at least 1 flight % completed at least 1 long flight Control Nosema Control Nosema Control Nosema Table 4-5 Percentages of drones surviving to the end of each experimental trial. Trial Treatment % surviving Control 33% Nosema 31% Control 35% Nosema 12% Control 45% Nosema 12% To address the first hypothesis, we compared the average duration of all, short and long flights across both treatment groups using a linear mixed model incorporating drone infection status (fixed effect), and bee, date and bee x date as random effects. However, the interaction factor bee x date was subsequently removed from several of the models for parsimony since its effect was not significant (p > 0.05). The average durations of all and long flights for infected drones were significantly shorter than the average durations for healthy drones in the second and third trials (Trial 1: All flights, p > 0.05; Long flights p > 0.05; Trial 2: All flights, p < 0.01; Long

72 flights, p < 0.05; Trial 3: All flights, p < 0.01, Long flights, p < 0.01) (Figures 6 and 7). The duration of short flights was not significantly different across treatment groups for the first and third trials (p > 0.05), but was significantly longer for infected bees in the second trial (p < 0.05) (Supplementary Materials, Figure 1). A summary of all model estimates is given in Table 2 of the Supplementary Materials. 61

73 62 A Average flight duration (Trial 1) Control Nosema Flight duration (min) Drone age (days) B Average flight duration (Trial 2) Control Nosema Flight duration (min) Drone age (days)

74 63 C Average flight duration (Trial 3) Control Flight duration (min) Drone age (days) Figure 4-6 Average flight duration for all completed drone flights. Data given for Trial 1 (A), Trial 2 (B) and Trial 3 (C). The number of flights taken by members of each treatment group on a given day are recorded at the base of each column.

75 64 A Average long flight duration (Trial 1) Control Nosema Flight duration (min) Drone age (days) B Average long flight duration (Trial 2) Flight duration (min) Control Nosema Drone age (days)

76 65 C Flight duration (min) Average long flight duration (Trial 3) Control Nosema Drone age (days) Figure 4-7 Average flight duration for all long ( 12 minutes) drone flights. Data given for Trial 1 (A), Trial 2 (B) and Trial 3 (C). The number of flights taken by members of each treatment group on a given day are recorded at the base of each column.to determine which factors predicted the length of inter-flight duration, linear mixed model analyses were used, incorporating infection status (fixed effect), prior flight length (covariate) and whether bees were taking a sequential long flight ( 12 minutes, covariate). Bee and date were incorporated as random effects. For all trials, prior flight length and whether drones were taking a sequential long flight on the same day were strong, positive predictors of inter-flight duration (Trials 1-3: p < 0.01). However, Nosema infection only significantly predicted longer inter-flight rests in the second trial (Trial 1: p >0.05; Trial 2: p < 0.05; Trial 3: p > 0.05). The average inter-flight durations for all flights across treatment groups and trials are included in Figure 2 of the Supplementary Materials and the average inter-flight durations between consecutive long flights are included in Figure 3 of the Supplementary Materials. All model estimates are given in Table 3 (Supplementary Materials). To determine if infected drones took fewer long flights on average than controls, we compared the number of long flights taken by each drone amongst all drones that took at least one long flight. Infected drones took fewer long flights on average for the second and third trial (Trial 1: Kruskal-Wallis Rank Sums X 2 (1) = 2.4, p > 0.12, control mean rank = 32.8, Nosema mean rank = 40.4; Trial 2: Welch ANOVA F(1) = 11.5, p < 0.001, control mean = 10.2, control sd = 5.8, Nosema mean = 6.0, Nosema sd = 4.2; Trial 3: Kruskal-Wallis Rank Sums X 2 (1) = 6.1, p < 0.02, control mean rank = 32.3, Nosema mean rank = 21.1) (Figure 8).

77 66 Average number of long flights/drone * * Control Nosema Control Nosema Control Nosema Rep. 1 Rep. 2 Rep. 3 Figure 4-8 Average number of long flights ( 12 minutes) taken by drones in each treatment group that took at least 1 long flight during each experimental trial. For all drones that succeeded in completing at least one long flight, infected drones took fewer long flights on average for the second and third trial (Trial 1: Kruskal-Wallis Rank Sums X2(1) = 2.4, p > 0.12, control mean rank = 32.8, Nosema mean rank = 40.4; Trial 2: Welch ANOVA F(1) = 11.5, p < 0.001, control mean = 10.2, control sd = 5.8, Nosema mean = 6.0, Nosema sd = 4.2; Trial 3: Kruskal-Wallis Rank Sums X2(1) = 6.1, p < 0.02, control mean rank = 32.3, Nosema mean rank = 21.1). The number of drones included in each treatment group are recorded at the base of each column.since Nosema infection in workers accelerates onset of foraging behavior, we determined whether Nosema similarly alters drone latency to begin taking flights. Using Kaplan Meir Survival Analysis, we compared the daily cumulative percentages of drones that completed their first successful flight across treatments and trials. Across trials, drones began flying when they were six days old, but the range of days by which all drones had completed taking their first flight varied with trial. Infection significantly increased the rate at which infected drones took their first flight in the first trial, though all drones that flew had taken their first flight within a 3-day time frame (Kaplan Meir Survival Analysis, Log-Rank, X 2 (1) = 4.0, p < 0.05; Figure 9A). There was also a non-significant trend for infected workers to fly earlier than controls in the third trial (Kaplan Meir Survival Analysis, Log-Rank, X 2 (1) = 2.8, p = 0.09; Figure 9C). Infection had no impact on flight latency in the second trial (Kaplan Meir Survival Analysis, Log-Rank, X 2 (1) = 0.21, p = 0.64; Figure 9C).

78 67 A Daily cumulative percent of drones taking first flight B Daily cumulative percent of drones taking first flight 120% 100% 80% 60% 40% 20% 0% 120% 100% 80% 60% 40% 20% 0% Trial 1 * Control Nosema Drone age (days) Trial 2 Control Nosema Drone age (days)

79 68 C Daily cumulative percent of drones taking first flight 120% 100% 80% 60% 40% 20% 0% Trial 3 Control Nosema Drone age (days) Figure 4-9 Daily cumulative percent of drones taking their first flight out of all drones that completed at least one flight. Data provided for Trial 1 (A), Trial 2 (B) and Trial 3. Infection significantly increased the rate at which infected drones took their first flight in the first trial (Kaplan Meir Survival Analysis, Log-Rank, X2(1) = 4.0, p < 0.05). There was also a non-significant trend for infected workers to fly earlier than controls in the third trial. Infection had no impact on flight latency in the second (Kaplan Meir Survival Analysis, Log-Rank, X2(1) = 0.21, p = 0.64) and third trials (Kaplan Meir Survival Analysis, Log-Rank, X2(1) = 2.8, p = 0.09).

80 Next, we determined whether the rates at which drones began taking longer flights differed between treatment groups across trials. A linear mixed model was used to determine the effect of treatment (fixed effect), flight number (all flights; covariate) and infection status x flight number (covariate) could predict subsequent flight length (bee and date were included as random effects). The model found that flight experience was a significant predictor of flight length in all three trials (p < 0.01). That is, the more flights a bee had taken, the more likely its next flight would be longer. For the second and third trials, the interaction factor infection status x flight number also emerged as a significant predictor of flight length (Trial 1, p < 0.05; Trial 2, p < 0.01). In other words, the rate at which control drones began taking longer flights was significantly faster than the rate for infected drones (Figure 10). All model estimates are included in Table 4 (Supplementary Materials). 69

81 70 A Flight duration (minutes) Control Nosema Flight number B Flight duration (minutes) Control Nosema Flight number Figure 4-10 Interaction between drone infection status, flight experience and flight length. Data given for A. Trial 1 and B. Trial 2. Model estimates for flight length (y) were plotted against total flight number (x) for each treatment group. Slopes significantly differed across treatment groups for each trial. (Trial 2: control slope > Nosema slope, p < 0.05; Trial 3: control slope > Nosema slope, p < 0.01).3.6 Impact of Nosema infection on drone gene expression: Relative expression of all candidate genes was compared across treatment groups and trials using a generalized linear model (GLM). Nosema co-infection significantly elevated expression of foxo (X 2 (1) = 11.3, p < <0.001) and usp (X 2 (1) = 6.6, p < 0.01) and suppressed expression of ilp1 (X 2 (1) = 6.39, p <0.02) and ilp2 (X 2 (1) = 18.4, p < <0.001) (Figure 11). Expression of vg (X 2 (1) = 0.06, p > 0.81), PGRP-S3 (Wilcoxin rank sum, 1-way Test, ChiSquare Approximation; X 2 (1) = 0.018, p > 0.89) and ECSIT (X 2 (1) = 0.23, p > 0.63) were marginally impacted by infection (Figure 11). Significant trial

82 effects (p < 0.05) were observed for ilp1 (X 2 (1) = 24.8, p <<0.001), ilp2 (X 2 (1) = 7.1, p < 0.008) and vg (X 2 (1) = 38.2, p <<0.001) only Control Nosema Control Nosema Control Relative mrna levels (arbitrary units; mean ± se) Nosema Control Nosema Control Nosema Control Nosema Control Nosema vg usp ilp1 ilp2 foxo ECSIT PGRP-S3 Figure 4-11 Effects of Nosema infection on drone gene expression patterns. Relative expression levels were measured in the abdominal fat bodies using quantatitative real time PCR. Relative expression between treatment groups only for each gene is provided for vg (vitellogenin), usp (ultraspiracle), foxo, ilp1 (insulin-like peptide 1), ilp2 (insulin-like peptide 2), ECSIT (evolutionarily conserved signaling intermediate in the Toll signaling pathway) and PGRP-S3 (pattern group recognition protein S3). Drone samples were assayed at 14 days post-inoculation (15 days old) and data were combined across experimental trials (control n = 20, Nosema n = 20). Nosema co-infection significantly elevated expression of foxo (X2(1) = 11.3, p < <0.001) and usp (X2(1) = 6.6, p < 0.01) and suppressed expression of ilp1 (X2(1) = 6.39, p <0.02) and ilp2 (X2(1) = 18.4, p < <0.001).3.7 Nosema infection confirmation: Microscopy and molecular results for all experiments indicated that controls were infection free. Drones in both sperm experimental replicates were infected with N. ceranae only. Drones utilized in EAG experiments were infected with N. ceranae only in the first trial and were co-infected with N. ceranae and N. apis in the second trial. Drones for both starvation assay trials were co-infected with N. ceranae and N. apis, whereas drones for all three metabolic rate assays were only infected with N. ceranae. Drones in the first observation trial were only infected with N. ceranae but drones in the second and third observation trials were co-infected. Finally, drones used in both molecular assays were co-infected (PCR gel images not included).

83 4.0 Discussion 72 Overall, Nosema ceranae infection had limited effects on some aspects of drone sexual maturation (sperm counts, EAG response to 9-ODA) but N. ceranae and N. apis co-infection significantly affected drone starvation rates, flight behavior and gene expression. Due to differences in infection types across experiments, data is cautiously interpreted. 4.1 Physiological assays N. ceranae infection did not alter adult drone sperm counts. In this study, each seminal vesicle contained 4.3 million ± 120,436 sperm (mean ± SE, data combined across both trials), giving a final estimate of approximately 8.6 million sperm/drone. This estimate falls within previously reported estimates of million sperm/drone [274, 275]. Spermatazoa are generated in the testes of immature drones and travel to the seminal vesicles early in adult drone life (see Box 1). Since drones were infected when they were 1 day old and Nosema infections take time to progress, it is perhaps not surprising that sperm migration was unaffected in experimental drones. Indeed, drones deprived of protein as adults still have live sperm percentages comparable to controls, suggesting that sperm viability is not sensitive to adult energetic status/nutrition [276], though pollen-restricted adult drones may suffer other copulationrelated complications identified in laboratory experiments [277]. Future studies may determine if sperm production and/or migration is hampered by microsporidian infections acquired during spermatogenesis in immature drones [102]. For example, Varroa mite parasitization of drone pupae can result in significantly lower sperm counts in adults [274] (but see [278]). Furthermore, even if drone sperm counts are not sensitive to Nosema infection, infection may still diminish sperm viability and fertility. Nosema infection also did not alter drone electroantennal responses to the queen pheromone compound 9-ODA. 9-ODA is a major component of queen mandibular pheromone and serves as important long-range attractant for queens during mating flights [279]. Thus, at least at the peripheral (antennal) level, infected and healthy drones appear equally responsive to potential mates. However, this data must be interpreted with caution since drones were singly infected with N. ceranae in the first trial and co-infected in the second trial. Future studies may replicate these findings. Furthermore, additional studies are needed to determine if infected drones are equally responsive to visual stimuli and/or if central processing of peripheral signals is consistent across infected and healthy drones. Though N. ceranae infection had negligible effects on some parameters related to drone sexual function, co-parasitization clearly imposed energetic costs. Indeed, co-infected drones starved faster than controls, mirroring findings from studies conducted N. ceranae infected workers [78]. However, compared with workers, drones starved much faster in general (nearly 100% mortality was observed within 14 hours or less for drones as opposed to ~90% mortality within 24 hours for workers). Furthermore, control drones starved ~1-2 hours later on average than infected drones, while control workers starved ~4-5 hours later on average that infected workers. These results highlight a lower overall tolerance to starvation in drones as compared with workers and also suggest that energetic stress induced by infection in drones is smaller relative to energetic costs of not having access to colony food stores and nestmate attendance as

84 compared with workers. However, differences in study design and infection levels could also modulate the starvation rates observed in these experiments Flight assays Since infected drones starved faster than controls, we suspected that energetic costs of infection were most likely to manifest during periods of energetic stress such as flight. However, laboratory assays found no differences in control and N. ceranae infected drone metabolic rates over 4 consecutive minutes of flight. Since drones generally embark on minute mating flights (see Box 2) in hopes of encountering a virgin queen we next characterized drone flight behavior in the field. Trials 2 and 3 indicated that co-infected drones take their first flight at the same time as controls, though there was a trend for infected drones to fly earlier in Trial 3. Trial 1 found that N. ceranae infected drones flew earlier than infected drones. However, the difference in average age of first onset for both treatment groups in Trial 3 was less than one day. This finding weakly suggests that infected drones may fly sooner, but contrasts with studies in workers where Nosema infection promotes precocious foraging [86-88] (but see [92] ). This caste difference is likely due to the chronic nature of Nosema infection and unique aspects of worker and drone life history. Drones begin orientation flights at ~5-7 days of age and subsequently progress to longer mating flights (as reviewed in [273]). In contrast, workers initiate foraging activity at between days of age (as reviewed in [167]). Since Nosema infection builds over time (pathogens complete their replication cycle within ~3 days [11]), caste differences in flight initiation may be a byproduct of longer incubation periods for workers as compared with drones before normal flight onset. Further analysis of drone flight length revealed differences across treatments. Infected drones flew for shorter periods of time on average in two out of the three trials. Furthermore, if all flights were divided into short (<12 minutes) and long ( 12 minutes) flights, the average length of long flights for infected drones was shorter than the average length for controls for two out of three trials. Also, out of all the drones that took long flights, infected drones also took fewer flights on average than control drones. Finally, infected drones rested for longer periods of time between sequential flights than controls in the second experimental trial. Across all three trials, prior flight length and whether drones were taking sequential long flights were the best predictors of how long drones would rest after taking a flight. However, there was a trend for infected drones to recuperate for longer periods of time between sequential long flights across all three trials starting at ~10-13 days post-infection (Appendix A, Figure 5). These findings on drone flight and inter-flight duration contrast sharply with findings in workers. Infected workers take fewer flights overall [92], but longer flights on average than control workers [92, 94]. Infected workers also spend shorter periods of time within the colony between trips [94]. These caste differences in flight behavior across treatment groups could again be the result of distinct caste tasks/goals. Authors investigating infected forager flight behavior suggest that longer foraging trips may be due to workers stopping to rest and/or to consume extra nectar since infection is energetically costly [92]. Drones have never been observed to visit flowers and return to colonies with empty crops suggesting that they do not feed outside (as reviewed in [273]), so shorter average flight durations for infected drones may be a direct result

85 of fatigue followed by potentially longer periods of time recuperating (and potentially eating more) inside the hive. Longer forager flight periods could also be a result of infection-induced disorientation, poorer flight performance and reduced homing ability as has been observed in workers [95, 96]. Indeed, in our study, lower percentages of infected drones were retrieved at the end of the second and third trials. Missing drones could have died, drifted or (improbably) mated (see Box 1). Finally, healthy drones began taking longer flights at a faster pace than infected drones in two of the experimental trials. Consistent with previous drone studies, prior flight number/experience, which is a proxy for drone age, was a positive predictor of the length of subsequent flights [280]. However, drone infection status interacted with previous flight number such that when drones were relatively young and had little flight experience, the average flight length was similar across treatment groups. As drones aged, the average durations of subsequent flights segregated, with healthy drones increasing their flight length at a faster rate (Figure 10). This divergence in flight lengths is again consistent with the chronic nature of Nosema parasitism, where infection costs are likely to accumulate over time. Overall, infection had significant effects on drone flight in the second and third trials but negligible impact in the first trial. Several factors may explain discrepancies in outcomes. First, drones in the first trial were only infected with N. ceranae, whereas drones in the second and third trials were co-infected. It is possible that co-infections are more damaging in drones. Alternatively, drones in the first trial may have originated from a colony that was more tolerant of or resistant to infection than the latter two. Tolerance mechanisms reduce the damage caused by pathogens and/or self-destruction caused by the host s own immune response to the infection, allowing tolerant individuals to sustain infection titers without incurring the same costs as less tolerant individuals. On the other hand, resistance traits actively limit parasite growth, reducing overall infection levels [139]. Indeed, average spore counts from surviving drones varied significantly across trials, with drones from Trial 1 having comparable spore counts to drones from Trial 2 but significantly fewer spores than drones from Trial 3 (Figure 5). There can be wide variability in colony resistance and/or tolerance to Nosema infection and some studies have identified genetic markers linked with infection resilience traits [60, 110, 111, 162]. Also, differences in nutrition may have affected drone performance. For example, caged infected workers fed on high quality diets generate more spores but also live longer than infected workers fed on poor diets [174, 176, 281]. Though colonies in these experiments were supplemented with MegaBee and sugar water, pollen and nectar sources changed with plant blooming period over the course of the summer, so the nutrients available to drones in their surrogate colonies (and during development in their natal nests) would have varied, potentially affecting the nutritional status and costs of infection in experimental drones. Across all observation study comparisons, the random effects bee and date frequently emerged as significant model factors (Appendix A, Tables 1-3) reflecting the large role that variability in individuals behavior and climatic factors had on observed outcomes. High variability in individuals behavior highlights the need for large sample sizes in observation studies so that infection signals can be detected over individuals flight tendencies. Similarly, date which incorporated both weather and time components had a large impact on flight behavior. Drones do not attempt flights during inclement weather and thus, not surprisingly, there were days when drones did not fly at all during experiments (e.g. Figure 6) due to poor flight conditions (i.e. it was cloudy, raining, windy, cool etc). Furthermore, as experiments advanced, drones sexually matured, their flight experience increased and their Nosema infections progressed. Given that all these factors were folded into date, it is not surprising that it generally 74

86 emerged as a significant variable across studies. Notably, our models did not explain all the variance in outcomes (residual variance always emerged as a significant factor, Supplementary Materials, Tables 2-4). Future studies that segregate the effects of weather and drone age may help account for residual variance. In addition, our studies were conducted by a single observer which introduced human error. For example, since flight length was only recorded to the minute, drones that flew for less than 1 minute were recorded as taking a 1-minute long flight Molecular assays Overall, starvation and observation assays indicated a cumulative chronic energetic cost of Nosema co-infection. To explore molecular mechanisms behind infection costs and potential implications for drone sexual maturation, energetic status and immune function, we examined expression of several candidate genes identified from earlier studies in workers [85]. We measured gene expression in drone fat body tissue (eviscerated abdomens) since this tissue regulates a number of metabolic, hormonal and immune processes [282, 283]. First we compared levels of vg (vitellogenin) and usp (ultraspiracle) across treatment groups. Vg is a large protein that is associated with worker nutritional and physiological status, and has a co-negative regulatory relationship with juvenile hormone (JH; high levels of Vg suppress JH and vice versa) in workers [284]. In workers, high vg and low JH levels are associated with workers in the nursing/brood care state. As nurses mature and transition to foraging, vg levels decrease and JH levels rise. Artificially altering vg and/or JH levels can cause corresponding changes in worker task orientation [ ]. Usp is a transcription factor that is sensitive to JH levels and regulates several molecular pathways associated with behavioral maturation in workers [288]. Some cage studies have found that N. ceranae infection suppresses worker vg levels [92, 107], giving rise to the hypothesis that infection promotes foraging behavior in part by dampening vg and/or elevating JH. Indeed, experimentally N. ceranae- or co-infected workers or have higher levels of JH in field and/or cage conditions [86, 289] and workers inoculated with N. ceranae have lower vg levels in the field (see Chapter 5). However, two cage studies [85, 108] and a field study have found that Nosema infection does not affect vg levels [86]. Also, usp is negatively correlated with vg expression, and greater worker usp expression is positively correlated with foraging behavior and Nosema infection [85, 288]. JH and Vg levels are correlated with important aspects of drone sexual maturation, but these hormones do not share the same developmental trajectory as in workers. Overall, drone Vg titers are much lower than worker titers across adulthood (reviewed in [290]), and drone Vg levels peak around 5 days of age before steadily declining [291]. Drone JH levels also peak at shortly before flight onset (~5 days of age) with variable levels through ~10 days of age before tapering (thus Vg and JH do not exhibit a negative relationship in adult drones). Artificially treating drones with a JH analogue (methoprene) promotes earlier flight initiation [292] and changes in drone diet and hive niche correlated with advanced drone age [293] suggesting that JH has parallel roles in regulating worker behavioral maturation and some behavioral aspects related to drone sexual maturation. However treatment with methoprene does not appear to accelerate drone sexual organ (testes, seminal vesicle, accessory gland) maturation [293].

87 In our study, Nosema co-infection did not impact drone vg levels but did increase usp levels in drones collected 14 days post-inoculation. Since overall Vg levels are low in 15 day-old drones [291], it is possible there may have been differences in vg expression at earlier timepoints (perhaps at 5 days, when Vg levels peak in drones). Alternatively, vg may not be sensitive to infection in drones as in workers. However, levels of usp were elevated in infected drones, suggesting that these drones are more sensitive to JH effects, and/or have higher levels of JH. We also compared expression of several genes involved in the insulin signaling pathway: ilp1 (insulin-like peptide 1), ilp 2 (insulin-like peptide 2), and foxo (fork-head box-containing protein, O subfamily). The insulin signaling pathway is a conserved hormonal pathway that regulates nutritional homeostasis in insects [294]. In worker honey bees, the insulin signaling pathway has also been co-opted in part to regulate behavioral maturation [295]. ilp1 and ilp2 are signaling peptides that bind to insulin receptors on target cell surfaces, potentiating a signaling cascade that can have metabolic and hormonal effects. Lower levels of ilp2 are associated with poor diet, foraging status or Nosema infection in worker abdominal fat body tissue [85, 270]. ilp1 is also associated with diet, with the highest ilp1 levels (abdominal fat body tissue) corresponding to a diet high in protein [271]. Finally, foxo is a transcription factor in the insulin signaling pathway that is elevated in Nosema co-infected workers [85] and workers fed on poor diets [270]. Interestingly, expression of both insulin-like peptides was significantly lower in infected drones while foxo expression was significantly higher, indicative of nutritional stress and consistent with the results in Nosema-infected workers [85]. Whether the insulin signaling pathway is also involved in drone sexual maturation has not, to our knowledge, been studied. Finally, we compared expression of ECSIT (Evolutionarily Conserved Signaling Intermediate in Toll pathways) and PGRP-S3 (Peptidoglycan Recognition Protein S3). The Toll signaling pathway is a conserved immune pathway that canonically responds to gram positive bacteria and fungi in Drosophila [283]. Several members of the Toll signaling pathway were differentially regulated in drones from a Nosema-tolerant honey bee strain [110] while ECSIT and other Toll members were regulated in Nosema infected workers [85]. Peptidoglycan recognition proteins are defense molecules involved in recognizing microbial invaders [283]. PGRP-S3 was differentially regulated by Nosema infection in workers in two microarray studies [85] (unpublished data, Judy Chen, USDA-ARS, Beltsville, MD) and one field study (Chapter 5). Neither of these immune genes was modulated by Nosema co-infection in drones. Differences in ECSIT and PGRP-S3 expression in drones and workers could cautiously be interpreted as castespecific responses to Nosema infection. However, studies directly comparing immune responses between drones and workers are needed to verify this hypothesis. Indeed, immune responses to Nosema infection in workers and queens are frequently sensitive to incubation period and may depend on a number of experimental conditions [108, 296] Infection type and conclusions Results from these molecular, physiological and behavioral studies suggest that Nosema infection has mixed proximate implications for drone physiology, but overall negative ramifications for drone fitness. Interestingly, co-infection significantly affected drone starvation rate, flight behavior and gene expression, but did not alter drone antennal sensitivity to 9-ODA in

88 one experimental trial. Also, N. ceranae infection alone did not affect drone sperm count, metabolic rate or drone antennal sensitivity to 9-ODA (in one trial). However, while not significant, the flight behavior of N. ceranae infected drones generally followed the same trends as co-infected drones. A possible interpretation of these findings is that N. ceranae infection in drones is not as damaging as co-infections. However, as drones were only infected with N. ceranae in experiments examining drone sperm count and drone metabolic rate while drones in other experiments were co-infected, it is possible that these physiological parameters are less sensitive to drone infection status. Indeed, as discussed earlier, spermatozoa are generated in immature drones while metabolic rate assays only measured drone CO 2 output over 4 minutes (a short time compared with mating flight duration). Future experiments directly testing the effects of single versus co-infections are needed to clarify potential differences in costs of single versus co-infections in drones. Taken together, however, these studies suggest negative effects of Nosema infection on drone fitness. N. ceranae infection alone does not alter drone sperm counts and antennal responses to 9-ODA and thus, infection theoretically would not prevent drones from locating a queen and delivering a normal quantity of sperm if they were able to catch her in flight. However, the energetic costs of infection may preclude drones from taking successful mating flights. N. ceranae and co-infections shorten drone life expectancy and co-infection also reduces the quantity of longer flights (12+ minutes) and the average length of flights, thereby reducing infected drones overall chances of mating success. In addition, Nosema co-infection reduces the rate at which drones begin taking longer flights (12+ minutes). Since longer flight times and flight length are correlated with flight experience (this study)/drone age [280], slower onset of longer flights in infected drones could be indicative of slower sexual maturation. Indeed, infected drones also had lower levels of usp which may interact with JH to regulate some aspects of drone sexual maturation [290]. Alternatively, energetic costs of Nosema infection, evidenced by starvation and molecular asays, may simply impede longer flights in infected drones. Future studies examining JH levels may clarify the role of Nosema (if any) in drone sexual maturation rate. Nosema infection in drones may have other negative consequences for colony fitness and performance. Drones appear to transmit Nosema to caged nestmates better than workers [100] and may facilitate disease-spread within a colony. In addition, our results indicate that infected drones starve faster than controls and thus are likely hungrier than controls as has been shown for parasitized workers [78]. Thus, infected drones may place supernumerary demands on worker attendants and colony food stores. It has long been debated whether the accelerated maturation of workers exposed to different stressors (including Nosema) is a strategic response, ensuring that the infected workers interact less with their nestmates and therefore reduce the spread of disease. Furthermore, foraging is the final behavioral stage of worker development, and the most risky - thus even workers with reduced longevity can benefit the colony by initiating foraging (e.g. [88, 297] but see [97]). Indeed, Nosema infected workers mature faster and spend more time outside of the colony as foragers. Our studies indicate that Nosema infection marginally affects rate of flight onset (a proxy for sexually mature behavior) but significantly decreases the rate at which drones take longer flights. Furthermore, infection reduces average flight length while showing a trend for extending inter-flight rests. Thus, drones appear to exhibit a distinct stress response from workers and this should be examined using other types of stressors. If drones do exhibit a different stress response, this may be because stress interacts differently with drone versus worker fitness: drone fitness is only increased if they mate, and thus any behaviors that improve mating 77

89 success (eg, delaying long flights until maturation is complete or nutritional stores are high) are favored. 78 Box 1: Drone biology In temperate climates, drones (male honey bees) are raised in the summer months during peak colony productivity and when swarming and subsequent colony founding is not prohibited by harsh weather conditions (see [167, 273, 290] for reviews of drone life history). Drones develop from haploid (unfertilized) eggs that are, under most circumstances, laid by the queen honey bee. Eggs hatch after 3 days and larvae mature to prepupal stages after 6 days and eclose as adults after 24 days. Drone sexual organs form during these preadult stages, and spermatogenesis is completed during pupation. However, after emerging as adults, drones are not yet physically or behaviorally sexually mature. Shortly after adult emergence, drone spermatozoa migrate from the testes to the seminal vesicles where they are stored until mating. During this time (5-6 days of age), drones can be experimentally induced to ejaculate, but will not release semen until they are ~8 days old (as reviewed in [167]). Over their first 4 days of life, drones reside inside their natal nest and are tended by workers. Between 5-7 days of age, however, drones leave the colony and undertake short afternoon orientation and/or cleansing flights where they presumably learn their colony s location and/or relieve their bowels ([298] and as reviewed in [273]). As drones age, their flight experience increases and their average flight durations lengthen [280]. During these longer mating flights, drones travel to specific geographic regions termed DCAs (Drone Congregation Areas) [273]. Why and how drones are attracted to DCAs is poorly understood though these localities are stable from one year to the next. However, drones do appear to prefer DCAs that are closer to their colony than those that are further away [299]. Once at a DCA, drones await the arrival of an eligible queen. If a queen arrives, drones are alerted to her presence by her sex pheromone which serves as a long-range attractant [279] and by visual cues at closer distances [300]. To mate, a drone must catch up with and mount a queen in flight and evert his endophallus into her vagina, depositing his sperm and associated seminal fluids. The process of everting the endophallus, however, is lethal and drones immediately die after a successful copulation (or an experimentally induced ejaculation). A virgin queen will mate with an average of drones in rapid succession [301, 302]. Despite virgin queens polyandrous behavior, drones have very poor odds of mating. Healthy colonies supplied with drone comb can produce up to 19,000-22,500 drones on average annually, while production of a virgin queen is a much rarer event (e.g. once or twice annually if at all) [303]. Thus, honey bees have a skewed operational sex ratio, and individual males chances of encountering and mating with a queen are low. However, studies indicate that some colonies are better than others at producing drones that successfully mate, suggesting that there are precopulatory factors that can enhance (or detract from) a male s mating success [302, 304]. Indeed, drones with greater wing symmetry are more likely to copulate [305]. Also, larger drones have greater sperm counts [275] and have greater chances of mating [304]. (However, small drone size in these experiments was achieved by rearing drones in worker cells (an unnatural situation), while the relatively larger males were reared in normal drone cells). Finally, drones with restricted pollen access during adult maturation have lower ejaculation rates under experimental conditions, suggesting that they may be less virile in the field [277].

90 Reported averages of drone adult longevity range from days, days to 36.2 days, with maximum life spans reaching 59 days (as reviewed in [167, 273]). High variability in experimental conditions likely influence study findings. Drones that do not return to their colonies most likely either died due to predation or age-related degeneration [306], since, as previously discussed, drones chances of mating success are low. Alternatively, drones frequently drift to other colonies which can complicate attempts to measure drone lifespan [307]. Finally drones may be forcefully removed from a colony by workers to conserve hive resources at the end of the summer/mating season (as reviewed in [273]). 79 Box 2: Adult drone sexual maturation and flight behavior In 1933, Howell & Usinger marked newly emerged drones and documented individuals flight behavior [298]. They noted not only that drones commence flight between 4-7 days of age, but also that flight duration was bimodally distributed. Howell & Usinger further observed that a large number of flights lasted 1-6 minutes while the overall mean flight duration was 27.3 minutes. Since many of the short (1-6 minute) flights comprised a majority of maiden flights for drones, the authors presciently suggested that drones take short orientation flights, followed by longer searching (for virgin queen) flights. This hypothesis was corroborated by later studies though estimates for orientation and mating (as opposed to searching ) flight durations flights varied. Nearly 40 years later, Witherell (1971) summarized findings from the available literature and added his own experimentally derived estimates noting: The distinction between mating and orientation flights seems somewhat arbitrary. Some writers base their distinction on flight duration, the longer flights presumably being for the purpose of mating (pg 611, [280]). Estimates of orientation flights reviewed by Witherell ranged from: 1-6 minutes, 1-8 minutes (avg 3.1 minutes), and 6-15 minutes. Estimates of average flight length reviewed and contributed by Witherell also varied: <50 minutes (avg 27 minutes), minutes, <50 minutes (with most ranging from minutes), or 32.6 ± 22.5 minutes. To date, there is no consensus in the literature on how to discriminate between flight types and there are no additional follow-up studies that we are aware of. Collectively, these studies demonstrate that the duration of short (orientation) and long (mating) flights vary across studies, likely due to factors varying across experiments (bee populations, weather, year, location, etc). Likewise, average flight lengths varied across our trials. For analysis purposes, we chose to segregate our flights based on a 12 minute cut-off, with all flights 12 minutes categorized as long flights. This cut-off point gave us average short and long flight values for control drone populations that resembled previous reports of orientation and mating flight lengths. Average short flight durations for all 3 trials ranged from minutes while long flight duration averages ranged from minutes in control bee populations (Appendix B, Table 1).

91 6.0 Acknowledgements 80 We would like to thank Drs. Tom Baker and Jim Marden for guidance and use of their EAG and respirometry equipment, Bernardo Nino and Mario Padilla for expert beekeeping and undergraduate researchers Megan Snyder and Jacqueline Patterson for experiment assistance. We are especially grateful to Weiyi Cheng and Jun Song from the Penn State University Statistics Consulting Center for their help with mixed model data analysis and to members of the Grozinger lab for critical reading of this manuscript. This material is based upon work supported by the National Science Foundation under Grant No. DGE to HLH. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. Additional funding was provided by USDA-AFRI (PI J Chen, copi: CMG).

92 Chapter 5 81 Conclusions and future directions Holly L. Holt and Christina M. Grozinger Department of Entomology, Center for Pollinator Research, Center for Chemical Ecology, Huck Institutes of the Life Sciences, Pennsylvania State University, University Park, USA This dissertation investigates basic and applied concerns of managing infectious diseases in an important pollinator species, Apis mellifera, using the microsporidian parasites Nosema apis and Nosema ceranae as model disease organisms. The need for a comprehensive Integrated Pest Management (IPM) plan for controlling Nosema parasites in honey bee colonies is discussed in Chapter 1. Novel, cost-effective methods for both detecting and treating Noseomosis are urgently needed in conjunction with sciencebacked recommendations for when treatments should be administered. Preventative measures including selective breeding may enhance colony tolerance and/or resistance to Nosema infection in addition to other colony stressors. Finally, improved understanding of Nosema pathology will augment disease control and may provide insight into microsporidian etiology in beneficial insects, including other species of bees afflicted by N. ceranae, and in emerging microsporidian diseases in humans. Chapters 2-4 include molecular, physiological and behavioral characterizations of honey bee transcriptional responses to non-replicating pathogens and infection with Nosema parasites. Chapter 2 [115] found that challenging workers with diverse immune stimulants (injection with saline, Sephadex beads or dead E. coli) could elicit both overlapping and pathogen-specific changes in global worker gene expression. Interestingly, many genes that were regulated by immune stimulation in this study [115] were not previously identified as members of canonical immune response pathways in honey bees [202]. These "non-canonical genes" identified in Chapter 2 provide an additional point of comparison for other studies evaluating changes in worker global gene expression profiles in response to challenge with taxonomically diverse pathogens. For example, co-infection with N. ceranae and N. apis modulates expression of a significant number of acute immune response genes (see Chapter 2),but not canonical immune genes (see Chapter 3 and [85]). Meanwhile, viral infection regulates some canonical immune genes but has little overlap with non-canonical genes or genes associated with microsporidian infection [308]. This handful of studies points to the breadth and specificity of honey bee genomic responses to distinct immune challenges. Chapter 3 investigates changes in local (midgut) and systemic (fat body) gene expression in workers parasitized with Nosema and identifies candidate molecular pathways likely regulating previously documented physiological and behavioral symptoms of microsporidian infection. Importantly, Nosema infection modulates expression of genes with overlapping nutritional and hormonal roles, including members of the insulin signaling pathway, which contribute both to changes in worker energetic status and behavioral state. Comparative analyses validated molecular findings from this study, indicating that infection drives patterns in gene expression associated with foraging and poor nutrition [270]. Furthermore, greater overlap in directional gene expression across Nosema and nutrition studies, as opposed to Nosema and behavioral state

93 studies, suggest that changes in worker metabolic/nutritional processes precede, and likely drive changes in worker behavioral state. Chapter 4, to date, provides the most comprehensive molecular, physiological and behavioral characterization of Nosema infection in drones. Nosema pathology, and indeed, disease pathology in general has received little attention in male honey bees compared with workers and queens. Due to differences in life history and genetic background, male honey bees may have caste-specific strategies for coping with infection (summarized in Chapter 4). Though results must be interpreted with caution due to differences in Nosema infection type, we found that inoculating drones with N. ceranae early in adulthood has limited effects on some aspects of drone sexual maturation (sperm counts do not differ between infected and control drones, and infected drone antennae are equally stimulated by a major component of queen sex pheromone). However, Nosema co-infection in drones, as in workers, is energetically costly and infected drones starve significantly faster than controls. Molecular studies show that expression of several candidate genes (identified in Chapter 3) involved in the insulin-signaling pathway (ilp1, ilp2, foxo) and those potentially involved in the Vg/JH hormonal network (usp) were similarly regulated in drones as in co-infected workers. These findings point to concordant energetic costs across casts and validate earlier gene expression studies [85]. Despite these similarities in caste responses to infection, behavioral studies documented divergent flight patterns across infected drones and foragers. While previous studies (summarized in Chapter 4) show that infected workers forage precociously, take longer foraging trips and shorter rests between trips, infected drones show no difference in latency to flight onset, but do take shorter mating flights on average and show a trend for taking longer rest periods between flights. These divergent behavioral responses to infection between castes may reflect optimized host evolutionary strategies for managing stress at the colony level. While additional studies are needed to evaluate this hypothesis, these experiments lay the groundwork for drone response to Nosema infection in natural colony settings. An additional goal of this dissertation was to directly compare worker and drone molecular responses to Nosema infection. Interestingly, however, drones and workers may differ in their susceptibility to Nosema apis and Nosema ceranae (Appendix C). Additional studies are needed not only to confirm putative caste-specific differences in susceptibility to Nosema spp. infection, but also to directly compare caste responses to the same infection type. Furthermore, workers infected with N. ceranae and housed in colonies showed different gene expression patterns than those predicted by former molecular studies housing workers in cages (Appendix C). Differences in findings across these and prior molecular studies point to the need for in-hive (as opposed to cage) studies to accurately characterize Nosema pathology. Indeed, few studies, aside from those outlined in Chapter 4 and Appendix C, have measured gene expression in workers retrieved from colony settings [86] (also, Dr. Judy Chen has compared genome-wide expression in workers collected directly from colonies-personal communication). Collectively, Chapters 1,3 and 4 point to the energetic demands of Nosema infection across castes as a driving force behind disease symptoms. Furthermore, the common observation that infected workers will consume additional food if allowed suggests that nutritionally supplementing infected colonies may mitigate some disease symptoms. Indeed, as summarized in Chapter 1 (see Section 5.3), good nutrition is vital to worker health and ability to cope with abiotic and biotic stressors. Unfortunately, cage and especially field studies examining the interaction between Nosema infection and diet are limited. Cage studies indicate that high quality diet can improve worker tolerance, but workers consequently carry greater spore loads [ ]. On the other hand, one field study found that nutrition supplementation cannot extend the life 82

94 expectancy of infected workers, indicating complex interactions between colony nutrition and individuals behavioral states, infection statuses, and individual life expectancies [179]. Since Nosema and many other stressors promote precocious foraging behavior, accelerated behavioral maturation can be interpreted as a strategic colony response to stress. However, if too many individuals become infected, colony division of labor may spiral out of control as younger and younger cohorts of workers must transition to foraging sooner to replace their sisters who foraged precociously and died prematurely [97]. Mathematical modelling suggest that the life expectancy of such stressed colonies can be extended through nutrient supplementation, but comprehensive field data examining the interaction between nutrition, Nosema infection and colony survival is lacking. However, the projected survival improvements of stressed colonies by nutrient supplementation (along with author recommendations to identify stressed colonies by monitoring age of foraging onset) offer hope that new treatment and monitoring technology can be developed to ameliorate global rates of colony loss due to diverse stressors [97]. As research progresses, there will be more that those working behind the bench can do to benefit beekeepers and bees. 83

95 84 Appendix A Reprint of Chapter 3 These materials are freely and openly available via

96 85

97 86

98 87

99 88

100 89

101 90

102 91

103 92

104 93

105 94

106 95

107 96

108 97

109 98

110 99

111 100

112 Appendix B 101 Supplementary Tables and Figures for Chapter 4 Table B-1 Average flight duration and flight number across observation trials N Mean (std.) N Mean (std.) Nosema Control Trial 1 Total flight length (9.28) (9.50) Mating flight (6.31) (6.53) Short flight (3.58) (3.37) Trial 2 Total flight length (7.07) (10.46) Mating flight (6.56) (6.76) Short flight (3.21) (3.32) Trial 3 Total flight length (6.00) (9.83) Mating flight (4.56) (6.22) Short flight (3.01) (2.99)

113 Table B-2 Model estimates and (standard errors) for transformed flight lengths 102 Flight type (transformation factor) Intercept Infection status Intercept variance (bee) Intercept variance (date) Intercept variance (bee*date) Residual variance Effect size Trial 1 All flights (0.7) 5.90** (.445) (.253).579** (.185) 1.669* (.859) NA 7.932** (.328) -.02 Long flights (-0.25).480** (.005).002 (.004).0001** (.00004).0002* (.0001) NA.001** (.0001).001 Short flights (0.6) 2.478** (.142) (.123).057 (.036).111* (.066) NA 1.271** (.085) -.01 Trial 2 All flights (0.5) 2.874** (.271) -.510** (.113).128** (.041).659* (.358).113** (.047) 1.319** (.062) -.11 Long flights (-0.6).163** (.007).025** (.004).0001** (.00004).0003 (.0003) NA.007** (.00005) -.01 Short flights (0.5) 1.821** (.154).192* (.076).051** (.017).201* (.107) NA.482** (.026).11 Trial 3 All flights (0.25) 3.388** (.312) -.698** (.094).036 (.025).934* (.451).091** (.036).703** (.041) -.16 Long flights (0.35) 2.136** (.035) -.168** (.021).002* (.001).008 (.005) NA.017** (.001) -.12 Short flights (1) 5.610** (.670).291 (.306) < * (.2.227).781 * (.463) 6.265** (.607) -.50 Note: * = p < 0.05 and ** = p < NA = interaction not included in the model because it was not statistically significant at the 0.05 level.

114 103 Table B-3 Model estimates and (standard errors) for transformed inter-flight duration. Trial (transformation factor) Intercept Infection status Prior flight length Sequent. mating flight Intercept variance (bee) Intercept variance (date) Residual variance Effect size Trial 1 (log(inter-flight)) 3.306** (.113).012 (.093) -.051** (.005) -.339** (.090).063** (.025).054* (.031).732** (.039).01 Trial 2 (-0.25).492** (.016) -.029* (.010).002** (.0004).093** (.010).001* (.0003).001 (.001).011** (.001).01 Trial 3 (-0.25).535** (.026) (.012).003** (.001).064** (.015).0003 (.0003).004 (.002).010** (.001).005 Note: * = p < 0.05 and ** = p < 0.01.

115 104 Table B-4 Model estimates and (standard errors) for transformed flight rate. Trial (transformation factor) Intercept Infection Number of flight Infection * Number of flight Intercept variance (bee) Residual variance Trial ** ** -.039*.292* 3.853** (0.6) (.163) (.231) (.011) (.017) (.089) (.157) Trial ** ** -.149**.550** 2.149** (0.6) (.166) (.264) (.012) (.018) (.140) (.110) Note: * = p < 0.05 and ** = p < 0.01.

116 105 A Average short flight duration (Trial 1) Flight duration (min) Control Nosema Drone age (days) B Flight duration (min) Average short flight duration (Trial 2) Control Nosema Drone age (days)

117 106 C Flight duration (min) Average short flight duration (Trial 3) Control Nosema Drone age (days) Figure B-1 Average flight duration for all short (<12 minutes) drone flights. Data given for Trial 1 (A), Trial 2 (B) and Trial 3 (C). The number of flights taken by members of each treatment group on a given day are recorded at the base of each column.

118 107 A Inter-flight duration (min) Average inter-flight duration (Trial 1) Control Nosema Drone age (days) B Inter-flight duration (min) Average inter-flight duration (Trial 2) Control Nosema Drone age (days)

119 108 C Inter-flight duration (min) Average inter-flight duration (Trial 3) Control Nosema Drone age (days) Figure B-2 Average inter-flight duration between all consecutive drone flights. Data given for Trial 1 (A), Trial 2 (B) and Trial 3 (C). The number of flights taken by members of each treatment group on a given day are recorded at the base of each column.

120 109 A Inter-flight duration (min) Average interval between consecutive long flights (Trial 1) Control Nosema B Inter-flight duration (min) Drone age Average interval between consecutive long flights (Trial 2) Control Nosema Drone age

121 110 C Inter-flight duration (min) Average interval between consecutive long flights (Trial 3) Control Nosema Drone age Figure B-3 Average inter-flight duration between all consecutive long ( 12 minutes) drone flights. Data given for Trial 1 (A), Trial 2 (B) and Trial 3 (C). The number of flights taken by members of each treatment group on a given day are recorded at the base of each column Appendix C Supplementary data for Chapter 5 An additional goal of this dissertation was to directly compare the impact of Nosema infection on drones and workers. To address this objective, we infected newly emerged workers from the same colonies that we obtained drones from in Chapter 4 (see Section 2.9). We also used the same spore isolates to inoculate drones and workers for each trial. However, drones and workers were infected on different days due to logistical limitations. Spores isolates were stored at 4 C between caste infections for each source colony. In the first trial, drones were infected first, and workers 3 days later, while in the second trial, workers were infected first and drones infected 8 days later. Workers (and drones) were released into surrogate colonies for the 14-day interim between inoculation and collection. Worker gene expression was evaluated as before (Chapter 4, Section 2.9) though qrt-pcr data was first log-transformed to meet conditions for normality and analyzed by GLM (normal distribution, identity link function). As with drones, data were combined across trials where appropriate. If normality conditions were violated, data were processed with Wilcoxin Rank Sum Tests.

122 Two interesting findings arose from this study. First, we found that in both experimental trials (using bees from two different source colonies and using 2 separate spore isolates), drones were co-infected with N. ceranae and N. apis, while workers only had N. ceranae infections as detected by PCR [123] (Figure 1). This finding suggests that there may be caste-specific differences in susceptibility to N. apis infections and/or N. ceranae may be more competitive in workers than in drones. Either scenario proposes complex intra-colony disease dynamics that may contribute to global patterns of Nosema spp. prevalence (see Chapter 1 Section 2.2). Indeed, experiments in workers indicate within host competition between N. apis and N. ceranae, with N. ceranae achieving greater competitive success in some instances [61]. A second important finding from this study was that many candidate genes that had previously been regulated by infection in caged workers at 14 days post-inoculation were not affected by infection in workers in these studies. As expected, vg was significantly suppressed (X 2 (1) = 13.9, p =0.0002) in infected workers across both trials (Figure 2A) while ilp2 was significantly suppressed in infected workers in the second trial (X 2 (1) = 11.6, p = ) (Figure 2B). PGRP-S3 was also significantly suppressed in infected workers in the first trial (Wilcoxin rank sum, 1-way Test, ChiSquare Approximation; X 2 (1) = 8.6, p = 0.003) (Figure 2C). However, ilp1, foxo, usp, and ECSIT were not significantly, differentially expressed across treatments (p > 0.05) (Figure 2A). For data jointly processed across trials, significant trial effects were observed for vg (X 2 (1) = 88.3, p < ), foxo (X 2 (1) = 90.4, p < ), usp (X 2 (1) = 44.2, p < ) and ECSIT (X 2 (1) = 85.2, p < ). No significant effects of trial and/or treatment were observed for ilp1 (X 2 (3) = 3.5, p = 0.32), though there was a non-significant trend for suppressed ilp1 expression in infected workers (Figure 2A). Differences in relative expression of candidate genes across these two studies may be due to a number of factors. First, in the previous microarray experiments, workers were caged, fed on sugar water and MegaBee and co-infected, while in these experiments, workers were reared in colonies on naturally collected forage and only infected with N. ceranae. Second, different populations of workers and parasite isolates were used across studies. Though all these factors may contribute to differences, rearing workers in colonies versus cages likely had a very large effect since workers reared in colonies would have been subject to chemical colony cues, social interactions at a much larger scale and likely enjoyed superior natural diets. The differences in findings across these and prior (Chapter 3) molecular studies point to the need for in-hive experiments where individuals infection costs may be buffered and/or temporally shifted due to colony environment. 111

123 Figure C-1 PCR confirming N. apis and N. ceranae infection status in a subset of drones (Chapter 4) and workers (Chapter 5) from molecular experiments. In well labels, I or II indicates Trial 1 or Trial 2, C or N indicates Control or Nosema treatment, and 1-3 indicates sample number. All control drones and workers were uninfected. Infected drones from both trials were parasitized by N. apis (band at 257 bp) and N. ceranae (band at 662 bp) [123]. Workers from both trials were only infected with N. ceranae. Negative controls for N. apis and N. ceranae trials are labeled as (H20Na and H20Nc). 112