Honey Bee Health in Maine Wild Blueberry Production.

Simple Summary: Wild blueberry is an important North American native crop that requires insect pollination. Migratory western honey bee colonies constitute the majority of commercial bees brought into Maine for pollination of wild blueberry. Currently many stressors impact the western honey bee in the U.S. We designed a two-year monitoring study (2014 and 2015) to assess the potential health of honey bee colonies hired for pollination services in wild blueberry fields. We monitored the colony health of 9 hive locations (3 hives / location) in 2014 and 9 locations (5 hives / location) in 2015 during bloom (May – June). Queen health status, colony population size, rate of population increase, and pesticide residues on pollen, wax, and honey bee workers were measured during bloom. In addition, each hive was sampled to assess levels of mite parasites, viruses, and Microsporidian and Trypanosome pathogens. Different patterns in colony health were observed over the two years. Factors predicting colony growth rate in both years were Varroa mite infestation and risk due to pollen pesticide residues during bloom. In addition, recently discovered parasites and pathogens were already observed in most these colonies suggesting that parasites and diseases spread rapidly and become established quickly in commercial honey bee colonies. Abstract: A two-year study was conducted in Maine wild blueberry fields ( Vaccinium angustifolium Aiton) on the health of migratory honey bee colonies in 2014 and 2015. In each year 3-5 colonies were monitored at each of 9 wild blueberry field locations during bloom (mid-May until mid-June). Colony health was measured by assessing percent worker and sealed brood rate of change from the beginning of bloom until the end of bloom. Potential factors that might affect colony health were queen failure or supersedure; pesticide residues on trapped pollen, wax comb, and bee bread; and parasites and pathogens. We found that Varroa mite and pesticide residues on trapped pollen were significant predictors of colony health as measured by the percent rate of change of sealed brood during bloom. These two factors explained 71% of the variance in colony health over the two years. Pesticide exposure was different in each year as were pathogen prevalence and incidence. We detected high prevalence and abundance of two recently discovered pathogens and one recently discovered parasite, the trypanosome Lotmaria passim Schwartz, the Sinai virus, and the phorid fly, Apocephalus borealis Brues. Contributions: methodology, collection, formal analysis, resources, F.D.; writing—original draft preparation, F.D.; writing—review and editing, F.D. B.E., J.L.; supervision, F.D.; project administration, F.D.; funding acquisition, F.D., B.E. All

worker population rate of increase, sealed brood population size, sealed brood population rate of increase, pathogen and parasite loads, and pesticide residues in trapped pollen, wax comb in the brood area, and stored bee bread.

Materials and Methods
Study site and colony health measures. This study was conducted in the major wild blueberry growing regions in Maine, Hancock and Washington counties, in the towns of Alexander, Aurora, Cherryfield, Columbia, Deblois, Jonesboro, and unorganized township T22, during the years 2014 and 2015. Beekeepers and blueberry growers were contacted for permission to sample hives in the field during the period of bloom (usually mid-May to mid-June). Several hive locations in wild blueberry fields were monitored each year. We monitored 9 hive locations (fields) (3 hives/location) in 2014 and 9 locations (5 hives/location) in 2015. In each field, hives were on wooden pallets with 4-6 hives per pallet and the numbers of hives per field ranged from . Colony sampling was conducted three times during bloom. In 2014, bloom was from 18 May-13 June and in 2015 bloom was from 20 May-16 June. Colony health was measured by sampling queen presence, egg laying and the presence of presence of supersedure cells, indicating preparation of queen supersedure. Sealed brood population was calculated by determining the percent area of wax comb on varying sized hive bodies with sealed brood and then converting to numbers of brood using published formulae [26]) Worker population size was calculated by determining the % area of wax comb on varying sized hive bodies with workers and then converted to numbers of workers using published formulae [26]). The health status sampling was performed twice, shortly after hive deployments in blueberry fields and just before bloom ended. A measure of population growth was calculated by determining the % rate of change from the first sample to the second sample of both sealed brood and workers (see equation 1). Previous studies have found that colony population size and rate of change is a reliable measure of colony health [27]. % rate of change = (population size at time t -population size at time t+1) * 100 [1]  were attached to the entrances without the trapping gate set for 24 hours, allowing colonies to adapt to the trap. On the following day the trapping gate was closed and pollen was trapped for 48 hr [28]. Wax comb (ca. 100 cm 2 ) was collected adjacent to the brood area of 3 hives/location during late bloom and pooled by location. In 2015 only, 5-10 gm/hive of bee bread (stored pollen mixed with nectar) was extracted from wax comb above the brood area, from 3 hives/location and pooled by location. Residue samples were transported at the end of the day of collection from the field to the laboratory in Orono, ME in insulated coolers containing ice-packs. Once at the laboratory samples were stored at -80 ºC. Samples were shipped overnight to the pesticide residue analysis laboratories when requested. In 2014 pesticide analyses of samples were conducted at the Connecticut

of 21
Agricultural Experiment Station, New Haven, CT., U.S. and in 2015 the analyses were conducted by the USDA National Science Laboratory, Gaston, N.C., U.S. The Connecticut Agricultural Experiment Station analytical chemistry laboratory used high pressure liquid chromatography analysis targeting 140 different pesticides and metabolites after extracting the residue targets using a modified QuECHERs procedure [29]. More details of the procedures can be found in Ostiguy et al. [30]. The USDA National Science Laboratories in Gastonia, NC. Screened for 200 agricultural pesticides and metabolites used gas chromatography and high-pressure liquid chromatography with mass spectrometry. This laboratory also utilized a modified QuECHERs procedure for extracting residue targets from the various matrices. Limits of detection (LOD) mostly ranged from 5 -25 ppb (maximum = 50 ppb) depending upon the matrix that residues were extracted from and the specific pesticide (eg. pyrethrin LOD = 50 ppb). Details of the procedures used at the National Science Laboratories can be obtained by contacting Dr. Jonathan Barber, Chemistry section supervisor (Jonathan.Barber@usda.gov).
Limits of detection (LOD) ranged between 0.5 and 20 ppb depending upon the matrix (pollen, wax, bee bread) that residues were extracted from. Most of the compounds had an LOD of less than 5ppb with 88 compounds at 1 ppb or less. However, the pesticides in the two laboratories' screens differed and the detection levels differed for many of the same pesticides that were in the screen. Because of this we chose to minimize laboratory bias by only selecting pesticides that both laboratories searched for and we only consid- when concentrations were at 10 ppb or higher. We realize that this reduced the number of pesticide detections and the overall total concentration of residues by eliminating detection of low concentrations over the two-year period, but it provided a consistent benchmark for making comparisons of exposure and toxic risk between years. We also corrected the detections as described above for bee bread even though this matrix was only sampled in 2015.
A quantitative measure of pesticide risk was calculated from the residue data to determine if levels of exposure observed from pollen, wax, and bee bread affected colony health. Contact Hazard Quotients were calculated using methods by previous authors [28,30], but because Oral LD50 estimates are: 1) less available, and 2) because contact and oral Hazard Quotients are highly correlated [28]; we only estimated contact Hazard Quotients for this study. To calculate the contact Hazard Quotient, lethal dose 50 th percentile values (LD50 in units of ppb) were compiled for all detected compounds based upon available literature and public databases (see [28,30] for information on databases). We used the LD50 values of parent pesticide compounds if LD50s were not available for metabolites [28,30]. Then we divided the concentration of each pesticide quantified in trapped pollen, wax comb, or bee bread for a given location by the contact LD50 estimated for honeybees. Contact LD50 values reported in terms of μg/bee were converted to ppb Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 5 May 2021 doi:10.20944/preprints202105.0037.v1 relative to body weight (ng pesticide per g bee) by multiplying each value by a factor of 10,000; this is an approximate equivalent to 1,000 ng per μg÷mean bee weight of 0.1 g [32]. An estimated Hazard Quotient of 1.0 suggests that the exposure level by contact will result in 50% mortality to colony populations. A Hazard Quotient greater than 1.0 represents an expectation of high proportions of mortality. Based upon these Hazard Quotients, we assessed risk both at the individual pesticide compound level, pesticide use-group level, and also additively across all pesticides detected, which provides a measure of total colony risk. This total colony Hazard Quotient assumes that effects due to pesticides are additive and this is most likely not the case based upon studies showing synergy among pesticides in honey bees [33,34]. However, we feel that our use of an additive Hazard Quotient is acceptable because we use total colony risk only as a relative measure of colony stress for comparing locations and years and not as an absolute estimate of acute mortality. Year and its interactions with fixed effects were included in the models as random effects. Because pesticide analysis was conducted on samples pooled over colonies within a location, the causal factor analysis was performed with location-level data using the pooled and averaged values by locations for all dependent and independent variables. All correlations and mixed models were estimated with JMP statistical software [38].   Table 1. Fungicides, herbicides, insecticides, and miticides were represented in the residues with the highest concentrations. The concentrations listed in Table 1 have high variation between hives/location as reflected by the large standard errors relative to the mean concentrations.     Figure 5 shows the percent composition by pesticide use-group of the calculated Hazard Quotient by year and route of exposure (ie. trapped pollen, wax comb, bee bread). For both years the risk to exposure of trapped pollen and wax comb is almost entirely due to miticides ( Fig. 5A-D), although insecticides contribute a measurable proportion of the total risk. Insecticides can contribute a disproportional amount of risk relative to concentration. As an example, in 2015, insecticides constituted 0.35% of total pesticide concentration (ppb) in wax comb, but 4.8% of the total Hazard Quotient ( Fig. 4D and 5D), and 10.1% of total pesticide concentration in bee bread, but 58.3% of the Hazard Quotient (Fig. 4E & 5E). This is due to the high proportion of total ppb contamination by miticides (Fig. 4A-D between years (P>0.05) (Fig. 6A-B). Phorid fly eggs and larvae (Apocephalus borealis) were only found in 2015, but they were common at all locations (mean infestation/loca-tion=5.5±1.0%) (Fig. 6B). The green dashed line in Figures 6A and 6B are the treatment threshold commonly recommended for Varroa Mite [39]. In 2014, five of the nine locations had colonies with Varroa mite infestation greater than the threshold of 3 mites/100 honey bees [39]. In 2015, none of the locations exceeded this threshold. In 2014, BQCV, DWV, and N. ceranae were detected at all nine locations, but virus incidences were at low to moderate levels relative to L. passim (Fig. 7A)). Sac brood virus and Trypanosome infections were detected at all but one location in 2014 (Fig 7A). Three viruses (KBV, IAPV, CBPV) were either at an extremely low prevalence and incidence or were absent in the honey bee populations we sampled. Because of this we did not plot these three viruses (Fig. 7A). In 2015, several viruses (ABPV, CPBV, and IAPV) were either not detected or were very low in prevalence and incidence (not plotted). Figure 7B

Discussion
Wild blueberry is an obligate insect pollinated plant, mostly dependent upon bees [5,40]. Migratory honey bee hives are used to supplement the pollination by native bee species [5,41]. There has been concern by both honey bee keepers and wild blueberry growers about the health of colonies that are brought in for pollination services each year.
In our study we found that colony health varies among years.
In migratory colonies brought to wild blueberry for pollination are fed sugar syrup during bloom and so we suspect that differences we observed between years was not due to starvation. Nutritional content of wild blueberry has been shown to be suboptimal for honey bees [43], but honey bees usually do not collect a high percentage of blueberry pollen while foraging and most of their dietary intake of pollen comes from other plant species surrounding wild blueberry fields [12]. We speculate that the nutritional quality of pollen would not fluctuate annually in a way that would explain the observed differences in colony health between 2014 and 2015.
We found that colony health over the two-year study period was best described by Varroa mite densities and the Hazard Quotient estimated from trapped pollen pesticide residues (both logarithmically transformed). The residue concentrations in 2014 were higher than 2015, but in both years the amount of miticide and miticide metabolites found in both pollen and wax comb was high. In this study we have assumed that Amitraz, Coumaphos, Fluvalinate, and their metabolites were all due to the use of these compounds as miticides to control Varroa mite. This is based upon the supposition that all of formulations of these compounds are currently registered for Varroa mite control but also because none of these pesticides were or are recommended for use in wild blueberry insect pest management [44] and as far as we know they have never been used by wild blueberry growers for crop pest management (Drummond pers. com.). The dominance of miticides in pollen and wax comb has been reported in a large-scale apiary study conducted in Spain [45]. These miticides are toxic to honey bees at high doses which were detected in this study [30]. In 2014, miticide levels in hives were exceptionally high and yet Varroa mite levels were also extremely high. This situation probably reflects Varroa mite resistance to Amitraz and Coumaphos miticides in the U.S. Resistance was first observed for both miticides in the late 1990s and early 2000s [46,47], although it can be seen that a decade and a half after the first reports of resistance, these miticides were still being heavily used.
Even though we found Varroa mite and the trapped pollen residue Hazard Quotient to be significant predictors of % colony population change during bloom one must still be cautious in concluding that the cause of the differences in colony health was due to only these two factors. Pesticide residues in the hive result in a complex dynamic and one measure, trapped pollen Hazard Quotient, may not adequately capture the mechanisms at play and subsequent health risk to the colony. We are aware that our measure of risk to honey bee colonies is crude. We only estimate contact risk based upon the LD50 response to pesticides in workers. Our approach did not capture toxicities of "inert" ingredients used in pesticide formulations which have been shown to have detrimental behavioral and physiological effects on honey bees [48]. It also did not capture the oral risk which can't be predicted from contact risk [30], or larval sensitivity to pesticide exposure [49], and synergy among mixtures of pesticides which will be the norm in the hive environment [50].
It is difficult to make predictions of the environment outside the hive with trapped pollen. Our initial assumption, along with other authors of several published studies [51], the foraging territory of the honey bee colony (in this study, blueberry fields). However, this might not be a valid assumption for all residues detected. We are suspicious that the high level of miticides that would be used for Varroa mite control detected in pollen would be an independent measure of floral contamination. Although, it could be the case that honey bee body surfaces contaminated with miticide after a recent miticide treatment contaminated pollen by direct body contact of the contaminated honey bees with floral surfaces or that contamination of pollen occurred when previously contaminated honey bees groomed the pollen off of their bodies and packed the pollen in their corbiculae [52]. Bee body contact with floral surfaces is the suggested mechanism of transfer of honey bee parasites and pathogens to native bees [53].
Wax comb residues were initially assumed to be predictive of colony health since they represent the integral of incoming contamination over time (minus degradation).
However, in our study, trapped pollen Hazard Quotients was a better predictor of colony health. This may be explained by difference in actual exposure (food vs contact through comb) which is difficult to measure but has been demonstrated with differences in outcomes to queens exposed to different sources of pesticides during development [54]. In addition, Hazard Quotients of trapped pollen and wax comb were not correlated (P=0.429).
The Hazard Quotients of bee bread, the processed food of larvae, was also expected to be a good predictor of colony health, especially sealed brood percent change, but we only had data from 2015 and so this metric could not be adequately tested. However, the Hazard Quotients of bee bread was significantly correlated with the Hazard Quotients of trapped pollen in 2015 (r=0.879, P=0.002) and so the risk due to contaminated bee bread may also be a good predictor of colony health in future studies.
Pesticide residues can have acute effects on individual honey bees and colony populations [55] or exposure can result in more chronic conditions [56]. Symptoms of pesticide exposure can be death of individuals in the colony [55], reduction in colony growth rate [57] reduction in queen productivity, increase in supersedure or queen loss [54], reduction in cognition and sensory modalities [58], and repellency of floral resources to foragers [59].
While colony losses have been shown to be caused by exposure to pesticides, especially insecticides, fewer studies have shown that miticides used to treat Varroa mite can have negative effects on colony population size. Johnson et al. [60] showed that interactions between miticides can result in highly toxic responses in honey bee workers. The use of Hazard Quotients has been used to assess exposure and potential colony effects [30,31], but a few studies have used these metrics with success to explain colony losses or declines in colony population size over time [61,62].
Honey bees in the U.S. often carry a high diversity and heavy load of pathogens and parasites [63]. We found this to be the case in both years of our study. Both tracheal mites and Varroa mites were abundant, but not equally across all locations. Molecular markers for 5 viruses were common with relatively high copy numbers, while markers for 4 viruses were either not detected or not prevalent and usually were represented by low copy numbers. The recently (2011) discovered Lake Sinai virus [64]  have found similar high prevalence and incidence along with evidence suggesting that virulence can be high [65,66].
We were surprised to observe infestation of the parasitic phorid, A. borealis in 2015 at all sites ranging from 2.2-10.6% parasitism (average 5.5%). We did not detect it in 2014.
In 2012, this parasite was detected in commercial colonies in South Dakota; San Francisco, California; and the Central Valley of California with parasitism levels at 12-38% [67]. It has been reported in the published literature in the U.S. since our study in 2015, but at low parasitism levels of 1-5% [68]. Whether this parasite is still common in commercial honey bee colonies in the U.S. is unknown. It has also been reported as a new parasite of the honey bee in Belgium [69] and Egypt [70]. Another new pathogen of fairly high prevalence and incidence (copy number) in our study was the trypanosome L. passim. This pathogen was described in the western honey bee in the U.S. in 2015 (71) and it has since been found to be detrimental to colony health [72]. With this rich diversity of pathogens and parasites, why was Varroa mite found to be the only significant causal factor in colony health? We speculate that because of the high level of co-occurrence of many of these pathogens in a colony and with the presence of the immune system compromising Varroa mite [73], pathogens become highly prevalent and abundant. Therefore, it is difficult to tease out a single causal pathogen agent and at a hierarchical level, high Varroa mite infestation in a colony represents a colony that has severe multiple pathologies of potentially different composition which ultimately can lead to a decline in colony health. The only constant or "Holy Grail" appears to be Varroa mite.

Conclusion
Health of migratory honey bee hives brought to wild blueberry for pollination was observed to vary over the two years that we conducted the study. Varroa mite infestation levels and pesticide residues in pollen (as measured by a Hazard Quotient) accounted for 57.8% of the variance in colony percent population growth rate during bloom. In general, pesticide residues other than miticides for the control of Varroa were common, but were not responsible for explaining a significant proportion of the variation in the percent change in sealed brood (P<0.498) when miticides were taken out of the Hazard Quotient).
Tracheal mite and many of the pathogens were common in both years, but were also not significant causal factors of colony health. Therefore, it appears that Varroa mite is the main factor responsible for colony health of migratory hives brought into pollinate wild blueberry. This is because the trapped pollen Hazard Quotient we identified as a causal factor is most likely the result of Varroa mite control prior to and during pollination.

Acknowledgements
We would like to thank all of the beekeepers and wild blueberry growers that allowed us access to hives and blueberry fields. We would also like to thank Ms. Judy Collins who helped coordinate the laboratory collection of data and sending off of samples to other laboratories for processing. We would also like to thank Wyman's Wild Blueberry