Colony Size, Rather Than Geographic Origin of Stocks

Copyedited by: OUP

Journal of Economic Entomology, XX(X), 2018, 1–9 doi: 10.1093/jee/toy377 Apiculture & Social Insects Research

Colony Size, Rather Than Geographic Origin of Stocks, Predicts Overwintering Success in Honey Bees Downloaded from https://academic.oup.com/jee/advance-article-abstract/doi/10.1093/jee/toy377/5251959 by guest on 11 January 2019 (Hymenoptera: Apidae) in the Northeastern United States

Mehmet Ali Döke,1,2,4 Carley M. McGrady,1 Mark Otieno,3 Christina M. Grozinger,1 and Maryann Frazier1

1Department of Entomology, Center for Pollinator Research, Huck Institutes of the Life Sciences, Pennsylvania State University, PA 16801, 2Current Address: Department of Biology, University of Puerto Rico, San Juan, PR, Phone: 814-441-2144, 3Agricultural Resource Management, Embu University College, Nairobi, Embu, Kenya, and 4Corresponding author, e-mail: [email protected]

Subject Editor: James Strange

Received 4 February 2018; Editorial decision 10 November 2018

Abstract Honey bees (Apis mellifera L.) are key pollinators of agricultural crops. However, approximately 30% of managed colonies die each winter in the United States. There has been great interest in breeding for ‘locally adapted stocks’ which survive winter conditions in a particular region. Here, we evaluate the impact of geographic origin of stock on colony weight, population size, and overwintering survival. Comparing four different U.S. honey bee stocks (two bred in southern and two bred in northern regions) under standard beekeeping practices in three different apiary locations in central Pennsylvania, we examined possible adaptation of these stocks to temperate conditions. We confirmed the genotypic difference among the stocks from different geographic origins via microsatellite analysis. We found that stock or region of origin was not correlated with weight, population size, or overwintering success. However, overwintering success was influenced by the weight and population size the colonies reached prior to winter where higher colony weight is a strong predictor of overwintering survival. Although the number of locations used in this study was limited, the difference in average colony sizes from different locations may be attributable to the abundance and diversity of floral resources near the honey bee colonies. Our results suggest that 1) honey bees may use similar strategies to cope with environmental conditions in both southern and northern regions, 2) colonies must reach a population size threshold to survive adverse conditions (an example of the Allee effect), and 3) landscape nutrition is a key component to colony survival.

Key words: overwintering, diapause, nutrition, landscape, behavior

In temperate climates, the winter season constitutes a great challenge adaptations can influence the long-term survival and success of for animals due to shortened photoperiod, dwindling resources, and local populations in their habitat (Reznick and Ghalambor 2001). extreme cold. Species inhabiting temperate regions have evolved In the case of honey bees, breeding for ‘locally adapted stocks’ strategies to survive the winter via molecular, physiological, and may significantly improve the survival and productivity of man- behavioral adaptations (van der Werf et al. 2009, Chen et al. 2014). aged honey bee colonies. Previous proteomic analyses of honey bee As a social insect, honey bees have evolved a variety of mechanisms stocks collected from around the world demonstrated significant to survive the winter, including distinct behavioral (decreased indi- differences in expression patterns of metabolic proteins accord- vidual activity, cessation of brood rearing, and formation of a ther- ing to geographic origin, though survival of these populations in moregulatory cluster) and physiological (altered endocrine profiles, a common, temperate environment was not assessed (Parker et al. increased nutrient stores, and longevity) features [reviewed in Döke 2010). Furthermore, a study spanning multiple regions in Europe et al. (2015)]. However, despite these adaptations, winter is still a demonstrated improved overwintering survival of colonies gener- very stressful period for honey bee colonies in temperate regions, ated from local stocks (Büchler et al. 2014, Hatjina et al. 2014). with ~30% average winter colony losses reported by beekeepers in In this 2-yr study, researchers followed colonies from 16 differ- the United States in the last decade (Kulhanek et al. 2017). ent genotypes that were placed in areas from six climatic regions Populations within a species can exhibit local adaptations to around Europe. Adult bee population and overwintering survival biotic and abiotic elements in their environment and these local was greatest when the genotypes were maintained in their home

2 Journal of Economic Entomology, 2018, Vol. XX, No. XX

range. These results demonstrate the existence of locally adapted into non-Russian colonies (Tarpy and Lee 2005), which is why we stocks in Europe. chose to use nucleus colonies for this group; this approach is consist- However, although European honey bees exhibit the clear, ent with standard beekeeping practices. However, since honey bee genetically distinct population structure necessary to create locally workers mature from egg to adult in 3 wk, and adult workers have adapted stocks (Estoup et al. 1995), U.S. honey bee populations a maximum lifespan of 6 wk in the summer (Winston 1987), all are not comparably structured. U.S. honey bee populations exhibit colonies should have consisted of worker offspring of their respec- genetic differences between vastly separated populations of east and tive queens by mid-July. Queens from different sources were marked west coasts (Delaney et al. 2009), but this has not been demonstrated with different colors. Each genetic stock was represented by 14–16 at a smaller geographic scale, or between Northern and Southern colonies that were equally distributed among three apiary sites (60 Downloaded from https://academic.oup.com/jee/advance-article-abstract/doi/10.1093/jee/toy377/5251959 by guest on 11 January 2019 United States. Honey bee colonies and queens are shipped across colonies in total). the United States by migratory beekeepers and commercial breeding We interviewed the queen bee breeders to obtain information

operations, which would serve to move alleles between distant loca- on their breeding practices. Breeder South1 (FL) has been using Pol- tions. Although genetic diversity of honey bees in North America line hygienic Italian (a name given to this stock by the commercial was thought to be limited due to bottleneck events created by intro- provider) queens bred for Varroa Sensitive Hygiene (VSH) behavior duction of bees to this continent (Cobey et al. 2012), heavy selection (Danka et al. 2016) with an outcrossing mating system since 2011.

by parasitic mites (Kraus and Page 1995), and consolidation of bee Breeder North1 (VT) has been selecting their own stock for local suc- breeders (Schiff and Sheppard 1995; 1996), an analysis of genetic cess (i.e., survival and productivity) which started from VSH queens diversity in managed honey bee colonies in Europe and United States purchased in 2004 with limited and sparse introductions from other

demonstrated that managed populations actually have higher levels regional beekeepers’ stocks. Breeder North2 (WV) is a member of of genetic diversity than their progenitor populations, likely due to the Russian Honeybee Breeders Association (RHBA) and has been increased admixture during breeding practices (Harpur et al. 2012). strictly using queens from the Russian stock since 2003. Both the Pol- In the United States, there is only one study in partial support of line and ARS Russian Bee stocks were developed through selective locally adapted stocks (MacGregor-Forbes 2014), but the research- breeding programs at USDA-ARS Honey Bee Breeding, Genetics and ers only used northern bred queens for requeening while maintain- Physiology Laboratory in Baton Rouge, LA (Rinderer et al. 2000; ing other colonies with their original southern bred queens (which Danka et al. 2016). We could not obtain information on breeding

were generated from commercial producers of ‘package’ bees rather practices of breeder South2 (TX). We investigated the genetic differ- than queen breeders) as a control. Under these circumstances, the entiation of the honey bee stocks used in this study through micros- observed increase in overwintering survival of the requeened versus atellite analysis (see Microsatellite Genotyping below). nonrequeened colonies can be due to the act of requeening itself, quality differences between the locally bred and package queens, Apiary Sites local adaptations of northern bred queens, or a combination of these Apiaries A (N40° 46′ 11.98″, W77° 40′ 41.84″) and B (N40° 47′ factors. 21.32″, W77° 40′ 49.36″) are located in a nonagricultural area of Here we test the hypothesis that honey bee breeding efforts in ~10 miles away from Pennsylvania State, University Park campus United States have produced locally adapted honey bee stocks with and within 1 mile of each other. Apiary C (N40° 45′ 32.84″, W77° higher overwintering success in climates where they were bred or 54′ 53.72″) is ~15 miles away from the first two, located in a more reduced overwintering success out of their climatic zone. We com- developed and agricultural landscape compared with the first two pared the overwintering success of colonies headed by queens from sites (see GIS results for details). two southern and two northern commercial honey bee breeders in apiaries located in central Pennsylvania, which experiences a tem- Management and Beekeeping Practices perate climate. We found that geographic origin of stocks did not Colonies were managed according to standard beekeeping practices. affect the overwintering survival. However, colony size and weight in Note that in a pilot study conducted in 2012 (data not shown), colo- fall, likely correlated with floral resources around the apiary, had the nies were not managed for Varroa mites or provided with supple- largest effect on the likelihood of colonies to successfully overwinter. mental nutrition, and only 5 out of 40 colonies survived (1 survivor colony for the two southern and one northern stock and 2 survivors Methods for the other northern stock). Thus, for this study, we focused on colony survival under standard beekeeping practices, rather than Honey Bee Stocks under ‘unmanaged’ conditions. Note that because of differences in

Package honey bees (~1.5 kg of worker bees and a mated queen) their biology and management practices, North2 (‘Russian’ stock) purchased from Gardener’s Apiaries (Baxley, GA) were installed into received different treatment, though this did not seem to affect their Langstroth hives in April 2013. Each colony was established using colony metrics relative to the other stocks (see Results). one package and maintained thereafter without splitting or combin- Packages were treated prior to installation with 25 ml of 3.5% ing them. Newly mated, laying queens purchased from one north- oxalic acid in 1:1 sucrose syrup to reduce Varroa mite populations

ern commercial breeder (based in Vermont—referred to as North1 (https://www2.gov.bc.ca/assets/gov/farming-natural-resources-and- below) and two southern commercial breeders (based in Florida and industry/agriculture-and-seafood/animal-and-crops/animal-produc-

Texas—referred to as South1 and South2, respectively) were installed tion/bee-assets/api_fs221.pdf). North2 colonies were exempt from into the colonies using standard apicultural practices in late May/ this initial treatment since they were delivered as nucleus colonies early June 2013. In addition, nucleus colonies with their associated with brood (it is less effective to treat colonies with oxalic acid when queens (~1.5 kg of worker bees, 5 combs, and any brood produced brood is present). At the time of installation, all package colonies by that queen) were purchased from a commercial breeder based in received a fumagillin treatment according to label instructions to

West Virginia (referred to as North2). Note that North2 consisted of reduce populations of Nosema, a microsporidian parasite of honey ‘Russian’ stock, and queens of these stocks are difficult to introduce bees (Holt and Grozinger 2016) and approximately 3 gallons of Copyedited by: OUP

Journal of Economic Entomology, 2018, Vol. XX, No. XX 3

sugar syrup (2:1, sucrose to water) over the course of the next 3 wk. dense thermoregulatory cluster in October, and cease brood rearing

North2 colonies were not fed initially. However, they were each sup- in November (Mattila et al. 2001). plemented with 2 gallons of sugar syrup at the end of September due to their low honey storage in comparison to the colonies of the Microsatellite Genotyping three other stocks, likely resulting from their late-season establish- We conducted DNA extraction, PCR amplification, and genotyping ment and relocation. of 56 individual honey bees (1 worker bee per colony except for the Colonies were regularly inspected for diseases, queen presence, 4 colonies that were lost before winter, 11–16 individuals per stock) and other issues that can affect survival. Varroa mite levels were using 10 microsatellite loci (A24, B124, AP66, A88, A28, A7, A113, determined monthly via either sugar roll test or sticky board method AP55, AP81, and AP43). We followed the standard protocols for Downloaded from https://academic.oup.com/jee/advance-article-abstract/doi/10.1093/jee/toy377/5251959 by guest on 11 January 2019 (Ostiguy and Sammataro 2000). August Varroa mite levels of North 2 genotyping honey bee colonies described in Delaney et al. (2009) (see colonies, which had not been treated with oxalic acid, as well as few Supplementary Material, Appendix A for details). Samples that were colonies within other stocks exceeded treatment threshold level of ambiguous at more than one locus were reprocessed to ensure none 20 mites daily drop on sticky boards (Sammataro et al. 2002), and a of our samples had greater than 35% missing data. We tested all formic acid treatment (Mite Away Quick Strips) was applied. Early loci for missing data, Hardy–Weinberg disequilibrium, null alleles, September screening showed that additional colonies were above and genotypic linkage disequilibrium (see Supplementary Material, the threshold and they were treated in the same fashion. After these Appendix B for details). treatments, mite loads of the four stocks were not significantly differ- We tested for genetic differentiation among the four stocks ent in October, 2013 [F(3, 55) = 1.67, P = 0.184; see Supplementary (South1, South2, North1, and North2) and two regions (Northern Material, Table D1 in Appendix D]. Moreover, mite loads were not and Southern) following methods described in López-Uribe et al. different among the three apiary sites [F(2, 55) = 0.39, P = 0.681; (2017). We used a pairwise comparison of three G-statistics (Nei’s see Supplementary Material, Table D1 in Appendix D] or different Gst, Hedrick’s Gst, and Jost’s Dest) with the function ‘diff_stats’ in the survival outcomes [F(1, 54) = 0.396, P = 0.532; see Supplementary R package mmod (Winter 2012). All three G-statistics determine the Material, Table D1 in Appendix D]. degree of genetic differentiation among stocks by comparing genetic diversity within and between a priori groups. Values close to 0 indi- Data Collection cate low differentiation, whereas values close to 1 indicate stocks Following the installation, colonies were inspected monthly from that are completely differentiated (Meirmans and Hedrick 2011). August to October. Colonies were weighed using a portable platform To understand the sources of variation for genetic differentiation scale, with the weight of equipment (boxes, frames, bottom board, among individuals, we performed an analysis of molecular variance inner cover, and top) subtracted from the gross weight to generate (AMOVA) between the Northern and Southern regions using indi- the ‘net weight’ of colonies. Thus, any reference to ‘weight’ hereaf- viduals as the sampling unit. For the AMOVA, we used the function ter refers to this net weight. The numbers of frames of adult bees, ‘poppr.amova’ of the R package poppr (Kamvar et al. 2014). brood, and food stores were performed using methods described in To further investigate genetic differentiation, we used discri- (Burgett and Burikam 1985, Page and Fondrk 1995). Assessment of minant analysis of principal components (DAPC) implemented in the number of frames of adult bees was made early in the morning the package adegenet for R (Jombart et al. 2010). This multivariate when the temperature was cool enough that little or no foraging has model-free approach does not assume Hardy–Weinberg Equilibirum begun but warm enough that bees are not clustered. Minimal smoke (HWE) to cluster individuals based on prior population informa- was used as each frame is removed and the coverage on each side is tion and is therefore useful when assessing genetic differentiation assessed as a value between 1 and 4, where 1= ¼, 2= ½, 3= ¾, and between nonnatural populations such as managed honey bee stocks. 4 = full frame. The values are added and then divided by 4 to gener- Using the DAPC analysis, we determined if there was genetic differ- ate the total sides of frames covered by bees and then by 2 to get entiation among stocks and/or between regions based on the ability the total number of frames covered by bees. This number can then to reassign individuals to their stocks of origin (K = 4), or to the be further converted to the number of bees depending on the size of ‘Northern vs Southern’ region (K = 2). We determined the number frames and knowing the average number of bees that cover a frame of principal components after two runs of the function ‘xvaldapc’. (e.g., ~1,200 adult bees can cover a single side of a deep frame). The first run of 30 replicates identified an approximate number of Assessment of brood and stores is performed in a similar fashion to principal components (PC) associated with the highest proportion of that of adult bees. This method works well when standard equip- successful outcome predictions. The second run of 1,000 replicates ment is used in all colonies. Counts can be adjusted to total deep or centered the cross-validation around the approximate PC identified medium depth frames, or number of cells (deep frames have 6,800 in the first run. We used the number of principal components associ- and mediums have 3,825 cells). Winter survival was assessed in April ated with the highest number of successful reassignments and the of 2014. Colonies which survived to this point in the year were con- lowest root mean square error. sidered survivors since in a real-life scenario, both the natural floral resources and supplementary feeding by beekeepers can practically GIS Analysis assure the colonies’ survival through the summer except in the rare The landscape surrounding apiaries was assessed using the National cases such as sudden queen loss, disease, animal attacks, or pesticide Agricultural Statistical Service (NASS) map layer 2012 and ground- applications. truthed in October 2014. The maps were uploaded in ArcGIS 10.1 For all the reported analyses, we only used the data collected and buffer layers created at increasing radii. These buffer layers during the October observations, since this was the latest reliable started from the center of the apiary and extended outwards to 250-, fall date for colony measurements and thus should provide the best 500-, 1,000-, 1,500-, and 2,000-m spatial scales to reflect the flight data for assessing the condition of the colonies as they entered the ranges of bees (Greenleaf et al. 2007, Joshi et al. 2016). The buff- overwintering phase. In the temperate North America, colonies start ers were then uploaded to Fragstats version 4.0 and proportions of producing their long-lived overwintering bees in September, form a each land cover/land use type around the apiary calculated for each Copyedited by: OUP

4 Journal of Economic Entomology, 2018, Vol. XX, No. XX

spatial scale. To simplify our results and reflect most relevant spatial different regions (see Supplementary Material, Tables C1–C3 in scale to honey bee foraging, only 2,000-m data are reported in the Appendix C). Overall estimates of genetic differentiation between

results (Couvillon et al. 2015). the two regions were 0.004 (95% CI: −0.004, 0.013) for Nei’s Gst,

0.023 (95% CI: −0.019, 0.064) for Hedrick’s Gst, and 0.014 (95%

Statistical Analyses of Honey Bee Health Metrics CI: −0.012, 0.040) for Jost’s Dest based on average heterozygosity. All statistical analyses (except microsatellite analysis, see above) For the AMOVA, we partitioned genetic variation among three were performed using JMP Pro 10 software (Cary, NC: SAS Institute levels of organization: individuals, stocks, and regions. The majority Inc.). A factorial (two-way) ANOVA was conducted to compare the of the variation, 98%, was found among individuals within stocks (ϕ = 0.010, P = 0.36). We found no differentiation between stocks main effects of stock and apiary location, and the interaction effect ST Downloaded from https://academic.oup.com/jee/advance-article-abstract/doi/10.1093/jee/toy377/5251959 by guest on 11 January 2019 within each region ( = −0.007, P = 0.70). The difference between between stock and apiary location on colony weights, adult, and ϕSR brood population sizes in October. ANOVA (one-way) was used for Northern and Southern regions was significant but small, accounting for 1.73% of the overall genetic variation ( = 0.017, P = 0.01; comparing means of October weights, and adult and brood popula- ϕRT tion sizes among different survival outcomes. Tukey’s HSD test was Table 1). performed as a post hoc for the ANOVA analysis to further evalu- Using DAPC, we were able to assign individuals to their stocks of ate differences between means. A Fisher’s exact test (two-sided) was origin with 70% success rate (Fig. 1A) and to their region of origin used for comparing survival ratios of different genotypes and loca- with 82% success rate (Fig. 1B), further supporting the genetic dif- tions. The Fisher’s exact test was chosen over standard chi-squared ferentiation between regional groups found in AMOVA. analysis due to differences in sample sizes of groups and small data set. A logistic regression model was used for examining the likeli- Impact of Stock and Stock Region of Origin With hood of October weight in predicting overwintering survival. Winter Survival Out of the 56 colonies that were alive in October, 39 survived to our April checkpoint in spring. Although survival rates of colonies Results with queens from the four different stocks were numerically differ-

Evaluating Genotypic Differences Among Stocks and ent (South1: 75%, South2: 57%, North1: 82%, and North2: 67%), Between Stock Region of Origin this difference was not statistically significant (P = 0.57, Fisher’s Over the summer, 4 of the 60 colonies were lost due to queen fail- exact test, two-sided). Similarly, there was no significant difference ure or replacement. We used standard protocols (Delaney et al. between the overwintering survival of the colonies when grouped by region of origin (Southern: 67% and Northern: 73%, P = 0.77, 2009) to genotype the remaining 56 colonies that survived (South1: Fisher’s exact test, two-sided). n = 14, South2: n = 16, North1: n = 11, North2: n = 15 colonies). Locus B124 was removed for 26.8% missing data across all samples (see Supplementary Material, Tables B1 and B2 in Appendix Association of Colony Metrics With Winter Survival B). In the remaining 9 loci, we found no deviations from Hardy– Colonies that survived the winter had significantly higher colony Weinberg equilibrium, nor any evidence of linkage disequilibrium weights [F(1, 54) = 31.77, P < 0.0001] as well as larger adult worker (see Supplementary Material, Tables B3–B5 in Appendix B). We did bee populations [F(1, 54) = 37.28, P < 0.0001] in October than colo-

find evidence of null alleles for locus AP81 in the North2 popula- nies that did not survive the winter. The mean weights of survivor and tion (see Supplementary Material, Tables B6 and B7 in Appendix B). nonsurvivor colonies were 29.59 ± 1.33 and 15.96 ± 2.02 kg, cor- Thus, we performed all genetic analyses (G-statistics, AMOVA, and responding to 18.44 ± 0.79 and 9.71 ± 1.19 frames of adult worker DAPC) with and without AP81 and found similar results. Therefore, bees, respectively. Since they had received supplementary nutrition we concluded that this isolated incident of null alleles was not driv- in the fall, we also examined the relationship between colony weight ing the differentiation we found and included AP81 in our final and survival within the Russian stock and found that survivors analysis. For the 9 loci considered in our study, the number of alleles (59.2 ± 5.5 kg, N = 10) were significantly heavier in October than per locus varied between 4 and 14 with expected heterozygosity per nonsurvivors [29.4 ± 7.8 kg, N = 5; F(1, 13) = 9.691, P = 0.008]. This locus between 0.25 and 0.84 (see Supplementary Material, Table A2 demonstrates that not only the colony weights within the Russian in Appendix A). stock were diverse (despite the supplemental feeding), but also their The three G-Statistics revealed that overall estimates of genetic survival depended on fall weight just as the other colonies involved differentiation among stocks were small and not significant [Nei’s in this study. The October brood population was not significantly

Gst = 0.002 (95% CI: −0.012, 0.016), Hedrick’s Gst = 0.008 (95% different between colonies that survived or did not survive the winter

CI: −0.036, 0.052), and Jost’s Dest = 0.005 (95% CI: −0.022, 0.032)]. [F(1, 54) = 0.052, P = 0.820; see Supplementary Material, Table D3 However, a pairwise comparison of the four stocks revealed that in Appendix D]. although there was no differentiation among stocks within the A logistic regression model evaluating individual weights of colo- same region, there was genetic differentiation between stocks from nies in October and their survival state in spring showed that colony

Table 1. Results of AMOVA testing for differentiation between Southern and Northern regions, and among stocks in A. mellifera

df σ2 % variance ϕ-statistic P value

Between regions 1 0.082 1.73 0.017 0.01 Between stocks within regions 2 −0.034 −0.72 −0.007 0.70 Within stocks (error) 52 4.675 98.99 0.010 0.36

There was a significant difference between Northern and Southern regions ϕ( RT = 0.017, P = 0.01). However, the stocks within regions were not different

(ϕSR = −0.007, P = 0.70). Bold values indicate statistically significant (P < 0.05). Copyedited by: OUP

Journal of Economic Entomology, 2018, Vol. XX, No. XX 5

weight in October is a strong predictor of the overwintering survival P = 0.084]. For means and standard errors, see Supplementary 2 [Χ (1) = 26.5, P < 0.0001, R2 = 0.3855; Fig. 2A]. To better visu- Material, Table D4 in Appendix D]. Colonies from south had signifi- alize the association of colony weight in October to overwintering cantly larger honey stores (9.32 ± 0.8 kg) than the ones from north survival, individual weights of colonies were plotted against their [6.27 ± 0.8 kg; F(1, 54) = 7.092, P = 0.010]. However, the differ- survival status with the additional information on location shown ence between the summer honey reserves of colonies from different below the horizontal axis. Although none of the colonies weighing regions did not affect the survival outcomes as noted above. less than 10 kg in October survived through the winter, survival rates of colonies of 10–19.9, 20–29.9, and 30+ kg are 33, 85, and 94%, Impacts of Apiary Location and Stock on Colony respectively (Fig. 2B). Metrics Downloaded from https://academic.oup.com/jee/advance-article-abstract/doi/10.1093/jee/toy377/5251959 by guest on 11 January 2019 We did not record the amount of honey stores in colonies in Overwintering survival was significantly different among the three October, since some of the colonies received supplementary feeding locations used as apiary sites (P = 0.0097, Fisher’s exact test, two- in September. However, honey stores in July were significantly greater sided) with Apiary A having the lowest survival rate (44%), Apiary in colonies that survived (8.7 ± 0.7 kg) versus the colonies that died B being intermediate (72%), and Apiary C having the highest sur- [6.1 ± 1.1 kg; F(1, 54) = 4.341, P = 0.042]. There was no differ- vival rate (90%). The main effect for apiary location on colony ence in honey stores in July according to stock [F(3, 52) = 2.342, weight in October was significant [F(2, 44) = 12.240, P < 0.0001,

Fig. 1. Membership probability plots. Membership probabilities from the DAPC. Each vertical line represents a single individual; the different colors in each line represent the proportion of each individual’s genotype that belongs to a particular stock or region. An individual represented by a single color would have a genotype that was unambiguously of a single stock or region. When groups are completely differentiated, all individuals in the same group are a single color, distinct from those of other groups. In the stock membership plot (A), South1 genotypes are green, South2 are pink, North1 are purple, and North2 are brown (When viewed in monochrome, from darkest to lightest shade are North2, South1, South2, North1). Stock DAPC was performed with 18 principal components and

3 discriminant functions. Overall assignment-success rate = 70%, South1 stock = 64%, South2 stock = 75%, North1 stock = 64%, North2 stock = 73%. In the regional membership probability plot (B), southern genotypes are green (gray in monochrome) and northern are brown (black in monochrome). Regional DAPC was performed with 10 principal components and 1 discriminant function. Overall assignment-success rate = 82%, Southern region = 80%, Northern region = 84.6%

Fig. 2. Overwintering success is significantly associated with colony weight. (A) A logistic regression model of October colony weights and overwintering survival. Probability of survival (vertical axis) is strongly predicted by colonies’ October weights (horizontal axis, in kg). Dashed curves represent the 95% confidence interval. (B) October weights of individual colonies predict their overwintering survival. Here, each colony is represented with a bar. Length of the bar corresponds to the weight of the colony. Blue (black in monochrome) and red (gray in monochrome) bars represent survivors and nonsurvivors at the end of winter, respectively. Colonies are separated into locations and within each location; colonies are listed according to weight in an ascending order. Locations are noted below the horizontal axis. Copyedited by: OUP

6 Journal of Economic Entomology, 2018, Vol. XX, No. XX

Fig. 3], whereas the main effect for stock was not [F(3, 44) = 1.692, (~33%), forest (~29%), soybean (~14%), grassland (~11%), and P = 0.183]. The interaction effect (stock × apiary location) was also several other types of crops present at lower percentages (Fig. 4). significant for October weight [F(6, 44) = 4.768, P = 0.0008]. The mean October weights of colonies at the different locations track survival rates, with Apiary A having the lowest mean colony weight Discussion (19.5 kg), Apiary B being intermediate (25.7 kg), and Apiary C hav- Our study demonstrates the importance of colony size in overwin- ing the highest mean colony weight (31.5 kg). Post hoc pairwise tering success. Previous research in Germany also found a positive Tukey tests found a significant difference between Apiary A and C, correlation between number of adult bees in the colonies and over- whereas B was intermediate. wintering survival (Genersch et al. 2010). Honey bees generate large Downloaded from https://academic.oup.com/jee/advance-article-abstract/doi/10.1093/jee/toy377/5251959 by guest on 11 January 2019 The effect of apiary location and stock, and their interaction on colonies with up to 50,000 individuals. In the fall, colonies produce adult population in October (colony size) were in line with their effect ‘winter bees’ which have distinct physiological features, including on colony weight. The main effect for apiary location on colony size an extended lifespan relative to summer bees [~8 mo vs 6 wk; Free in October was significant [F(2, 44) = 7.197, P < 0.002], whereas and Spencer-Booth 1959, Fukuda and Sekiguchi 1966, reviewed in the main effect for stock was not [F(3, 44) = 1.515, P = 0.224]. The Döke et al. (2015)]. The winter bees form a thermoregulatory clus- interaction effect (stock × apiary location) was also significant for ter once temperatures drop to 10°C (Phillips and Demuth 1914). October colony size [F(6, 44) = 3.282, P = 0.009]. Previous studies have demonstrated that larger thermoregulatory Neither main effects—stock and apiary location [F(3, clusters have lower per capita honey consumption to maintain the 44) = 1.487, P = 0.231 and F(2, 44) = 1.098, P = 0.343, respec- same temperatures with smaller clusters (Free and Racey 1968), tively]—nor their interaction [F(6, 44) = 0.687, P = 0.661] were which indicates an Allee effect (positive correlation between popula- significant for amount of brood in October. Similarly, neither main tion density and individual fitness) acting upon the winter cluster. effects—stock and apiary location [F(3, 44) = 1.429, P = 0.247 and Additionally, an Allee effect in summer colonies was shown, where F(2, 44) = 0.493, P = 0.614, respectively]—nor their interaction [F(6, colonies must reach a population threshold to survive and grow [see 44) = 0.445, P = 0.844] were significant for number of varroa mites also Allee effects reviewed in Courchamp et al. (1999) and Stephens in October. For means and standard errors of all colony metrics, see and Sutherland (1999)] and also observed in other social insects such Supplementary Material, Tables D1–D3 in Appendix D. as ants (Luque et al. 2013) and bumble bees (Bryden et al. 2013). Recent mathematical models of honey bee colony survival includ- Analysis of Landscapes Surrounding Apiaries ing Allee effect in their model parameters successfully explain the GIS analysis of the locations used in this study shows the difference sudden collapse of colonies (Dennis and Kemp 2016, Booton et al. in landscape in the vicinity of the three apiaries (2000-m radius). 2017). Thus, larger fall populations may lead to better thermoregu- Apiary A is surrounded mostly by forest (~79%) and developed/ lation and more efficient use of nutritional stores in the winter, and open spaces (~6%). Apiary B, despite being in close proximity of larger populations in the spring to support more efficient brood rear- Apiary A, is surrounded by a different landscape composed of forest ing and colony growth, although there may be an optimal size after (~33%), grassland (~32%), nonalfalfa hay (~16%), and developed/ which the colony gets too large and loses the capacity to efficiently open space (~11%). Apiary C, on the other hand, is surrounded by thermoregulate (Jeffree and Allen 1956, Harbo 1993). Interestingly, a mixture of natural and agricultural landscapes occupied by corn the improved survival of ‘local stocks’ in Europe may also reflect an

Fig. 3. Apiary location significantly affected colony weight and survival. Blue bars represent average colony weights in each apiary (see primary vertical axis on the left) and the red line follows the survival rates in same apiaries (see secondary vertical axis on the right), P values for each variable are below their respective vertical axis. There was a significant effect of location on colony weight in October at theP < 0.05 level for the three locations [F(2, 44) = 12.240, P < 0.0001]. Colonies from Apiary A were lightest, whereas colonies housed in Apiary B were intermediate, and colonies from Apiary C were the heaviest when compared in October. This trend follows respective overwintering survival rates of the colonies housed in these three locations (P = 0.0097, Fisher’s exact test, two-sided). Significantly different means are denoted by different letters (Tukey HSD test), and numbers at the base of each bar represent the sample size. Copyedited by: OUP

Journal of Economic Entomology, 2018, Vol. XX, No. XX 7

Our results also suggest that floral diversity and abundance around an apiary are positively correlated with the overwintering success of the colonies. Obviously, since only three apiary locations were used in this study, our results cannot be statistically validated. However, previous studies indicate that increased nearby floral diversity and abundance improve the nutritional quality and quantity of the bees’ diets, and thus improve brood rearing, adult longevity, and colony growth (Donkersley et al. 2014, Horn et al. 2015, Smart, 2015, Sponsler and Johnson 2015, Vaudo et al. 2015). Downloaded from https://academic.oup.com/jee/advance-article-abstract/doi/10.1093/jee/toy377/5251959 by guest on 11 January 2019 In our study, the apiary situated in an agricultural landscape sup- ported colony growth more than the two apiaries situated in forested landscapes; this is consistent with previous studies indicating that ‘edge’ habitat within and surrounding agricultural crops provides higher nutritional resources (Requier et al. 2015). Notably, since Fig. 4. The proportion of land cover/land use types surrounding apiaries at both adult population size and colony weight strongly correlated 2,000-m spatial scale. Different types of land cover around the apiaries are with overwintering success, it suggests that the landscape effect is represented with different colors (or different shades in monochrome view) not due to increasing the amount of food stores (which would be a in the stacked bars representing the relative abundance of them. component of the colony weight), but rather may reflect the ability of the landscape to support a larger adult population. effect of genotype of colony size. Workers in colonies headed by local In our study, we used standard beekeeping practices, which queens exhibit greater longevity in their original range, suggesting included obtaining Russian-stock colonies from splits/nuclei rather improved resilience to local stressor (Büchler et al. 2014, Hatjina than requeening packages, managing Varroa loads when these et al. 2014). This increased longevity should lead to increased colony reached treatment threshold levels in the fall, and providing sup-

size as well (Barron 2015), though this was not explicitly measured plemental nutrition to colonies (the North2 Russian stocks) that had in this study. low fall honey stores. In a previous study (2012), we did not manage Overall, our results show that, although there is evidence for low colonies and thus assessed their survival under more ‘natural’ condi- levels of genetic differentiation among stocks originating from dif- tions, and only 12% of our colonies survived (1 survivor colony for ferent regions, honey bee stocks bred in the northern United States the two southern and one northern stock and 2 survivors for the where the average January temperatures range between −17 and −4°C other northern stock, data not shown), compared with 63% survival and southern United States where the average January temperatures in our reported study. Thus, as reported in previous studies, manage- range between 5 and 18°C (https://www.ncdc.noaa.gov) did not per- ment of Varroa populations in particular is critical to the survivor- form differently (in terms of fall weight and population size) at the ship of colonies (Boecking and Genersch 2008, Genersch et al. 2010, same location in the Northeastern United States, which is in the tem- Guzmán-Novoa et al. 2010, van Dooremalen et al. 2012). It also perate climatic zone and experiences a harsh winter season with an suggests, unfortunately, that the commercial stocks used in this study average January temperature of −4°C (see Supplementary Material, that are selected for resistance to Varroa are not effective at reduc- Appendix E for daily temperature and precipitation information in the ing the negative impacts of Varroa infestation without additional research area through the study period). Differences in the overwin- management strategies; similar results were obtained in a recent tering success of colonies, however, were explained by colony weight study (DeGrandi-Hoffman et al. 2017). Perhaps, we should consider and adult populations. These parameters were significantly affected by using ‘tolerant’ instead of ‘resistant’ in reference to European honey the location and thus likely reflect the properties of the surrounding bee stocks selected for survival in the absence of Varroa treatment landscape. Thus, for the commercial honey bee stocks we tested, man- (Bahreini and Currie 2015). Our results also indicate that, despite agement practices by beekeepers in terms of selection of locations and our management interventions, there was still sufficient variation in ensuring appropriately sized colonies in the fall appear to be stronger colony weight and size to lead to differences in survival, and this was factors in determining colony success than queen genotype, though not related to stock (indeed, there was still significant correlation in overall queen quality is undoubtedly an important factor. weight and survival in Russian stocks, which all received supplemen- Genetic differentiation at the regional level but not between indi- tal nutrition in the fall). vidual stocks from the same region is intriguing as it may suggest Although we did not test queen quality here, comparing our that, despite different breeding populations and practices, stocks results with other studies indicated that this is another key fac- in particular regions are genoptypically similar, which would be tor influencing colony growth and survival. In a recent Northeast consistent with local adaptation to broader environmental condi- SARE–funded beekeeper study on the effects of requeening colonies tions. Genetic differentiation between regional groups of managed with locally produced queens in the Northeastern United States, honey bees in neutral markers such as microsatellites indicates MacGregor–Forbes (2014) reported much higher overwintering that although there is gene flow between regions—likely a result of success when packages obtained from Southern package produc- migratory beekeeping and commercial bee breeding operations— ers were requeened with northern-bred queens in mid-June (80% the population of U.S. honey bees is not a panmictic population, at survival) versus maintaining the queen that came in the package least as reflected by the four stocks that we evaluated. Therefore, a (20% survival). The results of our study suggest that requeening genome-wide investigation of honey bees from multiple relatively alone may be beneficial to colony survival (perhaps by breaking isolated sites and breeding operations in United States has the poten- the brood cycle and lowering Varroa levels; Boecking and Traynor tial to yield local adaptations in parts of the genome under selection, 2007) or that queen quality is an important variable in colony suc- which was beyond the scope of this study. However, if such local cess. Queens obtained in some commercial packages often exhibit adaptations do exist in the four stocks tested here, then they are not reduced egg production and likely increased supersedure (M. Frazier, influencing overwintering success. unpublished results). In our study, all the queens were obtained from Copyedited by: OUP

8 Journal of Economic Entomology, 2018, Vol. XX, No. XX

well-established commercial queen breeders who routinely assess (Parasitiformes: Varroidae)-infected honey bee (Hymenoptera: Apidae) the quality of queens prior to shipment. Thus, overall, obtaining colonies in a northern climate. J. Econ. Entomol. 108: 1945–1505. high-quality queens, regardless of their region of origin, can improve Barron, A. B. 2015. Death of the bee hive: understanding the failure of an overwintering success. insect society. Curr. Opin. Insect Sci. 10: 45–50. Boecking, O., and E. Genersch. 2008. Varroosis – The ongoing crisis in bee Colony weight and size in October, which were the strongest pre- keeping. J. fur Verbraucherschutz und Leb. 3: 221–228. dictors of success for the following winter, for three of the four honey Boecking, O., and K. Traynor. 2007. Varroa biology and methods of control. bee stocks we tested showed a similar pattern among three apiary Part I of three parts. Am. Bee J. 147: 873–878. sites we had—i.e., the means were similar between Apiaries A and Booton, R. D., Y. Iwasa, J. A. R. Marshall, and D. Z. Childs. 2017. Stress- B and much higher in Apiary C. However, one of the four stocks mediated Allee effects can cause the sudden collapse of honey bee colonies. Downloaded from https://academic.oup.com/jee/advance-article-abstract/doi/10.1093/jee/toy377/5251959 by guest on 11 January 2019

(South2) had lowest weight in Apiary C and highest in Apiary B. It is J. Theor. Biol. 420: 213–219. unclear to us why this specific stock acted differently from the rest Bryden, J., R. J. Gill, R. A. Mitton, N. E. Raine, and V. A. Jansen. 2013. Chronic we tested. The significant interaction observed in the effect of apiary sublethal stress causes bee colony failure. Ecol. Lett. 16: 1463–1469. location and honey bee stock on colony size and weight is worth not- Büchler, R., C. Costa, F. Hatjina, S. Andonov, M. D. Meixner, Y. Le Conte, ing because it underpins the complex relationship between genotype A. Uzunov, S. Berg, M. Bienkowska, M. Bouga, et al. 2014. The influence of genetic origin and its interaction with environmental effects on and environment. In the future, honey bee colonies of known genetic the survival of Apis mellifera L. colonies in Europ. J. Apic. Res. 53: background can be housed in a larger number of locations within 205–214. the same climatic area to further investigate this synergistic effect of Burgett, M., and I. Burikam. 1985. Number of adult honey bees (Hymenoptera: genotype and environment on honey bee survival. Apidae) occupying a comb: a standard for estimating colony populations. Honey bees (Apis mellifera L.) are widespread and found in every J. Econ. Entomol. 78: 1154–1156. continent except Antarctica. Thus, this species thrives in both tem- Chen, L., H. Xiang, Y. Miao, L. Zhang, Z. Guo, X. Zhao, J. Lin, and T. Li. perate and tropical climates. Although there are differences in levels 2014. An overview of cold resistance in plants. J. Agron. Crop Sci. 200: of key metabolic proteins in honey bee stocks originating from dif- 237–245. ferent geographic regions (Parker et al. 2010), it remains to be deter- Cobey, S., W. S. Sheppard, and D. R. Tarpy. 2012. Status of breeding practices mined if these proteomic differences underpin variation in responses and genetic diversity in domestic US honey bees. In: D. Sammataro and J. A. Yoder (eds.), Honey Bee Colony Health: Challenges and Sustainable to climatic conditions or other location-specific stressors (Genersch Solutions. CRC, Boca Raton, FL. 39–49. et al. 2010). Given the lack of evidence for local adaptation to north- Courchamp, F., T. Clutton-Brock, and B. Grenfell. 1999. Inverse density ern versus southern climates in our study, honey bees may use similar dependence and the Allee effect. Trends Ecol. Evol. 14: 405–410. molecular, physiological, and social strategies to successfully cope Couvillon, M. J., F. C. Riddell Pearce, C. Accleton, K. A. Fensome, S. with these vastly different climatic regimes. These strategies may K. L. Quah, E. L. Taylor, and F. L. W. Ratnieks. 2015. Honey bee foraging have developed early in their evolutionary history in the tropics, as distance depends on month and forage type. Apidologie. 46: 61–70. has been observed with similar adaptations related to diapause in Danka, R. G., J. W. Harris, and G. E. Dodds. 2016. Selection of VSH-derived Drosophila species (Zonato et al. 2017). Further studies are neces- “Pol-line” honey bees and evaluation of their Varroa-resistance character- sary to comprehensively investigate the mechanistic and evolution- istics. Apidologie. 47: 483–490. ary bases of these adaptive strategies. DeGrandi-Hoffman, G., F. Ahumada, R. Danka, M. Chambers, E. W. Dejong, and G. Hidalgo. 2017. Population growth of varroa destructor (Acari: Varroidae) in colonies of Russian and unselected honey bee (Hymenoptera: Supplementary Data Apidae) stocks as related to numbers of foragers with mites. J. Econ. Entomol. 110: 1–7. Supplementary data are available at Journal of Economic Delaney, D. A., M. D. Meixner, N. M. Schiff, and W. S. Sheppard. 2009. Entomology online. Genetic Characterization of Commercial Honey Bee (Hymenoptera : Apidae) Populations in the United States by Using Mitochondrial and Microsatellite Markers. Ann. Entomol. Soc. Am. 102: 666–673. Acknowledgments Dennis, B., and W. P. Kemp. 2016. How hives collapse: allee effects, ecological resilience, and the honey bee. PLoS One 11: e0150055. We would like to thank Ryan Reynolds, Sara Ashcraft, Bernardo Niño, and Döke, M. A., M. Frazier, and C. M. Grozinger. 2015. Overwintering honey Mario Padilla for expert beekeeping assistance, and valuable guidance from bees: biology and management. Curr. Opin. Insect Sci. 10: 185–193. Margarita López-Uribe and James Strange for the microsatellite analysis. Donkersley, P., G. Rhodes, R. W. Pickup, K. C. Jones, and K. Wilson. 2014. This study was funded by generous support from an anonymous donor to the Honeybee nutrition is linked to landscape composition. Ecol. Evol. 4: Penn State Center for Pollinator Research, a Honey Bee Health Improvement 4195–4206. Grant from the North American Pollinator Protection Campaign [to M.F. and van Dooremalen, C., L. Gerritsen, B. Cornelissen, J. J. van der Steen, F. van C.M.G], and a Langstroth Graduate Fellowship in Honey Bee Health and the Langevelde, and T. Blacquière. 2012. Winter survival of individual honey Apes Valentes Graduate Fellowship, from the Penn State Center for Pollinator bees and honey bee colonies depends on level of Varroa destructor infesta- Research [to M.A.D.]. tion. PLoS One 7: e36285. Estoup, A., L. Garnery, M. Solignac, and J. M. Cornuet. 1995. Microsatellite Authors’ Contributions variation in honey bee (Apis mellifera L.) populations: hierarchical genetic structure and test of the infinite allele and stepwise mutation models. M.F. and C.M.G. planned the study; M.F., M.A.D., and C.M.M performed Genetics. 140: 679–695. experiments; M.A.D. and C.M.M performed data analysis; M.O. performed Free, J. B., and P. A. Racey. 1968. The effect of the size of honeybee colonies GIS analysis; and all authors worked in preparing and editing the manuscript on food consumption, brood rearing and the longevity of the bees during prior to submission. winter. Entomol. Exp. Appl. 11: 241–249. Free, J. B., and Y. Spencer-Booth. 1959. The longevity of worker honey bees References Cited (Apis mellifera). Proc. R. Ent. Soc. Lond. A34: 141–150. Fukuda, H., and K. Sekiguchi. 1966. Seasonal change of the honeybee worker Bahreini, R., and R. W. Currie. 2015. Influence of honey bee genotype and longevity in Sapporo, North Japan, with notes on some factors affecting wintering method on wintering performance of Varroa destructor the life-span. Jpn. J. Ecol. 16: 206–212. Copyedited by: OUP

Journal of Economic Entomology, 2018, Vol. XX, No. XX 9

Genersch, E., W. Von Der Ohe, H. Kaatz, A. Schroeder, C. Otten, R. Büchler, versus northern requeened packages. https://projects.sare.org/sare_project/ S. Berg, W. Ritter, W. Mühlen, S. Gisder, et al. 2010. The German bee fne12-756/?ar=2014 monitoring project: a long term study to understand periodically high win- Mattila, H. R., J. L. Harris, and G. W. Otis. 2001. Timing of production of ter losses of honey bee colonies. Apidologie. 41: 332–352. winter bees in honey bee (Apis mellifera) colonies. Insectes Soc. 48: 88–93. Greenleaf, S. S., N. M. Williams, R. Winfree, and C. Kremen. 2007. Bee forag- Meirmans, P. G., and P. W. Hedrick. 2011. Assessing population structure: ing ranges and their relationship to body size. Oecologia. 153: 589–596. F(ST) and related measures. Mol. Ecol. Resour. 11: 5–18. Guzmán-Novoa, E., L. Eccles, Y. Calvete, J. Mcgowan, P. G. Kelly, and Ostiguy, N., and D. Sammataro. 2000. A simplified technique for counting A. Correa-Benítez. 2010. Varroa destructor is the main culprit for the varroa sticky boards. Apidologie. 31: 707–716. death and reduced populations of overwintered honey bee (Apis mellifera) Page, R., and M. Fondrk. 1995. The effects of colony level selection on the colonies in Ontario, Canada. Apidologie. 41: 443–450.

social-organization of honey-bee (Apis mellifera L) colonies – colony level Downloaded from https://academic.oup.com/jee/advance-article-abstract/doi/10.1093/jee/toy377/5251959 by guest on 11 January 2019 Harbo, J. R. 1993. Worker-bee crowding affects brood production, honey components of pollen hoarding. Behav. Ecol. Sociobiol. 36: 135–144. production, and longevity of honey bees (Hymenoptera: Apidae). J. Econ. Parker, R., A. P. Melathopoulos, R. White, S. F. Pernal, M. M. Guarna, and L. Entomol. 86: 1672–1678. J. Foster. 2010. Ecological adaptation of diverse honey bee (Apis mellifera) Harpur, B. A., S. Minaei, C. F. Kent, and A. Zayed. 2012. Management populations. PLoS One 5: e11096. increases genetic diversity of honey bees via admixture. Mol. Ecol. 21: Phillips, E. F., and G. S. Demuth. 1914. The temperature of the honeybee clus- 4414–4421. ter in winter. Bull. U.S. Dep. Agric. 93: 1–16. Hatjina, F., C. Costa, R. Büchler, A. Uzuno, M. Drazic, J. Filipi, L. Charistos, Requier, F., J. F. Odoux, T. Tamic, N. Moreau, M. Henry, A. Decourtye, and L. Ruottinen, S. Andonov, M. D. Meixner, et al. 2014. Population dynam- V. Bretagnolle. 2015. Honey bee diet in intensive farmland habitats reveals ics of European honey bee genotypes under different environmental condi- an unexpectedly high flower richness and a major role of weeds. Ecol. tions. J. Apic. Res. 53: 233–247. Appl. 25: 881–890. Holt, H. L., and C. M. Grozinger. 2016. Approaches and challenges to Reznick, D. N., and C. K. Ghalambor. 2001. The population ecology of con- managing nosema (Microspora: Nosematidae) parasites in honey bee temporary adaptations: what empirical studies reveal about the conditions (Hymenoptera: Apidae) colonies. J. Econ. Entomol. 109: 1487–1503. that promote adaptive evolution. Genetica. 112–113: 183–198. Horn, J., M. A. Becher, P. J. Kennedy, J. L. Osborne, and V. Grimm. 2015. Rinderer, T. E., L. I. De Guzman, J. Harris, V. Kuznetsov, G. T. Delatte, J. Multiple stressors: using the honeybee model BEEHAVE to explore how A. Stelzer, and L. Beaman. 2000. The release of ARS Russian honey bees. spatial and temporal forage stress affects colony resilience. OIKOS. 0: Am. Bee J. 140: 305–307. 001–016. Sammataro, D., M. Frazier, and N. Ostiguy. 2002. How to use a PSU/IPM Jeffree, E. P., and D. Allen. 1956. The influence of colony size and of nosema Varroa board. Am. Bee J. 142: 363–366. disease on the rate of population loss in honey bee colonies in winter. J. Schiff, N. M., and W. S. Sheppard. 1995. Genetic analysis of commercial honey Econ. Entomol. 49: 831–834. bees (Hymenoptera: Apidae) from the Southeastern United States. J. Econ. Jombart, T., S. Devillard, and F. Balloux. 2010. Discriminant analysis of prin- Entomol. 88: 1216–1220. cipal components: a new method for the analysis of genetically structured Schiff, N. M., and W. S. Sheppard. 1996. Genetic differentiation in the queen populations. BMC Genet. 11: 94. breeding popualtion of the Western United States. Apidologie. 27: 77–86. Joshi, N. K., M. Otieno, E. G. Rajotte, and S. J. Fleischer. 2016. Proximity Smart, Matthew. 2015. The influence of mid-continent agricultural land use to woodland and landscape structure drives pollinator visitation in apple on the health and survival of commercially managed honey bee (Apis mel- orchard ecosystem. Front. Ecol. Evol. 4: 38. lifera L.) colonies. Retrieved from the University of Minnesota Digital Kamvar, Z. N., J. F. Tabima, and N. J. Grünwald. 2014. Poppr: an R package Conservancy, http://hdl.handle.net/11299/171752. for genetic analysis of populations with clonal, partially clonal, and/or Sponsler, D. B., and R. M. Johnson. 2015. Honey bee success predicted by sexual reproduction. PeerJ 2: e281. landscape composition in Ohio, USA. PeerJ 3: e838. Kraus, B., and R. E. Page. 1995. Population growth of Varroa jacobsoni Oud Stephens, P. A., and W. J. Sutherland. 1999. Consequences of the Allee in mediterranean climates of California. Apidologie. 26: 149–157. effect for behaviour, ecology and conservation. Trends Ecol. Evol. 14: Kulhanek, K., N. Steinhauer, K. Rennich, D. M. Caron, R. R. Sagili, J. 401–405. S. Pettis, J. D. Ellis, M. E. Wilson, J. T. Wilkes, D. R. Tarpy, R. Rose, Tarpy, D. R., and J. Lee. 2005. Comparison of Russian and Italian honey bees. K. Lee, J. Rangel, and D. vanEngelsdorp. 2017. A national survey of man- North Carolina Cooperative Extension Service. http://digital.ncdcr.gov/ aged honey bee 2015–2016 annual colony losses in the USA, Journal of cdm/ref/collection/p249901coll22/id/2181 Apicultural Research, 56: 328–340. Vaudo, A. D., J. F. Tooker, C. M. Grozinger, and H. M. Patch. 2015. Bee nutri- López-Uribe, M. M., R. H. Appler, E. Youngsteadt, R. R. Dunn, S. D. Frank, tion and floral resource restoration. Curr. Opin. Insect Sci. 10: 133–141. and D. R. Tarpy. 2017. Higher immunocompetence is associated with van der Werf, J., H. Graser, R. Frankham, and C. Gondro. 2009. Adaptation higher genetic diversity in feral honey bee colonies (Apis mellifera). and fitness in animal populations. Springer Ebooks, USA. Conserv. Genet. 18: 659–666. Winston, M. L. 1987. The biology of the honey bee. Harvard University Press, Luque, G. M., T. Giraud, and F. Courchamp. 2013. Allee effects in ants. J. Cambridge, MA. Anim. Ecol. 82: 956–965. Winter, D. J. 2012. MMOD: an R library for the calculation of population MacGregor-Forbes, E. 2014. Sustainable Agriculture Research and Education differentiation statistics. Mol. Ecol. Resour. 12: 1158–1160. (SARE) Outreach, USDA - National Institute of Food and Agriculture Zonato, V., L. Collins, M. Pegoraro, E. Tauber, and C. P. Kyriacou. 2017. (NIFA) Annual Project (FNE12-756) Report: A comparison of strength Is diapause an ancient adaptation in Drosophila? J. Insect Physiol. 98: and survivability of honeybee colonies started with conventional 267–274.