Breeding for hygienic behavior in honey bees (Apis mellifera): a strong paternal effect

Abstract The haplodiploid sex determination of honey bees and the multiple mating of queens pose challenges in determining the genetic contribution of drones (male honey bees). This is especially important for breeding programs as, for example, when attempting to reinforce traits governing social immunity against pests and diseases. Here, we focused on breeding aiming at enhancing hygienic behavior, a trait that is known to reduce parasite load in honey bee colonies. To evaluate the contribution of drones versus queens to this trait, we conducted a two-step bidirectional selection program. First, we selected colonies with consistent phenotypes for low- or high-hygienic behavior (generation P). From those, we generated two types of daughter colonies (F1). One type originated from queens that had been artificially inseminated with selected drones that originated from queens of either low- or high-hygienic phenotype. The other type of colonies was set from naturally mated queens. We then compared the hygienic performance of the progeny colonies. In the next step, we used the F1 colonies (from either artificially inseminated or naturally mated queens) to produce naturally mated queens, which subsequently generated F2 colonies. These were then examined for the level of hygienic behavior. The results demonstrate the significant contribution of both parents to the phenotype of offspring. In particular, drones had a consistent and significant influence on the hygienic performance of the progeny throughout generations. These findings emphasize the great potential to propagate the hygienic trait in local populations by selecting lines for drones that carry the high-hygienic trait.


Introduction
The western honey bee, Apis mellifera L., the main pollinator of agricultural crops worldwide (Aizen & Harder, 2009;Gallai et al., 2009), has suffered considerable colony losses in recent years (Gray et al., 2019;2020). One of the main threats is believed to be infestation by the invasive ectoparasitic mite Varroa destructor that not only feeds on developing brood and adult bees but also serves as an efficient vector for several highly pathogenic viruses (Kurze et al., 2016;Traynor et al., 2020). Years of chemical treatments against Varroa, led to the development of resistance to most acaricides, which in turn led to the understanding that sustainable long-term honey bee management should involve generating stocks that are resistant to this parasite (Sammataro & Avitabile, 2011;Dietemann et al., 2012;Spivak & Danka, 2021).
Natural selection is a well-known evolutionary mechanism that stabilizes both parasite and host populations. The main mechanism depends upon selection against aggressive individuals from the parasite population and sensitive individuals from the host population. This apparent coevolution of Varroa destructor with the original host, the eastern honey bee (Apis cerana), led to many physiological and behavioral adaptations in the host bees (Rath, 1999;Wang et al., 2020), such as intensive grooming, entombing of infected drone brood (Peng et al., 1987;B€ uchler et al., 1992), hygienic behavior (B€ uchler et al., 2010) and rapid death of immature infected workers (Page et al., 2016). All of these mitigate the parasites' harmful effects and, overall in this system, rendering Varroa a less damaging parasite (Boot et al., 1999;Lin et al., 2018). Apis mellifera, being a close relative of A. cerana, is predicted to evolve, through natural selection, populations that can survive Varroa, and are further able to replace more susceptible colonies, unless treated against the pest (Blacqui ere et al., 2019;Rangel et al., 2020). Selection programs aimed at producing Varroa resistant honey bees in agricultural systems should not only prevent heavy colony losses in the process of selection but also maintain profitability in honey production and provide efficient pollination services.
Hygienic behavior has been comprehensively studied among the traits with the potential to promote honey bee resistance to Varroa. This behavior is defined as the rate of cell uncapping (hereafter termed uncapping) and removal of damaged brood (hereafter termed cleaning) (Lapidge et al., 2002;Oxley et al., 2010b;Behrens et al., 2011;Tsuruda et al., 2012). Rothenbuhler (1964) was the first to suggest that the inheritance of hygienic behavior is a complex genetic trait involving two loci: the first controls for the detection and uncapping of damaged brood cells and the second controls for the removal of damaged brood from the colony, traits that are not necessarily in linkage. Notwithstanding, more recent studies have shown that hygienic behavior is a quantitative trait, controlled by several loci (Lapidge et al., 2002;Oxley et al., 2010b;Behrens et al., 2011;Tsuruda et al., 2012;Boutin et al., 2015;Mondet et al., 2015). The accumulated evidence for strong genetic components of this behavior and its benefits in reducing the parasite load have driven efforts to selectively breed for this trait (Spivak & Danka, 2021).
Breeding of honey bees is complicated due to their haplodiploid sex determination, and the queen's extreme polyandry (mating with up to 20 males) that results in a relatively small genetic contribution of each paternal line to the colony (Estoup et al., 1994). A particular difficulty in honey bee breeding is their social organization, because the phenotype is not examined at the individual level, but rather by the average colony performance (Oxley & Oldroyd, 2010). Generally, honey bee breeding is based on naturally mated queens, which may mask the relative contribution of each parent. Indeed, studies on breeding for hygienic behavior that were based on naturally mated queens have obtained varied results, not only among breeding programs but also for different years within a given program (Pernal et al., 2012). In the case of the Russian honey bees (A. mellifera) from the Primorsky territory (Unger & Guzm an-Novoa, 2010), for example, it has been suggested that hygienic behavior is a predominantly maternal trait. On the other hand, Bigio et al. (2014) achieved the best hygienic scores through artificial insemination using selected stocks of both sexes. In addition, some authors have reported that environmental factors may also play an important role in the expression of hygienic behavior (Rosenkranz et al., 2010;Meixner et al., 2015;Nganso et al., 2017).
Queens pass on only half of their genes to diploid daughter queens, which subsequently mate with multiple drones. Consequently, the selected queen's direct genetic effect on the colony workers is only 25% (Collins, 1986;Brascamp & Bijma, 2014). On the other hand, honey bee drones are haploid and thus carry only maternal genes, emphasizing the fast and efficient manner by which they can magnify potential maternal traits (Laidlaw & Page, 1986). Thus, the maximum direct genetic effects of a selected queen can be achieved when she is the exclusive producer of drones that will be mated with a given (unrelated) queen. This would result in a daughter colony composed of workers with 50% of their genetic material from the queen of the sire colony.
Recently, drones' contribution to honey bee breeding has been addressed by Plate et al. (2019) who published a model simulating the power of selection in a drone controlled set-up. They have clearly shown that the selection based solely on queens in a large nonselected population is an insufficient way of breeding. Despite these theoretical considerations, only scant attention has been paid to date to the contribution of drones in breeding programs, mainly due to technical difficulties in controlling queen mating under natural conditions (Oxley et al., 2010a). Honey bee mating aggregation areas attract drones and queens from genetically variable populations over distances of up to 15 km (Jensen et al., 2006), which introduces great genetic variation. Therefore, selection for drones usually relies on one of three methods, all of which are technically difficult and time-consuming: (i) locating isolated mating stations such as on islands (Jensen et al., 2006), (ii) changing the time of mating of selected individuals, (Oxley et al., 2010a); and (iii) artificial insemination (Bigio et al., 2014).
In this study, we aimed to assess the efficacy of selection for the hygienic trait through the sexual offspring of selected queens. We compared the relative genetic contribution of drones versus queens to the hygienic behavior of daughter colonies. We further examined whether the trait is conserved for second filial generations, even when the daughter queens are naturally mated. To quantify the contribution of the drones to the trait in daughter and granddaughter colonies, we followed a two-step bidirectional selection-breeding scheme for low-and high-hygienic behavior.

Materials and methods
Honey bees: The research was conducted in the experimental apiary at the Volcani Center, Agricultural Research Organization (ARO), Israel. The source population of honey bees, basically Apis mellifera ligustica, was subjected to a bidirectional selection program for low and high hygienic behavior since 2012. Throughout the selection program, we collected data on colonies' hygienic behavior considering the parental phenotype (low or high hygienic behavior) and mating type (natural mating or artificial insemination). In 2015, we selected four colonies, thereafter termed generation P (see the breeding scheme in Figure 1). All the colonies were located in the same apiary, spread within a 500 meters radius.
Hygienic behavior was measured using a slightly modified "pin test" on 100 cells containing pink eye pupae as described in Seltzer et al. (2021). In brief, this test entails marking about 100 cells and piercing them with a #2 size entomological pin. The treated portion of the comb was photographed immediately after pinning and 24 hours after. The ratio of cells uncapped and completely cleaned was calculated by comparing the respective images.

The breeding scheme
For the first step of our experiment, during the spring and summer of 2015, we screened 122 colonies for consistent hygienic behavior. Consistently low-(L) and high-hygienic (H) colonies were defined, respectively, as those that had uncapped less than 0.40 (L) or more than 0.75 (H) of all pinned brood cells after 24 hours in three independent tests. Based on this screening, we selected four unrelated source colonies-two low-hygienic (L1; L2) and two high-hygienic (H1; H2). These were referred to as the parent generation (P) (Figure 1). Low hygienic colonies had an average cell uncapping and cleaning performance of 0.24 ± 0.04/0.48 ± 0.08 (L1) and 0.25 ± 0.2/0.42 ± 0.1 (L2), whereas high hygienic colonies had an average uncapping and cleaning performance of 0.81 ± 0.09/0.93 ± 0.009 (H1) and 0.85 ± 0.1/0.98 ± 00.03 (H2). Their daughter queens and their generated colonies are referred to as F 1 .
Naturally mated F1 queens: In October 2015, daughter queens F1 were introduced as queen cells to the colonies (n ¼ 58, Table 1) which later emerged and naturally mated with the unselected drone population in the apiary. Their progeny colonies were tested for hygienic behavior three times in 2016 starting in March at monthly intervals. The congregation area of drones mating with our queens is unknown and it is impossible to determine drones' origin and genetic profile. Nonetheless, a comparison of the hygienic performance between the progeny of naturally mated and artificially inseminated queens from low-and high-hygienic lines may allow us to characterize the contribution of the local drone population and examine its effects on selection strength.
Artificially inseminated F1 queens: Artificial inseminations (n ¼ 67) were performed between sexuals originating from selected, unrelated P colonies in April 2016. Daughter queens were inseminated with 8 ml of semen from drones from the same colony. The crosses and the number of replicates (colonies) for each cross are presented in Table 1. Eight weeks after egg-laying by the inseminated queen, Figure 1. Study design, including the breeding scheme and the assessment of hygienic behavior (phenotype) by pin test. (A) Generation P (white rectangle): In the first generation, four colonies were selected based on genetic origin and hygienic phenotype (either Low or High scores). (B) Generation F 1 : Queens that are descendants of generation P colonies were either artificially inseminated with drones from colonies of generation P to create 12 possible crosses (orange rectangle) or naturally mated with a non-selected drone population (green rectangle). Based on breeding technique, genetic origin, and phenotype, five colonies were selected as a source of queens and drones to establish the F 2 queen generation. (C) Generation F 2 : queens were again derived from these F1 colonies from both groups and naturally mated (green rectangles). The hygienic phenotype is shown in the F2 progeny colonies (blue rectangles).
the progeny colonies were tested for hygienic behavior at weekly intervals for three weeks during May-June.
In the second step, we selected queens from five colonies from generation F 1 based on their pedigree and their hygienic phenotype. Each of the selected five colonies was used to produce multiple daughter queens. Two queens were from the artificially inseminated pedigree (one low hygienic (L x L) and one high hygienic (H x H)) and three queens were from the naturally mated pedigrees (one L and two H). From these colonies, we reared queens (F 2 ), which were all subsequently naturally mated and then tested for hygienic behavior (Figure 1). We referred to naturally mated queens of the artificially inseminated (AI) pedigree as AI-Low (n ¼ 12) and AI-High (n ¼ 14), and colonies from the naturally mated (NM) pedigree as NM-Low (n ¼ 26) and NM-High (n ¼ 18). In total, from the F 2 generation we tested 70 colonies. All colonies were placed in the same apiary and tested simultaneously. Each colony was tested three times throughout the year for hygienic behavior as described above.
Statistical analysis: A mixed-model ANOVA was used to test hygienic behavior among different groups for two generations, defined by the proportion of uncapped and cleaned cells (proportion values were subjected to angular transformation prior to analysis of variance). Parental phenotypes (L or H) and their interaction served as fixed effects in the model. The colony was nested within the line and the line was nested within parental phenotypes. Post-hoc comparisons between parental phenotypes were performed using Tukey's HSD, p < 0.05.
Variable importance was further calculated to determine the relative contribution of the maternal and paternal phenotype of the same genetic background to hygienic behavior. The method estimates the variability in the predicted response based on a range of variations for each factor. If variation in the factor causes high variability in the response, then that effect is important relative to the model. These values reflect the amount of variability due to each factor alone and in combination with the other factors in the dependent variable (Saltelli, 2002).
All statistical tests were carried out using the JMP 14 Statistical Program (SAS, USA).
No difference was found in the hygienic performance of progeny colonies of naturally mated queens (uncapping, F (1,73) ¼ 25, p ¼ 0.89 and cleaning, F (1,73) ¼ 25, p ¼ 0.54, Figures 2A and 2C). Progeny colonies of naturally mated queens from an H Table 1. Summary of hygienic performance of colonies reared from the crosses between four selected colonies by artificial insemination and natural mating. The values are mean ± SE of ratio of uncapped cells (above) and cleaning (below). Numbers in parentheses represent the number of tested colonies. phenotypic source had an average (± SE) performance of 0.75 ± 0.03 uncapping and 0.52 ± 0.03 cleaning. Progeny colonies of queens from the L phenotypic source had an average (± SE) of 0.74 ± 0.03 uncapping and of 0.49 ± 0.04 cleaning.
To summarize the respective contributions of queens and drones to F1, we used a variable importance analysis of the hygienic behavior trait. The results show that for uncapping, the variable importance of drones was 0.74 while that of queens was only 0.27. For cleaning, the variable importance values were 0.65 vs 0.33 for drones and queens, respectively. This means that the drones' contribution was 2.7-fold and 1.96-fold greater than that of the queens for uncapping and cleaning behavior, respectively (Table S1, Supplementary material).
In the second step, we explored the persistence of the trait in the F 2 generation by comparing colonies generated by F 2 naturally mated queens (L or H), originating from pedigrees of either artificially inseminated or naturally mated queens. There was a significant effect of pedigree type (naturally mated vs. artificially inseminated) on the hygienic behavior of the resulting colonies (Figures 3A and 3B;uncapping: ANOVA,F (3,67.8 ¼9.5, p < 0.0001, cleaning: F (3,67.8) ¼5.8, p ¼ 0.0012). Tukey's HSD posthoc analysis revealed that the only significantly different progeny group was the artificially inseminated-Low hygienic colonies that performed hygienic behavior at the lowest rate of all other tested colonies (uncapping 0.49 ± 0.04; cleaning 0.29 ± 0.005). The highest proportion of hygienic behavior was performed by the progeny of artificially inseminated-High hygienic colonies (uncapping 0.8 ± 0.06; cleaning 0.61 ± 0.06). Artificially inseminated-High hygienic progeny did not differ significantly from naturally mated-High hygienic colonies (uncapping: 0.81 ± 0.03; cleaning 0.61 ± 0.06). Interestingly, progenies of both artificially inseminated-High hygienic-and naturally mated-High hygienic-did not differ from naturally mated-Low hygienic colonies (uncapping: 0.76 ± 0.06; cleaning 0.56 ± 0.06) (Figure 3).

Discussion
Hygienic behavior constitutes part of the natural social resistance mechanisms of honey bees to various pests and pathogens. Employment of bidirectional selective breeding enabled us to isolate and compare paternal and maternal contributions to hygienic behavior. Although hygienic behavior is known to be a heritable trait, selective breeding for this trait in honey bees remains challenging as heritability values vary (Milne, 1985;Harbo & Harris, 1999;Maucourt et al., 2020). Most honey bee breeding programs for improved hygienic behavior have been based on naturally mated queens (Kulincevic et al., 1992;Pernal et al., 2012;de Mattos et al., 2016;Kefuss et al., 2015) but some have used artificial Figure 2. Hygienic behavior in colonies F2 derived from parents of different phenotypes and mating types; the data are average proportions of uncapping (A and B) and cleaning (C and D) of pinned cells ± SE. The parental phenotypic source is noted on the x axis. In the case of artificially inseminated queens (B and D) a symbol of maternal phenotype precedes the paternal phenotype (queen X drone). In the case of naturally mated queens (A and C) only maternal phenotype is noted. Different letters above the bars indicate groups that are significantly different according to Tukey's HSD (p < 0.05).
insemination as Harbo and Harris (2001) or mating stations to ensure controlled mating as in the case of A.m. carnica breeding efforts in Europe over the years (Hoppe et al., 2020). Several breeding programs have reported low heritability values (e.g., Maucourt et al., 2020). However, heritability can be improved by increasing the genetic component in the selected trait. A few authors have noted the potential of using paternal lines and drone selection (Jandricic & Otis, 2003;Oxley et al., 2010aOxley et al., , 2010b, but the impact of drones on selection programs for hygienic behavior has been studied in only a handful of cases (P erez- Sato et al., 2009;Bigio et al., 2014).
Our findings have shown that progeny of low or high naturally mated queens did not differ in their hygienic behavior, and showed, on average, 0.74 uncapping and 0.68 cleaning. Moreover, the variable importance analysis revealed that the drones' contribution was two to three times greater than that of the queens. These results are consistent with the fact that drones are haploid and thus carry only maternal genes and thus not affected by multiple random mating. The strong effect of drones could explain why breeding programs that compared progeny of high-hygienic queens to those of the general population have demonstrated a weak effect of selection (Boecking et al., 2000;Pernal et al., 2012). Plate et al. (2019) published a model for simulating the efficacy of breeding based on naturally mated queens in a non-selected drone population. The model clearly demonstrates that in a large, unselected population there is little chance of successful breeding.
Honey bee mating involves the aggregation of several thousand sexually mature drones from different colonies within about a 5-km radius at specific congregation areas located up to 25 meters above the ground (Baudry et al., 1998;Jaff e & Moritz, 2010). Another important factor that contributes to the colony's hygienic phenotype is its genotypic composition (Arathi & Spivak, 2001) via affecting efficacy, persistency as well as the division of tasks among the hygienic workers. In naturally mated queens, we may often expect semen from more than 15 drones that are responsible for the corresponding number of workers' subfamilies. How many hygienic patrilines are sufficient to keep a high level of hygiene in the colony is not yet clear, though Arathi and Spivak (2001) data suggest that over 50% of the drones that a queen mates with should be from hygienic colonies to pass along the trait to the colony. . Hygienic behavior of F2 colonies derived from artificially inseminated and naturally mated queens. Hygienic behavior is expressed by the ratio of uncapped (A) and cleaned (B) cells to all pinned cells. Progeny of F2 colonies were divided according to their pedigrees that include the maternal mating type data: artificially inseminated (AI) or naturally mated (NM); maternal phenotype H queens (white bars), and L queens (black bars). Data are the mean ratio ± SE of cells' uncapping. Different letters above the bars indicate groups that are significantly different by ANOVA followed by Tukey's HSD (p < 0.05).
As the progeny of naturally mated queens from low and high hygienic lines showed similar and high levels of both uncapping and cleaning behaviors, it can be assumed that most males in the local population of the current study carry a high hygienic genotype. The occurrence of "high hygienic males" in the population is intriguing because to the best of our knowledge there were no selection efforts in the neighboring apiaries towards high hygienic performance, and even our selection was bidirectional, presumably also releasing "low hygienic males" to the local drone population. This phenomenon could be explained by the superiority of drones from the high hygienic genotypes at the local drone congregation area. In fact, Yañez et al. (2011) suggested that drone congregation at mating areas functions as a mechanism which ensures that queens mate only with those having better flight ability and higher responsiveness to the queen's cues. The direct impact of hygienic behavior on drone quality and responsiveness to queens remains to be studied. However, lower Varroa infestation in hygienic colonies has been shown in several studies (Robertson et al., 2014;Spivak & Danka, 2021;Hoppe et al., 2020;Seltzer et al., 2021), and this is expected to result in higher drone quality since it is well known that Varroa infestation reduces drone total protein content, flight ability, and sperm count (Glinski & Jarosz, 1984;Duay et al., 2002;Bubalo et al., 2005).
If indeed breeding for hygienic behavior can result in superior drones remains to be tested. Nevertheless, as previously mentioned, hygienic behavior is a valuable trait in apiculture, with apparently low or no costs to productivity (Seltzer et al., 2021), or on individual immunity (Erez et al., 2022). However, facilitating its spread in the local population remains a challenge due to both high queen polyandry (Tarpy et al., 2004) and complex mating behavior that acts against population homogeneity. According to Koeniger et al. (2005), drones preferentially visit the nearest drone congregation area (DCA) within a radius of 3.75 km, while virgin queens are prone to more distant flights (Utaipanon et al., 2019). Under these constraints, in the absence of isolated breeding areas and very high colony population density, the task seems entirely dependent on artificial insemination between selected sexuals. This is supported by our findings which show that the phenotype of the progeny from the artificially inseminated hygienic pedigree is consistent with their parental origin. Most importantly, the significant benefits of artificial insemination in F1 sustain an additional generation of queens. This is particularly apparent when comparing the hygienic phenotype of AI-low and NM-low in the F2 generation. AI-low remains low hygienic despite the fact that the queens were naturally mated and most of our nonselected drone population apparently came from colonies with high hygienic genotype, thus supporting a strong queen effect. The persistence of high hygienic traits in AI-high was also maintained but could not be visible due to similarities with naturally mated groups from both lines, thus supporting the additional impact of the local drones. However, the genotypic composition of any local drone population at DCA can vary since it depends on selection pressures on drones as discussed above. The significance of drone selection on breeding for hygienic behavior requires further study.
In conclusion, this work provides evidence for the importance of both parents, and particularly of drones in the process of selection for hygienic behavior. Moreover, this study suggests that naturally mated queens from an artificially inseminated hygienic pedigree can overcome the genetic influence of non-selected drones. But how long the trait persists based on natural drone selection without further reinforcement by artificial insemination requires further study. But even if only effective over two generations, this character may be implemented into breeding programs for commercial purposes in which virgin queens from the "elite" breeding population created by artificial breeding can be provided to the beekeeper for mating with males from any given population while preserving the "elite" parental phenotype.

Funding
We wish to thank the BARD IS5078-18 to VS for financing the final stages of this research and Coloss association for stimulating meetings that led to this research.