The global losses of honeybees are of major concern not just because of their economically important role as managed honey producers and pollinators of crop, but also because of their essential ecosystem service maintain floral biodiversity. The most important driver of global honeybee colony losses remains to be Varroa destructor, a parasitic mite of the brood and adult bees (Genersch et al., 2010). If an infested colony is not systematically treated with acaricides it dies within three years (Rosenkranz et al., 2010). V. destructor changed host from A. cerana to A. mellifera in Asia (Korea and Japan) where both honey bee species have been kept in contact for many years. Subsequently A. mellifera colonies and queen bees have been transferred back to Europe (Rinderer et al., 2001). Today, only Australia and some isolated oasis and islands are the only V. destructor free areas in the world (Roberts et al. 2015).In spite of the global losses of A. mellifera colonies due to Varroa infestation, various breeding strains have been identified, which may be resistant to mite infestations (Anderson & Trueman, 2000, Le Conte et al 2007, Kefuss et al 2016). For example tolerance against V. destructor in A. mellifera has been shown to be based on behavioural responses relating to nest hygienic behaviour (Harbo et al., 2009). The so called Varroa Sensitive Hygiene (VSH) lineage has been bred where workers remove parasitized pupae from the cells (Le Conte et al., 2011). Although this is a remarkable breeding success, the actual genetic mechanisms controlling this trait are difficult to address with an experimental approach. Workers with different genotypes need to interact with each other in the complex environment of the colony, which drives any genetic analyses even in times of full genome analyses to intractable complexity. In order to dissect the molecular mechanisms that drive the co-evolutionary resistance, a system where an individual parasite can interact with an individual host will already be a challenge. We here suggest to follow a different approach. There are several cases where the prevention of mite reproduction by the host provides true resistance of the honeybee because the mites do not activate their ovaries and remain sterile (Le Conte et al., 2007, Locke & Fries, 2011, Kefuss et al 2016). Clearly, understanding the underlying physiological processes that interfere with the mite-host larva crosstalk will be fundamental to comprehend this important step in host resistance. The female mite has only a very narrow time window (<12 h) for starting reproduction after the enclosure in the brood cell (Rosenkranz et al. 2010). Within five hours after the capping of the brood cell, the female mite consumes the first haemolymph meal and only few hours later oogenesis begins (Fig.1). If the female mite fails to initiate ovary activation at this stage it will remain infertile and have no offspring at all. After three days of cell capping the first (and only) haploid male2egg is laid followed by diploid female eggs in thirty hour intervals. The male mite mates with the female offspring and dies when the adult bee emerges. The females feed on adult bees or infest new brood cells. The ovary activation after infesting a brood cell depends on stimuli released by the bee larva. Hence a lack of such stimuli might be one way to prevent mite reproduction altogether (Frey et al 2013). In addition, we will study the simplest genetic system by focusing on the interaction between the haploid drone larvae with the reproduction of V. destructor in the capped brood cell. This test system is also important at the colony level, because V. destructor preferentially reproduces in the drone brood (Rosenkranz et al. 2010).(2) the elements of difficulty of the issueThe rich resources and broad knowledge on both honey bee and Varroa allows for comprehensively tackling its genetic underpinnings, which is greatly facilitated by the availability of both genomes. Hence, not only evolutionary models can be readily tested, also it is entirely possible to identify specific gene cascades involved using RNAseq and identify key genes with QTL analyses using high density SNP mapping (Stolle and Moritz 2013). The study becomes particularly appealing because we can compare the resistance processes of the novel host A. mellifera with those in the Asian honeybee A. cerana, the original host of V. destructor. For example Ji et al. (2014) already studied A. cerana workers' RNA expression profiles of entire colonies in relation of behavioural resistance to V. destructor. Unfortunately, the detected differences in expression patterns are of little value as they comprise individuals of various ages and physiological states not allowing for a functional analyses.(3) the limitations of current approachesDespite the importance of Varroa for apiculture not too much is known about genetic regulatory mechanisms of resistance in A. mellifera. The traits associated with behavioural resistance are usually complex colony level traits. They are not only complex because the behaviour as such is composed of different co-ordinated subtasks but also because the extreme polyandry of the queen. The colony is composed of ten to twenty subfamilies rendering the hive a "genotypic mess" (Schlüns et al. 2005).Fig. 1. Reproductive cycle of V. destructor. The critical time of mite ovary activation is on day 9 (bold arrow) just after capping of the brood cell.3Any colony level selection is therefore an exceedingly difficult task especially for behavioural traits that are influenced by temporal polyethism of the workers. Also, the genetic analyses of behavioural traits that are expressed as modular traits are difficult at best. For example hygienic behaviour consists of a sequence of different subcomponents (detection, opening and removing), which may be inherited independently and performed by different workers in the colony (Moritz & Fuchs 1998, Arechavaleta-Velasco et al. 2012). So it may not be surprising that in spite of great efforts to breed V. destructor resistant honey bees through controlled selective breeding programs on hygienic behaviour (e.g. in Germany "Varroaresistenz-Zucht"), it have been the simple, natural selection based programs that managed to produce resistant honey bee populations. Resistance to V. destructor has independently emerged as a consequence of natural selection on the island of Gotland (Sweden), in Avignon, and in Toulouse (France) (Le Conte et al., 2007, Locke & Fries, 2011, Kefuss et al 2016). In all cases the resistance to the parasite was not based on hygienic behaviour but on the inhibition of the mite's reproduction in the brood cell where it fails to produce offspring. Behrens et al. (2011) used a mapping population of drones to screen for genes that control the lack of mite reproduction in the resistant Gotland population. They found three candidate regions with moderate epistatic effects on chromosomes 4, 7, and 9. The explanatory power of the analysis however suffered from a rather low marker density (216 microsatellite markers). Considering the extremely high recombination frequency of up to 197 cM/Mb in A. mellifera (Liu et al. 2015) made the study prone to false positive markers. Lattorff et al. (2015) screened the candidate areas containing the QTL for inhibition of V. destructor reproduction at the population level. They found signatures of positive selection exactly within the significant QTL previously identified by Behrens et al. (2011). Although the selective sweep independently confirmed the region identified before, it could not conclusively identify specific loci contributing to the trait as these might also have resulted from stochastic population processes. Given the power of next generation genome mapping (e.g. Stolle and Moritz 2013) it is overdue to produce a high density full genome SNP map to provide a more comprehensive and probably a very different view of the resistance QTL. Indeed the high recombination rate is ideal for mapping with high density markers particularly if we can take advantage of the haploid genome of the drones for mining QTLs.D2. ObjectivesThe overall objective of this proposal is to2.1 unravel the genetic mechanisms underpinning the honey bee's resistance to V. destructor with(a) high density mapping and (b) transcriptomics of host and parasite2.2 implement a feasible and simple breeding program for Varroa resistance at the apiary level.4We will focus our search on the prevention of reproduction of the mite in the brood, a mechanism where resistant phenotype can be unambiguously identified at the individual level. We will use haploid drones as this allows for a unique test setting and powerful analyses of genetic mechanisms in the host. We will identify how specific individual host genomes negotiate with that of the parasite. We will test this in a comparative way using resistant and susceptible hosts. This will include already available resistant and susceptible A. mellifera breeds in Sweden, but also a population in Topa Mica (Romania) where colonies have survived for decades without any acaricide treatment. This population will also serve to establish a model regional/local breeding scheme to allow for breeding of resistant stock at the apiary level.Similar to (Behrens et al., 2011) we will produce a mapping population using the Swedish resistant strain, but now with high density SNP marker mapping covering the genome with ~20000 SNPs (e.g. Stolle and Moritz 2013) to identify important genomic regions that have most likely been missed in the old microsatellite study. In addition, we will construct a mapping population by producing hybrid queens from the Romanian resistant population, to test if similar genomic regions have been under selection and contribute to resistance as in the Gotland population. We will carefully follow the transcriptomic crosstalk between parasite and host during the critical phase of mite ovary activation comparing resistant and susceptible larvae (both worker and haploid drone larvae).The mapping studies will particularly profit from using haploid drones as host systems. Not only are they the most important individuals for mite reproduction in the colony, they also have the most simple genome that allows for straightforward interpretations. Complex dominance interactions between different alleles within a locus are lacking and even epistatic interactions among loci become easily tractable when there is only one set of chromosomes instead of two. In haploid genomes, even low sequence coverage is sufficient to provide reliable data on the phase of the loci as there is no need to reach 50:50 distribution to reliably detect heterozygous SNP loci in diploids. The recombination rate in honey bees is the highest recorded in plants and animals exceeding 197 cM/Mb at recombination hotspots (Liu et al., 2015). This extreme recombination rate is, however, a tremendous advantage in QTL mapping studies as fewer individuals will provide sufficient recombination events to identify loci linked with a trait. Since we are using ~20000 markers the average distance between two SNP markers is ~11.9 kb which is sufficient to saturate markers density so that the recombination map does not increase in size with additional markers. Indeed the combination of haploid drones with the high recombination genome renders A. mellifera an ideal model system for any QTL study if the traits are expressed in drones.5D3. Impact(1) the potential to significantly influence the scientific fieldOur experiments will comprehensively tackle the genetic basis of a single trait; the drone larvae's resistance to the V. destructor mite by inhibiting the mite's reproduction. The combination of high density SNP mapping with a tight sequence of transcriptome data showing changes in the cross talk between host and parasite when ovary activation takes place will reveal patterns of extremely swift host parasite co-evolution and is of profound general interest in evolutionary biology. This is only approach is only feasible within the life time of the 30 month project because full genome information is available and we use the phenotype of haploid drones as test organisms. This approach allows for dissecting additive gene effects but also epistatic effects with high precision.(2) The potential impact of the project in the scientific, social, economic or cultural environment and/or the applicative directions to be explored within the projectThe prevention of mite reproduction is a key resistance trait indeed, as it will not just serve as a model system to understand host-parasite interactions at the molecular level but also have profound implications for applied apiculture. It will open the way for honeybee breeders across the globe to use natural selection and abandon any treatment of the colony. However, the implementation into beekeeping requires a fundamental change in the current policy of treating against Varroa. Breeding for resistance must replace treatment. Wild honeybee populations in African honeybees are all resistant and nobody treats colonies south of the Sahara. It may well be that short term benefit of Varroa treatment is the very cause of the current dramatic global situation. Keeping the honey bee host highly susceptible at high density, creates ideal conditions to allow for both high transmission and high virulence in the mites. There is no longer a trade-off between virulence and transmission predicted under natural selection and resistance of the mites against acaricides enhances the problem on the long run. This project will therefore not only contribute to understanding the rapid coevolution between host and parasite, it may also open a sustainable alternative for beekeeping without acaricide treatment. The model breeding scheme will serve as a template not just for beekeepers in Romania but also around the globe to transform their operations from intensive acaricide treatment to breeding for and achieving resistant stock.6D4. Methodology4.1 SNP based high density QTL mapping of resistanceWe will take advantage of existing susceptible (S) and resistant (R) honeybee strains, which already have been identified in Sweden and Romania. We will produce hybrid queens (S/R) that produce hundreds of drone offspring with resistant and susceptible phenotypes providing large mapping populations. This focus on drones is not just practical for the genetic analyses. V. destructor has a tenfold preference to infest drone brood. As the drones' development lasts 24 days in contrast to 21 days of a worker bee, the mites produce much more offspring per reproductive cycle. Therefore, any traits of resistance to V. destructor that are expressed in the drone brood will have a strong influence on V. destructor epidemics. If selection operates on haploid drones, resistance gene(s) will spread much faster in the population (Kidner & Moritz 2015) than any behavioural trait of workers. The latter have no offspring and selection can only operate very indirectly at the colony level.The true power of this methodology lies in the resources available:1) full genomes for both A. mellifera and V. destructor (Weinstock et al 2006, Cornman et al. 2010) and SNP markers can easily be assigned to map positions of genes.2) The use of a strictly binary phenotype of the drones. As cells will be individually inspected, we can discriminate resistant drones (no mite offspring) from susceptible ones (with mite offspring).4.2 RNA expression analysis (RNA-seq)In addition to identifying key QTL we will also identify the role of these genes in the full gene cascades involved in resistance. We will therefore study the transcriptomic crosstalk between the A. mellifera host and the V. destructor parasite just before and after the brood capping comparing susceptible and resistant A. mellifera drone larvae of those queens that also produced the mapping population. In addition worker and drone larvae of the original susceptible and resistant stock will also be tested to verify that the same mechanisms are operational in workers and drones. This will be done on the Romanian breed if these honey bees also are resistant based on preventing mite ovary activation (otherwise we will use the Swedish breed as contingency measure)Two hour time intervals starting from before capping of the cell until the onset of oogenesis will provide dense data sets for rigorous analysis of the gene activity in host and parasite. We cover the entire critical phase of mite ovary activation with eight RNA extraction data points. We will conduct the sampling as a snapshot on a single comb. The queen lays eggs starting from the centre and proceeds towards the outer edge of the comb. We will cage the queen of a highly infested colony on an empty comb. So we know the start of first egg laying. When all the different developmental stages of bee brood are available we will dip the entire comb into liquid nitrogen. After this snap freezing we will screen the cells and collect the samples that are relevant in our experimental design. After7RNA extraction, we will first screen for specific gene expression of ovary activation related pathways (e.g. juvenile hormone, vitellogenin; Cabrera Cordon et al. 2013) by using quantitative PCR. Having identified time points with strong differences between target genes, we will subject those samples to RNAseq using the Nextgen Sequencing service (e.g. BGI). This will allow us to identify the full interplay of gene cascades of host and parasite by keeping the considerable expenses of RNAseq low. The R package Bioconductor will be used to analyse the expression profiles (Gentleman et al. 2004) to search for covariances between gene pathways of both host and parasite.