Honey bee predisposition of resistance to ubiquitous mite infestations

Abstract

Host-parasite co-evolution history is lacking when parasites switch to novel hosts. This was the case for Western honey bees (Apis mellifera) when the ectoparasitic mite, Varroa destructor, switched hosts from Eastern honey bees (Apis cerana). This mite has since become the most severe biological threat to A. mellifera worldwide. However, some A. mellifera populations are known to survive infestations, largely by suppressing mite population growth. One known mechanism is suppressed mite reproduction (SMR), but the underlying genetics are poorly understood. Here, we take advantage of haploid drones, originating from one queen from the Netherlands that developed Varroa-resistance, whole exome sequencing and elastic-net regression to identify genetic variants associated with SMR in resistant honeybees. An eight variants model predicted 88% of the phenotypes correctly and identified six risk and two protective variants. Reproducing and non-reproducing mites could not be distinguished using DNA microsatellites, which is in agreement with the hypothesis that it is not the parasite but the host that adapted itself. Our results suggest that the brood pheromone-dependent mite oogenesis is disrupted in resistant hosts. The identified genetic markers have a considerable potential to contribute to a sustainable global apiculture.

Full text

1
SCIENTIFIC REPORTS | (2019) 9:7794 | https://doi.org/10.1038/s41598-019-44254-8
www.nature.com/scientificreports
Honey bee predisposition of
resistance to ubiquitous mite
infestations
Bart J. G. Broeckx
1, Lina De Smet2, Tjeerd Blacquière3, Kevin Maebe4, Mikalaï Khalenkow5,
Mario Van Poucke1, Bjorn Dahle6,7, Peter Neumann8, Kim Bach Nguyen9, Guy Smagghe
4,
Dieter Deforce
10, Filip Van Nieuwerburgh
10, Luc Peelman1 & Dirk C. de Graaf
2,5
Host-parasite co-evolution history is lacking when parasites switch to novel hosts. This was the case
for Western honey bees (Apis mellifera) when the ectoparasitic mite, Varroa destructor, switched hosts
from Eastern honey bees (Apis cerana). This mite has since become the most severe biological threat
to A. mellifera worldwide. However, some A. mellifera populations are known to survive infestations,
largely by suppressing mite population growth. One known mechanism is suppressed mite reproduction
(SMR), but the underlying genetics are poorly understood. Here, we take advantage of haploid drones,
originating from one queen from the Netherlands that developed Varroa-resistance, whole exome
sequencing and elastic-net regression to identify genetic variants associated with SMR in resistant
honeybees. An eight variants model predicted 88% of the phenotypes correctly and identied six
risk and two protective variants. Reproducing and non-reproducing mites could not be distinguished
using DNA microsatellites, which is in agreement with the hypothesis that it is not the parasite but the
host that adapted itself. Our results suggest that the brood pheromone-dependent mite oogenesis is
disrupted in resistant hosts. The identied genetic markers have a considerable potential to contribute
to a sustainable global apiculture.
e ubiquitous ectoparasitic mite Varroa destructor, an invasive species from Asia, is the most important bio-
logical driver of global losses of honey bee, Apis mellifera, colonies1. A. mellifera is not the original host however
as this mite occurred rst in the Asian honey bee, Apis cerana1. Whereas A. cerana and V. destructor have a long
history of co-evolution, this is not the case for A. mellifera2. is short period of co-evolution has le A. mellifera
vulnerable2. To avoid colony collapse, treatment with acaricides seemed the best option and this is what mite con-
trol strategies have relied on for more than ve decades. However, scattered observations in aected regions show
that untreated honey bee colonies can survive mite infestations3–5. is prompted several initiatives aimed at
breeding these V. destructor-tolerant or -resistant bees (VR)6–8. ‘Tolerance’ is a defence strategy whereby the host
can limit the harm caused by a given parasite burden, whereas ‘resistance’ refers to the ability of the host to limit
the actual parasite burden itself9. Both social and individual traits have been implicated in the defence against
the V. destructor-mite. Varroa sensitive hygiene (VSH) is a social behaviour trait that consists of three compo-
nents (detection, opening and removal of infested and damaged pupae) that are each inherited independently9
and is expressed by the adult worker honey bees. It was formerly called ‘suppressed mite reproduction’ (SMR)
as reproductive mites were eliminated by hygienic behaviour. However, the term SMR is no longer used for the
social, hygienic behaviour as true SMR has been observed in several populations. As such, SMR is currently
dened as a trait where mites fail to produce ospring in honey bee pupae by a not yet identied mechanism. is
1Department of Nutrition, Genetics and Ethology, Ghent University, B-9820, Merelbeke, Belgium. 2Department of
Biochemistry and Microbiology, Ghent University, B-9000, Ghent, Belgium. 3Wageningen University & Research,
6708PB, Wageningen, The Netherlands. 4Department of Plants and Crops, Ghent University, B-9000, Ghent,
Belgium. 5Honeybee Valley, Ghent University, B-9000, Ghent, Belgium. 6Norwegian Beekeepers Association,
NO-2040, Kløfta, Norway. 7Department of Animal and Apicultural Sciences, Norwegian University of Science
and Technology, NO-1432, Ås, Norway. 8Institute of Bee Health, University of Bern, 3097, Bern, Switzerland.
9Department of Functional and Evolutionary Entomology, Gembloux Agro-Bio Tech, University of Liege, B-5030,
Gembloux, Belgium. 10Department of Pharmaceutics, Ghent University, B-9000, Ghent, Belgium. Bart J. G. Broeckx,
Lina De Smet, Luc Peelman and Dirk C. de Graaf contributed equally. Correspondence and requests for materials
should be addressed to B.J.G.B. (email: Bart.Broeckx@ugent.be)
Received: 11 January 2019
Accepted: 7 May 2019
Published: xx xx xxxx
OPEN
Content courtesy of Springer Nature, terms of use apply. Rights reserved

2
SCIENTIFIC REPORTS | (2019) 9:7794 | https://doi.org/10.1038/s41598-019-44254-8
www.nature.com/scientificreports
www.nature.com/scientificreports/
phenomenon of true resistance emerged independently from natural selection on the island Gotland (Sweden5)
and in Avignon (France4), and there are subtle dierences between these two distinct bee populations in how they
succeed in reducing the reproductive success of the V. destructor mites. In more detail, these resistance pheno-
types are delayed oviposition and actual infertility of the mother mites, respectively10. SMR can be expressed both
in worker and drone brood, which are called worker and drone brood resistance (DBR), respectively. Whereas
both types of resistance are benecial, drone brood has a longer pupation time relative to worker brood, which
gives the V. destructor parasite more time to reproduce. As a consequence, any disturbance of mite reproduction
in drone brood will aect mite population dynamics signicantly9.
Given the burden on the honey bee population, unravelling the genetic architecture of DBR has recently
gained much interest for several reasons11. Firstly, identifying the exact molecular mechanisms might lead to a
better understanding of the host pathogen interaction and new eradication strategies. Secondly, the identication
of genetic markers associated with the phenotype can also immediately be used to selectively breed colonies that
are more resistant.
Here, we investigate the possibility to identify DBR-associated markers by comparing VR and
V. destructor-sensitive drones (VS). One of the diculties when performing these kind of case-control studies is
however the risk of spurious associations that might arise due to population stratication12,13. An ideal solution is
having access to one population where all cases and controls are equally related. As within a colony, it is one queen
that gives rise to many equally related drones, we aimed to create one colony containing both VR and VS drones
at equal frequencies by selective breeding.
When successful, both reproducing and non-reproducing mites will be identified in that colony. While
non-reproducing mites can be a true consequence of host adaptation, which is the desired mechanism,
non-reproduction might however also be a consequence of these mites being a distinct subgroup of V. destructor
parasites lacking the ability to reproduce. While we hypothesize that non-reproduction is a consequence of
host adaptation, this will rst be investigated using microsatellites of both reproducing and non-reproducing
V. destructor mites. Upon conrmation of DBR, the identication of markers can be initiated.
Previously, attempts to identify quantitative trait loci all relied on coarse (with at most 3000 markers) and
indirect (i.e. linkage disequilibrium-based) mapping strategies, oen with inconsistent results14–16. e down-
side of these indirect approaches is that they always require further steps downstream (e.g. ne mapping, candi-
date gene sequencing). Contrary to this approach, whole exome sequencing (WES) has the potential to identify
disease-causing mutations directly if they fall within target regions. But even if causal mutations reside outside
target regions, WES variants discovered during sequencing can be used as tagvariants to identify regions asso-
ciated with the phenotype17. Based on previous experiences18,19, we hypothesize that WES will allow mapping at
a far higher resolution relative to previous approaches. As such, we aimed to develop and evaluate the rst WES
design for the honey bee and immediately assessed its performance to identify variants associated with DBR.
While WES is a technological step forward, association studies, including the coarse mapping studies already
performed for DBR, still oen use single variant association tests in their data analysis. Given the phenotype and
in line with previous studies, a multifactorial, complex inheritance mode was however expected for DBR14–16. As
this involves the combination of an (unknown) number of loci and single-marker tests only analyze the marginal
eect between a phenotype and a variant, these single-marker tests ignore important information when multiple
variants are associated with complex phenotypes. In this study, we use a novel analysis method, called elastic-net
penalized regression, that allows joint modelling of variants and hypothesize it outperforms the standardly used
single-marker tests.
Finally, to allow a broad use of markers identied in one single colony, it is important that these markers seg-
regate in the general population. We investigate this hypothesis by performing a population screening.
Results
Creating a hybrid VR/VS colony. Virgin queens from honey bee stocks selected for VR were selected from
dierent locations in Europe. ree populations became VR aer several years being le untreated against the
V. destructor-mite in the Netherlands (Amsterdam Water Dunes20), France (Toulouse21) and Norway (Østlandet
Region22). From Belgium, virgin sister queens from a breeding queen with the highest recorded Varroa-index
for Belgium in 2014 were selected. is Varroa -index combines measurements of the V. destructor population
dynamics and hygienic behaviour (further explained in materials and methods)8.
ese queens were next crossed twice with local VS drones (Fig.1). In more detail, this implies single VS
drone articial insemination of VR queens in the parental generation (P), leading to hybrid VR/VS queens in
the rst lial generation (F1). ese F1 hybrid queens carrying both VR- and VS-associated alleles were then
mated naturally with local VS drones. As only the queens give rise to drones in the subsequent generation, this
allows segregation of both VR and VS alleles in the second lial (F2) haploid drone brood generation (Fig.1). To
eliminate the eect of environment as much as possible, all F2 colonies were reared at the same location. All ve
populations (the four VR hybrids and the control strain) were successfully bred and maintained.
Screening for the DBR phenotype. Individual host brood cells were opened at the adequate age and
assessed for V. destructor-mite infestation (as described in more detail in materials and methods). is screen-
ing revealed that the percentage of non-reproducing mites in these four hybrid VR populations was 51.0% (the
Netherlands), 37.5% (France), 31.6% (Norway) and 14.0% (Belgium), respectively, whereas this percentage
was 19.0% in a local VS control strain (TableS1). e colony from the Netherlands had the highest frequency
non-reproducing mites. is was also the only colony where this frequency was signicantly higher relative to
the local control strain (P = 0.01), which indicates that the DBR phenotype from this colony can be transmitted
to subsequent generations, i.e. it is suggestive for a genetically-based resistance mechanism. Furthermore, since
the queens of the F1 generation were VR/VS hybrids, the outcome of F2 DBR typing for this colony was even
Content courtesy of Springer Nature, terms of use apply. Rights reserved

3
SCIENTIFIC REPORTS | (2019) 9:7794 | https://doi.org/10.1038/s41598-019-44254-8
www.nature.com/scientificreports
www.nature.com/scientificreports/
Figure 1. Overview of the breeding experiment and phenotypical screening. (A) Origin of the honey bee
populations used in this study. Numbers 1–4 were selected for Varroa destructor-resistance; number 5 was
the Varroa-sensitive control: 1. Østlanded Region, Norway; 2. Amsterdam Water Dunes, e Netherlands;
3. Kapellen, Belgium; 4. Toulouse, France; 5. Ghent, Belgium. (B) Crossing scheme to obtain hybrid Varroa-
resistance/Varroa -sensitive colonies. For one hypothetical locus associated with the phenotype, the allele
associated with drone brood resistance (DBR) is coloured green, while the opposite (undesired) allele is
coloured blue. e resulting F2 drones were phenotyped for DBR. In reality, the situation is more complex as
several variants, as mentioned in Table1, were found to be associated with DBR. P = parentalis; F = lialis;
Q = queen; D = drone. (C) Outcome of the screening for the DBR phenotype in the dierent crossed
populations. is graph depicts the percentage of non-reproducing mites for each of the 5 colonies. To assess
whether this DBR phenotype segregated at signicantly dierent frequencies relative to the control population
(blue), a Fisher exact test was performed. Only for the Amsterdam Water Dunes colony (green), this result was
signicant at the 0.01 level. Due to the very high DBR prevalence, the Amsterdam Water Dunes bee colony was
used in the subsequent exome sequencing. Other p-values can be found in TableS1. *P = 0.01.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

4
SCIENTIFIC REPORTS | (2019) 9:7794 | https://doi.org/10.1038/s41598-019-44254-8
www.nature.com/scientificreports
www.nature.com/scientificreports/
unexpectedly high (expected maximum score was 50%) and corresponds to what can be expected in a Mendelian
mode of inheritance, which makes it ideal for subsequent association studies. As such, this Dutch colony was
selected and used further downstream in the WES part of this study.
Varroa destructor-mite genotyping. In order to verify whether the observed DBR phenomenon is due to
host adaptations and not a parasite-mediated eect, 56 mother mite specimens originating from drone brood cells
with non-reproducing mites (=DBR phenotype) or with reproducing mites were genotyped at 13 microsatellite
loci (TableS2)23. All 13 loci amplied successfully, which led to a total of 23 dierent alleles, varying from one to
eight per locus (TableS2). Per colony, the eective number of alleles ranged from 1.145 to 1.502 (TableS3). e
observed heterozygosity HO values were very low ranging from 0.000 to 0.077. e mean expected heterozygosity
HE was 0.130, with values ranging from 0.075 to 0.186 within each colony. As expected, the inbreeding coecients
were very high (mean: 0.788) demonstrating clear evidence of inbreeding between the V. destructor-mites within
a colony (TableS3). Furthermore, no signicant dierences in these genetic parameters were found between
reproducing and non-reproducing mites from a colony (Paired t-tests, df = 7, P > 0.05), nor when comparing all
reproducing and non-reproducing mites (Independent t-tests, df = 6, P > 0.05). e Evanno method identied
ΔK = 2, which implies that K (or the number of populations) equals two or one (as ΔK cannot be used to distin-
guish those two values)24. Based on the results from the individual mites, K was set to one as each mite belonged
for approximately 50% to both groups (Fig.2). is implies that these mites could not be genetically distinguished
from each other. Based on these results, the observed DBR seems to represent the real and desired genetic resist-
ance phenotype and is not a consequence of dierences between mites that might mimic it.
Whole exome sequencing. To identify genetic variants associated with the DBR phenotype, a WES design
targeting all exons of the A. mellifera genome (Amel.4.5)25 was developed. is design contained 26,184,643 base
pairs (bp) divided over 81,571 regions. Sixty-four drones (32 DBR positive; 32 DBR negative), originating from
the colony from the Netherlands (i.e. from the colony that diered signicantly from the local control strain
in terms of the number of non-reproducing mites), were selected and Illumina sequenced. In these drones, a
Scaold Location Gene G ene name Variant RNA level Protein level Eect β% Present (n)
Group1.41 909712 GB54921 mucin-12 isoform X1 T > C GB54921-RA:r.144 A>GGB54921-PA:p.
A47 = (or Ala47=)Silent 0.23 91% (42/46)
Group1.41 909762 GB54921 mucin-12 isoform X1 C > T GB54921-RA:r.94 G>AGB54921-PA:p.
V32I (or Val32Ile) Missense 0.22 89% (41/46)
Group10.23 545027 GB48382 solute carrier family 22
member 21 C > T GB48382-RA:r.987 G>AGB48382-PA:p.
A328 = (or Ala328=)Silent −0.06 78% (36/46)
Group15.14 757132 GB50526 sodium-coupled
monocarboxylate
transporter 1 C > T GB50526-RA:r.1662G>AGB50526-PA:p.
P554 = (or Pro554=)Silent −0.15 9% (4/46)
Group15.19 133198 GB50114 dynein beta chain, ciliary T > C GB50114-RA:r.1662A>GGB50114-PA:p.
P3167 = (or Pro3167=)Silent 0.26 73% (34/46)
Group3.15 494900 GB47018 uncharacterized protein
LOC724886 isoform X2 G > A GB47018-RA:r.1824C>UGB47018-PA:p.
L608 = (or Leu608=)Silent 0.30 87% (40/46)
Group9.12 805359 GB53345 uncharacterized protein
LOC100578770 T > C GB53345-RA:r.37 A>GG GB53345-PA:p.
M13V (or Met13Val) Missense 0.02 98%(45/46)
Group9.12 888963 GB53340 spectrin beta chain
isoform X1 A > C GB53340-RA:r.4143U>GGB53340-PA:p.
V1381 = (or Val1381=)Silent 0.09 96% (44/46)
Table 1. Variant description together with the eect size (β) and allelic frequency analysis giving the number of
colonies where the mutations were found (from the total of 46 colonies that were sampled). e intercept of the
model equals −0.39 (model settings: α = 0.9, λ = 0.18).
Figure 2. Bayesian clustering of the Varroa destructor-mites originating from eight beehives. Each vertical line
stands for an individual specimen, while the colors are indicative for the proportion a specimen belongs to a
certain group.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

5
SCIENTIFIC REPORTS | (2019) 9:7794 | https://doi.org/10.1038/s41598-019-44254-8
www.nature.com/scientificreports
www.nature.com/scientificreports/
median coverage of 64.5x was obtained, while >97% of the 26 Mb target bp pairs were covered at ≥10x (Fig.S2;
TableS4). In terms of regions, the median number of regions that were entirely sequenced at ≥10x was 91.3%
(Fig.S3; TableS4). e sequencing reproducibility was high, with close to 75% of the regions entirely sequenced
in all samples. Similar to other WES designs, we found that sequencing performance was inversely correlated with
target region length and low complexity/highly repetitive nucleotide composition, whereas GC nucleotide com-
position of a target region and quality of the target region (reected by the proportion of ambiguous nucleotides
in the target region) aected sequencing performance far less (Fig.S4; TableS5)19. e aforementioned results
demonstrate that the novel WES design outperforms several WES designs in other species18,19.
Next, we looked at the number of variants that can be used for the association study. Overall, more than
140,000 variants with a call rate of at least 99% were discovered inside the exons. Both in terms of the number
of markers (50 times higher), as with respect to the resolution (the median and mean distance between markers
were 71 bp (rst quartile: 18 bp – third quartile: 342 bp) and 1562 bp, respectively (Fig.S5)), this is by far the most
dense study on DBR conducted up until now14–16.
Single-marker tests and elastic-net penalized regression. Whereas single-marker tests are most
commonly used, previous studies on this phenotype have shown that these tests oen lack the necessary power
to obtain signicant results15,16. e same result was observed here as no single variant reached genome-wide sig-
nicance (Fig.S6). is was no surprise as for the proposed complex inheritance of the phenotype, single-marker
tests are far from ideal as these tests only analyze the eect of an individual variant on a phenotype, while ignoring
all the others26–29.
A potential solution is to model all variants jointly. In genetic studies, including the one presented here,
there are however additional complications that have to be taken into account26–30. Firstly, the number of vari-
ants (=parameters p) oen exceed the sample size n by far (oen referred to as “the small n, large p” problem).
Secondly, both due to the high dimensionality (i.e. the large number of variants) as the actual physical linkage
between variants, dierent variants can be correlated with each other, leading to the multicollinearity problem.
Where standard multiple regression cannot deal with these issues, the so-called elastic-net penalized regres-
sion that combines the strengths of LASSO in terms of parameter selection and ridge regression to deal with
correlated variants, solves all these problems26–28. Based on the combination of cross-validation and stringent
cut-os, elastic-nets have shown to control the number of false positives (oen perfect, i.e. no single false positive
result) while at the same time identifying a large number of true associated variants28. In this study, the same strict
methodology was used, which resulted in eight variants (eight SNPs) in seven dierent genes that were found to
be associated with the DBR phenotype (Table1). From these eight mutations, two were missense, altering the
amino-acid composition, while the remaining six were silent mutations. Altogether, this eight variants model
predicted 88% of the initial sixty-four phenotypes correctly (56/64) (Table2) and identied six risk (i.e. associ-
ated with a lower resistance to V. destructor mites) and two protective (i.e. associated with a higher resistance to
V. destructor mites) variants.
So far, previous (coarse) mapping studies on DBR in the Swedish bee population identied loci at chromo-
some four, seven and nine15 and the chromosomes two, three and een, respectively16. In our study, SNPs at
chromosome nine and chromosome een were retained in the nal model. While we believe this substantiates
the results found in this study, at the same time, it is clear that, in line with the (partially diering) DBR pheno-
type, both populations (the Swedish one from the previous studies versus our Dutch population from this study)
also developed a unique way of dealing with the same parasite.
Allelic frequency analysis in the Belgian honey bee population. As a nal step, a stratied sampling
strategy was used to evaluate the allelic frequencies of the previously identied associated variants in the general
bee population in Belgium. Sanger sequencing, each time two bees per colony for a total of 46 colonies, revealed
a widespread distribution of all variants throughout the bee colonies (Table1). is indicates that these variants
are not colony specic, which facilitates their use in centrally coordinated population-wide selection programs.
Furthermore, on average, the risk mutations were found in more colonies (89%; 41/46 colonies) relative to the
protective mutations (43%; 20/46 colonies), which is no surprise given the widespread V. destructor-sensitivity in
the Belgian honey bee population. Finally, in agreement with the expectations for complex diseases, there was a
clear negative correlation (Spearman’s correlation coecient = −0.38) between the prevalence of the mutations
in the population and the eect size (i.e. the absolute magnitude of the β coecients of the dierent variants as
depicted in Table1)31.
Hypothesized biological involvement of the identied genes. As mentioned and in agreement with
literature, the DBR phenotype was expected to segregate in a complex manner, requiring the involvement of sev-
eral genes14–16. In our study, variants in seven genes (i.e. mucin-12 isoform X1, solute carrier family 22 member
Model
Truth
Control Aected
Control 29 5
Aected 3 27
Table 2. Contingency table comparing the predicted phenotypes based on the eight variant model (“Model”)
relative to the observed phenotypes (“Truth”) of the 64 drones used in the whole exome sequencing experiment.
Fiy-six out of 64 drones were correctly classied (88%).
Content courtesy of Springer Nature, terms of use apply. Rights reserved

6
SCIENTIFIC REPORTS | (2019) 9:7794 | https://doi.org/10.1038/s41598-019-44254-8
www.nature.com/scientificreports
www.nature.com/scientificreports/
22, sodium-coupled monocarboxylate transporter-1, dynein beta chain, spectrin beta chain isoform X1 and two
uncharacterized proteins; Table1) were found and both synonymous and non-synonymous variants were iden-
tied. It goes without saying that these variants can both be actual phenotype-causing variants or might just be
markers associated with it. While traditionally non-synonymous mutations are oen focused on, recent studies,
however, strongly support the potential role that synonymous mutations on (complex) phenotypes might have32–
34. In addition, the involvement of the dynein beta chain in the DBR phenotype was remarkable, as it represents
a cytoskeletal motor protein involved in intracellular retrograde transport along the microtubules of eukaryotic
cilia and agella35. Among insects, cilia/agella are present in the sperm tail (agellar) and in mechano- and che-
mosensory neurons (ciliary) only36,37. Moreover, the ciliary form of dynein that we identied here can exclusively
be related with sensory functions of the honey bee antenna, most probably through its olfactory neurons.
In honey bees, brood care is secured through a few behavioural sequences (feeding, brood cell capping, ther-
moregulation) that are initiated by brood pheromones38. e sensing of these cuticular hydrocarbon pheromones
occurs on the insects’ antennae, more in particularly by specialized olfactory sensillae39. It has been demon-
strated that the V. destructor-mite synchronizes its reproduction with the ontogenic development of the honeybee
larvae and that the mite’s oogenesis is triggered by volatiles of the larval cuticle40,41. We hypothesize that the
DBR phenotype was obtained by two phenomenons. Firstly, the variant of the dynein beta chain found in DBR
positive bee colonies causes a better pheromone sensing by an improved intracellular transport in the olfactory
neurons. Secondly, the other variants cause a diminished production of the cuticular hydrocarbon brood pher-
omones by a reduced general or tissue-targeted (integumentum) metabolism. Here, the dierent transporters
might play an important role42. Consequently, the pheromone release falls to a level that is no longer able to
initiate oogenesis in the mite, although it permits normal chemical communication - and thus brood care - by
the adult bees. e hypothesis links several of the genes from which a variant is associated with the DBR phe-
notype, but so far any experimental evidence is lacking. However, a similar strong involvement of the olfactory
sensing has been demonstrated by transcriptome studies43–45 of the VSH phenotype that renders bees resistant to
V. destructor-mite infestations by hygienic behaviour. In that case again intracellular transport and vesicle traf-
cking in the antennae is one of the underlying biological mechanisms44,46. In addition, in the Gotland VR bees
a glucose-methanol-choline oxidoreductase was found to be putatively involved in changing volatiles emitted by
the bee larvae14.
Discussion
Selection for more resilient bees oers a sustainable solution for the main driver of global honey bee colony losses,
i.e. the infestation by the ectoparasitic V. destructor mite. Although the voluntary denial of any treatment has
already led several times to V. destructor-tolerant or -resistant bees, this approach is still dicult for the average
beekeeper because of the high losses that may go with it. at is why unravelling how natural selection shapes
resilient honey bee populations at both the phenotypic and genetic level, represents a crucial lever to support the
classical selection programs through breeding value estimation or marker-assisted selection. e present work
developed novel state-of-the-art methodologies for honey bees to discover genetic variants associated with the
given trait.
Firstly, we have developed a WES design for the honey bee. In terms of performance, this newly developed
WES design excels in comparison with WES designs from other species19. Furthermore, relative to for example
RESTseq16,47 and the 44 K GWAS SNP assay48, far more variants are found, which should improve the identica-
tion of disease-associated variants, while relative to whole genome sequencing, it is a more cost-ecient approach
until sequencing prices drop further. Focusing on exons, it is important to stress that variants outside the target
regions are potentially missed, especially for those located at a distance of exonic variants. Overall however, this
WES design fullled our expectations and we are condent that, for the time being, it will be a valuable tool for
further genetic studies of honey bees.
Secondly, we evaluated elastic-net regression and demonstrated that it outperforms single-marker association
tests. Combined with WES, it was capable of discovering variants associated with complex traits. To our knowl-
edge, this is the rst time that either WES or elastic-nets have been used in the honey bee.
In our approach, we specically chose to sequence drones derived from one colony for two main reasons.
By sequencing equally related bees, the problem of spurious associations due to population stratication was
avoided. In addition, while SMR has been reported to have developed naturally in several populations, the under-
lying mechanism is not entirely the same in every population10. As such, there is a high risk that the genetic con-
tributors dier as well, which, in turn, would have reduced the power to detect an association. is is avoided by
focusing on one population. A potential downside is that this might lead to a reduced generalizability, i.e. that (at
least) some variants associated with the phenotype are unique for that specic population. One option for future
studies is thus to redo this experiment by combining drones from dierent populations.
Also in terms of the population study, we deliberately chose to look for the variants in the Belgian honey bee
population. e main reason was a direct evaluation of the potential applicability of the results in that population.
Other options that might be pursued in the future are for example looking for the variants in public datasets49,50
and other VR populations to investigate how widespread these mutations are.
With respect to the phenotype, the present study describes a very promising trait, i.e. DBR, in an already
well-studied resilient honey bee population that was established in the Amsterdam Water Dunes in the
Netherlands by natural selection20. V. destructor has a strong preference for drone brood51, making drone
brood removal or drone brood ‘cutting’ a common intervention in beekeeping practice and part of a biotech-
nical V. destructor control strategy. As the post-capping developmental stage of drones lasts two days longer
than for worker bees (14 days instead of 12 days), more mature V. destructor-mites emerge from a drone brood
cell. e Cape honey bee (A. m. capensis) has an innate resistance against the V. destructor-mite by shortening
the post-capping developmental time to only 9 days on average52. e failure of mite reproduction in the DBR
Content courtesy of Springer Nature, terms of use apply. Rights reserved

7
SCIENTIFIC REPORTS | (2019) 9:7794 | https://doi.org/10.1038/s41598-019-44254-8
www.nature.com/scientificreports
www.nature.com/scientificreports/
phenotype has a similar eect on the V. destructor population dynamics, and thus is an alternative way to get
resistance against the mite by avoiding massive mite reproduction in the male brood. Moreover, the costs of this
trait for the colony is relatively low when compared to hygienic behaviour, where up to 32.4% of the pupae are
removed from the brood53. As the identied variants associated with DBR were widespread in the natural bee
population, we believe the road to marker assisted selection is open.
Material and Methods
Honey bee populations and crossing. Virgin queens from honey bee stocks selected for VR originated
from dierent locations in Europe. From Belgium, we also took virgin sister queens from breeding queen with
code 57-584-11612-2012 from Guido Haagdorens, a participant of the Belgian branch of the Beebreed program
(Länderinstitut für Bienenkunde Hohen Neuendorf, Germany). is breeding queen was chosen because it gave
the highest recorded Varroa-index for Belgium in 2014. e Beebreed program relies on breeding value esti-
mation and the Varroa-index combines measurements of the V. destructor population dynamics and hygienic
behaviour8. It requires estimations of the daily mite-fall in spring, the size of the phoretic mite population by the
powdered sugar method in summer and the clearance of dead pupae by the pin-test54.
Six virgin queens of each honey bee stock were crossed twice with local VS drones. In the parental generation,
we performed articial insemination at a queen age of 9–10 days under CO2 treatment (this treatment was also
given 1 day earlier for 6 min). For each insemination fresh semen was taken from another drone from the same
colony (code EXP10) and injected in the queen’s main oviduct (1.5 µl semen + 0.5 µl dilution buer; dilution
buer contained 0.2 M NaCl, 5 mM glucose monohydrate, 0.67 mM L-lysine, 0.57 mM L-arginine, 0.68 mM
L-glutaminic acid, 0.02 M Trisma HCl, 0.03 M Trisma base and 2.5 mg/ml dihydrostreptomycine). Ten to twelve
days aer introducing the queens in 3-frame hives, we checked for oviposition and nine days later for brood cap-
ping. At that moment and when insemination was proven to be successful, we started queen rearing.
From each genetic stock only one colony was chosen for queen rearing. At least four one-day-old larvae from
each colony were graed to articial queen cells and then transferred to a cell building colony. Once the queen
cells were sealed they were transferred to an incubator. e emerging queens were introduced in small mating
nuclei colonies, that were treated with oxalic acid (V. destructor-treatment) prior to queen introduction. In this
rst lial generation virgin queens were allowed to mate naturally. Again we checked for oviposition and cell
capping. A second mite treament was done just before winter.
In spring, all colonies were moved to the same apiary (campus Sterre) in order to do the testing in the same
environment. Drone brood frames were introduced in two colonies per genetic stock. However, only one colony
was used for determination of the phenotype.
Determination of the phenotype. When drone brood cells were capped, the frames were transferred to
an incubator and kept at 34 °C. By doing so, we could avoid that our measurements were inuenced by hygienic
behaviour of the adult bees. Ten days later the drone brood and the V. destructor-mites were killed by freezing.
is simplies the determination of the phenotype as mites will no more escape when drone brood cells are
opened. We examined drone brood cells for the presence of a mother mite only (non-reproducing) or a mother
mite with her progeny (reproducing) in the presence of red-eyed drone pupae. Both drone pupae and mother
mites were stored at −80 °C for subsequent WES and V. destructor genotyping of selected samples, respectively.
A Fisher exact test was used to compare the frequency of the VR and the VS drones in the colonies derived from
the VR populations relative to the local VS control population. Signicance was set at α ≤ 0.05/4 (Bonferroni
correction for multiple testing). is was performed in R v3.4.2 (“Short Summer”).
Varroa destructor-mite genotyping. Fiy-six V. destructor specimens were genotyped with 13 micro-
satellite loci which already have proven to give reliable signals23 (TableS2). Ten loci were originally developed
for V. destructor: seven loci (VD146, VD163, VD001, VD151, VD015, VD112 and VD114) by Cornman et al.55,
and three loci (VD305, VD306 and VD307) by Solignac et al.56. e nal 3 loci (VJ275, VJ294, and VJ292) were
developed for V. jacobsoni57.
Individual DNA was obtained from whole V. destructor mites following a Chelex (InstaGene Matrix, BioRad)
DNA extraction method as described in Maebe et al.58. In short, 200 µl Chelex and 10 µl proteinase K was added
to an individual mite (sliced with a sterile blade) and incubated for 2 h at 56 °C. Aer a second incubation step of
15 min at 96 °C, the supernatants of 180 µl (DNA) was frozen in −20 °C until further use.
By multiplex PCR, these 13 microsatellites were amplified by HotstarTAq DNA Polymerase (QIAgen,
Belgium) in a total volume of 10 µl. e PCR mix consisted out of 1 µl template DNA, PCR buer (1x), 0.2 µM
dNTP’s, 0.1 µM forward primer, 0.5 µM reverse primer, 0.5 µM dierent labelled forward M13 primers and 1 unit
Taq polymerase. Fluorescent labelling of the PCR products was done using a tailed-primer approach59. In this
approach a universal M13-primer (=‘t a il’, 5′-GAGTTTTCCCAGTCACGAC-3′) was coupled to a VIC, 6-FAM,
PET or NED uorescent label (TableS2). To allow the incorporation of the tail during PCR, the same sequence as
the tail was built-in at the 5′-end of all forward primers (see also60). Furthermore, the normal annealing temper-
ature of 60 °C was decreased with 2 °C to 58 °C. e other PCR conditions were: a rst denaturation step at 95 °C
for 15 min, then 30 cycles of 30 s denaturation at 95 °C, 30 s annealing at 58 °C and 30 s extension at 72 °C, followed
by a nal extension step of 10 min at 72 °C. Visualization of the PCR products was done by capillary electro-
phoreses on an ABI3730xl sequencer (Applied Biosystems) with help of a 500 LIZ standard (Genescan, Applied
Biosystems). e fragments were scored manually with the Peak Scanner v1.0 soware (Applied Biosystems). As
a quality control, the amplication of 16 randomly chosen samples was repeated.
For the mites originating from the same beehive, several genetic parameters were determined with the pro-
gram GenAlEx 6.561 including: Nei’s unbiased expected heterozygosity (HE), the observed heterozygosity (HO)
and the eective number of alleles (Ne) as parameters of genetic diversity, and also the inbreeding coecient (Fis).
Content courtesy of Springer Nature, terms of use apply. Rights reserved

8
SCIENTIFIC REPORTS | (2019) 9:7794 | https://doi.org/10.1038/s41598-019-44254-8
www.nature.com/scientificreports
www.nature.com/scientificreports/
Paired t-tests were performed over all loci to search for dierences between reproducing and non-reproducing
mites within a colony, while possible dierences over all colonies between reproducing and non-reproducing
mites were tested with independent t-tests. ese tests were performed in SPSS.
Hence, the soware Structure v2.3.362 was used to perform a Bayesian approach to determine the number of
populations (or groups) within the dataset. In this analysis, the number of populations (K) was estimated from
1 to 8, and this was repeated 9 times. Each K-value was calculated with a burn in of 1,000,000 iterations and
500,000 MCMC data collecting steps. Hence, the free online program Structure Harvester v0.6.9363, was used to
determine the best value of K, and the program Distruct v1.1 soware was used for graphical visualization of the
population structure.
Development of exome design. e reference genome (Amel_4.5_scaolds.fa) and corresponding anno-
tation (amel_OGSv3.2.g3.gz) were obtained from the Hymenoptera Genome Database64. A bed le containing
all the exons (including the UTRs) was created with the bedr (v1.0.4) and seqinr (v3.4–5) R-packages. Our design
was processed by the Roche Nimblegen custom design group (Madison, USA). Using an SSAHA algorithm, cap-
turing baits were developed based on our design and the reference genome. Design settings for the baits allowed
ve or fewer single-base insertions, deletions or substitutions between the baits and the genome. Each bait itself
was allowed to match up to 20 locations in the genome. Regions under 100 bp were padded to 100 bp to increase
capturing eciency. Aer approval, the baits were generated and provided as SeqCap Developer Library.
DNA extraction. Sixty-four pupae (32 coming from a drone brood cell with a non-reproducing mite; 32
with a reproducing mite), all derived from one colony, were subsequently used for DNA extraction. Each pupae
was individually homogenized in a total of 1 ml RLTPlus buer by mechanical agitation in a TissueLyser for 90 s
at 30 Hz, in the presence of 4 metal beads and glass beads. One third of the sample was used to isolate DNA and
RNA with the ALLPREP DNA/RNA isolation kit from Qiagen following the manufacturers recommendations.
e RNA was eluated in 100 µl RNAse free water while the DNA was eluated in 80 µl EB buer.
Sample preparation and sequencing. Following DNA extraction, a picogreen assay was performed and
1 μg of every sample was subsequently treated with RNase I (ermoFisher Scientic). e DNA was next frag-
mented on a Covaris S2 System in a 130 μl volume (aim: 300 bp fragments, settings: duty cycle: 10%, intensity: 4,
cycles per burst: 200, time: 80 s). Depending on the yield aer DNA-extraction, between 500 ng and 1 μg of the
fragmented DNA was used as input for the library preparation. Samples were end repaired, A-tailed and ligated
with TruSeq adapters using the reagents from the KAPA library prep kit according to the manufacturer’s protocol.
Size selection was performed on a 2% E-Gel (Invitrogen Life Technologies) (G4010-02), fragments were selected
with an insert size around 200–700 bp. ereaer, the pre-capture LM-PCR was performed on the samples for 8
cycles as prescribed in the SeqCap EZ library protocol. e concentration of each PCR product was determined
using Quant-iT™ PicoGreen™ dsDNA Assay Kit (Life Technologies). Sixteen times four samples were equimol-
arly pooled to obtain a total DNA input of 1250 ng. e pooled library was hybridized for 19 hours and 30 minutes
with the baits (SeqCap Developer Library). e hybridized library was washed and the captured and pooled
DNA was recovered. Aer a nal amplication (LM-PCR, 13 cycles), the quality of the library was checked using
the High Sensitivity DNA chip (Agilent). To check the fold enrichment aer capturing, a qPCR is performed
as a quality control step before sequencing. e used primers are shown in TableS6. An additional qPCR was
performed to determine the quantity of the library to ensure optimal cluster densities. Two times twenty-four
samples and one time sixteen samples were sequenced per lane on the NextSeq 500 PE 75 bp.
Sequencing data-analysis. The reads were aligned to the reference genome (Amel_4.5) using BWA
v0.7.1565. Duplicate reads were marked with Picard tools v2.1.1. Using the GATK v3.8-0, variants were called
according to the GATK Best Practices66.
Variant ltering. From the total list of putative variants, only those were retained that 1/passed the “hard”
quality lter suggested from the GATK Best Practices, 2/that had at least 2 dierent alleles segregating in the
population, 3/fell within the target regions and 4/with a call rate of 99%. Filtering was performed with VCFtools
v0.1.14 and custom R-scripts67. Retrieval of the gene function was done by BLAST searching.
Single-marker tests and elastic-net penalized regression. For the single marker tests, a Fisher exact
test was conducted for each variant. e Bonferroni correction for multiple testing was applied. Elastic-nets
penalized regression was next performed with the glmnet v2.0–12 R-package in R v3.4.2 (“Short Summer”)27.
For the model selection, potential parameters were the 140,151 variants and no covariates were added. To obtain
the optimal lambda (i.e. the penalty) and alpha (i.e. the balance between more “lasso”-like and “ridge”-regression
like behaviour), a leave-one-out cross-validation was performed for alpha ranging from 0 to 1 (in steps of 0.1).
Lambda was set at the stringent MSE+1SE threshold28.
Allelic frequency analysis in the Belgian honey bee population. Worker bees were sampled at the
apiaries from breeders involved in the Flemish beekeeping program. irty breeders sampled only a single col-
ony; two others sampled four and twelve colonies, respectively. e apiaries are distributed throughout Flanders,
the northern part of Belgium. Two individual worker bees from 46 dierent colonies were sampled from dierent
apiaries. Each single bee was homogenized in 0.5 ml of 100 mM NaCl; 20 mM Tris-HCl, pH 8; 25 mM EDTA, pH
8; 0.5% SDS by mechanical agitation in a TissueLyser for 90 s at 30 Hz, in the presence of metal beads and glass
beads. Aer homogenization 20 µl/ml Proteinase K (20 mg/ml) was added and incubated for 1 hour at 37 °C. First,
an equal volume of phenol:chloroform was added and centrifuged at 12,000 g for 10 min at 4 °C. e supernatant
was then extracted with an equal volume of chloroform:isoamyl alcohol (24:1) and centrifuged at 12,000 g for
Content courtesy of Springer Nature, terms of use apply. Rights reserved

9
SCIENTIFIC REPORTS | (2019) 9:7794 | https://doi.org/10.1038/s41598-019-44254-8
www.nature.com/scientificreports
www.nature.com/scientificreports/
10 min at 4 °C. e DNA was precipitated using two volumes of ethanol and centrifuged at 12,000 g for 30 min at
4 °C. e precipitated DNA was nally washed with 0.5 ml 70% ethanol. e DNA was dissolved in 100 µl DNase/
RNase free water.
Primer pairs were designed with Primer-BLAST68. Primers were chosen in regions that were free of second-
ary structures (Mfold) and are listed in TableS769. PCR amplicons were analyzed via agarose gel electropho-
resis. Sequencing reactions were performed using the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied
Biosystems, Foster City, CA, USA) with the individual PCR primers as sequencing primers and run at Eurons
Genomics (Ebersberg, Germany). Sequence analysis was performed with BioEdit v7.2.6.
e allele frequency was determined at the colony level, followed by a comparison of the average allele fre-
quency for risk and protective alleles and an evaluation of the correlation between the eect size and the allele
frequency (Spearman correlation).
Data Availability
All the data is available upon request.
References
1. osenranz, P., Aumeier, P. & Ziegelmann, B. Biology and control of Varroa destructor. J. Invertebr. Pathol. 103, S96–S119 (2010).
2. ruitwagen, A., van Langevelde, F., van Dooremalen, C. & Blacquière, T. Naturally selected honey bee (Apis mellifera) colonies
resistant to Varroa destructor do not groom more intensively. J. Apic. Res. 56, 354–365 (2017).
3. De Jong, D. & Soares, A. E. E. An isolated population of Italian bees that has survived Varroa jacobsoni infestation without treatment
for over 12 years. Am Bee J 137, 742–745 (1997).
4. Le Conte, Y. L. et al. Honey bee colonies that have survived Varroa destructor. Apidologie 38, 566–572 (2007).
5. Fries, I., Imdorf, A. & osenranz, P. Survival of mite infested (Varroa destructor) honey bee (Apis mellifera) colonies in a Nordic
climate. Apidologie 37, 564–570 (2006).
6. Büchler, . Varroa tolerance in honey bees - Occurrence, characters and breeding. Be e Wor ld 75, 54–70 (1994).
7. Harbo, J. . & Hoopingarner, . A. Honey bees (Hymenoptera: Apidae) in the United States that express resistance to Varroa
jacobsoni (Mesostigmata: Varroidae). J. Econ. Entomol. 90, 893–898 (1997).
8. Bienefeld, . Breeding Success or Genetic Diversity in Honey Bees? Be e Wo rld 93, 40–44 (2016).
9. urze, C., outtu, J. & Moritz, . F. A. Parasite resistance and tolerance in honeybees at the individual and social level. Zoology 119,
290–297 (2016).
10. Loce, B., Le Conte, Y., Crauser, D. & Fries, I. Host adaptations reduce the reproductive success of varroa destructor in two distinct
european honey bee populations. Ecol. Evol. 2, 1144–1150 (2012).
11. Zaar, E., Jávor, A. & usza, S. Genetic bases of tolerance to Varroa destructor in honey bees (Apis mellifera L.). Insectes Soc. 61,
207–215 (2014).
12. Lewis, C. M. Genetic association studies: design, analysis and interpretation. Brief. Bioinform. 3, 146–153 (2002).
13. Price, A. L., Zaitlen, N. A., eich, D. & Patterson, N. New approaches to population stratication in genome-wide association
studies. Nat. Rev. Genet. 11, 459–63 (2010).
14. Lattor, H. M. G., Buchholz, J., Fries, I. & Moritz, . F. A. A selective sweep in a Varroa destructor resistant honeybee (Apis
mellifera) population. Infect. Genet. Evol. 31, 169–176 (2015).
15. Behrens, D. et al. ree QTL in the honey bee Apis mellifera L. suppress reproduction of the parasitic mite Varroa destructor. Ecol.
Evol. 1, 451–458 (2011).
16. Conlon, B. H. et al. e role of epistatic interactions underpinning resistance to parasitic Varroa mites in haploid honey bee (Apis
mellifera) drones. J. Evol. Biol. 31, 801–809 (2018).
17. Broecx, B. J. G. et al. An exome sequencing based approach for genome-wide association studies in the dog. Sci. Rep. 7, 15680
(2017).
18. Broecx, B. J. G. et al. Development and performance of a targeted whole exome sequencing enrichment it for the dog (Canis
Familiaris Build 3.1). Sci. Rep. 4, 5597 (2014).
19. Broecx, B. J. G. et al. Improved canine exome designs, featuring ncNAs and increased coverage of protein coding genes. Sci. Rep.
5, 12810 (2015).
20. Panziera, D., van Langevelde, F. & Blacquière, T. Varroa sensitive hygiene contributes to naturally selected varroa resistance in honey
bees. J. Apic. Res. 56, 635–642 (2017).
21. efuss, J., Vanpouce, J., Bolt, M. & efuss, C. Selection for resistance to Varroa destructor under commercial beeeeping
conditions. J. Apic. Res. 54, 563–576 (2015).
22. Oddie, M. A. Y., Dahle, B. & Neumann, P. Norwegian honey bees surviving Varroa destructor mite infestations by means of natural
selection. PeerJ 5, e3956 (2017).
23. oberts, J. M. ., Anderson, D. L. & Tay, W. T. Multiple host shis by the emerging honeybee parasite, Varroa jacobsoni. Mol. Ecol.
24, 2379–2391 (2015).
24. Evanno, G., egnaut, S. & Goudet, J. Detecting the number of clusters of individuals using the soware STUCTUE: A simulation
study. Mol. Ecol. 14, 2611–2620 (2005).
25. Elsi, C. G. et al. Finding the missing honey bee genes: Lessons learned from a genome upgrade. BMC Genomics 15, 1–29 (2014).
26. Wu, T. T., Chen, Y. F., Hastie, T., Sobel, E. & Lange, . Genome-wide association analysis by lasso penalized logistic regression.
Bioinformatics 25, 714–721 (2009).
27. Friedman, J., Hastie, T. & Tibshirani, . egularization Paths for Generalized Linear Models via Coordinate Descent. J. Stat. Sow.
33 (2010).
28. Waldmann, P., Mészáros, G., Gredler, B., Fuerst, C. & Sölner, J. Evaluation of the lasso and the elastic net in genome-wide
association studies. Front. Genet. 4, 1–11 (2013).
29. Arbet, J., McGue, M., Chatterjee, S. & Basu, S. esampling-based tests for Lasso in genome-wide association studies. BMC Genet. 18,
1–15 (2017).
30. Fan, J. & Lv, J. A Selective Overview of Variable Selection in High Dimensional Feature Space. Stat. Sin. 20, 101–148 (2010).
31. Manolio, T. A. et al . Finding the missing heritability of complex diseases. Nature 461, 747–53 (2009).
32. Brule, C. E. & Grayhac, E. J. Synonymous Codons: Choose Wisely for Expression. Trends Genet. 33, 283–297 (2017).
33. Fu, J. et al. Codon usage aects the structure and function of the Drosophila circadian cloc protein PEIOD. G enes Dev. 1761–1775
(2016).
34. McCarthy, C., Carrea, A. & Diambra, L. Bicodon bias can determine the role of synonymous SNPs in human diseases. BMC
Genomics 18, 1–11 (2017).
35. oberts, A. J. Emerging mechanisms of dynein transport in the cytoplasm versus the cilium. Biochem. Soc. Trans. 46, 967–982 (2018).
36. Han, Y.-G., wo, B. H. & ernan, M. J. Intraaggelar Transport Is equired in Drosophila to Dierentiate Sensory Cilia but Not
Sperm. Curr. Biol. 13, 1679–1686 (2003).
Content courtesy of Springer Nature, terms of use apply. Rights reserved

10
SCIENTIFIC REPORTS | (2019) 9:7794 | https://doi.org/10.1038/s41598-019-44254-8
www.nature.com/scientificreports
www.nature.com/scientificreports/
37. Mencarelli, C., Lupetti, P. & Dallai, . New insights into the cell biology of insect axonemes. Int. Rev. Cell Mol. Biol. 268, 95–145
(2008).
38. Le Conte, Y., Arnold, G., Trouiller, J., Masson, C. & Chappe, B. Identification of a brood pheromone in honeybees.
Naturwissenschaen 77, 334–336 (1990).
39. Sharma, . . et al. Cuticular Hydrocarbon Pheromones for Social Behavior and eir Coding in the Ant Antenna. Cell Rep. 12,
1261–1271 (2015).
40. Garrido, C. & osenranz, P. Volatiles of the honey bee larva initiate oogenesis in the parasitic mite Varroa destructor. Chemoecology
14, 193–197 (2004).
41. Frey, E., Odemer, ., Blum, T. & osenranz, P. Activation and interruption of the reproduction of Varroa destructor is triggered by
host signals (Apis mellifera). J. Invertebr. Pathol. 113, 56–62 (2013).
42. Halestrap, A. Monocarboxylic acid transport. Compr Physiol 3, 1611–1643 (2013).
43. Navajas, M. et al. Dierential gene expression of the honey bee Apis mellifera associated with Varroa destructor infection. BMC
Genomics 9, 1–11 (2008).
44. Mondet, F. et al. Antennae hold a ey to Varroa-sensitive hygiene behaviour in honey bees. Sci. Rep. 5, 10454 (2015).
45. Scannapieco, A. C. et al. Expression analysis of genes putatively associated with hygienic behavior in selected stocs of Apis mellifera
L. from Argentina. Insectes Soc. 64, 485–494 (2017).
46. Hu, H. et al. Proteome analysis of the hemolymph, mushroom body, and antenna provides novel insight into honeybee resistance
against varroa infestation. J. Proteome Res. 15, 2841–2854 (2016).
47. Stolle, E. & Moritz, . F. A. ESTseq - Ecient Benchtop Population Genomics with ESTriction Fragment SEQuencing. PLoS One
8, 4–8 (2013).
48. Spötter, A., Gupta, P., Nürnberg, G., einsch, N. & Bienefeld, . Development of a 44 SNP assay focussing on the analysis of a
varroa-specic defence behaviour in honey bees (Apis mellifera carnica). Mol. Ecol. Resour. 12, 323–332 (2012).
49. Wallberg, A. et al. A worldwide survey of genome sequence variation provides insight into the evolutionary history of the honeybee
Apis mellifera. Nat. Genet. 46, 1081–1088 (2014).
50. Harpur, A. B. A. et al. Integrative Genomics eveals the Genetics and Evolution of the Honey Bee’s S ocial Immune System. Genome
Biol. Evol. evz018 (2019).
51. Fuchs, S. Preference for drone brood cells by Varroa jacobsoni Oud in colonies of Apis mellifera carnica. Apidologie 21, 193–199
(1990).
52. Moritz, . F. A. & Hänel, H. estricted development of the parasitic mite Varroa jacobsoni Oud. in the Cape honey bee Apis
mellifera capensis Esch. 1. J. Appl. Entomol. 97, 91–95 (1984).
53. Boecing, O., Bienefeld, . & Drescher, W. Heritability of the Varroa-specic hygienic behaviour in honey bees (Hymenoptera:
Apidae). J. Anim. Breed. Genet. 117, 417–424 (2000).
54. Büchler, . et al. Standard methods for rearing and selection of Apis mellifera queens Standard methods for rearing and selection of
Apis mellifera queens. J. Apic. Res. 52, 1–30 (2015).
55. Cornman, S. . et al . Genomic survey of the ectoparasitic mite Varroa destructor, a major pest of the honey bee Apis mellifera. BMC
Genomics 11, 602 (2010).
56. Solignac, M. et al. Characterization of microsatellite marers for the apicultural pest Varroa destructor (Acari: Varroidae) and its
relatives. Mol. Ecol. Notes 3, 556–559 (2003).
57. Evans, J. D. Microsatellite loci in the honey bee parasitic mite Varroa jacobsoni. Mol. Ecol. 9, 1436–1438 (2000).
58. Maebe, . et al. Microsatellite analysis of museum specimens reveals historical dierences in genetic diversity between declining and
more stable Bombus species. PLoS One 10, 1–15 (2015).
59. Schuele, M. An economic method for the uorescent labeling of PC fragments. Nat. Biotechnol. 18, 233–234 (2000).
60. Maebe, ., Meeus, I., De ie, J. & Smagghe, G. Quantitative trait loci for light sensitivity, body weight, body size, and morphological
eye parameters in the bumblebee, Bombus terrestris. PLoS One 10, 1–21 (2015).
61. Peaall, . & Smouse, P. E. GenALEx 6.5: Genetic analysis in Excel. Population genetic soware for teaching and research-an update.
Bioinformatics 28, 2537–2539 (2012).
62. Pritchard, J. ., Stephens, M. & Donnelly, P. Inference of Population Structure Using Multilocus Genotype Data. Genetics 155,
945–959 (2000).
63. Earl, D. A. & vonHoldt, B. M. STUCTUE HAVESTE: A website and program for visualizing STUCTUE output and
implementing the Evanno method. Conserv. Genet. Resour. 4, 359–361 (2012).
64. Elsi, C. G. et al. Hymenoptera Genome Database: Integrating genome annotations in HymenopteraMine. Nucleic Acids Res. 44,
D793–D800 (2016).
65. Li, H. & Durbin, . Fast and accurate long-read alignment with Burrows-Wheeler transform. Bioinformatics 26, 589–595 (2010).
66. Van der Auwera, G. A. et al. From FastQ Data to High-Condence Variant Calls: e Genome Analysis Toolit Best Practices
Pipeline. Curr. Protoc. Bioinforma. 43, 11.10.1–11.10.33 (2013).
67. Danece, P. et al. e variant call format and VCFtools. Bioinformatics 27, 2156–2158 (2011).
68. Ye, J. et al. Primer-BLAST: A tool to design target-specic primers for polymerase chain reaction. BMC Bioinformatics 13, 134 (2012).
69. Zuer, M. Mfold web server for nucleic acid folding and hybridization prediction. Nucleic Acids Res. 31, 3406–3415 (2003).
Acknowledgements
This work was supported by FOD Volksgezondheid, Veiligheid van de Voedselketen en Leefmilieu (Cel
Contractueel Onderzoek) RT13/4. e computational resources (Stevin Supercomputer Infrastructure) used for
the analysis of the WES data were provided by the VSC (Flemish Supercomputer Center), funded by Ghent
University, FWO and the Flemish Government – department EWI. We would like to thank Sarah De Keulenaer,
Ellen De Meester and Sylvie Decraene from NXT GNT (http://www.nxtgnt.com/) for performing the library
preparation and sequencing, Dominique Vander Donckt and Linda Impe for performing the PCR and Sanger
sequencing in the population study, Ruben Van Gansbeke for the administrative support, Marleen Brunain for
DBR phenotyping and all beekeepers who participated in this study, especially Guido Haagdorens. DCdG and
his team are grateful for the availability of locations for experimental apiaries oered by the municipality of
Merelbeke and the faculties of Sciences, Veterinary Medicine, Economics and Business Administration, and Arts
and Philosophy of Ghent University.
Author Contributions
D.C.d.G., L.P., B.J.G.B., L.D.S. conceived and designed the experiments. B.J.G.B., L.D.S., T.B., K.M., M.K., M.V.P.,
B.D., P.N., K.B.N., G.S., D.D., F.V.N., L.P. and D.C.d.G. performed the experiments, analyzed and interpreted the
data. All authors contributed, improved and approved the paper.
Content courtesy of Springer Nature, terms of use apply. Rights reserved

11
SCIENTIFIC REPORTS | (2019) 9:7794 | https://doi.org/10.1038/s41598-019-44254-8
www.nature.com/scientificreports
www.nature.com/scientificreports/
Additional Information
Supplementary information accompanies this paper at https://doi.org/10.1038/s41598-019-44254-8.
Competing Interests: e authors declare no competing interests.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional aliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Cre-
ative Commons license, and indicate if changes were made. e images or other third party material in this
article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons license and your intended use is not per-
mitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
© e Author(s) 2019
Content courtesy of Springer Nature, terms of use apply. Rights reserved

1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com