Genome‐wide scans between two honeybee populations reveal putative signatures of human‐mediated selection

Summary

Human‐mediated selection has left signatures in the genomes of many domesticated animals, including the European dark honeybee, Apis mellifera mellifera, which has been selected by apiculturists for centuries. Using whole‐genome sequence information, we investigated selection signatures in spatially separated honeybee subpopulations (Switzerland, n = 39 and France, n = 17). Three different test statistics were calculated in windows of 2 kb (fixation index, cross‐population extended haplotype homozygosity and cross‐population composite likelihood ratio) and combined into a recently developed composite selection score. Applying a stringent false discovery rate of 0.01, we identified six significant selective sweeps distributed across five chromosomes covering eight genes. These genes are associated with multiple molecular and biological functions, including regulation of transcription, receptor binding and signal transduction. Of particular interest is a selection signature on chromosome 1, which corresponds to the WNT4 gene, the family of which is conserved across the animal kingdom with a variety of functions. In Drosophila melanogaster, WNT4 alleles have been associated with differential wing, cross vein and abdominal phenotypes. Defining phenotypic characteristics of different Apis mellifera ssp., which are typically used as selection criteria, include colour and wing venation pattern. This signal is therefore likely to be a good candidate for human mediated‐selection arising from different applied breeding practices in the two managed populations.

The Western honeybee, Apis mellifera, is the most economically valuable pollinator for agriculture (Gallai et al. 2009). Its domestication began at least 3000 years ago in the Near East (Crane 1999). Today managed honeybees are selected mostly for specific characteristics suitable for apiculture such as docility, productivity and swarming behaviour (Crane 1999). Another important criterion is the breeding of specific honeybee subspecies. Currently, more than 27 subspecies are recognized, differing in morphology and behaviour (Meixner et al. 2013). The European dark honeybee, Apis mellifera mellifera, has been selected by apiculturists for a few centuries based on various characteristics, including colour, hair length and wing morphology (Ruttner 1988). In particular, the cubital index, which measures the ratio between two vein segments that are split by a cross vein, is used for pure race breeding (Ruttner 1988). The pattern of the fore wing veins is heritable and specific for each breed of honeybees and therefore widely applied for breeding purposes (Ruttner 1988).

In a previous study, we identified genetic substructures in two geographically isolated A. m. mellifera populations from Switzerland and France (Parejo et al. 2016). The samples from Switzerland originated predominantly from conservatories, where conservation breeding efforts for A. m. mellifera began in the 1970s. The introduction of non‐native honeybees, such as the Carniolan bee, A. m. carnica, preferred by apiculturists due to their docile nature and higher productivity, threatens the genetic composition of the native type through introgression (Parejo et al. 2016). To distinguish native from introduced honeybees, breeders have typically referred to wing morphology, in particular the cubital index, although recently DNA‐testing based on microsatellite markers is increasingly being used. The second population originated from Savoy, France, from a conservation breeding centre established in 1997. Here, selection was much more recent and based mainly on wing morphological parameters. Genetic diversity of this population is slightly higher than in the Swiss population (Parejo et al. 2016), potentially relating to the longer selective pressure against introgression from introduced bees in Switzerland. Thus, putative signals of selection could relate to differences in current and historical breeding regimes and the efficiency of breeding efforts to purge introduced alleles.

We investigated selection signatures between the two subpopulations using 2 924 632 SNPs identified from whole‐genome sequence information of 56 A. m. mellifera drones (Switzerland, n = 39 and France, n = 17) (Parejo et al. 2016; see Table S1 for further information on sample origin and data availability). To confidently identify selection signals, we calculated three different test statistics in non‐overlapping windows of 2 kb: (i) fixation index, F _ST (Weir & Cockerham 1984), (ii) cross‐population extended haplotype homozygosity, XP‐EHH (Sabeti et al. 2007), and (iii) cross‐population composite likelihood ratio, XP‐CLR (Chen et al. 2010) (Fig. S1). Subsequently, we estimated the composite selection score (CSS) following Randhawa et al. (2014), which combines different test statistics based on a joint fractional rank. Finally, we applied a false discovery rate (FDR = 0.01) to CSS. Details on the calculations of these test statistics can be found in Appendix S1.

In total, we identified six putative sweep regions (A–F) distributed across five chromosomes including eight genes (Fig. 1, Table 1). Unfortunately, the honeybee genome is still not very well annotated, such that in sweep regions A, E and F only uncharacterized loci are located.

Figure 1.

Analysis of genome‐wide selection signatures identified six putative sweep regions (A–F). Composite selection scores combined F_ST, XP‐EHH and XP‐CLR test statistics in windows of 2 kb across the 16 honeybee chromosomes. The grey dashed line indicates the genome‐wide significance threshold (false discovery rate = 0.01).

Table 1.

Candidate genes located within and around (±2 kb) the six sweep regions

Sweep region	BeeBase/Gene ID	Gene name	Molecular function	Biological function	Cellular component	Drosophila ortholog
A	GB53211/LOC100576626	Uncharacterized LOC100576626
	GB54829/LOC408614	Uncharacterized LOC408614		Neuropeptide signaling pathway	Secretory granule	FBgn0041707, 7B2
B	GB44787/WNT4 also known as LOC552374	Wnt family member 4, also known as protein Wnt‐1	Receptor binding	Multicellular organism development, Wnt signaling pathway	Extracellular region, proteinaceous extracellular matrix	FBgn0010453, WNT4
C	GB55364/LOC725294	Tyrosine‐protein phosphatase 99A‐like	Protein tyrosine phosphatase activity, protein binding	Protein dephosphorylation	Integral component of membrane	FBgn0004369, PTP99A
D	GB50478/LOC724717	Uncharacterized LOC724717 (Tyrosine‐protein phosphatase gamma)			Integral component of membrane	FBgn0038749, Xport‐A
	GB50402/LOC412801	Mediator of RNA polymerase II transcription subunit 12‐like	Guanyl‐nucleotide exchange factor activity	Small gtpase mediated signal transduction	Intracellular	FBgn0037188
E	GB55726/LOC107964680	Uncharacterized LOC107964680
F	GB47148/LOC552326	Uncharacterized LOC552326				FBgn0032913

Yet, of particular note is the most significant window on chromosome 1 (sweep region B), which covers the WNT4 gene. Acting as intercellular signals, wnt proteins confer polarity and asymmetry to cells that are proliferating and thereby give shape to tissues (Loh et al. 2016). In Drosophila melanogaster, WNT4 alleles have been associated with differential wing, wing hair, wing margin bristle, cross vein and abdominal phenotypes (Swarup & Verheyen 2012; Gramates et al. 2016). To have a better idea of the differences in this region between the two populations, we further examined the haplotype block in the most significant window and compared that to haplotypes found in other honeybee subspecies. Forty‐one SNPs describing four major haplotype blocks were found within this window (Fig. S2b). Even though sample size in the French population was lower (n = 17), we observed higher haplotype diversity (Fig. S2a), whereas in the Swiss population one haplotype (1a) was dominant (28 out of 39 drones). Moreover, we identified one additional haplotype (5Clin) in the French population that is predominantly present in C‐lineage bees (Parejo et al. 2016; D. Wragg & A. Vignal, unpublished data). These findings suggest that conservation breeding efforts have not entirely purged all foreign alleles from the French population, whereas the longer and intensive selective pressure in the Swiss population has led to reduced haplotype diversity. Given that the honeybee has been selected on wing morphological characteristics, the signal found in this gene could thus be human‐mediated and a result of differently applied breeding practices within the two subpopulations.

Among the eight genes located in the significant sweep regions, two [LOC725294 (GB55364) and LOC724717 (GB50478)] encode tyrosine‐protein phosphatases and are found on two different chromosomes. These enzymes are key regulatory components in signal transduction pathways by regulating enzyme activity and controlling cell growth and differentiation (Tonks 2006). Therefore, given their relevance, the selection signal found in the two tyrosine‐protein phosphatase genes have the potential to manifest in differential phenotypes. However, further research is needed to identify the trait(s) associated with these genes in the honeybee.

Finally, sweep region D also entails a gene [LOC412801 (GB50402)] that is involved in the regulation of transcription of RNA polymerase II‐dependent genes. Acting as a co‐activator in the mediator complex, it is vital to regulatory mechanisms with a broad and dynamic range of functions (Malik & Roeder 2010).

In conclusion, we identified six sweep regions across the genome including eight genes, of which four have unknown functions and four are annotated for important molecular and biological functions. Collectively, these findings suggest that differential selective pressures are acting on these genes in these two closely related populations. However, it needs to be mentioned that selection signature analyses can reveal only putative candidate genes whose functional relevance on phenotypic differences remains to be tested. The strongest selection signal was found in WNT4, a gene affecting wing vein patterns in D. melanogaster. In addition, A. m. mellifera has been intensely selected on wing veins for decades. This further evidence in the case of WNT4 makes it a plausible candidate gene for wing venation patterns in A. mellifera and an exemplification of human‐mediated selection in the Western honeybee.

Supporting information

Figure S1 Manhattan plots of the three employed statistics to infer selection signatures (F _ST, XP‐EHH, and XP‐CLR) in windows of 2 kb.

Figure S2 (a) Haplotype distribution of sweep region B in the Swiss and French populations, and (b) corresponding haplotype blocks per individual.

Table S1 Further information on sample origin and data availability.

Appendix S1 Materials and methods.