Data from: Landscape and climate-associated selection in the native and widespread bumblebee, Bombus terrestris
Data files
Sep 29, 2025 version files 15.47 MB
-
Dataset_1_chunks.mcmc.xml
179.20 KB
-
Dataset_1_chunks.mcmcTree.xml
12.72 KB
-
Dataset_1_chunks.out
730.31 KB
-
Dataset_1_imputed.lfmm
4.07 MB
-
Dataset_1.lfmm
4.07 MB
-
README.md
3.87 KB
-
SNP_dataset_1.vcf.gz
545.28 KB
-
SNP_dataset_2.vcf
5.85 MB
Abstract
Anthropogenic land-use and climate change pose novel selection pressures on bees, yet their evolutionary responses in terms of morphological or physiological adaptations remain unclear. While adaptive responses are expected, these may be constrained by gene flow when changes in selection pressures are spatially heterogeneous. The buff-tailed bumblebee (Bombus terrestris) is a widespread species that copes well with anthropogenic land-use and climate change, suggesting high adaptive capacity or phenotypic plasticity. Here, we genotyped populations of native B. terrestris in south and central Sweden using RADseq to investigate genetic structure and local adaptation across a paired design of agricultural landscapes with high and low land-use complexity along a geographic climate gradient. We expected to find genetic structure reflective of regional barriers to gene flow, and molecular evidence for local adaptation to differing landscape and climate conditions. We found genetic structure separating southern Sweden from more northern regions, with a negative Tajima’s D indicating a potential population expansion, likely northwards and inland into forested areas, consistent with observational data indicating a range shift. We found weak but significant evidence for local adaptation to climate and land use, specifically to agricultural land cover, including genes under putative selection linked to insecticide resistance. Signatures of selection were also identified in relation to latitude, temperature, and urban land cover, with other candidate SNPs associated with olfaction and immune response. Our results suggest that B. terrestris successfully responded to anthropogenic land-use and climate changes, likely due to its generalist traits, enabling phenotypic adaptation to changing environments.
Dataset DOI: 10.5061/dryad.f4qrfj76b
Description of the data and file structure
Wild B. terrestris workers were sampled from 19 study sites, organized into nine matched pairs based on landscape complexity (n=9), in addition to one unmatched study site. Within each pair, landscapes were characterized by either low or high availability of open semi-natural habitat (simple versus complex landscapes, respectively), and study site pairs were arranged along a temperature gradient spanning southern and central Sweden. Collected B. terrestris were extracted for DNA using a custom HotSHOT protocol, followed by identification the COI mitochondrial gene and the PCR product was submitted to Macrogen Europe for Sanger sequencing. Samples identified as B. terrestris were re-extracted for DNA and prepared into genomic libraries for RAD sequencing at the Diversity Arrays Technology sequencing (DArTseq) facility (Canberra, Australia). Samples were individually sequenced on an Illumina Hiseq2500 by DArTseq.
Files and variables
File: SNP_dataset_1.vcf.gz
Description: Dataset 1 consist of 6702 SNPs and was generated for selection detection analyses. It was filtered with the dartR package and with the following parameters: monomorphic SNPs were removed, and the dataset was filtered for a depth of coverage between 10-200, followed by filtering of call rate by loci of >85% and removal of SNPs with <100% reproducibility. A minor allele frequency (MAF) filter of >0.01 was applied to reduce potential sequencing errors. This dataset was used for fineRADstructure, Latent Factor Mixed Modelling analysis (LFMM) and multivariate Redundancy Analysis (RDA).
File: SNP_dataset_2.vcf
Description: Dataset 2 consists of 4546 SNPs and was generated for neutral genetic structure analyses. It was filtered identical to Dataset 1 with the additional filters: SNPs were further filtered for being out of Hardy Weinberg (HWE), p-value threshold of 1e-6 using Plink v.1.9 and removed from the dataset. Lastly, SNPs were filtered for linkage disequilibrium using the --indep-pairwise 50 (window size), 5 (step size), and 0.5 (r^2 threshold) option in Plink v.1.9. The dataset was used for calculating the expected (HE), observed (HO) heterozygosity, allelic richness (AR), the inbreeding coefficient (FIS), pair-wise FST and Tajima’s D for each site. In addition, it was used for the Isolation-by-distance (IBD) analysis, PCA, DAPC, ADMIXTURE and EEMS analyses.
For both datasets, full-siblings has been removed from the dataset (--relatedness2 filter in VCFtools) and samples with high percentage of missing data (25 %).
These files below were generated from Dataset 1 for selection detection analyses and used for fineRADstructure analysis, Latent Factor Mixed Modelling analysis (LFMM) and multivariate Redundancy Analysis (RDA).
Input files for fineRADstructure:
File: Dataset_1_chunks.out
Description: RADpainter output file with the calculate co-ancestry matrix.
File: Dataset_1_chunks.mcmc.xml
Description: fineRADstructure output mcmc file which has the individuals assigned to populations.
File: Dataset_1_chunks.mcmcTree.xml
Description: final fineRADstructure output tree building file.
Input files for LFMM and RDA analysis:
Before running LFMM, missing genotypes were first imputed, using ancestry coefficients (snmf program), using the impute function within the LEA package.
File: Dataset_1.lfmm
Description: input file for imputing missing genotypes.
File: Dataset_1_imputed.lfmm
Description: output file from running the snmf program and the input file for the LFMM and RDA analyses.
The study design was based on 19 study sites, organized into nine matched pairs based on landscape complexity (n=9), in addition to one unmatched study site. Within each pair, landscapes were characterized by either low or high availability of open semi-natural habitat (simple versus complex landscapes, respectively), and study site pairs were arranged along a temperature gradient spanning southern and central Sweden.
Study site pairs were matched to represent similar temperatures across the temperature differences of the geographical gradient (from the southernmost to northernmost pairs). The design crossed three geographical regions in Sweden: Skåne (region 1), Västra Götaland (region 2) and Mälardalen (region 3).
We collected B. terrestris workers during two time points at each study site in the summer of 2019 (once in June and once in July). Bombus terrestris workers were sexed both in the field and subsequently confirmed in the lab by counting the number of segments on the flagellum of the antennae under a dissecting microscope (males 11 segments, females 10 segments after the pedicel). Only female worker bees were retained in the study.