Assessing the impact of pesticide exposure on honey bees (Apis mellifera) is crucial for understanding the drivers of pollinator decline and the mechanisms underlying environmental adaptation. Honey bees play a key role in ecosystem functioning and agricultural productivity, but they are increasingly threatened by chemical stressors that can affect their health, behaviour, and genetic resilience.
Directly measuring exposure across large geographic scales is often logistically and financially challenging; therefore, environmental proxies—such as modelled pesticide use grids and land-use data—could facilitate this type of analysis by providing spatially explicit estimates of potential stressors. These proxies allow researchers to approximate exposure across landscapes, integrate multiple sources of environmental information, and combine them with genomic data to identify genetic variants associated with environmental pressures, even when direct measurements are limited.
To quantify pesticide exposure, we used two complementary approaches. The first measured direct pesticide exposure using INSIGNIA-EU’s APIStrips (absorb pesticide in-hive strip) to quantify real-world pesticide residues. The second estimated the landscape-scale exposure integrating modelled pesticides-exposure grids from PEST-CHEMGRIDS to quantify the proportion of agricultural fields within 3km radius of each colony using Corine Land Cover.
Leveraging these measured and proxy-based pesticide exposures, we analysed 102 whole-genome sequences across 33 European countries as part of the Better-B project. To identify candidate single-nucleotide polymorphisms (SNPs), three complementary Genomic–Environment Association (GEA) approaches were applied: SAMBADA (a spatial analysis tool), LFMM (latent factor mixed models), and Redundancy Analysis (RDA).
Even though the APIStrip measurements yielded a greater number of SNPs, both approaches converged on key detoxification and stress-response genes, including a transcript variant of sushi von Willebrand factor type gene. This overlap highlights the biological relevance of proxy-based signals, even with reduced statistical power, and supports their utility for large-scale genomic studies where direct sampling is infeasible.
