Contrasting selection on GxE and main effect loci for soybean grain yield

In soybean, breeding programs are designed to select for maximizing yield across environments in the program's region. Genetic loci with positive effects across environments for yield (main effects) will have genetic diversity surrounding these loci reduced through strong positive selection. However, yield is a complex trait with a complex genetic architecture. In addition to loci with main effects, there are loci that have a significant interaction with the environment (GxE loci). We combined multienvironment yield trials with whole-genome resequencing to evaluate how artificial selection has shaped genomic regions underlying main and GxE variance. By integrating sliding window estimates of variance explained with Tajima's D and weighted Fst (fixation index), we quantified how diversity and divergence vary across effect types. Regions with higher diversity and lower divergence were significantly associated with greater GxE variance but not main variance. This pattern indicates relaxed selection at loci involved in environmental responsiveness. Conditionally neutral loci showed strong signatures of directional selection and contributed substantially to total genetic variance. Permutation-based comparisons revealed both positive and negative selection among GxE loci. Enrichment of beneficial alleles was strongest at conditionally neutral regions. These results reveal that elite soybean breeding has maintained adaptive potential at many GxE loci while intensifying selection on a subset of conditionally neutral effects. They also demonstrate that testing environments differ in the types of GxE effects they enrich. This provides breeders with a framework to understand how environmental conditions shape genetic responses for yield and other important traits.