Functional genomic map of local adaptation in sorghum to guide allele mining

Genomic data from genebanks could be exploited to find alleles adapted to target environments for resilience breeding, but it can be difficult to prioritize among the thousands of accessions and millions of genomic variants. There are competing hypotheses for the molecular basis and architecture of local adaptations: e.g. whether cis- regulatory versus amino acid changing variants are more important; or whether small-effect, low pleiotropy versus large-effect, high pleiotropy variants are more important. Here, we compare a range of variant types and genomic contexts thought to influence effect size, pleiotropy, and selection for their role in local adaptation in 443 whole genome resequenced African sorghum landraces. We used genotype-environment associations (GEAs) as evidence of local adaptation. We found that GEA were particularly enriched in the vicinity of genes and depleted elsewhere. However, enrichment was strongest in likely cis -regulatory contexts: accessible chromatin, unmethylated regions, and in transposable elements close to genes. Near genes, there were clear peaks in GEAs at the transcription start site, where mutations are demonstrated to have the largest expression effects. Additionally, GEAs in accessible chromatin and unmethylated regions were better predictors of genetic variation in response to experimental drought than comparable loci. Having tested hypotheses about the variants underlying local adaptation, we can now use this knowledge of the importance of cis -regulatory variation in the search for new environmentally-adaptive alleles for plant improvement.