Data from: Linking genotype to phenotype in a changing ocean: inferring the genomic architecture of a blue mussel stress response with genome-wide association
Kingston, Sarah E. et al. (2017), Data from: Linking genotype to phenotype in a changing ocean: inferring the genomic architecture of a blue mussel stress response with genome-wide association, Dryad, Dataset, https://doi.org/10.5061/dryad.2d8b5
A key component to understanding the evolutionary response to a changing climate is linking underlying genetic variation to phenotypic variation in stress response. Here we use a genome-wide association approach (GWAS) to understand the genetic architecture of calcification rates under simulated climate stress. We take advantage of the genomic gradient across the blue mussel hybrid zone (Mytilus edulis and Mytilus trossulus) in the Gulf of Maine (GOM) to link genetic variation with variance in calcification rates in response to simulated climate change. Falling calcium carbonate saturation states are predicted to negatively impact many marine organisms that build calcium carbonate shells - like blue mussels. We sampled wild mussels and measured net calcification phenotypes after exposing mussels to a “climate change” common garden, where we raised temperature 3°C, decreased pH by 0.2 units, and limited food supply by filtering out planktonic particles > 5 μm, compared to ambient GOM conditions in the summer. This climate change exposure greatly increased phenotypic variation in net calcification rates compared to ambient conditions. We then used regression models to link the phenotypic variation with over 170,000 single nucleotide polymorphism loci (SNPs) generated by genotype by sequencing to identify genomic locations associated with calcification phenotype, and estimate heritability and architecture of the trait. We identified at least one of potentially 2-10 genomic regions responsible for 30% of the phenotypic variation in calcification rates that are potential targets of natural selection by climate change. Our simulations suggest a power of 13.7% with our study's average effective sample size of 118 individuals and rare alleles, but a power of > 90% when effective sample size is 900.